GATE Production and Industrial Engineering (PI): Engineering Mathematics (matrices, limits, vector calculus, differential equations, Laplace transforms, complex variables, probability, trapezoidal and Simpson's rules), previous-year questions solved, part 2
Part 2 of 2 of the Engineering Mathematics hub: 11 more official GATE PI questions on complex variables, probability and statistics and numerical methods, each solved step by step and checked against the official answer key. Part 1 covers linear algebra: matrices, determinants and eigenvalues, calculus and vector calculus and differential equations and Laplace transforms.
Questions from official GATE PI papers (IISc Bengaluru 2024, IIT Roorkee 2025, IIT Guwahati 2026); solutions by Skill-Lync. · Updated 19 Nov 2026 · 11 of 24 questions on this page
Analytic function f = u + iv: which of four partial-derivative equations are true
- (A)
- (B)
- (C)
- (D)
Answer: (A) and (B) and (D) Official key: A;B;D
- A function is analytic when it satisfies the Cauchy-Riemann equations: and .
- Differentiate the first with respect to and the second with respect to , then add. The mixed derivatives of cancel and . So (A) is true: is harmonic.
- Do the same the other way round (first with respect to , second with respect to , then subtract): . So (B) is true: is harmonic too.
- Substitute the Cauchy-Riemann equations into (D): . So (D) is true. It says the gradients of and are perpendicular, so the curves constant and constant cross at right angles.
- (C) mixes in with in . Test it on , where and : and , so the sum is . (C) is false.
- The true statements are (A), (B) and (D).
Trap: assuming that any pair of second derivatives adds to zero. Only the Laplacian of one function at a time ( alone or alone) does. We tested five analytic functions (, , , and ) at 40 random points each with finite differences: (A), (B) and (D) were zero every time, and (C) failed at once (for at it gave ).
Harmonic functions describe steady-state heat conduction and ideal fluid flow in two dimensions, so analytic functions are used to map an awkward tool or part outline onto a simple one where the temperature or flow field is easy to solve. The perpendicular crossing of the two families of curves is how flow lines and constant-potential lines are drawn.
Identify the analytic function among four complex expressions in x and y
- (A)
- (B)
- (C)
- (D)
Answer: (A) Official key: A
- A function is analytic (complex differentiable) where and have continuous first partial derivatives that satisfy the Cauchy-Riemann equations and .
- (A) , : and . Both hold everywhere. In fact (A) is .
- (B) gives . It is , the conjugate, which is not analytic.
- (C) gives but . Fails.
- (D) gives but . Fails: is not a function of alone.
Check done in code: the largest Cauchy-Riemann mismatch at 200 random points was 2e-09 for (A) and at least 13.7 for each of the others.
Analytic functions give harmonic real and imaginary parts, which is why conformal mapping solves potential flow and steady heat conduction problems in awkward shapes. It is the classical hand method behind flow-net and field plots taught for those problems.
Value of 1 divided by i to the n for an even positive integer n
where , and is an even positive integer.
- (A) or
- (B) or
- (C)only
- (D)only
Answer: (A) Official key: A
- Rule: powers of repeat in a cycle of four: , , , , then it starts again.
- An even can be written , so . This is when is even () and when is odd ().
- The reciprocal of is and the reciprocal of is , so takes the values or , which is (A).
- Check with two cases: gives , and gives . Both signs occur, so (C) is false, and the value is real, so (B) and (D) are false.
Trap: the reciprocal of itself is (), so it is tempting to answer with . That holds for odd only. We multiplied out for every even up to 200: was always exactly or , and both occurred.
Rotating machinery and AC electrical supplies are analysed with complex phasors, where multiplying by i turns the phasor a quarter turn, so the cycle of four in the powers of i is the same as four quarter turns of a vibrating or rotating part.
Value of K for the pdf f(x) = K/4 on 0 to 1 and 0 elsewhere
Answer: 4 Official key: 4 to 4
- A probability density function must integrate to over all values of . This is the normalisation condition: total probability is .
- The density is zero outside , so only that interval counts: .
- The density is constant, so the integral is its height times the width: .
- Hence . The height of the density is , a uniform distribution on .
Trap: stopping at and answering . The question asks for , not for the height . We swept every integer from to and summed the area under the density over with 2000 midpoint slices: only gives an area of (and a non-negative density).
Every probability model of a process, such as the spread of a dimension, a cycle time or a time to failure, has to integrate to one before it can be used for yield or reliability numbers. Fitting a density to shop-floor data always starts with this normalising constant.
Seven cards numbered 1 to 7: probability the card is greater than 4 given it is odd, in percent
Answer: 50 Official key: 50 to 50
- Conditional probability: . Here is the event that the card is odd and is the event that the card is greater than .
- Odd cards: , so .
- Odd cards greater than : and , so .
- .
Shortcut: once we know the card is odd, the sample space shrinks to the four odd cards, all equally likely, and two of them exceed . Trap: using the whole box and answering (cards ), or writing when the blank asks for a percentage. We enumerated the seven cards with exact fractions () and ran a seeded Monte Carlo of 200 000 draws, keeping the odd ones: 49.8.
