
Design of Machine Elements
Size a shaft by the ASME code, find the critical bolt in an eccentric bracket, put a fluctuating stress on the Goodman diagram, design a spring that will not surge and pick a bearing for 20,000 hours
26 chapters, 26 interactives, few words. Every chapter is mapped to the Anna University, VTU, SPPU, Mumbai University and KTU units it covers: the design process and preferred numbers, factor of safety and the theories of failure, eccentric loading and curved beams, stress concentration, fatigue and the endurance limit, Soderberg, Goodman and Gerber, shafts by strength, rigidity and critical speed, keys and couplings, cotter and knuckle joints, power screws, bolted, welded and riveted joints, helical and leaf springs, flywheels, connecting rods, belts and chains, spur, helical, bevel and worm gears, clutches and brakes, journal bearings, rolling bearings, and thick cylinders, seals and gaskets.
- AUME3591Design of Machine Elements
- VTUBME602Machine Design
- SPPU302043Design of Machine Elements
- MUMEC601Machine Design
- KTUPCMET602Machine Design
By Skill-Lync · Updated 17 Nov 2026 · 26 chapters · 26 interactives · free, no sign-up
The design process, materials, standards and preferred numbers
Which syllabus covers thisAU ME3591 Unit IVTU BME602 Module 1SPPU not in syllabusMU MEC601 Module 1KTU PCMET602 Module 1
What are the steps in designing a machine element, and what kinds of design are there?
A design starts with a need, which is turned into a clear problem statement with loads, speeds, life, space and cost. Next come synthesis (sketching a mechanism or arrangement), analysis of forces and stresses, choice of material, sizing of each part, and detailed drawings with tolerances. A prototype is then tested and the design revised before production. Most work is adaptive design (a known product adapted to new loads), some is developmental (a known principle with new materials or methods) and a little is new design. Sizes come from rational calculation, from empirical rules, or from both.
How do aesthetics, ergonomics and manufacturing shape a design?
Aesthetics decides how the product looks: proportion, colour and clean lines help a machine sell. Ergonomics fits it to the user: handle heights, lever forces a hand can apply, readable dials and controls placed where they are expected. Manufacturing decides details. A casting needs uniform wall thickness, generous fillets so the metal cools evenly, draft on faces so the pattern comes out, and no isolated thick sections that shrink and leave cavities. A forging needs a sensible parting line, draft, and no thin, deep ribs. Designing for the process from the start is cheaper than fixing it later.
How is a material chosen for a machine part?
List what the part must do and pick properties to match: strength (yield and ultimate), stiffness (modulus), toughness for shock, hardness for wear, fatigue strength for cyclic loads, plus cost, machinability, weldability and availability. Plain carbon steels suit most shafts and pins; alloy steels suit highly stressed gears; grey cast iron suits frames and housings because it is cheap, casts well and damps vibration, but it is weak in tension. Indian designations follow IS standards: for example 40C8 is a plain carbon steel with about 0.4 % carbon and 0.8 % manganese, and FG 200 is grey cast iron with a minimum tensile strength of 200 MPa.
Why do designers use standards and codes?
A standard fixes sizes, materials, tests and methods so that parts from different makers fit and behave alike. Bolts, keys, bearings, gears and steels are bought to standard sizes, which makes them cheap and replaceable. Codes add design rules for safety, such as the ASME code for shafts and pressure vessels. In India the Bureau of Indian Standards publishes IS standards, many identical to ISO ones: IS 1076 Part 1 on preferred numbers is identical to ISO 3. A design that uses standard parts and standard sizes is quicker to make, cheaper to stock and easier to service.
What are preferred numbers, and how do you use them?
Preferred numbers are the Renard series of ISO 3. Series Rn splits each decade into n equal ratios of 10^(1/n): R5 steps by about 1.58, R10 by 1.26, R20 by 1.12 and R40 by 1.06, with the values rounded (R10: 1, 1.25, 1.6, 2, 2.5, 3.15, 4, 5, 6.3, 8). Because the steps are ratios, a range of sizes or capacities covers the field evenly with few models. Round a computed size up to the next preferred number, never down. In the figure, 37 mm becomes 40 mm in R10 or R20, but 37.5 mm in R40.
A strength calculation gives a 37 mm dimension. Which R20 preferred size do you specify?
R20 × 10 runs 31.5, 35.5, 40. Rounding down to 35.5 would leave the part under strength, and 37.5 belongs to R40, not R20. So 40 mm, 8.1 % above the computed size.
Worked examples
Rounding a computed size up · A calculated shaft step diameter is 37 mm. Find the next preferred size in R10, R20 and R40 and the oversize in each
- Decade 10 to 100: multiply the series values by 10
- R10: 31.5, 40 → 40 mm; R20: 35.5, 40 → 40 mm; R40: 35.5, 37.5 → 37.5 mm
- Oversize: R10 and R20 40/37 − 1 = 8.1 %; R40 37.5/37 − 1 = 1.35 %
- Rounding down to 35.5 mm would leave the part 4.1 % under size, so always go up
Answer: 40 mm in R10 or R20 (+8.1 %), 37.5 mm in R40 (+1.35 %)
A range of machine sizes · A maker wants a range of hydraulic presses from 10 t to 100 t with as few models as possible but even coverage. Use R5
- R5 values × 10: 10, 16, 25, 40, 63, then 100
- Step ratios: 16/10 = 1.6, 25/16 = 1.5625, 40/25 = 1.6, 63/40 = 1.575, 100/63 = 1.587
- Exact R5 ratio 10^(1/5) = 1.5849; the product of the five steps is exactly 10
- R10 would give 11 models (10, 12.5, 16, ... 100) with steps of about 1.26
Answer: Six presses: 10, 16, 25, 40, 63 and 100 t, each about 1.6 times the last
How rough is the rounding, and derived series · Compare each R10 value with its exact value 10^(i/10), then form the derived series R10/3
- Largest deviation: 1.6 against 10^0.2 = 1.5849, +0.95 %
- 3.15 against √10 = 3.1623: −0.39 %
- Every R10 value is within 1 % of the exact geometric value
- R10/3 takes every third term: 1, 2, 4, 8, 16 ...; exact step 10^0.3 = 1.9953, practically doubling
Answer: Rounding moves no R10 value by more than 0.95 %; R10/3 gives a doubling series 1, 2, 4, 8
Factor of safety and the theories of failure
Which syllabus covers thisAU ME3591 Unit IVTU BME602 Module 1SPPU 302043 Unit 1MU MEC601 Module 1KTU PCMET602 Module 1
In what ways can a machine part fail?
A part fails when it can no longer do its job, which is not always breaking. Ductile parts fail by yielding: permanent set that ruins fits and alignment. Brittle parts, such as cast iron, fracture with little warning. Slender parts buckle; shafts can deflect too much for gears and bearings to work; surfaces wear, pit or seize; parts under cyclic load fail by fatigue at stresses well below yield; hot parts creep. A ductile material in tension shows yield before it breaks, while a brittle one breaks near its ultimate strength, so the failure criterion depends on the material.
What are the factor of safety and the service factor?
The factor of safety n = failure stress / working stress. For ductile materials the failure stress is the yield strength Syt; for brittle ones the ultimate strength Sut. It covers scatter in strength, approximate loads, simplified analysis and the cost of failure. Typical values: about 1.5 to 2 for ductile steel under steady, well-known loads, 3 to 4 for brittle materials, higher under shock (typical ranges, not code values). A service factor scales the rated load instead: a 10 kW motor driving a machine with moderate shock at a typical service factor of 1.5 is designed for 15 kW.
Which theories predict yielding of ductile materials, and how do they differ?
Maximum shear stress (Tresca, Guest): yield when the largest shear stress reaches Syt/2, so n = Syt/(σmax − σmin) using all three principal stresses. Distortion energy (von Mises, Hencky): yield when the energy of shape change reaches its value in a tension test, so n = Syt/σ′ with σ′ = √(σ1² − σ1σ2 + σ2²) = √(σx² − σxσy + σy² + 3τxy²). On the σ1-σ2 plane Tresca is a hexagon inside the von Mises ellipse. They agree for uniaxial and equal biaxial stress and differ most in pure shear (0.5 Syt against 0.577 Syt). Tests on steel fit von Mises best; Tresca is simpler and slightly conservative.
Which theories suit brittle materials such as cast iron?
Brittle materials are far stronger in compression than in tension: grey cast iron is typically Sut 200 MPa and Suc 700 MPa. Maximum normal stress (Rankine) says fracture occurs when σ1 reaches Sut or σ2 reaches −Suc: a square on the σ1-σ2 plane, unsafe when one stress is tensile and the other compressive. Coulomb-Mohr joins (Sut, 0) and (0, −Suc) with a straight line in that quadrant: 1/n = σA/Sut − σB/Suc. Modified Mohr keeps the Rankine line down to σB = −Sut, then runs to (0, −Suc); it matches cast-iron tests best and is the usual choice.
What are the strain theories, and which theory should you use?
Maximum principal strain (Saint-Venant) says failure occurs when the largest strain equals the strain at yield in a tension test. Total strain energy (Haigh, Beltrami) uses all the stored energy: σ1² + σ2² − 2νσ1σ2 = Syt², an ellipse that depends on Poisson's ratio. Both are mainly of historical interest: they predict that hydrostatic pressure, which stores energy but causes no yield, would cause failure. Rule of thumb: ductile parts, distortion energy (or maximum shear for a quick conservative check); brittle parts, modified Mohr or Coulomb-Mohr.
A ductile steel element (Syt = 300 MPa) is in pure shear, τ = 100 MPa. Which theory gives the larger factor of safety?
Tresca: σ1 = 100, σ2 = −100, n = 300/200 = 1.5. Von Mises: σ′ = √3 × 100 = 173.2 MPa, n = 1.73. Pure shear is where the hexagon sits furthest inside the ellipse.
Worked examples
Shaft surface under bending and torsion · At the surface of a ductile steel shaft (Syt = 300 MPa, typical) σx = 80 MPa from bending and τxy = 60 MPa from torsion. Find n by maximum shear stress and by distortion energy
- σ1, σ2 = 40 ± √(40² + 60²) = 40 ± 72.11 → 112.11 and −32.11 MPa (σ3 = 0)
- Tresca: σmax − σmin = 112.11 − (−32.11) = 144.22 MPa; n = 300/144.22 = 2.08
- Von Mises: σ′ = √(80² + 3 × 60²) = √17,200 = 131.15 MPa; n = 300/131.15 = 2.29
Answer: n = 2.08 by maximum shear stress and 2.29 by distortion energy
Cast-iron bracket: three brittle theories · Grey cast iron, Sut = 200 MPa and Suc = 700 MPa (typical), with σx = 50, σy = −150 and τxy = 40 MPa
- Centre −50, radius √(100² + 40²) = 107.70: σA = 57.70, σB = −157.70 MPa
- Maximum normal: n = min(200/57.70, 700/157.70) = 3.47
- Coulomb-Mohr: 1/n = 57.70/200 + 157.70/700 = 0.5138, n = 1.95
- Modified Mohr: |σB| > σA, so 1/n = (700 − 200)(57.70)/(700 × 200) + 157.70/700 = 0.4314, n = 2.32
Answer: Rankine 3.47 (unsafe guess), Coulomb-Mohr 1.95, modified Mohr 2.32
Four theories, one stress state · Ductile steel, Syt = 300 MPa, ν = 0.3 (typical), principal stresses σ1 = 150 and σ2 = −100 MPa
- Maximum normal: n = 300/150 = 2.00
- Maximum shear: n = 300/(150 + 100) = 1.20
- Distortion energy: σ′ = √(150² + 150 × 100 + 100²) = 217.9; n = 1.38
- Strain energy: √(150² + 100² + 2 × 0.3 × 150 × 100) = 203.7; n = 1.47
Answer: Maximum normal stress gives 2.00, far from the yield-based answers (1.20 to 1.47): with stresses of opposite sign, never use it for a ductile part
Combined and eccentric loading, crane hooks and curved beams
Which syllabus covers thisAU ME3591 Unit IVTU BME602 Module 1SPPU 302043 Unit 1MU not in syllabusKTU PCMET602 Module 1
How are direct and bending stresses combined?
When a load P acts along the axis, every fibre carries P/A. A bending moment M adds My/I, tension on one side and compression on the other. Both are normal stresses on the same section, so they simply add: σ = P/A ± M y/I. The extreme fibre on the tension side governs a ductile part; for a brittle part check the compressive side as well. For a rectangle, a compressive load stays inside the middle third (e ≤ h/6) if the whole section is to stay in compression, which matters for masonry and cast-iron columns.
What happens when normal and shear stresses act together?
Bending plus torsion on a shaft, or a bracket loaded off its axis, gives a normal stress σ and a shear stress τ at the same point. Neither alone tells you whether the part is safe. The principal stresses σ1, σ2 = σ/2 ± √((σ/2)² + τ²) and the maximum shear stress √((σ/2)² + τ²) do; they come from Mohr's circle (see the Strength of Materials book). Feed them to a theory of failure: for a ductile shaft, von Mises gives σ′ = √(σ² + 3τ²).
How do you design a part loaded off its axis, such as a C-frame or an offset link?
Move the load to the centroid of the critical section and add the moment it creates: a load P at a distance e becomes P through the centroid plus a moment P e. The section then carries σ = P/A ± P e c/I. Offsetting a 10 kN pull by 60 mm on a 20 × 40 mm link raises the peak from 12.5 MPa to 125 MPa, ten times higher. That is why punch frames, clamps and offset links are deep in the plane of the moment and why designers keep the line of action close to the centroid wherever they can.
Why is a curved beam different from a straight one?
In a curved bar the inner fibres are shorter than the outer ones, so the same rotation of a section strains them more. The bending stress is hyperbolic, not linear: σ = M(rn − r)/(A e r). The neutral axis moves towards the centre of curvature, to rn = A/∫dA/r, a distance e = R − rn inside the centroid. The inner fibre carries much more stress than straight-beam theory predicts, the outer fibre less. The figure shows the stress across the section, the shifted neutral axis, and the straight-beam line for comparison.
How is a crane hook analysed, and why is its section trapezoidal?
The load hangs from the inside of the hook on a line through the centre of curvature. Cut the hook horizontally through that centre: the section carries a direct tension W and a moment W × R, where R is the radius of the centroid. That moment opens the hook, so the inner fibre takes direct plus bending tension, the largest stress in the hook. A trapezoid with its wide side inside puts more material where the stress is highest and moves the centroid inwards, which lowers the moment arm R; I-sections and T-sections do the same in large frames.
If you used the straight-beam formula W/A + Mc/I for the inner fibre of a crane hook, how would it compare with the curved-beam value?
Curvature concentrates bending stress on the short inner fibres. For the hook in example 2 the straight-beam value is 96 MPa but the curved-beam value is 129 MPa, 1.34 times higher.
Worked examples
Offset link · A link of 20 × 40 mm rectangular section (40 mm in the plane of the offset) carries a pull of 10 kN whose line of action is 60 mm from the centroid of the critical section
- A = 20 × 40 = 800 mm²; I = 20 × 40³/12 = 106,667 mm⁴; c = 20 mm
- Direct: 10,000/800 = 12.5 MPa; bending: 10,000 × 60 × 20/106,667 = 112.5 MPa
- σmax = 12.5 + 112.5 = 125 MPa (tension, side nearer the load); σmin = −100 MPa
Answer: Peak stress 125 MPa in tension and −100 MPa on the far edge, ten times the 12.5 MPa of an axial pull
Crane hook with a trapezoidal section · Critical section of a hook: trapezoid 60 mm wide inside, 20 mm outside, 90 mm deep; inner radius 60 mm; load 50 kN on a line through the centre of curvature
- A = (60 + 20) × 90/2 = 3600 mm²; R = 60 + 90(60 + 40)/(3 × 80) = 97.5 mm
- ∫dA/r = [(60 × 150 − 20 × 60)/90] ln(150/60) − 40 = 39.412 mm; rn = 3600/39.412 = 91.34 mm; e = 6.16 mm
- M = 50,000 × 97.5 = 4.875 × 10⁶ N·mm
- Inner: σi = 50,000/3,600 + M(91.34 − 60)/(3,600 × 6.16 × 60) = 128.8 MPa
- Outer: σo = 13.9 + M(91.34 − 150)/(3,600 × 6.16 × 150) = −72.1 MPa
- Straight-beam theory: inner 96.0, outer −101.0 MPa
Answer: Inner fibre 129 MPa tension, outer fibre −72 MPa compression; straight-beam theory underestimates the inner stress by 25 %
C-frame with a rectangular section · The curved back of a C-frame has a 30 mm wide × 60 mm deep rectangular section with inner radius 60 mm. The jaws push apart with 15 kN on a line 150 mm beyond the centre of curvature
- A = 1,800 mm²; R = 90 mm; rn = 60/ln(120/60) = 86.56 mm; e = 3.438 mm
- Moment arm R + 150 = 240 mm; M = 15,000 × 240 = 3.6 × 10⁶ N·mm
- σi = 15,000/1,800 + 3.6 × 10⁶ × (86.56 − 60)/(1,800 × 3.438 × 60) = 266 MPa
- σo = 8.33 + 3.6 × 10⁶ × (86.56 − 120)/(1,800 × 3.438 × 120) = −154 MPa
- Straight beam: 8.33 + 3.6 × 10⁶ × 30/540,000 = 208 MPa
Answer: Inner fibre 266 MPa against 208 MPa by straight-beam theory (1.28 times); outer fibre −154 MPa
Stress concentration and notch sensitivity
Which syllabus covers thisAU ME3591 Unit IVTU BME602 Module 1SPPU 302043 Unit 4MU MEC601 Module 3KTU PCMET602 Module 1
What is the theoretical stress concentration factor Kt?
