GATE Mechanical: Probability and Complex Variables (Engineering Mathematics), previous-year questions solved
Probability and complex variables carried 16 marks across the five keyed GATE Mechanical papers from 2022 to 2026, mostly as 2-mark questions: 4 on probability (counting, Poisson, expected value) and 5 on complex variables (analytic functions and contour integrals by residues). Each is solved here with the exam shortcut and the trap behind the wrong options.
Questions from official GATE ME papers (IIT Kharagpur 2022, IISc Bengaluru 2024, IIT Roorkee 2025, IIT Guwahati 2026); solutions by Skill-Lync. · Updated 16 Nov 2026 · 9 questions
Two unbiased coins are tossed: what is the probability of at least one head?
- (A)0.25
- (B)0.5
- (C)0.675
- (D)0.75
Answer: (D) Official key: D
- 'At least one' is fastest through the complement: the only outcome with no head is TT, with probability .
- : option (D).
Listing the four equally likely outcomes HH, HT, TH, TT gives the same answer: three of the four contain a head. The traps are the neighbouring events: exactly one head is (B), two heads is (A).
The same complement rule sizes redundant systems. If each of two independent pumps is available with probability , at least one is available with probability : two pumps that are each up 90% of the time give 99% availability. Reliability engineers use this for parallel redundancy in cooling, hydraulic and braking systems, as long as the failures really are independent (a shared power supply breaks that assumption).
Three students chosen from a council of 2 + 3 + 5: probability exactly two share a school?
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- Total ways to choose 3 of 10: .
- Pick the school that supplies the pair, then one student from the other 8 (or 7, or 5): .
- Probability : option (A).
Check with the complement: all three from different schools ; all three from the same school (engineering has only two students). . We also listed all 120 possible groups: 79 have exactly two from one school.
Trap: forgetting the 10 all-arts groups in the complement gives , option (C).
This is sampling without replacement, the hypergeometric model behind acceptance sampling: draw a few parts from a small lot and count how many are defective. Because each part drawn changes what is left in the lot, the binomial formula is only an approximation when the lot is small; quality engineers use the hypergeometric distribution for exactly this case.
Poisson random variable with P(X = 1) = P(X = 2): what is P(X = 3)?
Answer: 0.18 Official key: 0.17 to 0.19
- Poisson with mean : .
- : , so ( would make both probabilities zero).
- .
Check: with , . Trap: dropping the gives and , outside the key range of 0.17 to 0.19.
Poisson counts random, independent events at a steady average rate: defects per sheet of steel, breakdowns per month on a machine line, customer arrivals per hour. With a constant failure rate, the number of failures in a period is Poisson, which is how maintenance planners decide how many spares to stock: enough that the chance of running out in the lead time is acceptably small.
X is uniform on [0, 1] and Y = ln(X + 1): what is the expected value of Y?
Answer: 0.39 Official key: 0.38 to 0.39
- The expected value of a function of a random variable is that function averaged against the density: .
- Integrate by parts (or substitute ): , so
- .
Trap: plugging the mean into the function, , is wrong and falls outside the key range of 0.38 to 0.39. Because is concave, the average of is always smaller than of the average (Jensen's inequality). A Monte Carlo check with 200 000 random values of gives .
Monte Carlo tolerance analysis works exactly like the check above: sample every dimension from its distribution, compute the assembly gap for each sample and average. For a linear stack, the mean gap is just the gap at the mean dimensions; for a nonlinear one (angles, radii, clearances inside a square root) it is not, which is the same trap as . Our tolerance stack-up calculator runs a Monte Carlo check alongside worst case and RSS.
w = ln z = ln(x + iy): where is it analytic, and what are its conjugate functions?
- (A) is analytic everywhere except at
- (B) is non-analytic everywhere
- (C)The conjugate functions of are and
- (D)The conjugate functions of are and
Answer: (A) Official key: A
- Write : , so and .
- Cauchy-Riemann: and . Both hold at every point except the origin, where is undefined, and there. So option (A), and (B) is false.