Any abrupt change of section, such as a shoulder, hole, groove, keyway or thread, disturbs the flow of stress, and the stress peaks at the root of the change. Kt = peak elastic stress / nominal stress. It depends only on geometry and the type of loading, not on the material. The nominal stress must be defined with it, usually on the reduced (net) section. A small round hole in a very wide plate under tension has Kt = 3; an elliptical hole or a sharp crack is far worse, because Kt rises as the root radius shrinks.
Where do Kt values come from, and what controls them?
From elasticity solutions, photoelastic tests and finite element studies, published as charts (Peterson) and curve fits (Pilkey, Roark). For a stepped shaft the two ratios that matter are r/d, the fillet radius over the smaller diameter, and D/d, the size of the step. A bigger fillet radius always helps; a deeper step makes things worse until D/d is about 2. For a plate with a central hole Kt falls from 3 for a tiny hole to about 2 as the hole fills the width, measured on the net section. The figure uses Pilkey's fits.
What is notch sensitivity, and how is Kf found?
Under fatigue loading a real material feels less than the full Kt, especially at small radii, because the stress gradient is steep and a small volume is highly stressed. Notch sensitivity q runs from 0 (no effect) to 1 (full Kt), and the fatigue stress concentration factor is Kf = 1 + q(Kt − 1). The Neuber (Kuhn-Hardrath) form q = 1/(1 + √a/√r) uses a material constant √a that falls as Sut rises: high-strength steels are more notch sensitive. For 600 MPa steel √a is about 0.38 √mm, so a 4 mm fillet gives q ≈ 0.84.
When must Kt or Kf be applied?
Ductile material, static load: usually ignore Kt, because the small peak region yields and passes load to its neighbours before the part as a whole yields. Brittle material, static load: apply Kt, since nothing yields to relieve the peak. Any material, fatigue load: apply Kf to the alternating stress (and, in common practice, to the mean stress). Grey cast iron is an exception: its graphite flakes already act as internal notches, so an extra notch adds little and q is low.
How can a designer reduce stress concentration?
Use the largest fillet radius the mating part allows; where a bearing needs a sharp shoulder, use an undercut fillet or a relief groove so the radius can be larger. Add relief notches or extra small holes on either side of an unavoidable notch so the stress flows round it more smoothly. End keyways with a sled-runner shape instead of a sharp end-milled step. Chamfer or radius the edges of holes. Put notches where the stress is low. Then add compressive residual stress by shot peening or rolling the fillet.
A small circular hole is drilled in a very wide plate under 100 MPa tension. What is the peak stress at the edge of the hole?
For a small hole in a wide plate Kt = 3 (Kirsch's solution), so the edge of the hole on the line across the load carries 3 × 100 = 300 MPa, whatever the hole's size.
Worked examples
Stepped shaft in bending · A shaft steps from 60 mm to 40 mm with a 4 mm shoulder fillet and carries a bending moment of 400 N·m. Steel Sut = 600 MPa
- σnom on the small diameter = 32 × 400,000/(π × 40³) = 63.66 MPa
- Step h = (60 − 40)/2 = 10 mm; h/r = 2.5; 2h/D = 0.333. Pilkey's fit (2 ≤ h/r ≤ 20): Kt = 1.743
- Neuber constant at 600 MPa: √a = 0.377 √mm; q = 1/(1 + 0.377/√4) = 0.842
- Kf = 1 + 0.842(1.743 − 1) = 1.625
Answer: Peak elastic stress 111 MPa (Kt = 1.74); for fatigue use Kf = 1.63, i.e. 103 MPa
Plate with a central hole · A plate 80 mm wide and 10 mm thick with a 20 mm central hole carries a 60 kN pull
- d/W = 20/80 = 0.25; Kt = 3 − 3.14(0.25) + 3.667(0.25)² − 1.527(0.25)³ = 2.420
- Check with Heywood's formula 2 + (1 − d/W)³ = 2.422
- Net-section nominal stress = 60,000/((80 − 20) × 10) = 100 MPa (gross section would give 75 MPa)
- σmax = 2.420 × 100 = 242 MPa
Answer: Peak stress 242 MPa at the hole edge; using the gross-section stress by mistake would give 182 MPa, an underestimate
Strong steel, sharp notch · Two steels, Sut = 400 and 1,200 MPa, each with a notch of Kt = 2.0, at root radii of 1 mm and 4 mm
- √a: 0.569 √mm at 400 MPa; 0.134 √mm at 1,200 MPa
- r = 1 mm: q = 0.637 and 0.882; Kf = 1.64 and 1.88
- r = 4 mm: q = 0.778 and 0.937; Kf = 1.78 and 1.94
Answer: The 1,200 MPa steel feels almost all of Kt (Kf 1.88 even at r = 1 mm), so stronger steels need smoother fillets to gain fatigue strength
Fatigue, the S-N curve and the corrected endurance limit
Which syllabus covers thisAU ME3591 Unit IVTU BME602 Module 1SPPU 302043 Unit 4MU MEC601 Module 3KTU PCMET602 Module 1
How does a part fail in fatigue?
Repeated stress, even well below yield, causes slip in a few badly oriented grains at the surface, usually at a notch, scratch or inclusion. After many cycles a crack starts there, then grows a little each cycle, leaving beach marks and fine striations on the fracture face. When the remaining section can no longer carry the load it breaks suddenly, leaving a rough final-fracture zone. There is no visible warning and no gross yielding, which is why most broken shafts, springs, bolts and crankshafts are fatigue failures.
What does the S-N curve show?
Test specimens at a fully reversed stress amplitude S and count the cycles N to failure; plot S against log N. Below about 10³ cycles is low-cycle fatigue, where the stress is near the ultimate strength and plastic strain controls life. Above that is high-cycle fatigue. Steels show a knee near 10⁶ to 10⁷ cycles; below that stress, the endurance limit, life is practically infinite. Aluminium alloys have no knee, so designers quote a fatigue strength at a stated life such as 5 × 10⁸ cycles. The Engineering Materials book shows the curves for both.
How is the endurance limit of a real part estimated?
Start from the rotating-beam value of a small polished specimen: for steel Se′ ≈ 0.5 Sut, up to 700 MPa for Sut above 1,400 MPa. Then multiply by the Marin factors: surface ka = a Sut^b (rough surfaces and strong steels lose most), size kb (1.24 d^−0.107 up to 51 mm, 1 in axial loading), load kc (1 bending, 0.85 axial, 0.59 torsion), temperature kd, reliability ke (0.897 for 90 %, 0.814 for 99 %). A machined 40 mm shaft of 600 MPa steel at 90 % reliability keeps only 186 of the 300 MPa.
How is the life found when the stress is above the endurance limit?
Between 10³ and 10⁶ cycles the S-N curve of steel is close to a straight line on log-log axes. Join f Sut at 10³ cycles (f = 0.9 is the usual estimate for steel) to the corrected Se at 10⁶. The line is S = a N^b with a = (f Sut)²/Se and b = −(1/3) log(f Sut/Se). Invert it for the life at an amplitude σa: N = (σa/a)^(1/b). Because the line is shallow, a small rise in stress cuts life sharply: 10 % more stress cuts the life almost in half.
How are shock and impact loads handled?
A load applied suddenly, even from zero height, produces twice the static deflection and stress. A weight W dropped from a height h onto a stop produces a deflection δ = δst(1 + √(1 + 2h/δst)), from the energy balance W(h + δ) = kδ²/2, and the stress rises by the same impact factor. A stiff part has a small δst and so a large impact factor: making a bolt or rod more flexible, for example by turning its shank down to the thread root, lowers the impact stress. In design, shock is also covered by service factors on the load.
An as-forged steel part is upgraded from Sut = 600 MPa to Sut = 1,200 MPa. What happens to its corrected endurance limit?
Se′ doubles, but the forged-surface factor ka = 272 Sut^−0.995 nearly halves. ka × Se′ goes from 140.4 to 140.9 MPa. A stronger steel only pays off in fatigue if the surface is machined or ground.
Worked examples
Corrected endurance limit of a shaft · A machined 40 mm steel shaft, Sut = 600 MPa, rotating bending, room temperature, 90 % reliability
- Se′ = 0.5 × 600 = 300 MPa
- ka = 4.51 × 600^−0.265 = 0.828; kb = 1.24 × 40^−0.107 = 0.836
- kc = 1 (bending); kd = 1 (room temperature); ke = 0.897 (90 %)
- Se = 0.828 × 0.836 × 1 × 1 × 0.897 × 300 = 186.2 MPa
Answer: Se ≈ 186 MPa, only 62 % of the polished-specimen value
Finite life above the endurance limit · The shaft of example 1 runs at a fully reversed stress amplitude of 250 MPa (Kf already included)
- 10³-cycle strength: 0.9 × 600 = 540 MPa
- a = 540²/186.2 = 1,566 MPa; b = −(1/3) log(540/186.2) = −0.1542
- N = (250/1,566)^(1/−0.1542) = 1,47,697 cycles
- At 3,000 rpm that is 49 minutes of running
Answer: About 1,48,000 cycles: a 34 % overstress above Se cuts life from infinite to under an hour at 3,000 rpm
Falling weight on a rod · A 500 N weight slides down a vertical steel rod (1 m long, 20 mm diameter, E = 207 GPa, typical) and drops 20 mm onto a collar at its lower end
- A = π × 20²/4 = 314.2 mm²; static stress W/A = 1.59 MPa
- δst = WL/(AE) = 500 × 1,000/(314.2 × 207,000) = 0.00769 mm
- Impact factor = 1 + √(1 + 2 × 20/0.00769) = 73.1
- Impact stress = 73.1 × 1.59 = 116 MPa; extension 0.562 mm
- Same load applied suddenly (h = 0): factor 2, stress 3.18 MPa
Answer: A 20 mm drop raises the stress 73 times, from 1.59 to 116 MPa
Fluctuating stresses: Soderberg, Goodman, Gerber and cumulative damage
Which syllabus covers thisAU ME3591 Unit IVTU not in syllabusSPPU 302043 Unit 4MU MEC601 Module 3KTU PCMET602 Module 1
How is a fluctuating stress described?
Any cycle between σmax and σmin is a steady mean stress plus an alternating amplitude: σm = (σmax + σmin)/2 and σa = (σmax − σmin)/2. Fully reversed bending of a rotating shaft has σm = 0; a bolt or spring that never unloads has a large mean and a small amplitude. The stress ratio R = σmin/σmax is −1 for fully reversed and 0 for repeated (zero to maximum) loading. For a notched ductile part, multiply both components by Kf before going to a design diagram (common practice).
How do Soderberg, Goodman and Gerber account for the mean stress?
A tensile mean stress lowers the amplitude a part can survive. On the σm-σa diagram each criterion joins Se on the σa axis to a static strength on the σm axis. Soderberg runs straight to Syt: safe against fatigue and yield, but conservative. Goodman runs straight to Sut: the usual design line. Gerber is a parabola to Sut, which follows test data for steels best. ASME-elliptic is a quarter ellipse to Syt. For a load line from the origin, n is how far the point can move before it reaches the line.
What is the modified Goodman diagram?
Goodman alone ignores yield: with a high mean stress the point can sit inside the Goodman line yet beyond Syt on the first cycle. The modified Goodman diagram adds the Langer line σa + σm = Syt, so the safe region is bounded by Goodman at low mean stress and by yield at high mean stress. The design factor is the smaller of the two: n = min(nGoodman, Syt/(σa + σm)). Compressive mean stresses help fatigue, so on that side only yield and the endurance limit limit the design.
How are combined fluctuating stresses handled?
A shaft often sees reversed bending with a steady torque, so its alternating and mean parts mix normal and shear stresses. Combine each set separately with von Mises: σa′ = √(σa² + 3τa²) and σm′ = √(σm² + 3τm²), each with its own fatigue factor already applied. Then use σa′ and σm′ on the Goodman (or Gerber) diagram as if they were simple stresses. Check first-cycle yield with the maximum von Mises stress, √((σa + σm)² + 3(τa + τm)²).
What is Miner's rule for varying amplitudes?
Real parts see blocks of different amplitudes. Palmgren-Miner assumes each cycle at amplitude σi uses up 1/Ni of the life, where Ni is the life at that amplitude from the S-N line. Failure is predicted when the damage D = Σ ni/Ni reaches 1. Cycles below the endurance limit are taken to do no damage. It ignores the order of loading (a high-then-low sequence is in fact worse), so designers often limit D to below 1, for example 0.7, as a margin.
A stress cycles between 50 and 250 MPa. What are its mean stress and amplitude?
σm = (250 + 50)/2 = 150 MPa and σa = (250 − 50)/2 = 100 MPa; the stress ratio is 50/250 = 0.2.
Worked examples
One cycle, five criteria · A notched steel link sees 50 to 250 MPa (Kf already applied). Sut = 600, Syt = 420 MPa (typical), corrected Se = 200 MPa
- σm = 150, σa = 100 MPa
- Soderberg: 1/n = 100/200 + 150/420 → n = 1.17
- Goodman: 1/n = 100/200 + 150/600 = 0.75 → n = 1.33
- Gerber: (150n/600)² + 100n/200 = 1 → 0.0625n² + 0.5n − 1 = 0 → n = 1.66
- ASME-elliptic: n = 1/√(0.5² + (150/420)²) = 1.63; first-cycle yield: 420/250 = 1.68
Answer: Design n = 1.33 by modified Goodman (Goodman 1.33 is below the yield value 1.68); Soderberg 1.17, Gerber 1.66
Bending and steady torque together · A rotating shaft carries fully reversed bending σa = 80 MPa and steady torsion τm = 60 MPa (fatigue factors applied). Sut = 600, Syt = 420, Se = 200 MPa
- Alternating von Mises: σa′ = √(80² + 3 × 0²) = 80 MPa
- Mean von Mises: σm′ = √(0² + 3 × 60²) = 103.92 MPa
- Goodman: 1/n = 80/200 + 103.92/600 = 0.5732 → n = 1.74
- First-cycle yield: σmax′ = √(80² + 3 × 60²) = 131.1 MPa; n = 420/131.1 = 3.20
Answer: n = 1.74 against fatigue (Goodman) and 3.20 against yield: fatigue governs
Miner's rule over a duty cycle · A steel part (Sut = 600 MPa, Se = 200 MPa) sees, in each block: 2,000 cycles at 400 MPa, 10,000 at 300 MPa and 30,000 at 250 MPa (fully reversed). Use the S-N line from 0.9 Sut at 10³ to Se at 10⁶
- a = 540²/200 = 1,458 MPa; b = −(1/3) log(540/200) = −0.1438
- N at 400, 300, 250 MPa: 8,062, 59,613, 2,11,846 cycles
- Damage per block: 2,000/8,062 + 10,000/59,613 + 30,000/2,11,846 = 0.2481 + 0.1677 + 0.1416 = 0.5574
- Blocks to failure = 1/0.5574 = 1.79
Answer: About 1.8 blocks (75,344 cycles in all); the 2,000 cycles at 400 MPa do 45 % of the damage
Shaft design for strength: combined bending and torsion, ASME code
Which syllabus covers thisAU ME3591 Unit IIVTU BME602 Module 2SPPU 302043 Unit 2MU MEC601 Module 3KTU not in syllabus
What loads does a shaft carry, and how are bending and torsion combined?
A shaft transmits torque T from a motor to a gear or pulley, and the gear forces, belt pulls and part weights bend it between the bearings, giving a bending moment M. At the surface the bending stress σ = 32M/(πd³) and the shear stress τ = 16T/(πd³) act together. The maximum shear stress theory, used for ductile steel shafts, gives τmax = 16√(M² + T²)/(πd³): the shaft behaves as if it carried one equivalent torque Te = √(M² + T²). The maximum normal stress theory gives an equivalent bending moment Me = (M + Te)/2, used for brittle materials.
What does the ASME code add to that, and how is the diameter found?
A rotating shaft sees reversed bending every turn and often shock, so the ASME code multiplies M by a combined shock and fatigue factor Kb and T by Kt. For rotating shafts: steady load 1.5 and 1.0, minor shock 1.5-2.0 and 1.0-1.5, heavy shock 2.0-3.0 and 1.5-3.0. The allowable shear is 30 % of yield but not over 18 % of ultimate, cut by 25 % for a keyway: 114 MPa, or 85.5 MPa keyed, for steel of 380/640 MPa (illustrative). The diameter is rounded up to a standard size: steps of 5 mm from 25 to 60 mm, 10 mm to 110 mm.
When is a hollow shaft worth it?
Most of the bending and torsional stress is carried by the outer material; the core near the axis does little. Boring out the centre to k = di/do = 0.6 removes 36 % of the area but only 13 % of the strength (1 − k⁴ = 0.870). For the same equivalent torque the outside diameter grows by only 4.7 % and the shaft weighs 30 % less, with a higher stiffness per kilogram and so a higher critical speed. The price is machining or buying tube, so hollow shafts appear where weight or inertia matters: propeller shafts, machine-tool spindles with a through-bore, aircraft.
What if the shaft also carries an axial load?
Helical gears, worms and propellers push the shaft along its axis. An axial force F adds a direct stress 4F/(πd²) to the bending stress, so the full ASME equation puts it inside the bending term: Kb M + αFd(1 + k²)/8. The factor α is 1 for tension; for a long shaft in compression it is larger, to allow for column buckling. Because d appears on both sides, the diameter is found by trial or iteration. Indian BIS practice for transmission shafting follows the same maximum shear stress approach with its own factors.
What are shafts made of, and how is the design checked?
Most shafts are plain carbon steels (about 0.3-0.45 % carbon, hot-rolled or cold-drawn), with alloy steels such as nickel-chromium-molybdenum grades for heavily loaded or hardened shafts. Cold drawing gives a better finish and higher strength but leaves residual stresses that can bend the bar when keyways are cut. The strength design gives a first diameter; the shaft is then checked for torsional and lateral rigidity, critical speed and fatigue at each shoulder and keyway, and the largest diameter wins.
A keyed shaft carries M = 420 N·m and T = 199 N·m. You change the loading from steady (Kb 1.5, Kt 1.0) to heavy shock (Kb 2.5, Kt 2.25). The ASME diameter grows by about:
d grows as the cube root of the equivalent torque: from 34.0 mm to 40.8 mm, 20 %. The shock factors raise Te by 73 %, but the cube root tames it.
Worked examples
Shaft diameter by the ASME code · 15 kW at 720 rpm; a pulley 150 mm from the left bearing pulls 4 kN radially; bearings 500 mm apart; minor shock (Kb 1.75, Kt 1.25); keyed; steel Syt 380, Sut 640 MPa (illustrative)
- T = 60 × 15,000/(2π × 720) = 198.9 N·m
- RA = 4,000 × 350/500 = 2,800 N; M = 2,800 × 150 = 420 N·m
- τ = min(0.30 × 380, 0.18 × 640) × 0.75 = 85.5 MPa
- Te = √((1.75 × 420)² + (1.25 × 198.9)²) = 775.9 N·m
- d = (16 × 7,75,929/(π × 85.5))^(1/3) = 35.9 mm → next standard size 40 mm
Answer: d = 35.9 mm, use a 40 mm shaft
The same shaft made hollow · The loads above, hollow with di/do = 0.6
- 1 − k⁴ = 1 − 0.6⁴ = 0.8704
- do = 35.89/(0.8704)^(1/3) = 37.6 mm → 40 mm outside, di = 0.6 × 40 = 24 mm
- Mass per length against the computed solid shaft: 37.59² × 0.64 / 35.89² = 0.702
- Against a solid 40 mm bar the 40/24 tube removes 36 % of the metal
Answer: 40 mm outside, 24 mm bore; about 30 % lighter for the same strength
A spur gear at mid-span · 10 kW at 500 rpm through a 200 mm pitch-diameter spur gear (20° pressure angle) at the middle of a 400 mm span; steady load (Kb 1.5, Kt 1.0); τ = 85.5 MPa keyed
- T = 191.0 N·m; tangential force Ft = 2T/D = 1,910 N
- Radial force Fr = Ft tan 20° = 695 N; resultant 2,032 N
- M = R L/4 = 2,032.4 × 400/4 = 203.2 N·m
- Te = √((1.5 × 203.2)² + 191.0²) = 359.7 N·m
- d = 27.8 mm → 30 mm
Answer: d = 27.8 mm, use 30 mm
Shaft rigidity and critical speed
Which syllabus covers thisAU ME3591 Unit IIVTU BME602 Module 2SPPU 302043 Unit 2MU not in syllabusKTU not in syllabus
What is torsional rigidity, and why can it decide the size?
A shaft twists by θ = TL/(GJ). Too much twist upsets timing (camshafts, machine-tool feeds) and lets long line shafts wind up and vibrate, so the twist is limited, often to about 0.25° per metre for line shafting (a typical figure). Because θ falls with d⁴, rigidity can need a fatter shaft than strength: 10 kW at 1,500 rpm needs only 20 mm for 40 MPa of shear (illustrative), but 37.0 mm to keep the twist under 0.25°/m. A 40 mm shaft twists 0.18°/m.
What is lateral rigidity, and what limits it?
Gear and pulley forces bend the shaft. Too much deflection pulls gear teeth out of proper mesh, and too much slope at a bearing makes the races misalign, loading the edges of rollers or the end of a journal. Deflection and slope are found by beam methods (Macaulay, superposition or integration) with the moment of inertia πd⁴/64. Design offices set limits per application, for example a small fraction of a millimetre at a gear and about a thousandth of a radian of slope at a rolling bearing (typical values; check the gear and bearing makers' data).
What is the critical speed, and how do Rayleigh and Dunkerley estimate it?
A disc is never perfectly balanced. As the shaft spins, the out-of-balance force bends it, and at the shaft's natural frequency in bending the deflection grows without limit: the critical or whirling speed. Rayleigh's method uses the static sag y under each weight: ωc² = gΣWy/ΣWy². It slightly overestimates. Dunkerley's method adds the inverse squares of each disc's own frequency and of the bare shaft's: it slightly underestimates. A 20 kg disc in the middle of a 40 mm × 800 mm shaft gives 3,335 rpm alone and 3,051 rpm with the shaft's own mass.
How far from the critical speed should a shaft run?
With unbalance e, the whirl radius is r = e(ω/ωc)²/(1 − (ω/ωc)²): it grows sharply near ωc, and beyond it the disc whirls with its heavy side inward, so the deflection falls back towards e. Most shafts run well below the first critical speed, with a margin of at least about 20 % (a common rule). Some, like turbine rotors and long propeller shafts, run above it and pass through the critical quickly while speeding up and slowing down, with damping to limit the peak. Stiffer shafts, shorter spans and lighter discs raise the critical speed.
A disc sits in the middle of a light shaft. You double the shaft diameter. The critical speed:
The static sag falls with d⁴ (16 times less) and the critical speed goes as 1/√sag, so it rises 4 times. With the shaft's own mass included it rises a little less, because the shaft gets four times heavier too.
Worked examples
Twist of a line shaft · 10 kW at 1,500 rpm; steel G = 79.3 GPa; twist limit 0.25°/m (typical)
- T = 60 × 10,000/(2π × 1,500) = 63.66 N·m
- For d = 40 mm: J = π × 0.04⁴/32 = 2.5133e-07 m⁴; θ/L = 63.66/(79.3×10⁹ × 2.5133e-07) = 3.194e-03 rad/m = 0.183°/m
- Smallest d for 0.25°/m: d⁴ = 32T × 180/(π² G × 0.25) → d = 37.0 mm
- Strength alone (τ = 40 MPa, illustrative) needs only 20.1 mm
Answer: 0.18°/m at 40 mm; rigidity needs 37.0 mm against 20 mm for strength
Critical speed of one disc, with the shaft's mass · Steel shaft 40 mm, bearings 800 mm apart, a 20 kg disc at mid-span; E = 207 GPa, ρ = 7,850 kg/m³
- I = π × 0.04⁴/64 = 1.2566e-07 m⁴; EI = 26,012 N·m²
- Sag under the disc: y = WL³/(48EI) = 196.2 × 0.8³/(48 × 26,012) = 0.0805 mm
- Disc alone: ω = √(g/y) = 349.2 rad/s = 3,335 rpm
- Shaft alone: m̄ = 9.865 kg/m; ωs = π²√(EI/(m̄L⁴)) = 792 rad/s = 7,562 rpm
- Dunkerley: 1/n² = 1/3,335² + 1/7,562² → n = 3,051 rpm
Answer: About 3,051 rpm; running at 1,500 rpm leaves a margin of 2.0 times
Two discs: Rayleigh, Dunkerley and the exact answer · Steel shaft 45 mm, span 900 mm; 30 kg at 250 mm and 20 kg at 600 mm from the left bearing; shaft mass ignored
- EI = 207×10⁹ × π × 0.045⁴/64 = 41,667 N·m²
- Influence coefficients (m/N): a11 = 2.347e-07, a22 = 2.880e-07, a12 = 2.192e-07
- Static sags: y1 = 0.1121 mm, y2 = 0.1210 mm
- Rayleigh: ω² = g(30 y1 + 20 y2)/(30 y1² + 20 y2²) → 2,779.2 rpm
- Dunkerley: discs alone 3,599 and 3,979 rpm → 2,669 rpm
- Exact (2 × 2 eigenvalue of the flexibility matrix times the masses): 2,779.0 rpm
Answer: Dunkerley 2,669 ≤ exact 2,779.0 ≤ Rayleigh 2,779.2 rpm: Rayleigh is almost exact, Dunkerley 4 % low
Keys, splines and the Kennedy key
Which syllabus covers thisAU ME3591 Unit IIVTU BME602 Module 2SPPU 302043 Unit 2MU MEC601 Module 3KTU not in syllabus
What kinds of key are there?
A key is a removable steel piece that locks a hub to a shaft so torque passes between them. Sunk keys sit half in a shaft keyway and half in the hub: rectangular (wider than tall, the common choice) or square. Saddle keys sit on the shaft without a keyway, holding by friction, for light torques. A feather key is fixed to one part and lets the hub slide along the shaft. A Woodruff key is a half-disc that rocks to align itself, used on small and tapered shafts. Tangential and Kennedy keys sit in pairs at right angles for heavy, reversing torques.
How is a sunk key sized?
The width and height come from a table by shaft diameter (DIN 6885 gives 12 × 8 mm for shafts over 38 up to 44 mm). Only the length is designed. The torque appears as a force F = 2T/d at the shaft surface. It can shear the key across its width b, or crush the half of its height pressed against the hub. For a 40 mm shaft carrying 500 N·m with a key of Syt 380 MPa and a factor of safety 2, shear needs 21.9 mm and crushing 32.9 mm, so the next standard key is 36 mm long.
Parallel or taper sunk key?
A parallel key has the same height along its length; the hub is held axially by a shoulder, a circlip or a set screw, and the fit is easy to assemble and dismantle. A taper key has a slope of 1 in 100 on its top face and is driven in, wedging the hub tight against the shaft, so it also takes some axial load and stops rattling under reversals. A gib head on the end gives a puller something to grip. The wedge pushes the hub off-centre slightly, so taper keys are avoided on high-speed, precisely balanced parts.
How does a Kennedy key carry torque?
Two square keys sit at 90° to each other with their diagonals along the shaft surface, each driven like a taper key. Each carries a tangential force T/d; on its slanted face the normal force is √2 times that, so the crushing stress is √2T/(dbl) and the shear on the diagonal plane, √2 b wide, is T/(√2 dbl). An 80 mm shaft carrying 3,000 N·m on 20 mm keys needs 22.1 mm for shear at 60 MPa and 26.5 mm for crushing at 100 MPa (illustrative allowables).
When are splines used instead of keys?
A splined shaft has several key-like teeth cut integral with it, so the torque is shared by many faces, the shaft is not weakened by a single deep keyway, and a hub can slide along it, as in gearboxes. Capacity is set by the bearing pressure on the tooth flanks: T = p × (number of teeth × tooth height × length) × mean radius, with only part of the teeth bearing because of errors. Six splines between 28 and 34 mm, 40 mm long, at 6.5 MPa with 75 % bearing (illustrative values) carry 54 N·m. Involute splines self-centre and are stronger at the root.
A rectangular key (b wider than h) of one material is lengthened until it is safe. Which mode sets its length?
With τ = Syt/(2 FoS) and σc = Syt/FoS, the two lengths are in the ratio l_crush/l_shear = b/h. For a 12 × 8 key that is 1.5: 32.9 against 21.9 mm. A square key is equally strong in both.
Worked examples
Length of a sunk key · 40 mm shaft, 500 N·m; key steel Syt 380 MPa, factor of safety 2
- DIN 6885: shafts over 38 up to 44 mm take a 12 × 8 mm key
- τ = 0.5 × 380/2 = 95 MPa; σc = 380/2 = 190 MPa
- F = 2 × 5,00,000/40 = 25,000 N
- Shear: l = 25,000/(12 × 95) = 21.9 mm; crushing: l = 2 × 25,000/(8 × 190) = 32.9 mm
- Next standard length 36 mm, shorter than a 60 mm hub (1.5d)
Answer: 12 × 8 × 36 mm key
A square key instead · The same shaft and torque, square key 12 × 12 mm
- Shear length is unchanged: 21.9 mm
- Crushing: l = 2 × 25,000/(12 × 190) = 21.9 mm, now equal to the shear length
- Next standard length 22 mm
Answer: 12 × 12 × 22 mm: 14 mm shorter than the rectangular key
A Kennedy key · 80 mm shaft, 3,000 N·m; two 20 mm square keys; τ = 60 MPa, σc = 100 MPa (illustrative)
- Tangential force on each key = T/d = 37,500 N
- Shear: l = T/(√2 d b τ) = 30,00,000/(1.414 × 80 × 20 × 60) = 22.1 mm
- Crushing: l = √2 T/(d b σc) = 26.5 mm
- Crushing governs; round up to a standard 28 mm
Answer: About 26.5 mm, use 28 mm keys
Rigid flange and bush-pin flexible couplings
Which syllabus covers thisAU ME3591 Unit IIVTU BME602 Module 2SPPU 302043 Unit 2MU MEC601 Module 3KTU not in syllabus
Rigid or flexible: which coupling when?
A coupling joins two shafts so they turn together. Rigid couplings (a muff or sleeve keyed to both shafts, a split clamp muff, or two bolted flanges) need the shafts accurately in line and pass every shock straight through. Flexible couplings accept small errors in alignment and soak up shock: the bush-pin type uses rubber bushes on steel pins; jaw and tyre couplings use elastomer inserts; gear couplings use crowned teeth. An Oldham coupling joins parallel shafts with a sliding cross disc, and a universal (Hooke's) joint joins shafts at an angle.
How is a rigid flange coupling designed?
Each shaft has a keyed cast-iron flange; fitted bolts on a pitch circle carry the torque in shear. Typical proportions scale from the shaft d: hub 2d across and 1.5d long, pitch circle 3d, flange 0.5d thick, outside 4d, with a protecting rim over the bolt heads. Checks: bolt shear, bolt crushing on the flange, flange shear where it meets the hub, and the hub as a hollow shaft. For 37 kW at 720 rpm with a service factor 1.5, the shaft is 50 mm and four fitted M10 bolts carry 31 MPa of shear (40 MPa allowed, illustrative).
How does a bush-pin flexible coupling work?
Steel pins fixed in one flange carry rubber bushes (over brass sleeves) that sit in oversized holes in the other flange. The torque passes through the rubber, which deflects a little, taking up misalignment and damping shocks. The pin force F = 2T/(nD1) must not press the rubber harder than about 1 MPa (typical), which sets the bush length. The pin is a cantilever loaded at mid-bush, so bending usually sets its diameter. For 15 kW at 960 rpm (service factor 1.5) six 11 mm pins with 27 × 26 mm bushes give 94 MPa bending.
How are muff and clamp couplings checked?
A muff coupling is a cast-iron sleeve keyed over both shaft ends, with typical proportions of about 2d + 13 mm outside and 3.5d long. It is checked as a hollow shaft in torsion: τ = 16TD/(π(D⁴ − d⁴)), against the low shear strength of cast iron, and the key is designed as usual. A clamp coupling is a sleeve split along its length and bolted round the shafts: the bolts squeeze the halves on, and friction carries most of the torque, so it can be fitted without moving the shafts axially.
What do Oldham and universal joints allow, and what is the catch?
An Oldham coupling has a centre disc with tongues at 90° sliding in slots in both hubs, so it accepts a parallel offset at the cost of sliding friction and wear. A Hooke's joint lets shafts meet at an angle β, but the output speed swings twice per turn between ω cos β and ω/cos β. At 20° and 1,440 rpm that is 1,353 to 1,532 rpm, a swing of 12.4 %. Two joints phased correctly with equal angles cancel the swing, as in a car's propeller shaft.
A flange coupling's bolts are set by shear. You go from 4 bolts to 8 on the same pitch circle. The bolt diameter needed:
The torque per bolt halves, and the bolt area goes as d², so d falls by 1/√2: about 29 %.
Worked examples
Rigid flange coupling · 37 kW at 720 rpm, service factor 1.5; τ = 40 MPa for shaft and bolts, bolt crushing 80 MPa, cast-iron flange shear 15 MPa (illustrative); 4 fitted bolts; typical proportions
- Design torque = 1.5 × 60 × 37,000/(2π × 720) = 736 N·m
- Shaft: d = (16T/(π × 40))^(1/3) = 45.4 → 50 mm
- Hub 100 mm, pitch circle 150 mm, flange 25 mm thick
- Bolt: db = √(8T/(4π × 40 × 150)) = 8.84 mm → M10; shear 31.2 MPa
- Crushing 9.8 MPa; flange shear at the hub 1.87 MPa; hub torsion 4.0 MPa: all well inside
Answer: 50 mm shafts, four M10 fitted bolts on a 150 mm pitch circle
Bush-pin flexible coupling · 15 kW at 960 rpm, service factor 1.5; 6 pins; rubber pressure 1 MPa, pin bending 100 MPa (illustrative); brass sleeve 2 mm and rubber 6 mm thick; 5 mm gap; pitch circle 2d + Db + 12 mm
- T = 223.8 N·m; shaft d = 35 mm
- Try d1 = 11 mm: Db = 11 + 16 = 27 mm, D1 = 70 + 27 + 12 = 109 mm
- Pin force F = 2T/(6 × 109) = 684 N; bush length = 684/(27 × 1) = 25.3 → 26 mm
- Pin bending: M = 684 × (13 + 5) = 12,320 N·mm; σ = 94.3 MPa (10 mm pins would give 129 MPa)
Answer: Six 11 mm pins with 27 mm × 26 mm rubber bushes on a 109 mm pitch circle
Speed swing of a Hooke's joint · Input 1,440 rpm steady, shafts at 20°
- Lowest output speed = 1,440 × cos 20° = 1,353.2 rpm
- Highest = 1,440/cos 20° = 1,532.4 rpm
- Swing = 179 rpm, 12.4 % of the mean, twice per turn
Answer: 1,353 to 1,532 rpm
Cotter joints, knuckle joints and levers
Which syllabus covers thisAU ME3591 Unit IIIVTU not in syllabusSPPU 302043 Unit 1MU MEC601 Module 2KTU not in syllabus
How can a socket-and-spigot cotter joint fail?