- The conjugate (harmonic) pair is and . (C) has the wrong real part and an imaginary part, , that is not even harmonic. (D) puts an angle where the real part should be a logarithm.
We checked the Cauchy-Riemann equations numerically at 58 random points and at 20 of them. One fine point the exam ignores: because is only defined up to multiples of , a single-valued branch of (the principal value, say) also needs a cut from the origin, usually along the negative real axis, and is analytic everywhere except on that cut. The answer the key expects is the textbook one: the only singular point is .
In potential flow, is the complex potential of a source of strength at the origin: the real part is the velocity potential (circles of equal potential) and the imaginary part is the stream function (radial streamlines). The one point where is not analytic, , is the source itself. Adding sources, sinks, vortices and a uniform stream is how classical aerodynamics builds the flow around cylinders and airfoils.
Analytic f(z) with real part cosh x cos y and zero imaginary part on y = 0: what is f(z)?
- (A)
- (B)
- (C)
- (D)
Answer: (D) Official key: D
- Expand the last option: . Its real part is , and its imaginary part is zero at .
- is analytic everywhere (it is built from and ). An analytic function is fixed by its real part up to an imaginary constant, and the condition at makes that constant zero. So : option (D).
- Milne-Thomson gets there without guessing: with , ; put to get , so , and .
Trap: option (A) also has real part and zero imaginary part on , but its imaginary part is , and differs from : it fails the Cauchy-Riemann equations, so it is not analytic. We checked all four options at 40 random points; only (D) passes all three conditions.
Any analytic function hands you two harmonic functions at once, and that is the basis of 2D potential flow: the real part is a velocity potential and the imaginary part a stream function that automatically satisfy Laplace's equation. Conformal maps built from analytic functions, such as the Joukowski transform , turn the flow around a circle into the flow around an airfoil, a classic way to calculate the lift of an airfoil by hand.
Contour integral of 6z/(2z⁴ − 3z³ + 7z² − 3z + 5) around only the pole z = i
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- Since is a pole, must divide the denominator:
- Residue at the simple pole , using : which is .
- Residue theorem: : option (A).
Trap: forgetting the in gives , option (B). We integrated numerically around a circle of radius 0.5 about (the nearest other pole is 0.85 away) and got , which is .
Residues are how inverse Laplace transforms are done in practice. The response of a spring-mass-damper or a control loop is the sum of the residues of its transfer function times the input, one term per pole; the partial-fraction (Heaviside) expansion taught in control courses is exactly this residue calculation. The poles' locations then tell you the natural frequencies and damping of each mode.
(1/2π) ∮ dz/((z − i)(z + 4i)) around the circle |z| = 2
Answer: 0.2 Official key: 0.2 to 0.2
- Poles at (, inside the circle) and (, outside).
- Residue at : .
- , so the value is .
Trap: counting both poles. Their residues, and , cancel, so a contour enclosing both gives 0. Direct numerical integration around gives ; around it gives less than , zero to rounding.
Which poles sit inside a closed contour is the stability question in control engineering. The Nyquist criterion runs a contour around the whole right half of the s-plane and counts how often the open-loop frequency response encircles −1; by the argument principle, a cousin of the residue theorem, that count reveals closed-loop poles in the right half-plane, which mean an unstable system.
∮ z³/((z² + 4)(z² − 4)) dz = 2πi n on the unit circle: what is n?
Answer: 0.0 Official key: -0.1 to 0.1
- The integrand blows up only where or : at and . All four poles have , outside the unit circle.
- Inside and on the function is analytic, so by Cauchy's theorem the integral is zero: .
Trap: computing residues without checking where the poles are. Each of the four residues is , so a contour enclosing all four (radius 3, say) gives , and enclosing only would give 0.5. Numerical integration around the unit circle gives , zero to rounding.
The unit circle in this question is the same contour that decides stability in digital control and signal processing. A causal discrete-time system (a digital filter, or a controller running on a microcontroller) is stable exactly when every pole of its z-domain transfer function lies inside the unit circle; a pole outside it, like the four here, makes the response grow without bound.