A cotter joint connects two rods in tension or compression with no rotation between them, like a piston rod to a crosshead. One rod ends in a socket, the other in a spigot that fits inside, and a tapered flat cotter is driven through a slot in both. Each part can fail in its own way: the rods in tension; the spigot and the socket in tension across the slot, where metal is cut away; the cotter in double shear and in bending; the spigot and socket crushing against the cotter; and the spigot end, socket end and collar shearing out. Each mode gives one dimension.
How is a knuckle joint sized?
A knuckle joint links two rods with a pin, so they can swing in one plane: an eye on one rod sits between the two jaws of a fork on the other. With typical proportions (eye 2d outside, jaws 0.75D and eye 1.25D thick, D the rod), the pin is checked in double shear and in bending. Bending usually wins: for 50 kN at 50 MPa shear the pin needs 25.2 mm for shear but 37.5 mm for bending at 80 MPa, so it becomes 38 mm. The eye and the fork are then checked in tension, shear and crushing.
How are hand and foot levers designed?
A lever multiplies force: effort × effort arm = load × load arm about the fulcrum. The arms are sized for bending at the boss, where the moment is largest, usually as a rectangle about two to three times as deep as it is wide (typical). The fulcrum pin carries the vector sum of the effort and the load, and is sized for bearing pressure on its bush and for shear. For hand and foot levers the effort is set by what a person can apply comfortably, and the boss is keyed or clamped to its shaft.
What does a safety-valve lever do?
In a lever safety valve a dead weight on a long arm holds the valve shut until the pressure force, p times the valve area, overcomes it. For a 60 mm valve at 1.2 MPa, 80 mm from the fulcrum, a weight on a 600 mm arm must be 452 N; the fulcrum pin takes the remaining 2,941 N. The arm is sized for bending at the valve, M = W(l − a) = 235 N·m.
What is a bell-crank lever?
A bell-crank has its two arms at an angle, usually 90°, so it turns the direction of a force as well as scaling it, as in railway signals and governors. With a 1 kN load on a 150 mm arm and the effort on a 300 mm arm, the effort is 500 N, and the fulcrum pin carries the vector sum: √(1,000² + 500²) = 1,118 N, more than either force.
A knuckle joint with typical proportions carries 50 kN. Compared with the pin sized for double shear alone, the pin needed when bending is checked is:
Shear needs 25.2 mm; bending under M = (P/2)(b/4 + a/3) = 415 N·m needs 37.5 mm, 49 % more.
Worked examples
Knuckle joint for 50 kN · Axial pull 50 kN; σt = 80, τ = 50, σc = 100 MPa (illustrative); typical proportions d0 = 2d, a = 0.75D, b = 1.25D; whole mm
- Rod: D = √(4 × 50,000/(π × 80)) = 28.2 → 29 mm; jaws a = 22 mm, eye b = 37 mm
- Pin for double shear: √(2 × 50,000/(π × 50)) = 25.2 mm
- Bending: M = 25,000 × (37/4 + 22/3) = 4,14,583 N·mm → d = (32M/(π × 80))^(1/3) = 37.5 mm
- Pin 38 mm, eye outside 76 mm; bending stress 77.0 MPa
- Eye tension = shear = 50,000/(37 × 38) = 35.6 MPa; eye crushing 35.6; fork tension 29.9, crushing 29.9 MPa
Answer: Rod 29 mm, pin 38 mm, eye 76 mm outside and 37 mm thick, jaws 22 mm
Socket-and-spigot cotter joint for 50 kN · The same load and allowables; cotter thickness t = 0.31 d2, spigot collar 1.5d (typical); whole mm
- Rod d = 29 mm. Spigot: tension across the slot needs 36.3 mm, crushing on the cotter needs 40.2 mm → d2 = 41 mm, t = 13 mm
- Socket across the slot: π/4(d1² − 41²) − 13(d1 − 41) = 50,000/80 = 625 → d1 = 52 mm
- Socket collar d4 = 41 + 50,000/(13 × 100) = 80 mm
- Cotter: shear needs 38.5 mm, bending M = 25,000 × (39/6 + 41/4) = 4,18,750 N·mm needs 49.2 mm → b = 50 mm
- Spigot end a = 13 mm, socket end c = 13 mm, spigot collar 49 × 14 mm; the highest check is cotter bending at 99 %
Answer: Spigot 41 mm, socket 52 mm and 80 mm at the collar, cotter 13 × 50 mm
Lever safety valve · Valve 60 mm diameter, pressure 1.2 MPa, valve 80 mm and dead weight 600 mm from the fulcrum; arm rectangular, depth 2 × width, bending 70 MPa; fulcrum pin 12 × 18 mm (illustrative)
- Valve force = 1.2 × π × 60²/4 = 3,393 N
- Moments about the fulcrum: W = 3,393 × 80/600 = 452 N
- Fulcrum reaction = 3,393 − 452 = 2,941 N; pin bearing pressure = 2,941/(12 × 18) = 13.6 MPa
- Bending at the valve: M = 452 × 520 = 2,35,242 N·mm; Z = 2b³/3 → b = 17.1 mm
Answer: Dead weight 452 N; arm 18 × 36 mm at the valve
Power screws and the screw jack
Which syllabus covers thisAU not in syllabusVTU not in syllabusSPPU 302043 Unit 3MU MEC601 Module 2KTU not in syllabus
Which thread forms are used for power screws?
Power screws turn rotation into a large, slow push: screw jacks, vices, presses, lathe lead screws. The square thread has flanks at right angles to the axis, so there is no wedging: lowest friction and highest efficiency, but it is hard to cut and wear cannot be taken up. The trapezoidal (metric, 30°) and Acme (29°) threads are stronger at the root, easy to machine, and a split nut can close up to take up wear; they waste a little more in friction. The buttress thread has one near-square flank and one sloping flank, for heavy loads in one direction only.
How much torque does it take to raise and lower a load?
Unroll one turn of thread and it is an incline: base π dm, rise equal to the lead l, helix angle tan α = l/(π dm). Pushing the load up the incline against friction angle φ needs a torque W(dm/2)tan(φ + α); letting it down needs W(dm/2)tan(φ − α). A trapezoidal thread wedges, so its effective friction is μ/cos 15°. The thrust collar adds μc W Rc both ways. For 50 kN on a 50 × 8 square thread, μ 0.12, the thread needs 203 N·m to raise and the collar another 225 N·m.
What sets efficiency and self-locking?
Efficiency is the useful work W l divided by the work put in, 2πT, per turn. For the thread alone it is tan α/tan(φ + α). If φ > α the load cannot run down by itself: the screw is self-locking, which a jack needs. A self-locking screw can never be more than 50 % efficient (at μ = 0.12 the bound is 49.3 %). The best efficiency, (1 − sin φ)/(1 + sin φ) = 79 %, comes at α = 45° − φ/2, a very steep helix that would overhaul. Collar friction often halves the overall efficiency.
How is a screw jack designed?
Choose a thread from the core stress: the screw is a short column in compression W/(π dc²/4) plus torsion from the thread torque, combined by a failure theory and checked for buckling if it is long. The nut is softer bronze; the number of threads in contact comes from a bearing pressure of about 10-18 MPa (typical) on the thread annulus, then the nut threads are checked in shear. The handle length follows from the torque and a comfortable effort, and is sized for bending. A C-clamp is the same screw with a frame checked as a curved beam.
What are differential, compound and ball screws?
A differential screw has two threads of the same hand but different pitches, say 5 and 4 mm: one turn moves the load by the difference, 1 mm, giving very fine adjustment and a large force. A compound screw uses opposite hands, so the movement is the sum, for fast closing. A recirculating ball screw replaces sliding with balls rolling in a groove and returning through a tube: efficiency is high (about 90 %, typical), there is little wear and no stick-slip, so CNC machines use it, but it is not self-locking and needs a brake for vertical loads.
A screw jack is self-locking (it holds the load with the handle let go). Its thread efficiency can be at most about:
Self-locking needs φ ≥ α. At the limit α = φ the efficiency is tan φ/tan 2φ = (1 − tan² φ)/2, just under 50 % (49.3 % at μ = 0.12). Real jacks with collar friction run at 15-30 %.
Worked examples
Screw jack with a square thread · Load 50 kN; square thread 50 × 8 mm, single start; μ = 0.12; collar μc = 0.15 at Rc = 30 mm (illustrative); handle 500 mm; nut pressure 12 MPa (typical)
- dm = 50 − 4 = 46 mm; tan α = 8/(π × 46) → α = 3.17°; φ = atan 0.12 = 6.84°: self-locking
- Thread torque to raise = 50,000 × 23 × tan(10.01°) = 203.0 N·m; collar 0.15 × 50,000 × 30 = 225 N·m
- Total 428 N·m; handle force 856 N
- Efficiency = 50,000 × 8/(2π × 4,28,011) = 14.9 % (thread alone 31.4 %)
- To lower: 73.8 + 225 = 299 N·m
- Nut: 50,000/(12 × π/4 × (50² − 42²)) = 7.2 → 8 threads, 64 mm tall
Answer: 428 N·m to raise (856 N on the handle), 14.9 % efficient, nut 64 mm
The same jack with a trapezoidal thread · Tr 50 × 8 (ISO 2904: d2 = 46, d3 = 41 mm), everything else as above
- μ' = 0.12/cos 15° = 0.1242; φ' = 7.08°
- Thread torque to raise = 208.0 N·m (+4.9 over square)
- Efficiency 14.7 %; thread alone 30.6 %
- Nut on the ISO core 41 mm: 7 threads, 56 mm
Answer: Only 2.4 % more thread torque than the square thread, with a stronger, easily made thread
Best helix angle and the self-locking limit · Square thread, μ = 0.12, no collar friction
- φ = 6.84°
- Best helix α = 45° − φ/2 = 41.6° → η max = (1 − sin φ)/(1 + sin φ) = 78.7 %
- Self-locking limit α = φ: η = tan φ/tan 2φ = 49.3 %
- The jack above (α = 3.17°) runs at 31.4 % on the thread
Answer: At most 79 %, and below 49 % for any self-locking screw
Threaded fasteners and eccentrically loaded bolted joints
Which syllabus covers thisAU ME3591 Unit IIIVTU not in syllabusSPPU 302043 Unit 5MU MEC601 Module 2KTU not in syllabus
What sets the strength of a bolt in tension?
A bolt breaks at its threads, so its tensile strength is the allowable stress times the tensile stress area, a little more than the core (minor-diameter) area. ISO metric threads are written M16 × 2: 16 mm major diameter, 2 mm pitch; the coarse pitch is assumed if none is given. The stress area is the area of a circle whose diameter is the mean of the pitch and minor diameters, about 157 mm² for M16, against 201 mm² for the plain 16 mm shank. Sizing a bolt therefore means finding the area needed and picking the next size up from the stress-area table.
Why are bolts tightened, and how much of the outside load reaches the bolt?
Tightening stretches the bolt and squeezes the parts together with a preload Fi. When an outside load P pulls the joint apart, the bolt stretches a little more and the parts unload a little. The share reaching the bolt is C = kb/(kb + km), small because the clamped parts are stiffer than the slender bolt. With Fi = 20 kN, C = 0.25 and P = 12 kN (illustrative), the bolt carries 23 kN while the clamp force falls to 11 kN; the joint only opens at P = Fi/(1 − C) = 26.7 kN. Preload also keeps the nut from working loose and the bolt from fatiguing.
What are bolts of uniform strength and locking devices?
Under shock, the energy a bolt can absorb depends on how much of its length is highly stressed. A plain bolt is stressed most at the threads and lightly along the thicker shank, so the shank is turned down to the core diameter, or bored hollow, until its stress matches the threads: a bolt of uniform strength, used for cylinder-head studs and connecting-rod bolts. Vibration can loosen nuts, so joints use locking devices: a jam (lock) nut tightened against the main nut, a castle or slotted nut with a split pin, spring or tab washers, nylon-insert nuts, wire locking, or a thread-locking adhesive.
How does an eccentric load in the plane of the bolts share out?
Move the load P to the centroid G of the bolt group, adding a moment P·e. The direct load gives every bolt the same primary shear P/n. The moment turns the bracket about G, so each bolt resists with a secondary shear proportional to its distance r from G, at right angles to r: P·e·r/Σr². The two add as vectors. The critical bolt is the one where both point the same way, usually a far corner on the load side. In the default bracket, 6 bolts take 3.33 kN each directly, the corners another 12.7 kN, and the worst bolt carries 14.5 kN.
What if the load pulls the bracket away from the wall?
A bracket bolted to a wall with the load at distance e tends to tilt about its bottom edge. Each bolt stretches in proportion to its height l above that edge, so it carries P·e·l/Σl² in tension: the top row works hardest. Every bolt also carries the direct shear P/n. Tension and shear combine through the maximum principal or maximum shear stress theory. A circular base, such as a pillar crane on a flange, is handled the same way with l measured from the tangent at the tilting edge; Σl² = n(R0² + R²/2) for n equal bolts on radius R, R0 being the edge distance.
A bracket's load stays the same but its eccentricity doubles. What happens to the bolt forces?
Primary shear P/n does not depend on e. The secondary part P·e·r/Σr² is proportional to e. In the 3 × 2 bracket at e = 250 mm the corner bolt's secondary shear is 12.7 kN and the primary 3.33 kN; at 500 mm the secondary becomes 25.4 kN.
Worked examples
Critical bolt in a bracket loaded in its own plane · Six bolts in 3 rows × 2 columns at 80 mm pitch; P = 20 kN at 250 mm from the group centroid; allowable tensile stress 100 MPa (typical), shear 50 MPa (half, by the maximum shear stress theory)
- Primary shear = 20,000/6 = 3,333 N on every bolt
- Σr² = 4 × (40² + 80²) + 2 × 40² = 35,200 mm²
- Corner radius r = √(40² + 80²) = 89.44 mm; secondary = 20,000 × 250 × 89.44/35,200 = 12,705 N
- At the two corners on the load side the components add: resultant = 14,505 N
- Stress area needed = 14,505/50 = 290 mm² → M24 (353 mm²)
Answer: Critical bolt force 14.5 kN; use M24 bolts
Bracket on a wall: tension in the top bolts · Four bolts, two at 40 mm and two at 200 mm above the tilting edge; P = 15 kN at 300 mm from the wall; allowable tensile stress 100 MPa (typical)
- Σl² = 2 × 40² + 2 × 200² = 83,200 mm²
- Top-bolt tension = 15,000 × 300 × 200/83,200 = 10,817 N
- Direct shear = 15,000/4 = 3,750 N
- Maximum shear force on the area: √((10,817/2)² + 3,750²) = 6,581 N
- Area for τ ≤ 50 MPa: 6,581/50 = 131.6 mm² (the principal-stress rule gives 119.9 mm²) → M16 (157 mm²)
Answer: Top bolts carry 10.82 kN tension and 3.75 kN shear; M16
Circular flange tilting about its edge · Six equal bolts on a 150 mm radius; the flange tilts about a tangent 200 mm from its centre; P = 10 kN at 500 mm; allowable tensile stress 100 MPa (typical)
- Σl² = n(R0² + R²/2) = 6 × (200² + 150²/2) = 3,07,500 mm² (the same for any orientation)
- Worst case: one bolt farthest from the edge, l = 350 mm: F = 10,000 × 500 × 350/3,07,500 = 5,691 N
- Turned half a pitch, the farthest bolt is at 200 + 150 cos 30° = 329.9 mm: 5,364 N, so the first case governs
- Area needed = 5,691/100 = 56.9 mm² → M10 (58.0 mm²)
Answer: Maximum bolt tension 5.69 kN; M10 bolts
Welded joints: butt, fillet and eccentric loads
Which syllabus covers thisAU ME3591 Unit IIIVTU BME602 Module 3SPPU 302043 Unit 5MU MEC601 Module 2KTU PCMET602 Module 2
What kinds of welded joint are there, and how are they shown on drawings?
A butt weld joins plates edge to edge; thick plates are bevelled first into a V, double-V, U or J groove so the weld can reach the root. A fillet weld fills the corner of a lap or T joint and has a roughly triangular section. Plug and slot welds fill holes in a lapped plate; spot and seam welds join sheets by resistance heating. On a drawing, an arrow points at the joint from a reference line; the weld symbol (a triangle for a fillet, a V for a V-groove) sits below the line for the arrow side and above it for the other side, with the size on its left and the length on its right.
Why is a fillet weld designed on its throat?
A fillet weld with equal legs s breaks across its narrowest section, the throat, which is s cos 45° = 0.707 s. Its strength per unit length is therefore 0.707 s τ, whichever way the load runs. Parallel (side) fillets carry the load along their length in shear; transverse (end) fillets are in fact somewhat stronger, but the common design practice treats both as shear on the throat. A full-penetration butt weld is designed like the plate: t × L × σ. With E60xx electrodes the permitted shear is 0.30 × 427 = 128 MPa, so an 8 mm fillet carries 725 N per mm of length.
How are the welds of an angle section balanced?
An angle's centroid is not in the middle of the connected leg, so equal welds along both edges would make the load eccentric. The weld lengths are chosen so their combined force passes through the centroid: take moments about it, l_heel × y_heel = l_toe × y_toe, with the total length fixed by the load. For a 100 × 100 × 10 angle the centroid is 28.7 mm from the heel, so the weld along the heel, closer to the centroid, is about 2.5 times as long as the one along the toe.
How do you design a weld group with the load off-centre, in its plane?
Treat each weld as a line. The load P at e from the weld group's centroid G gives a direct (primary) shear P/L per mm of weld and a twisting (secondary) shear P·e·r/Ju per mm, at right angles to the radius r, where Ju is the polar moment of the lines about G per unit throat. Add them as vectors at each end of each line; the largest is the critical point. Divide by 0.707 × allowable stress to get the leg size. The C-shaped weld in the interactive needs a 4.5 mm leg for 30 kN at 200 mm, so a 5 mm fillet; the secondary part dominates, so the leg grows almost in step with e.
What changes when the load bends the weld instead of twisting it?
When a bracket is welded to a column along its top and bottom edges and loaded out of the weld plane, the welds see a bending moment P·e as tension and compression on the throat, plus direct shear. Treat the lines as a beam section of unit width: for two horizontal welds b long and d apart the unit section modulus is b·d, so σ = P e/(t b d) and τ = P/(2 t b). The maximum shear stress theory then gives √((σ/2)² + τ²), which must stay within the allowable shear.
How is a ring of fillet weld round a shaft checked in torsion?
A shaft welded to a plate by a fillet all round carries its torque T as shear on a thin ring of throat 0.707 s. Treated as a line, the ring has Ju = 2πr³, so τ = T r/(0.707 s × 2πr³) = T/(1.414 π s r²). For a 60 mm shaft, a 6 mm fillet and 800 N·m the shear is 33 MPa, well inside 128 MPa. If the shaft also carries a bending moment, the weld adds a bending stress M/(0.707 s π r²) and the two are combined as above.
Two parallel fillet welds 8 mm in leg carry 160 kN with E60xx electrodes. Roughly how long must each weld be?
Each mm of an 8 mm fillet carries 0.707 × 8 × 128.1 = 725 N, so 160 kN needs 221 mm of weld in all: about 110 mm on each side.
Worked examples
Lap joint with two parallel fillet welds · Load 160 kN; 8 mm fillet welds along both edges of the lapped plate; E60xx electrodes, permitted shear 0.30 × 427 MPa
- Allowable shear = 0.30 × 427 = 128.1 MPa
- Throat = 0.707 × 8 = 5.656 mm; strength per mm = 5.656 × 128.1 = 724.5 N/mm
- Total length = 1,60,000/724.5 = 220.8 mm
- Each weld = 110.4 mm, say 115 mm (add a little for the start and stop craters)
Answer: Two welds about 115 mm long
Balancing the welds of an angle · Equal angle 100 × 100 × 10 mm carrying 150 kN along its centroid, connected by its leg to a gusset with 6 mm fillets along the heel and the toe; E60xx
- Centroid from the heel: (100 × 10 × 5 + 90 × 10 × 55)/1,900 = 28.68 mm
- Strength per mm = 0.707 × 6 × 128.1 = 543.4 N/mm; total length = 1,50,000/543.4 = 276.0 mm
- Moments about the centroid: l_heel × 28.68 = l_toe × 71.32
- l_heel = 276.0 × 71.32/100 = 196.9 mm; l_toe = 276.0 × 28.68/100 = 79.2 mm
Answer: About 197 mm along the heel and 80 mm along the toe
C-shaped weld with an eccentric load · Two horizontal welds b = 100 mm, one vertical d = 150 mm (a C open to the load); P = 30 kN at 200 mm to the right of the weld centroid; E60xx
- x̄ = b²/(2b + d) = 100²/350 = 28.57 mm from the vertical weld; L = 350 mm
- Ju = (8b³ + 6bd² + d³)/12 − b⁴/(2b + d) = 17,87,202 mm³
- Primary = 30,000/350 = 85.7 N/mm (upward on the plate)
- Worst point: the free ends of the horizontal welds, 71.4 mm right of and 75 mm from G; secondary = 30,000 × 200 × 103.6/17,87,202 = 348 N/mm
- Vector sum = 412 N/mm; leg s = 412/(0.707 × 128.1) = 4.5 mm
Answer: s = 4.54 mm: use 5 mm fillet welds
Bracket welded along two edges and loaded in bending · Two horizontal fillet welds 120 mm long, 200 mm apart (top and bottom of the bracket); P = 60 kN at 250 mm from the column face; E60xx
- M = 60,000 × 250 = 1,50,00,000 N·mm; unit section modulus Zu = b d = 24,000 mm²
- Per mm of throat: σ·t = M/Zu = 625 N/mm; τ·t = P/(2b) = 250 N/mm
- Maximum shear: √((625/2)² + 250²) = 400.2 N/mm
- Throat t = 400.2/128.1 = 3.12 mm; leg s = 3.12/0.707 = 4.42 mm
Answer: Use 5 mm fillet welds
Riveted joints, boiler joints and bonded joints
Which syllabus covers thisAU ME3591 Unit IIIVTU BME602 Module 3SPPU not in syllabusMU not in syllabusKTU PCMET602 Module 2
What are rivets made of, and what kinds of riveted joint are there?
A rivet is a plain shank with one head, put through matching holes and closed by forming a second head, hot for steel structures and boilers, cold for small rivets. Heads may be snap (round), pan, countersunk or conical. Materials must be ductile: mild steel for structures and boilers, aluminium alloys for aircraft skins, copper and brass for small work. In a lap joint the plates overlap; in a butt joint they meet edge to edge under one or two cover straps. Rows may be in line (chain) or staggered (zig-zag). Riveting is permanent and does not heat the plate, which is why it survives in aircraft and old bridges.
What are caulking and fullering?
Boilers and tanks must not leak, and a riveted seam is not tight on its own. In caulking, the edge of the plate is first bevelled at about 75-80°, then a blunt caulking tool is hammered along it, upsetting the edge so it presses hard on the plate below; the rivet heads are often caulked too. Fullering does the same job with a fullering tool as wide as the bevelled edge, which pushes the whole edge down and is less likely to damage the lower plate. Both only seal; strength still comes from the rivets and the plate.
How can a riveted joint fail, and what is its efficiency?
Over one pitch length p, the joint can fail by tearing the plate between holes, (p − d) t σt; by shearing the rivets, n (π/4) d² τ, with double-strap butt rivets in double shear counted as 1.875 × single shear under the Indian Boiler Regulations; or by crushing the plate or rivet, n d t σc. The weakest mode is the joint strength. Efficiency is that divided by the strength of the solid plate, p t σt. Unwin's d = 6√t gives a first rivet size for plates over about 8 mm, and the pitch that makes tearing equal shearing balances the joint.
How are boiler joints designed?
In a boiler shell the hoop stress, p D/(2t), acts across the longitudinal seam, twice the longitudinal stress across the circumferential seams. So the longitudinal joint is a strong double-strap butt joint, and the shell thickness is t = p D/(2 σt η_l) plus a corrosion allowance, using the efficiency of that joint. The circumferential joint only has to carry the end thrust (π/4) D² p and is usually a lap joint, sized by the number of rivets. Boiler regulations also fix minimum plate and strap thicknesses, the maximum pitch, the distance between rows and a margin of about 1.5 d.
How are structural joints and riveted brackets handled?
Structural joints in trusses and frames join members of uniform strength, so the rivets are arranged to load the plate as evenly as possible, often in a diamond (Lozenge) pattern that takes out only one hole at the first row. A riveted bracket with an off-centre load is analysed exactly like a bolt group in shear: primary shear P/n on every rivet plus secondary shear P·e·r/Σr² from the moment, added as vectors, with the critical rivet at a far corner. Rivets are assumed to fill their holes, so the shank, not the thread, carries the shear.
How does a bonded (adhesive) lap joint carry load?
An adhesive joint carries load as shear in a thin glue layer. The simple check is the average shear stress P/(b L), b the width and L the overlap: 3 kN on a 25 × 12.5 mm overlap gives 9.6 MPa. But the plates stretch as they pass load into the glue, so the shear is not uniform: it peaks at the two ends of the overlap and is low in the middle (shear lag). Because the load is also offset, the ends peel. Making the overlap longer adds little strength once the middle is lightly loaded; tapering the plate ends, a fillet of adhesive, or a stiffer adherend lowers the peaks.
A double-strap butt joint has rivets in double shear. How does its rivet shearing strength compare with the same rivets in a lap joint?
Each rivet is cut at two planes, but the two halves do not share the load perfectly, so the regulations credit 1.875 × the single-shear strength. For two 24 mm rivets at 60 MPa that is 101.8 kN instead of 54.3 kN.
Worked examples
Single-riveted lap joint · Plates 10 mm thick; σt = 80, τ = 60, σc = 120 MPa (typical); rivet hole by Unwin; pitch 45 mm
- Unwin: d = 6√10 = 18.97 mm, take 19 mm
- Shearing Ps = (π/4) × 19² × 60 = 17,012 N; crushing Pc = 19 × 10 × 120 = 22,800 N
- Pitch for equal tearing and shearing: p = 19 + 17,012/(10 × 80) = 40.3 mm; we use 45 mm
- Tearing Pt = (45 − 19) × 10 × 80 = 20,800 N
- Efficiency = 17,012/(45 × 10 × 80) = 47.3% (shearing governs)
Answer: η ≈ 47%, limited by rivet shear
Double-strap butt joint, double riveted · t = 16 mm, d = 24 mm (Unwin), p = 90 mm, two rivets per pitch in double shear (1.875 factor); σt = 80, τ = 60, σc = 120 MPa (typical)
- Tearing Pt = (90 − 24) × 16 × 80 = 84,480 N
- Shearing Ps = 2 × 1.875 × (π/4) × 24² × 60 = 1,01,788 N
- Crushing Pc = 2 × 24 × 16 × 120 = 92,160 N
- Solid plate = 90 × 16 × 80 = 1,15,200 N; η = 84,480/1,15,200 = 73.3% (tearing governs)
- Balanced pitch would be 24 + 1,01,788/(16 × 80) = 103.5 mm, but then crushing (92,160 N) would govern
Answer: η = 73.3%, tearing of the plate governs
Boiler shell and its circumferential seam · Shell 1,500 mm in diameter, steam at 1.2 MPa, σt = 80 MPa, longitudinal joint efficiency 80%, corrosion allowance 1 mm (illustrative); circumferential lap joint with 24 mm rivets at τ = 60 MPa
- t = 1.2 × 1,500/(2 × 80 × 0.80) + 1 = 15.06 mm → 16 mm
- End thrust = (π/4) × 1,500² × 1.2 = 21,20,575 N
- One rivet in single shear: (π/4) × 24² × 60 = 27,143 N
- Rivets needed = 21,20,575/27,143 = 78.1 → 79, in practice rounded to suit the pitch round the shell
Answer: 16 mm shell; about 79 rivets in the circumferential seam
Helical springs: Wahl factor, deflection, surge and concentric springs
Which syllabus covers thisAU ME3591 Unit IVVTU not in syllabusSPPU 302043 Unit 6MU MEC601 Module 6KTU PCMET602 Module 3
What kinds of spring are there, and what are they made of?
Helical compression springs push back when squeezed (valve springs, suspension coils); helical extension springs pull back and have hooks and an initial tension wound in; torsion springs resist twisting (clothes pegs, door hinges); spiral springs store energy in a flat coiled strip (clocks); Belleville washers are dished discs that give large forces in little space; leaf springs follow in the next chapter. Common materials are cold-drawn carbon wire, oil-tempered wire, music wire for small springs, chrome-vanadium and chrome-silicon alloy steels for fatigue and heat, stainless steel against corrosion, and phosphor bronze where the spring must conduct or be non-magnetic.
Why is the stress in a coil spring more than the torsion formula gives?
A compressed coil spring is a wire twisted by a torque P·D/2, so the torsion formula gives τ = 8 P D/(π d³). Two effects add to it: the direct shear from P itself, and the curvature of the coil, which crowds stress onto the inside of each turn. The Wahl factor K = (4C − 1)/(4C − 4) + 0.615/C, with spring index C = D/d, covers both. At C = 7 it is 1.21; at C = 4 it climbs to 1.40, which is why tightly wound springs with C below about 4 are avoided. The interactive's default spring, 6 mm wire on a 42 mm mean diameter at 900 N, reaches 540 MPa.
How do end types, solid length, free length and buckling fit together?
Only the active coils deflect. Squared (closed) ends add about one dead coil at each end, and grinding them flat makes the spring sit square. Total coils are Na + 0, 1, 2 or 2 for plain, plain-ground, squared and squared-ground ends; solid length is d(Nt + 1), or d·Nt when the ends are ground. Free length is solid length plus the working deflection plus a clash allowance, commonly 15% of the deflection, so the coils never touch in service. A long, slender spring can buckle sideways: with ends on flat parallel plates, steel springs are safe while the free length is below about 5.2 D.
How do springs combine in series, in parallel and as a concentric pair?
Springs in series share the same force and their deflections add, so 1/k = 1/k1 + 1/k2: a 20 and a 30 N/mm spring in series give 12 N/mm. In parallel they share the same deflection and their forces add, k = k1 + k2 = 50 N/mm. A concentric (nested) pair is two springs in parallel inside each other, used where space is short, as in heavy valve gear and railway buffers. Made of the same material with the same index and the same solid length, both take the same stress and deflection, and they share the load in proportion to d². They are wound in opposite hands so the coils cannot interlock.
What is spring surge?
A spring has mass, so a sudden push at one end travels along the coils as a wave instead of compressing them all at once. If the spring is driven near its own natural frequency, as a valve spring can be by its cam, the waves build up, coils clash, and the stress far exceeds the static value. With both ends fixed, the lowest natural frequency is f = (d/(2π Na D²))√(G/(2ρ)); the default spring's is about 123 Hz. Designers keep it 15 to 20 times above the highest forcing frequency, raise it with a smaller D or fewer coils, or add friction dampers.
How are springs designed for fluctuating loads, and what does shot peening do?
Most springs cycle between a minimum and a maximum force, so they see a mean and an alternating shear stress. With 300-900 N on the default spring, τm = 360 MPa and τa = 180 MPa (Wahl factor on both). These are put on a Goodman-type line in shear (chapter 6), using torsional endurance limits for spring wire. Fatigue cracks start at the surface on the inside of the coil, so shot peening, blasting the surface with small steel balls, leaves a layer of compressive residual stress that markedly raises fatigue strength. Resilience, the energy stored, is ½Pδ: 23 J for the default spring at 900 N.
Keep the wire and load fixed but double the spring index C (so D doubles). What happens to the deflection?
δ = 8 P D³ N/(G d⁴): deflection goes with D³, so doubling D multiplies it by 8. The stress roughly doubles too (8PD/πd³, with a slightly smaller Wahl factor).
Worked examples
Checking a compression spring · Wire d = 6 mm, mean diameter D = 42 mm, 10 active coils, squared and ground ends, P = 900 N; steel G = 81.37 GPa, E = 207 GPa (typical)
- C = 42/6 = 7; K = (28 − 1)/(28 − 4) + 0.615/7 = 1.2129
- τ = 1.2129 × 8 × 900 × 42/(π × 6³) = 540 MPa
- δ = 8 × 900 × 42³ × 10/(81,370 × 6⁴) = 50.58 mm; k = 17.79 N/mm
- Nt = 12; Ls = 12 × 6 = 72 mm; Lf = 72 + 1.15 × 50.58 = 130.2 mm
- Buckling limit (flat parallel ends, α = 0.5): 218 mm, so Lf = 130 mm is safe
- Surge frequency = 123 Hz: keep the forcing below about 8 Hz
Answer: τ = 540 MPa, δ = 50.6 mm, free length 130 mm, no buckling
A concentric pair · Outer spring d = 8 mm, C = 6, 10 active coils; total load 3,000 N; inner spring with the same index and solid length, radial clearance (do − di)/2
- di = do(C − 2)/C = 8 × 4/6 = 5.333 mm; Di = 6 × 5.333 = 32.0 mm
- Same solid length: Ni = 10 × 8/5.333 = 15 coils
- Load split as d²: Po = 3,000 × 64/(64 + 28.44) = 2077 N; Pi = 923 N
- Stress in both: K × 8 × 2077 × 48/(π × 8³) = 621 MPa (the same for the inner spring)
- Clearance = (8 − 5.333)/2 = 1.33 mm, and the outer coil's inside radius minus the inner coil's outside radius is the same 1.33 mm
Answer: Inner spring 5.33 mm wire, 32 mm mean diameter, 15 coils; loads 2077 N and 923 N
Springs in series and in parallel · Two springs, 20 N/mm and 30 N/mm, carrying 300 N
- Series: 1/k = 1/20 + 1/30, k = 12 N/mm; deflection = 300/12 = 25 mm (each carries 300 N)
- Parallel: k = 20 + 30 = 50 N/mm; deflection = 300/50 = 6 mm
- In parallel the stiffer spring carries more: 30 × 6 = 180 N and 20 × 6 = 120 N
Answer: 25 mm in series, 6 mm in parallel
Leaf springs, nipping and rubber springs
Which syllabus covers thisAU ME3591 Unit IVVTU not in syllabusSPPU 302043 Unit 6MU MEC601 Module 6KTU PCMET602 Module 3
How is a semi-elliptic leaf spring built?
A laminated spring is a stack of flat steel leaves of equal width, clamped at the middle by a centre bolt and U-bolts to the axle. The longest, the master leaf, is rolled into an eye at each end, one pinned to the frame and the other to a swinging shackle so the spring can lengthen as it flattens. One or more extra full-length leaves lie under it; the rest are graduated, each shorter than the one above. Rebound clips keep the leaves together when the spring is pulled back. Rubbing between leaves adds friction damping. Quarter-elliptic and cantilever springs use the same leaves fixed at one end.
Why do the full-length leaves carry more stress than the graduated ones?
Each half of a semi-elliptic spring of span 2L, carrying 2W at the middle, is a cantilever L long with W at its end. The graduated leaves together act like a triangular plate, a beam of uniform strength; the full-length leaves act like a rectangular cantilever, which bends more easily near the tip. To deflect equally the full leaves take a larger share of the load, Wf = 3nfW/(3nf + 2ng), and their stress, 18WL/(bt²(3nf + 2ng)), is 50% higher than the graduated leaves' 12WL/(bt²(3nf + 2ng)). Deflection is 12WL³/(Ebt³(3nf + 2ng)).
What is nipping?
To even out the stresses, the full-length leaves are bent to a larger radius than the graduated leaves, leaving a gap, the nip C, at their ends before assembly. Tightening the centre bolt closes the gap and pre-stresses the stack: the full leaves start with a little reverse bending and the graduated leaves with a little extra. Choose C so that at full load every leaf has the same stress 6WL/(nbt²). Then C = 2WL³/(Enbt³) and the bolt load needed to close it is Pi = 2nf ng W/(n(3nf + 2ng)). The spring's stiffness does not change, only the share of stress.
How do rubber springs work?
Rubber stretches a great deal with little force and loses some energy every cycle, so rubber mounts both carry a machine and damp its vibration. They are usually bonded between metal plates and loaded in shear, where the behaviour is closest to linear: a block of area A and thickness h has a shear stiffness k ≈ G A/h. G for engineering rubbers is roughly 0.4-1.5 MPa depending on hardness (typical). Rubber is not loaded in plain compression without care, because the bulging sides make it much stiffer, and it creeps, ages in oil, ozone and heat, and softens at high strain.
A leaf spring has 2 full-length and 8 graduated leaves, without nipping. How does the stress in the full leaves compare with the graduated leaves?
18WL against 12WL over the same denominator: the full leaves always run 1.5 times the graduated-leaf stress, whatever the number of leaves. Nipping is the cure.
Worked examples
Leaf thickness for a nipped spring · Span 1,100 mm (L = 550 mm), central load 20 kN (W = 10 kN per half), 10 leaves 75 mm wide, nipped so all leaves have equal stress; allowable 450 MPa (illustrative)
- Equal stress after nipping: σ = 6WL/(n b t²)
- t = √(6 × 10,000 × 550/(10 × 75 × 450)) = 9.89 mm
- Take t = 10 mm
- Stress at full load = 6 × 10,000 × 550/(10 × 75 × 10²) = 440 MPa
Answer: t = 10 mm; every leaf at 440 MPa
Stresses, deflection and nip · The same spring: L = 550 mm, W = 10 kN, b = 75 mm, t = 10 mm, nf = 2 extra full-length leaves (with the master), ng = 8 graduated; E = 207 GPa
- 3nf + 2ng = 6 + 16 = 22
- Without nipping: σf = 18 × 10,000 × 550/(75 × 100 × 22) = 600 MPa, σg = 400 MPa
- Deflection δ = 12 × 10,000 × 550³/(2,07,000 × 75 × 1,000 × 22) = 58.5 mm
- Nip C = 2 × 10,000 × 550³/(2,07,000 × 10 × 75 × 1,000) = 21.4 mm
- Centre-bolt load Pi = 2 × 2 × 8 × 10,000/(10 × 22) = 1,455 N
Answer: Without nip: 600 and 400 MPa; with a 21.4 mm nip and 1.45 kN bolt load: 440 MPa in all leaves; δ = 58 mm
Rubber mounts under a machine · A 2 kN machine on four bonded rubber blocks 50 × 50 mm, 20 mm thick, loaded in shear; G = 0.8 MPa (typical)
- One block: k = G A/h = 0.8 × 2,500/20 = 100 N/mm
- Four in parallel: 400 N/mm
- Static deflection = 2,000/400 = 5 mm
- Natural frequency f = (1/2π)√(g/δ) = (1/2π)√(9,810/5) = 7.05 Hz
Answer: 5 mm static deflection, natural frequency about 7.0 Hz
Flywheels for punching presses and engines: energy, size and rim stress
Which syllabus covers thisAU ME3591 Unit IVVTU not in syllabusSPPU not in syllabusMU MEC601 Module 5KTU not in syllabus
What do the fluctuation of energy and the coefficient of fluctuation of speed mean?
A flywheel stores energy when the driving torque exceeds the load and gives it back when it falls short; its speed rises and falls in step. The maximum fluctuation of energy ΔE is the largest swing in that stored energy over a cycle. The coefficient of fluctuation of speed Cs = (ωmax − ωmin)/ω compares the speed swing with the mean speed: about 0.002-0.02 for generators and engines, 0.1-0.2 for punches and shears (typical). Since ½I(ωmax² − ωmin²) = ΔE, the flywheel needs I = ΔE/(ω² Cs). A faster flywheel needs far less inertia.
Where does ΔE come from for an engine?
An engine's torque swings widely through each cycle: high on the power stroke, negative on compression. The turning-moment diagram plots that torque against crank angle; the mean torque line cuts it into loops above and below. The areas of the loops are energy surpluses and deficits, and ΔE is the largest difference between the highest and lowest points of the running total. The Automobile Engineering book's flywheel interactive builds this diagram for single- and multi-cylinder engines. Here we take ΔE as given and size the wheel: 1.8 kJ at 1,500 rpm with Cs = 0.02 needs I = 3.65 kg·m².
How is a punching-press flywheel sized?
A punch works only during a short part of each stroke but needs a large force then: Fmax = π d t τu. The force falls roughly linearly as the punch goes through the plate, so the energy per hole is ½ Fmax t. A motor sized for that peak would sit idle most of the time; instead a small motor runs steadily, supplying the energy over the whole cycle, and the flywheel provides the deficit while punching, ΔE = E(1 − θ/360), θ being the crank angle while the punch is in the plate. A step-by-step integration with the falling force gives a ΔE a few per cent higher, so round I up.
Solid disc or rimmed flywheel?
Inertia comes from mass far from the axis. A rim of mass m at mean radius R has I ≈ m R²; a solid disc of radius R has only ½ m R², so for the same I and outer radius a disc is twice as heavy. Rimmed flywheels, a heavy rim joined to the hub by arms or a thin web, therefore suit large slow machines such as presses and old steam engines; the rim is usually taken to give about 90% of the inertia, the hub and arms the rest. Solid discs, simpler and safer at high speed, are used on small engines and where space is tight.
What limits a flywheel's speed?
Each bit of a spinning rim pulls outward, and the rim holds itself together by hoop tension. For a thin free rim the hoop stress is σ = ρ v², independent of the rim's cross-section: only the rim speed matters. Cast iron, with ρ = 7,200 kg/m³ and an allowable around 6 MPa (illustrative, since cast iron is weak in tension), is limited to about 29 m/s. The arms add bending: the rim expands more than the arms allow, and the arms also carry the torque as cantilevers from the hub. Steel or welded rims allow higher speeds; bursting flywheels have caused deaths, so the speed limit is not optional.
A flywheel's speed is doubled (same ΔE and Cs). How does the moment of inertia it needs change?
I = ΔE/(ω² Cs): doubling ω cuts the required inertia to a quarter. That is why presses gear the flywheel up rather than mount it on the slow crankshaft, as long as the rim speed (σ = ρv²) stays safe.
Worked examples
Flywheel for a punching press · 20 mm holes in 12 mm plate, τu = 350 MPa (typical mild steel); 20 strokes a minute; crank radius 50 mm (illustrative); flywheel at 300 rpm, Cs = 0.1, rim mean radius 400 mm, rim giving 90% of I, cast iron 7,200 kg/m³
- Fmax = π × 20 × 12 × 350 = 2,63,894 N; energy per hole = ½ × 2,63,894 × 12 = 15,83,363 N·mm = 1583 J
- Crank angle while punching: 50(1 − cos θ) = 12 → θ = 40.5°
- ΔE = 1583 × (1 − 40.5/360) = 1405 J; motor power = 1583 × 20/60 = 528 W before losses
- ω = 2π × 300/60 = 31.42 rad/s; I = 1405/(31.42² × 0.1) = 14.24 kg·m²
- Rim mass = 0.9 × 14.24/0.4² = 80.1 kg; rim speed 12.6 m/s, hoop stress 7,200 × 12.57² = 1.14 MPa
Answer: I ≈ 14.2 kg·m², rim about 80 kg, motor about 0.53 kW plus losses
Engine flywheel from a given ΔE · ΔE = 1.8 kJ from the turning-moment diagram; 1,500 rpm; speed to stay within ±1% (Cs = 0.02); rim mean radius 0.25 m giving all the inertia; cast iron
- ω = 2π × 1,500/60 = 157.08 rad/s
- I = 1,800/(157.08² × 0.02) = 3.648 kg·m²
- Rim mass = 3.648/0.25² = 58.4 kg; section = 58.4/(7,200 × 2π × 0.25) = 5,160 mm², e.g. about 72 × 72 mm
- Rim speed = 157.08 × 0.25 = 39.3 m/s; σ = 7,200 × 39.27² = 11.1 MPa: too high for cast iron (about 29 m/s at 6 MPa), so use steel or a smaller radius
Answer: I = 3.65 kg·m², 58 kg rim, but the 39 m/s rim speed needs steel
Disc against rim for the same inertia · I = 3.648 kg·m², outer radius 0.3 m
- Solid disc: I = ½ m R² → m = 2 × 3.648/0.09 = 81.1 kg
- Thin rim at 0.3 m: m = 3.648/0.09 = 40.5 kg
- The disc is twice as heavy for the same smoothing, but has no arms to fail and a lower stress at the bore
Answer: 81 kg disc against 41 kg rim
Connecting rods and crankshafts
Which syllabus covers thisAU ME3591 Unit IVVTU not in syllabusSPPU not in syllabusMU not in syllabusKTU not in syllabus
What forces act on a connecting rod?
The rod links the piston to the crank. The largest load is the gas force (π/4)D²p near top dead centre on the power stroke, which puts the rod in compression. The inertia of the piston and the reciprocating part of the rod, mR ω² r (cos θ + cos 2θ/n) with n = l/r, opposes the gas force on the down-stroke and adds to it on the up-stroke; at TDC on the exhaust stroke it is the only load and pulls the rod in tension, which is what the big-end cap bolts carry. The rod's own swinging adds a sideways inertia bending, whipping. Piston and bearing friction is small and usually neglected.
Why does a connecting rod have an I-section 4t wide and 5t deep?
In compression the rod is a strut and can buckle two ways. In the plane of the crank's rotation its ends are free to turn on the pins (hinged, length L); at right angles the pins hold it square (treated as fixed, length L/2). Fixed ends are four times as stiff, so the section should be about four times stiffer in the plane of rotation: Ixx ≈ 4 Iyy. The standard I-section, flanges 4t wide, depth 5t, flange and web t thick, gives A = 11t², Ixx = 419t⁴/12, Iyy = 131t⁴/12, a ratio of 3.2, close enough, and is light for its stiffness. It is forged with the web in the plane of rotation.
How is the section sized?
The rod is not a long slender column, so Euler's formula overestimates its strength; the Rankine-Gordon formula blends crushing and buckling: Pcr = σc A/(1 + a(L/k)²), with a = 1/7500 for mild steel. Set Pcr about the xx axis equal to the gas force times a factor of safety, usually 5 to 6 for engine rods to cover the inertia and fatigue loads, and solve for t. A 100 mm bore at 5 MPa with a 300 mm rod and σc = 330 MPa needs t ≈ 8.3 mm: a 33 × 41 mm section. Then check yy, which comes out slightly stronger.
What is whipping, and how big is it?
As the crank turns, each point of the rod moves on a path between the piston's straight line and the crankpin's circle, so its sideways acceleration grows from zero at the small end to ω²r at the big end. That inertia load bends the rod like a beam with a triangular load. Taking it as simply supported, the bending moment peaks at L/√3 from the small end: Mmax = m ω² r L/(9√3), m being the rod's mass. It grows with the square of speed, so it matters in fast engines, but it peaks around 65-70° of crank angle, not at TDC with the gas load, so the two are checked separately.
How is a crankshaft checked?
A crankshaft is examined in two positions. At top dead centre the gas force is largest and acts along the crank, so the crankpin, webs and journals are in bending and direct stress only: for a centre crank between two bearings, each takes half the force and the crankpin's bending moment is (F/2)·b, b being the bearing-to-pin distance. At the angle of maximum torque, about 25-35° after TDC, the force has a tangential component that also twists the shaft, so the journal sections are checked for combined bending and torsion, as in the shaft chapter. Crankpin length comes from the allowable bearing pressure, F/(d l).
For a 4t × 5t connecting rod, which way does it buckle first if the ends are hinged in the plane of rotation and fixed at right angles to it?
(L/k)² about xx is L²/(3.17t²) = 0.315 L²/t²; about yy with L/2 it is L²/(4 × 0.99t²) = 0.252 L²/t². The xx axis is more slender, so it governs; in the worked example the yy buckling load is 1.1% higher.
Worked examples
Sizing the I-section · Bore 100 mm, maximum gas pressure 5 MPa, rod length 300 mm, FoS 6, compressive yield 330 MPa, Rankine a = 1/7500
- Gas force F = (π/4) × 100² × 5 = 39,270 N; design load = 6F = 2,35,619 N
- kxx² = 419/132 t² = 3.1742 t²; 330 × 11t² = 2,35,619 (1 + 300²/(7500 × 3.1742 t²))
- Rearranged: 3,630 t⁴ − 2,35,619 t² − 8,90,743 = 0 → t² = 68.49, t = 8.28 mm
- Section 33.1 mm wide × 41.4 mm deep, flange and web 8.3 mm
- Check yy (L/2, kyy² = 0.992 t²): Pcr = 2,38,115 N > 2,35,619 N
Answer: t ≈ 8.28 mm, say 8.5 mm: 34 × 42.5 mm
Whipping stress at 3,000 rpm · The rod above (t = 8.28 mm, L = 300 mm), steel 7,800 kg/m³ (typical), crank radius 75 mm, 3,000 rpm
- Shank mass m = 7,800 × 11 × (8.28 × 10⁻³)² × 0.3 = 1.763 kg
- ω = 2π × 3,000/60 = 314.2 rad/s
- Mmax = 1.763 × 314.2² × 0.075 × 0.3/(9√3) = 251.1 N·m
- Zxx = 419t³/30 = 7,917 mm³; σ = 2,51,145/7,917 = 31.7 MPa
Answer: About 32 MPa of bending at the critical crank angle, on top of a much smaller axial load there
Big-end cap bolts · Reciprocating mass 1.6 kg (illustrative), 3,000 rpm, crank radius 75 mm, rod 300 mm, two bolts, allowable tensile stress 80 MPa (illustrative, high FoS for repeated load)
- Inertia force at TDC on the exhaust stroke: mR ω² r (1 + r/l) = 1.6 × 314.2² × 0.075 × 1.25 = 14,804 N
- Stress area per bolt = 14,804/(2 × 80) = 92.5 mm²
- From the ISO coarse table: M12 has 84.3 mm², too small; M16 has 157 mm²
Answer: Two M16 bolts
Crankpin of a centre crank at TDC · Gas force 39,270 N on the crankpin; main bearings 200 mm apart, crankpin midway; allowable bending 75 MPa and bearing pressure 10 MPa (illustrative)
- Each bearing takes 19,635 N; M = 19,635 × 100 = 19,63,495 N·mm
- d = (32M/(π × 75))^(1/3) = 64.4 mm
- Crankpin length for 10 MPa: l = 39,270/(64.4 × 10) = 61 mm
Answer: Crankpin about 65 mm diameter and 62 mm long
Flat belts, V-belts and roller chains
Which syllabus covers thisAU not in syllabusVTU not in syllabusSPPU not in syllabusMU MEC601 Module 5KTU PCMET602 Module 4
How does a flat belt transmit power, and why does it slip and creep?
A belt drives by friction. The driver pulls the belt in on the tight side (tension T1) and lets it out on the slack side (T2); the torque on each pulley is (T1 − T2) × radius. Creep is unavoidable: the belt stretches more under T1 than under T2, so it shrinks slightly as it runs round the driver and the driven pulley turns a little slower than the speed ratio suggests. Slip is different: when the torque asks for more than friction can give, the whole belt slides over the pulley. Leather, rubberised fabric and polyamide-cored belts are common, and the arc of contact on the small pulley sets the grip.
Where do the tension ratio and the centrifugal tension come from?
Take a short piece of belt on the pulley. The pull grows along the arc because each piece is gripped by friction μ times the pressure under it, and that pressure comes from the tension pressing the belt onto the rim. Integrating gives T1/T2 = e^(μθ). A running belt also carries centrifugal tension Tc = m v² (m in kg/m, v in m/s), which stretches the belt everywhere but does not press it onto the pulley, so the ratio applies to T − Tc. A 0.5 kg/m belt at 10 m/s already carries 50 N that does no work.
Why does a V-belt grip better, and how many belts does a drive need?
A V-belt sits on the sides of a groove of angle 2β (34-40°). The wedge multiplies the normal force by 1/sin β, so the ratio becomes e^(μθ/sin β): with a 40° groove that is μ multiplied by 2.92. One belt with T1 at its allowable value carries (Tmax − m v²)(1 − 1/ratio) v. The design power is the motor power times a service factor for the kind of driver and load, and the number of belts is design power ÷ power per belt, rounded up. Belt makers publish ratings per section; the interactive uses an illustrative belt.
Is there a best belt speed?
Yes. With T1 held at Tmax, power = (Tmax − m v²)(1 − 1/ratio) v. At low speed the belt is slow; at high speed centrifugal tension eats the allowable pull. Setting the derivative to zero gives maximum power when Tc = Tmax/3, that is v = √(Tmax/3m). For a 1,250 N, 0.5 kg/m flat belt that is 28.9 m/s. Many drives run slower than this, because bearings, pulley rims and belt life suffer at high speed.
How is a roller chain different, and what is chordal action?
A roller chain cannot slip, so the speed ratio is exact and there is no initial tension. But it wraps the sprocket as a polygon of z sides, not a circle. As each link seats, the chain line rises and falls between the pitch radius R and R cos(180°/z), so the chain speed pulses by 1 − cos(180°/z): 1.7% with 17 teeth, 0.8% with 25. That is why designers avoid very small sprockets. Selection starts from the design power and speed of the small sprocket, uses the maker's rating chart to pick the pitch, then the number of links and the lubrication.
A belt's allowable tension is fixed. At what speed does it carry the most power?
Power ∝ (Tmax − m v²) v; its maximum is at m v² = Tmax/3. For a 1,250 N, 0.5 kg/m belt: v = √(1,250/1.5) = 28.9 m/s. At m v² = Tmax the belt carries nothing at all.
Worked examples
Power of an open flat-belt drive · Driver 200 mm at 960 rpm, driven 500 mm, centre distance 1,500 mm, μ = 0.3, belt 100 × 5 mm, 0.5 kg/m, allowable 2.5 MPa
- θ = π − 2 sin⁻¹(300/3,000) = 2.9413 rad = 168.5°
- L = 3,000 + π(700)/2 + 300²/6,000 = 4,114.6 mm
- Tmax = 2.5 × 100 × 5 = 1,250 N; v = π × 0.2 × 960/60 = 10.053 m/s; Tc = 0.5 × 10.053² = 50.5 N
- e^(0.3 × 2.9413) = 2.4166, so T2 = 50.5 + (1,250 − 50.5)/2.4166 = 546.9 N
- P = (1,250 − 546.9) × 10.053 = 7.07 kW
Answer: About 7.07 kW, with a 4,115 mm belt
Number of V-belts · 18.5 kW motor, service factor 1.3, pulleys 180 and 540 mm, C = 900 mm, 1,440 rpm, groove 40°, μ = 0.28; one belt (illustrative): 0.19 kg/m, Tmax 600 N
- Design power = 18.5 × 1.3 = 24.05 kW
- θ = π − 2 sin⁻¹(360/1,800) = 156.9° = 2.7389 rad; v = 13.57 m/s; Tc = 35.0 N
- Ratio = e^(0.28 × 2.7389/sin 20°) = 9.41
- Per belt = (600 − 35.0)(1 − 1/9.41) × 13.57 = 6.85 kW
- Belts = 24.05/6.85 = 3.51, round up; belt length 2,967 mm before choosing the nearest standard length
Answer: 4 belts
Chordal action of a chain sprocket · Chain of 15.875 mm pitch; sprockets of 17 and 25 teeth; small sprocket at 600 rpm
- Pitch diameters: 15.875/sin(180°/17) = 86.4 mm; 15.875/sin(180°/25) = 126.7 mm
- Mean chain speed (17 teeth) = z p N/60,000 = 17 × 15.875 × 600/60,000 = 2.70 m/s
- Speed variation = 1 − cos(180°/17) = 1.70% with 17 teeth
- With 25 teeth: 1 − cos 7.2° = 0.79%
Answer: 1.7% speed ripple with 17 teeth, 0.8% with 25
Spur gears: Lewis equation, dynamic load and wear
Which syllabus covers thisAU not in syllabusVTU BME602 Module 3SPPU not in syllabusMU not in syllabusKTU not in syllabus
What are the basic spur gear terms?
The pitch circle is the imaginary rolling circle: d = z m, where z is the number of teeth and m the module in mm. Circular pitch p = π m. A standard 20° full-depth tooth has addendum m above the pitch circle and dedendum 1.25 m below it. The pressure angle φ is the angle of the line along which teeth push each other. Speed ratio = zg/zp. A pinion with too few teeth interferes with the gear: for a rack and 20° full-depth teeth the limit is 2/sin²φ = 17.1, so pinions are usually 18 teeth or more.
What does the Lewis equation assume, and what is the form factor?
Lewis treated a tooth as a cantilever loaded at its tip by the whole tangential force, and fitted a parabola of uniform strength inside the tooth profile; the weakest section is where the parabola touches the root fillets. That gives σ = Pt/(b m Y), where Y is the Lewis form factor, a number set by tooth shape only. Y grows with the number of teeth (0.230 at 12 teeth, 0.308 at 20, 0.479 for a rack, 20° full depth), which is why the smaller pinion is usually the weaker wheel. Face width b is commonly 9.5-12.5 times the module.
How does pitch-line speed enter the bending design?
Real teeth are not perfect, so each pair of teeth meets with a small impact, and the faster the gears turn the worse it gets. Barth's velocity factor reduces the allowable load: 6/(6 + v) for cut or milled teeth, 50/(50 + √(200 v)) for hobbed or shaped teeth, and √(78/(78 + √(200 v))) for shaved or ground teeth, with v in m/s. The service (overload) factor Ko adds the roughness of the driver and the load, for example 1.25 for an electric motor driving a machine with moderate shocks. Effective load = Ko Pt / Kv must not exceed the beam strength.
What is Buckingham's dynamic load?
Buckingham worked out the extra force from tooth errors directly. A profile error e makes the gears accelerate and decelerate as teeth engage; the increment depends on the error, the tooth stiffness and the speed. The deformation factor C = 11,860 e N/mm for steel on steel 20° full-depth teeth, with e from a table by module and quality: 0.050 mm for first-class commercial gears up to module 4, 0.025 mm carefully cut, 0.0125 mm precision. The total dynamic load Fd = Pt + Fi must stay below the beam strength and the wear strength.
How is the wear (surface) strength found?
Teeth also fail by pitting: the contact between two teeth is like two cylinders pressed together, with radii d sin φ/2 at the pitch point, and Hertz contact stress repeated millions of times starts surface cracks. Buckingham's wear strength is the tangential load that raises the contact stress to the surface fatigue strength σsf, about 2.8 BHN − 69 MPa for steel. Wear strength rises with the square of σsf, so surface hardening is the usual cure. A pinion that is strong in bending is often too soft for wear.
Keep the material, teeth and the face width at 10 modules, and double the module. What happens to the Lewis beam strength?
S_b = [σ] b Y m with b = 10 m, so S_b = 10 [σ] Y m²: twice the module gives four times the strength (and twice the pitch diameter).
Worked examples
Lewis bending check of a pinion · 20-tooth pinion, module 5 mm, face 50 mm, SAE 1045 WQT ([σ] = 456 MPa, 205 BHN), 15 kW at 960 rpm, hobbed teeth, Ko = 1.25
- d = 20 × 5 = 100 mm; v = π × 0.1 × 960/60 = 5.027 m/s
- Pt = 15,000/5.027 = 2,984 N
- Y(20) = 0.30769; Sb = 456 × 50 × 0.30769 × 5 = 35,077 N
- Kv = 50/(50 + √(200 × 5.027)) = 0.612; Peff = 1.25 × 2,984/0.612 = 6,096 N
- FoS = 35,077/6,096 = 5.75
Answer: Safe in bending with a factor of safety of 5.8
Buckingham's dynamic load · Same pinion, carefully cut teeth (e = 0.025 mm for module 5), steel on steel
- C = 11,860 × 0.025 = 296.5 N/mm; C b + Pt = 296.5 × 50 + 2,984 = 17,809 N
- Fi = 9.84 × 5.027 × 17,809 / (9.84 × 5.027 + 0.4696 × √17,809) = 7,856 N
- Fd = 2,984 + 7,856 = 10,840 N
- Sb/Fd = 35,077/10,840 = 3.24
Answer: Fd = 10.8 kN, about 3.6 times the transmitted load; bending is still safe (3.2)
Wear strength and the hardness needed · Same pair, gear ratio 3, both wheels 205 BHN, φ = 20°, Cp = 191 √MPa
- σsf = 2.8 × 205 − 69 = 505 MPa; I = sin 20° cos 20°/2 × 3/4 = 0.1205
- Sw = 50 × 100 × 0.1205 × (505/191)² = 4,213 N
- Sw/Fd = 4,213/10,840 = 0.39: the teeth would pit
- Needed σsf = 191 √(10,840/(50 × 100 × 0.1205)) = 810 MPa, so BHN = (810 + 69)/2.8 = 314
Answer: Wear governs: harden both wheels to at least about 320 BHN
Helical, bevel and worm gears
Which syllabus covers thisAU not in syllabusVTU BME602 Module 4SPPU not in syllabusMU not in syllabusKTU not in syllabus
What changes when the teeth are helical?
Helical teeth are cut at a helix angle ψ (typically 15-30°), so contact starts at one end of a tooth and spreads across the face: smoother, quieter and stronger than spur gears. Two modules now exist. The normal module mn (in the plane square to the teeth) is the cutter's module; the transverse module mt = mn/cos ψ sets the pitch diameter d = z mt. The tooth force has three parts: tangential Pt, radial Pt tan φn/cos ψ and an axial thrust Pt tan ψ that the bearings must carry. Herringbone (double-helical) gears cancel the thrust.
What is the formative (virtual) number of teeth?
Cut a helical gear square to its teeth and the pitch circle shows up as an ellipse. At the tooth in contact the ellipse is flatter than the real pitch circle: its radius of curvature is d/(2 cos²ψ). A spur gear with that radius would have z/cos³ψ teeth, and the tooth there behaves like a tooth of that spur gear. So the Lewis form factor of a helical gear is read at the formative number z/cos³ψ, not at z. A bevel gear does the same along its back cone: formative teeth z/cos δ.
How are straight bevel gears sized?
Bevel gears join intersecting shafts, usually at 90°, and roll like two cones. Pitch angles follow from the teeth: tan δp = zp/zg, δg = 90° − δp. The cone distance A0 = (m/2)√(zp² + zg²) is the length of the cone's slant. Teeth shrink toward the apex, so the face width is kept to about A0/3 and the Lewis beam strength is multiplied by (1 − b/A0). Forces act at the mean radius; on the pinion the radial force is Pt tan φ cos δp and the axial thrust Pt tan φ sin δp, which becomes the radial force on the gear.
How efficient is a worm drive, and when does it lock itself?
A worm is a screw with z1 starts meshing with a wheel, giving ratios of 10-100 in one step between crossed shafts. The diametral quotient q = d1/m sets the worm size and the lead angle: tan γ = z1/q. The teeth slide, at vs = v1/cos γ, so friction decides the efficiency. Efficiency rises with the lead angle. With the wheel driving, the efficiency is (cos φn − μ cot γ)/(cos φn + μ tan γ); if μ ≥ tan γ cos φn it is zero or negative and the drive is self-locking. Running friction is low, so many single-start worms still back-drive: do not count on them as brakes.
What are worms made of, and why does heat matter?
The usual pair is a case-hardened and ground steel worm running on a phosphor-bronze wheel: the soft bronze runs in, holds the oil film and resists seizure under heavy sliding. Cast iron wheels suit slow, light drives. Worms are cylindrical (single enveloping: the wheel wraps the worm) or double enveloping (both wrap each other, more contact and capacity). Every watt lost to friction, (1 − η) times the input power, heats the oil in the gearbox housing, so worm drives are often rated by how much heat the housing can shed rather than by tooth strength.
A worm drive is changed from a single-start worm to a two-start worm with the same q and friction. What happens to its efficiency?
The lead angle grows from 5.71° to 11.31°, so less of the sliding is wasted: with μ = 0.035 the efficiency rises from 72.6% to 83.7% (and the speed ratio halves).
Worked examples
Helical pinion: geometry and forces · 24 teeth, normal module 4 mm, helix angle 20°, φn = 20°, torque 150 N·m
- mt = 4/cos 20° = 4.257 mm; d = 24 × 4.257 = 102.16 mm
- Formative teeth = 24/cos³20° = 28.9 (read Y at about 29 teeth)
- Pt = 2 × 150,000/102.16 = 2,937 N
- Pr = 2,937 × tan 20°/cos 20° = 1,137 N; Pa = 2,937 × tan 20° = 1,069 N
Answer: d = 102.2 mm; forces 2,937 N tangential, 1,137 N radial, 1,069 N axial
Straight bevel pinion: strength at the mean radius · 18 and 36 teeth at 90°, module 5 mm, face 30 mm, SAE 1040 ([σ] = 380 MPa), pinion torque 90 N·m
- δp = tan⁻¹(18/36) = 26.57°; A0 = 2.5 √(18² + 36²) = 100.6 mm (A0/3 = 33.5 mm, so 30 mm is fine)
- Formative teeth: 18/cos 26.57° = 20.12, 36/cos 63.43° = 80.5; Y(20.12) = 0.3085
- Sb = 5 × 30 × 380 × 0.3085 × (1 − 30/100.6) = 12,343 N
- Mean radius = 45 − 15 sin 26.57° = 38.29 mm; Pt = 90,000/38.29 = 2,350 N
- Pr = 2,350 tan 20° cos 26.57° = 765 N; Pa = 2,350 tan 20° sin 26.57° = 383 N
Answer: Static beam strength 12.3 kN against 2.35 kN transmitted (ratio 5.3 before speed and service factors)
Worm efficiency, heat and back-driving · Worm module 6 mm, q = 10, 1,440 rpm, μ = 0.035, φn = 20°; compare two starts with one start; 10 kW input
- Two starts: γ = tan⁻¹(2/10) = 11.31°; d1 = 60 mm, v1 = 4.52 m/s, vs = 4.61 m/s
- η = (cos 20° − 0.035 × 0.2)/(cos 20° + 0.035 × 5) = 83.7%; heat = 10 × (1 − 0.837) = 1.63 kW
- Wheel driving: η' = (cos 20° − 0.035 × 5)/(cos 20° + 0.035 × 0.2) = 80.8%: it back-drives
- One start: γ = 5.71°, η = 72.6%, η' = 62.5%: still back-drives
- Self-locking needs μ ≥ tan γ cos φn = 0.094 (one start), more than twice this μ
Answer: Two starts: 84% efficient, 1.6 kW of heat; neither worm is self-locking at μ = 0.035
Plate clutches, band brakes and block brakes
Which syllabus covers thisAU not in syllabusVTU BME602 Module 5SPPU not in syllabusMU MEC601 Module 6KTU not in syllabus
Uniform pressure or uniform wear: which torque formula do you use for a plate clutch?
A new clutch, with flat faces, presses evenly: uniform pressure, friction radius 2/3 (ro³ − ri³)/(ro² − ri²). After running in, the outer part, sliding faster, wears more until wear is uniform; then pressure × radius is constant, the peak pressure sits at the inner radius and the friction radius is (ro + ri)/2. Uniform wear gives slightly less torque, so designers use it for a safe answer. A clutch with n pairs of rubbing faces carries n times the torque of one pair: 2 for a single plate gripped on both sides. The automobile book's clutch interactive shows the same torque on a car clutch.
When are multi-plate and cone clutches used?
When the diameter is limited, more plates give more torque: alternate plates are splined to the shaft and to the housing, and the number of friction pairs is the number of plates minus one. Wet multi-plate clutches run in oil, which cools them but lowers μ. A cone clutch has conical faces of semi-cone angle α; the axial force W presses the faces with a normal force W/sin α, so it gives torque μ W Rm/sin α for a mean radius Rm. Make α too small and the cone wedges in and will not release cleanly.
How does a band brake work, and why does the drum direction matter?
A steel band lined with friction material wraps the drum through an angle θ. As in a belt, the tensions at the two ends differ by e^(μθ), and the braking torque is (T1 − T2) r. The tight end is the one where the drum surface enters the band; reverse the drum and the ends swap. In a simple band brake one end is anchored at the lever fulcrum; attaching the slack end to the lever needs far less force. The interactive takes moments about the fulcrum for each layout and direction.
What makes a differential band brake self-locking?
In a differential band brake both ends go to the lever, on opposite sides of the fulcrum: the slack end at b2, on the side of the operating force, and the tight end at b1 on the other side. Then P l = T2 b2 − T1 b1: the tight end helps the operator. If b2 ≤ b1 e^(μθ), the force needed is zero or negative and the brake applies itself once touched. That is useful as a backstop on a hoist, which must hold the load against reverse rotation, but dangerous in a brake that has to be released.
When is a block brake self-energising?
A single block (shoe) brake presses one shoe on the drum with a lever. For a short shoe, the normal force N gives friction μN, and braking torque μ N r. The friction force acts along the drum surface, at a height c from the lever pivot. If it turns the lever the same way as the applied force P, it helps: the brake is self-energising and P = N (a − μc)/l. Turn the drum the other way and the friction resists: P = N (a + μc)/l. If μc ≥ a the shoe locks itself on. Putting the pivot on the friction line (c = 0) makes the brake the same in both directions.
In a simple band brake, the free end of the band goes to the lever and the other end to the fulcrum. Which end should be on the lever?
P l = T b for the end on the lever. With the drum and band of example 2 (T1 = 2,890 N, T2 = 890 N, b = 150 mm, l = 500 mm) the slack end needs 267 N and the tight end 867 N.
Worked examples
Single-plate clutch: uniform wear vs uniform pressure · Faces 300 mm outside and 200 mm inside diameter, both sides of the plate active, μ = 0.3, spring force 5 kN, 1,500 rpm
- Uniform wear: friction radius = (150 + 100)/2 = 125 mm; T = 2 × 0.3 × 5,000 × 0.125 = 375 N·m
- Uniform pressure: radius = 2/3 × (150³ − 100³)/(150² − 100²) = 126.67 mm; T = 380 N·m
- Peak pressure (uniform wear, at the inner radius) = 5,000/(2π × 100 × 50) = 0.159 MPa
- Power at 1,500 rpm = 375 × 2π × 1,500/60 = 58.9 kW
Answer: Design on 375 N·m (uniform wear), good for 59 kW at 1,500 rpm
Differential band brake · Drum radius 200 mm, wrap 270°, μ = 0.25, braking torque 400 N·m; lever 500 mm; slack end 150 mm on the force side, tight end 40 mm on the other side
- e^(0.25 × 3π/2) = 3.2482; T1 − T2 = 400,000/200 = 2,000 N
- T2 = 2,000/(3.2482 − 1) = 889.6 N; T1 = 2,889.6 N
- P = (889.6 × 150 − 2,889.6 × 40)/500 = 35.7 N
- Self-locking would need b2 ≤ 40 × 3.2482 = 129.9 mm
Answer: Only 36 N at the lever; shorten b2 below 130 mm and it locks
Block brake in both directions · Drum radius 150 mm, μ = 0.3, braking torque 120 N·m; shoe 200 mm from the pivot, force 500 mm from it; friction line 80 mm from the pivot
- N = 120,000/(0.3 × 150) = 2,666.7 N; friction μN = 800 N
- Friction resisting P: P = 2,666.7 × (200 + 0.3 × 80)/500 = 1,195 N
- Friction helping P (drum reversed): P = 2,666.7 × (200 − 24)/500 = 939 N
- Self-locking only if c ≥ a/μ = 200/0.3 = 667 mm
Answer: 939 N self-energising, 1,195 N the other way; no risk of locking
Lubrication and hydrodynamic journal bearings
Which syllabus covers thisAU ME3591 Unit VVTU BME602 Module 5SPPU not in syllabusMU MEC601 Module 4KTU not in syllabus
What are the lubrication regimes of a sliding bearing?
Plot friction against the bearing characteristic number μN/p (viscosity × speed ÷ unit load) and you get the Stribeck curve. At low values the film is thinner than the surface roughness: boundary lubrication, where metal touches through thin adsorbed layers and friction is high. In the middle, mixed lubrication, part film and part contact. At high values a full hydrodynamic film separates the surfaces and friction drops to its lowest, then rises slowly as the oil is sheared harder. A bearing should run well to the right of the friction minimum, so that a dip in speed or a rise in load does not break the film.
How do lubricants and bearing materials differ?
Mineral oils are graded by kinematic viscosity at 40 °C (ISO VG 32, 46, 68, 100 ...: the number is the mid value in mm²/s). Viscosity falls steeply with temperature: a typical VG 68 oil drops from 68 mm²/s at 40 °C to 8.5 at 100 °C, which the Walther equation fits. Dynamic viscosity = kinematic × density. Bearing materials must be soft enough to embed dirt and conform, yet strong: white metal (babbitt, tin or lead based) for low loads, bronzes, aluminium alloys and porous oil-impregnated sintered bronze for small, hard-to-reach bearings, and polymers such as PTFE and nylon for dry running.
How does a journal bearing build up pressure, and what is the Sommerfeld number?
At rest the journal sits on the bottom of the bearing. As it turns, it drags oil into the narrowing gap ahead of the contact; the oil cannot escape fast enough, pressure builds and lifts the journal, which floats slightly off-centre with a minimum film thickness h0. The Sommerfeld number S = (r/c)² μ Ns/p (Ns in rev/s) gathers all the operating variables into one dimensionless group. Petroff's analysis of a centred journal gives the friction f = 2π² (μ Ns/p)(r/c), which is good for lightly loaded bearings. McKee's measured friction adds an end-leakage term of about 0.002 for L/D of 0.75-2.8.
At what temperature does a self-contained bearing run?
Friction heat Hg = f W v must leave through the housing: Hd = C A B (T0 − Ta), with A ≈ 20 D L the housing area, C about 15.3 W/m²°C for average practice (11.4 in still air, 33.5 with air at 2.5 m/s) and B 0.5-0.83 for oil-ring and oil-bath bearings. But the heat generated depends on the viscosity, which depends on the temperature, so the answer is where the two curves cross. A hotter oil is thinner, generates less heat and runs closer to boundary lubrication. Oil temperature should stay below about 93-123 °C.
What do the Raimondi and Boyd charts give, and what is a hydrostatic bearing?
Raimondi and Boyd solved Reynolds' equation by computer for bearings of L/D = 1/4, 1/2, 1 and infinite length and plotted the results against the Sommerfeld number: minimum film thickness h0/c and eccentricity, friction variable (r/c) f, oil flow, side leakage, temperature rise and peak pressure. A designer computes S, reads the charts, and iterates on the oil temperature. A hydrostatic bearing pumps oil in under pressure from outside, so it carries load even at zero speed: used for very heavy, slow machines such as telescopes and large machine-tool slides.
A lightly loaded journal bearing doubles its speed, oil viscosity unchanged. What does Petroff's equation say about the friction coefficient?
f = 2π² (μ Ns/p)(r/c) is proportional to speed: the oil is sheared twice as fast. The heat generated, f W v, rises four times, so in practice the oil warms, thins and the rise is smaller.
Worked examples
Petroff friction and power loss · Journal 60 mm, length 60 mm, c/r = 0.001, 1,500 rpm, load 4 kN, oil viscosity 20 mPa·s
- p = 4,000/(0.06 × 0.06) = 1.111 MPa; Ns = 25 rev/s
- S = (1,000)² × 0.02 × 25/1,111,111 = 0.45
- f = 2π² × (0.02 × 25/1,111,111) × 1,000 = 0.0089
- v = π × 0.06 × 25 = 4.712 m/s; heat = 0.0089 × 4,000 × 4.712 = 167 W
Answer: f ≈ 0.0089, about 167 W of heat
Viscosity at running temperature · ISO VG 46 oil: 46 mm²/s at 40 °C, 6.8 at 100 °C, 870 kg/m³; oil runs at 70 °C
- Walther: log log(ν + 0.7) = A − B log T, T in kelvin; fit A and B through the two data points
- B = [log log 46.7 − log log 7.5]/[log 373.15 − log 313.15] = 3.684
- At 343.15 K: ν = 14.85 mm²/s
- μ = ν ρ = 14.85 × 10⁻⁶ × 870 = 0.0129 Pa·s = 12.9 mPa·s (density kept at its 15 °C value)
Answer: About 14.8 mm²/s, 12.9 mPa·s at 70 °C: a third of its 40 °C value
Heat balance of a ring-oiled bearing · Journal 80 mm, L/D = 1, c/r = 0.0012, 10 kN, 600 rpm, VG 68 oil; air at 30 °C, C = 15.3 W/m²°C, B = 0.6
- p = 10,000/(0.08 × 0.08) = 1.562 MPa
- Heat dissipated per °C = 15.3 × 20 × 0.08 × 0.08 × 0.6 = 1.175 W/°C
- Try 40 °C: oil 68 mm²/s, Petroff heat 157 W but the housing sheds only 12 W: hotter
- Bisect until generated = dissipated: T0 = 69.5 °C, μ = 17.5 mPa·s, heat 46 W
- Check S = 0.078; read h0/c and the friction variable from the Raimondi-Boyd chart for L/D = 1 at this S
Answer: The oil settles near 69 °C, well below the 93 °C limit
Rolling contact bearings: equivalent load, life and selection
Which syllabus covers thisAU ME3591 Unit VVTU not in syllabusSPPU not in syllabusMU MEC601 Module 4KTU not in syllabus
Which rolling bearing for which job?
Deep-groove ball bearings take radial load plus moderate thrust both ways and run fast: the default choice. Angular-contact ball bearings take heavy thrust in one direction, usually in pairs. Cylindrical roller bearings carry large radial loads but little or no thrust. Taper roller bearings take combined radial and thrust loads and are mounted in opposed pairs, as in wheel hubs. Spherical roller and self-aligning ball bearings tolerate shaft misalignment and deflection. Thrust ball and roller bearings take axial load only. Rollers carry more than balls of the same size because they touch along a line, not at a point.
How do you read a designation like 6208?
In the common ISO scheme the first digit gives the type (6 = single-row deep groove ball), the next the dimension series (2 = light, 3 = medium: same bore, bigger and stronger), and the last two the bore: 00 = 10 mm, 01 = 12, 02 = 15, 03 = 17, and from 04 up the bore is 5 × the code, so 6208 has a 40 mm bore. The catalogue then gives the outside diameter and width (6208: 80 × 18 mm), the basic dynamic load rating C (29.1 kN in one maker's catalogue) and the static rating C0 (17.9 kN). Suffixes add shields, seals and clearance.
What are the load ratings and the L10 life?
Rolling bearings fail by fatigue of the raceways: after enough cycles a flake spalls off. Identical bearings under identical load scatter widely in life, so ratings are statistical. L10 is the life that 90% of a large group reach. The basic dynamic load rating C is the load that gives an L10 of one million revolutions; life then falls as the cube of load for ball bearings (10/3 for rollers). The static rating C0 is the load that causes a permanent dent just large enough to matter, used for slow or stationary bearings.
How are radial and axial loads combined?
The equivalent dynamic load P is the pure radial load that would give the same life: P = X Fr + Y Fa. For a deep-groove ball bearing the catalogue first computes f0 Fa/C0 and reads e from a table. If Fa/Fr ≤ e the thrust is small and P = Fr. Otherwise X = 0.56 and Y comes from the same table, falling from 2.30 to 1.00 as the relative thrust grows. The required capacity for Lh hours at N rpm is then C = P (60 N Lh/10⁶)^(1/3), and you pick the smallest bearing that fits the shaft and has at least that C.
How do you design for a reliability above 90%, or for a varying load?
Multiply the L10 life by the reliability factor a1: 0.62 for 95%, 0.44 for 97% and 0.21 for 99% in the ISO 281 table used by many design tools (the 2007 edition recalculated these slightly). For a duty cycle of loads Pi for ni revolutions, the bearing feels the cubic mean P = (Σ Pi³ ni/Σ ni)^(1/3), the load that uses up the same fatigue life (Miner's rule). Short peaks barely matter if they are rare; long heavy spells dominate because of the cube.
The load on a ball bearing is doubled. What happens to its L10 life?
L10 = (C/P)³: twice the load gives 1/2³ = 1/8 of the life. A roller bearing falls to 1/2^(10/3) ≈ 1/10.
Worked examples
Life of a 6208 under combined load · Fr = 2.5 kN, Fa = 0.8 kN, 1,450 rpm, inner ring rotating; 6208: C = 29.1 kN, C0 = 17.9 kN, f0 = 14.0
- f0 Fa/C0 = 14.0 × 800/17,900 = 0.626; between table rows 0.345 and 0.689: e = 0.253, Y = 1.762
- Fa/Fr = 0.32 > e, so X = 0.56: P = 0.56 × 2,500 + 1.762 × 800 = 2,809 N
- L10 = (29,100/2,809)³ = 1,112 million revolutions
- Hours = 1,112 × 10⁶/(60 × 1,450) = 12,776 h; at 95% reliability × 0.62 = 7,921 h
Answer: About 12,800 h (L10), 7,900 h at 95% reliability
Selecting a bearing for 25,000 hours · Pure radial load 4 kN, 1,000 rpm, 25,000 h at 90% reliability; candidates 6210 (35.0 kN), 6308 (40.5), 6309 (53.0), 6310 (62.0)
- Revolutions = 60 × 1,000 × 25,000 = 1,500 million
- C needed = 4 × (1,500)^(1/3) = 4 × 11.447 = 45.8 kN
- 6210 (35.0) and 6308 (40.5) fall short; 6309 has 53.0 kN
- Check: (53/4)³ = 2,326 million rev = 38,770 h
Answer: 6309 (45 mm bore), with about 39,000 h of L10 life
Cubic mean load over a duty cycle · 6309 (C = 53 kN) at 1,000 rpm: 3 kN for 40% of the revolutions, 5 kN for 40%, 8 kN for 20%
- P = (0.4 × 3³ + 0.4 × 5³ + 0.2 × 8³)^(1/3) = (10.8 + 50 + 102.4)^(1/3) = 5.465 kN
- L10 = (53/5.465)³ = 912 million rev = 15,204 h
- Miner check: Σ (fraction × L10 / life at each load) = 1.000
- The 8 kN spell is only 20% of the time but uses 63% of the life
Answer: Equivalent load 5.46 kN, about 15,200 h
Thick cylinders, seals and gaskets
Which syllabus covers thisAU ME3591 Unit VVTU not in syllabusSPPU not in syllabusMU MEC601 Module 1KTU PCMET602 Module 4
When is a cylinder thick, and how does the stress vary through the wall?
The thin-cylinder formula σ = p d/2t assumes the hoop stress is the same right through the wall. Once the wall is more than about a tenth of the bore radius, it is not. Lamé's solution, from equilibrium of a ring element and compatible strains, gives σt = A + B/r² and σr = A − B/r²: the hoop stress is greatest at the bore and falls outward, and the radial stress runs from −p at the bore to zero outside. Closed ends add a uniform axial stress p ri²/(ro² − ri²). The Strength of Materials book's cylinder interactive plots these stresses; here we size the wall.
Lamé, Clavarino or Birnie: which thickness formula?
Each sets a failure criterion at the bore, the worst point. Lamé's equation limits the largest principal stress: right for brittle materials such as cast iron. Clavarino's limits the largest principal strain with closed ends (the axial stress is included); Birnie's does the same for open ends, such as a hydraulic ram cylinder without end caps. For ν = 0.3, Clavarino's wall is thinner than Lamé's once the allowable stress exceeds three times the pressure; Birnie's, with no axial stress to help, is always a little thicker. Note the limit: whatever the thickness, the bore hoop stress never falls below p, so p must stay under the allowable stress.
How much does the bore grow, and how is a thick cylinder strengthened?
The bore expands by its hoop strain times the diameter: Δd = d [σt − ν(σr + σl)]/E at the bore. This matters for a piston fit or a press-fitted liner. Because the inner layers do most of the work, very high-pressure vessels are pre-stressed: two cylinders shrunk one over the other (compound cylinders), or a single tube autofrettaged by over-pressurising it once so the bore yields and is left in residual compression. Both lower the working hoop stress at the bore.
How do O-rings, lip seals and mechanical seals work?
An O-ring is an elastomer ring squeezed in a groove: the squeeze seals at low pressure, and fluid pressure pushes it against the groove wall so it seals harder as pressure rises. Static O-rings seal covers and flanges; dynamic ones need smooth surfaces and backup rings at high pressure to stop extrusion into the gap. A radial lip seal has a flexible lip held on a rotating shaft by a garter spring, running on a thin oil film: the standard oil seal for gearboxes. A mechanical face seal uses two flat rings, one rotating, one stationary, pressed together: for pumps that must not leak.
How are gasket bolt loads found?
A gasket fills the roughness between two flanges, but only if the bolts first crush it into the surfaces: seating load Wm2 = π b G y, where y is the minimum seating stress. In service the pressure pushes the flanges apart with π G² p/4 while the gasket must keep a compression of m times the pressure: Wm1 = π G² p/4 + 2π b G m p. The bolts must carry the larger. ASME VIII-1 suggests m and y by gasket type: rubber below 75 Shore A m = 0.5, y = 0; compressed fibre 2.0 and 11.0 MPa; spiral wound 3.0 and 68.9 MPa; soft-steel ring joint 5.5 and 124.1 MPa. The effective width b is b0 up to 6.35 mm, then 0.5√b0 in inches.
Lamé: an allowable stress of 100 MPa, an internal pressure of 120 MPa. How thick must the wall be?
The bore hoop stress is p (ro² + ri²)/(ro² − ri²), which is always more than p however large ro gets. With p above the allowable stress, a thicker wall cannot help: use a stronger material, a compound cylinder or autofrettage.
Worked examples
Wall of a hydraulic cylinder · Bore 100 mm (ri = 50 mm), 25 MPa, closed ends, steel with allowable 100 MPa, ν = 0.3
- Lamé: t = 50 [√(125/75) − 1] = 14.55 mm
- Clavarino (closed ends): t = 50 [√((100 + 0.4 × 25)/(100 − 1.3 × 25)) − 1] = 50 [√(110/67.5) − 1] = 13.83 mm
- For comparison: Birnie (open ends) 15.97 mm, thin formula 25 × 50/100 = 12.5 mm
- Choose 15 mm (ro = 65 mm): bore hoop stress = 25 (65² + 50²)/(65² − 50²) = 97.5 MPa, outside 72.5 MPa
Answer: Clavarino gives 13.8 mm for a ductile steel; use a 15 mm wall
Growth of the bore · The cylinder above at 25 MPa with a 15 mm wall, E = 207 GPa (typical steel)
- σt = 97.46 MPa, σr = −25 MPa, σl = 25 × 50²/(65² − 50²) = 36.23 MPa at the bore
- E εt = 97.46 − 0.3 (−25 + 36.23) = 94.09 MPa
- Δd = 100 × 94.09/207,000 = 0.0455 mm
Answer: The 100 mm bore grows by about 45 µm
Bolt load for a spiral-wound gasket · Gasket reaction diameter G = 300 mm, effective width 6 mm (basic width under 6.35 mm, so b = b0), m = 3.0, y = 68.9 MPa, 2 MPa, 12 bolts, allowable bolt stress 138 MPa (illustrative)
- Operating: Wm1 = π/4 × 300² × 2 + 2π × 6 × 300 × 3 × 2 = 2,09,230 N
- Seating: Wm2 = π × 6 × 300 × 68.9 = 3,89,620 N: seating governs
- Bolt area = 3,89,620/138 = 2,823 mm², 235 mm² per bolt
- M20 bolts (tensile stress area 245 mm²) are enough
Answer: Seating load 390 kN; 12 × M20
Keep going
- Strength of Materials interactive book
- Engineering Materials and Metallurgy interactive book
- Automobile Engineering interactive book
- How to calculate fatigue life in product development
- How CAD and CAE are changing machine design
- PG Program in CAE
- Finite Element Analysis interactive book
- All Skill-Lync interactive books
Recommended reading
- Design of Machine Elements, V. B. Bhandari
- Shigley's Mechanical Engineering Design, R. G. Budynas, J. K. Nisbett
- Machine Design: An Integrated Approach, R. L. Norton
- Fundamentals of Machine Component Design, R. C. Juvinall, K. M. Marshek
- Design Data: Data Book of Engineers, PSG College of Technology
- Machine Design Data Handbook, K. Lingaiah
How this book was made
Chapters follow the official syllabi listed above. Every interactive's formulas and every worked-example number were checked by script. Explanations, animations and examples are original to Skill-Lync. Topics informed by MIT OpenCourseWare 2.72 (Elements of Mechanical Design), CC BY-NC-SA. All explanations, animations and examples are original to Skill-Lync; MIT does not endorse this page.