Conditional probability is how an inspector updates the odds once a part is known to belong to a subgroup, for example the share of defects that come from one machine given the lot is from the night shift. Counting only inside the known subgroup is the same shortcut used in sampling plans and warranty analysis.
Probability that the second ball drawn is red from a bag of 5 red, 7 green and 3 blue balls
- (A)
- (B)
- (C)
- (D)
Answer: (B) Official key: B
- The balls are drawn one after another without replacement, so the second draw depends on the first. Split on what the first ball was (law of total probability).
- First ball red, probability ; then 4 red remain among 14: .
- First ball not red, probability ; then still 5 red among 14: .
- Add: . This is option (B).
- Check by listing: of the 210 ordered pairs of distinct balls, 70 have a red second ball, and 70/210 = 1/3. A simulation of 200000 draws gave 0.334.
Insight: with no information about the first ball, every ball is equally likely to be in second place, so the answer is the same as for the first draw, . Trap: is only the probability that both are red.
Sampling without replacement is how inspectors draw parts from a lot: the chance that the second part pulled is defective equals the lot defect fraction, whatever the order. The same reasoning is used when parts are picked from a bin for audit.
Which of 0, 10.3, -0.9 and 2 are possible values of a correlation coefficient
- (A)
- (B)
- (C)
- (D)
Answer: (A) and (C) Official key: A;C
- Rule: the correlation coefficient is . By the Cauchy-Schwarz inequality , so always. is a perfect rising straight line, a perfect falling line, and means no linear relation.
- (A) lies in . Possible: for example and give .
- (B) is greater than . Impossible.
- (C) lies in . Possible: a strong falling trend. We built a three point data set whose correlation is exactly .
- (D) is greater than . Impossible. The possible values are (A) and (C), so the answer is (A) and (C).
Trap: a value of looks odd to some students, but a negative sign only means that falls as rises; only the size is limited, to at most . We also computed for 20000 random data sets: every value fell between and .
Quality engineers use the correlation coefficient to see whether a process setting such as cutting speed, temperature or pressure moves together with a measured result such as surface roughness or part size, and a value near 0 tells them that the setting is not the lever to pull.
Probability that all 4 inserts taken from 5 coated and 10 uncoated are uncoated
The probability of the milling cutter having all uncoated inserts is ______ (rounded off to two decimal places).
Answer: 0.15 Official key: 0.14 to 0.16
- Rule: when a group is taken at once from a mixed box, every group of 4 inserts is equally likely. Probability (groups that match) (all groups), counted with combinations .
- All groups of 4 from 15 inserts: .
- Groups with all 4 uncoated, chosen from the 10 uncoated: .
- 0.15, inside the official range 0.14 to 0.16.
Same answer by drawing one at a time without putting back: . Common slip: treating each draw as if the insert were put back, , which is far too large because the box has only 10 uncoated inserts and each one removed lowers the chance. We listed all 1365 possible groups of 4 and counted 210 with no coated insert (210/1365 = 0.1538); counting every ordered draw (32760 of them) gave 5040/32760, the same value; and 200000 simulated draws gave 0.153.
Acceptance sampling and kitting work the same way: when a few items are drawn from a mixed lot, hypergeometric counting gives the chance that all of them are good or all are defective, which is how sampling plans for incoming tooling and parts are set.
Trapezoidal rule for the integral of y(x) from 0 to 4 using a table of five values
Answer: 24.5 Official key: 24.5 to 24.5
- The trapezoidal rule replaces the curve between each pair of neighbouring points by a straight line and adds the trapezium areas. With equal spacing : .
- Here and the table has strips: , , and the inside values are .
- .
Check strip by strip: . Trap: using Simpson's rule, which the question does not ask for and which gives , or forgetting the factor on the inside values. We coded the composite trapezoidal formula and the sum of the four trapezium areas from the table: both gave .
Process data such as force, power or temperature is logged at discrete instants, and the trapezoidal rule is the standard way to turn it into a total, for example energy used in a cycle or heat input along a weld. It needs only the logged table and no formula for the curve.
Value of n when Simpson's 1/3 rule with 4 subintervals gives n/3 for the integral of x squared minus 2x
Answer: 2 Official key: 2 to 2
- Simpson's 1/3 rule fits parabolas through pairs of strips: for 4 subintervals, with .
- Ordinates of at : .
- Weighted sum of the ordinates: .
- Integral .
- So and .
- Check: the exact integral is (0.666667), and a 10000-panel Simpson sum gives 0.666667.
Insight: Simpson's rule is exact for any polynomial up to degree 3, so for this quadratic even 2 subintervals would give the same . Trap: the trapezoidal rule with 4 strips gives , not .
Simpson's rule integrates measured profiles and logged curves, for example the area under a force-displacement record to get forming energy, or a surface roughness trace sampled at equal spacing. Equal spacing and an even number of intervals are the practical conditions.
Closest approximation to the integral of exp(-x squared) from 0 to 1 by numerical integration
Note: Use numerical integration
- (A)
- (B)
- (C)
- (D)
Answer: (C) Official key: C
- Rule: the integral of has no elementary antiderivative, so it is estimated numerically. The simplest rule is the trapezoidal rule: replace the curve by a straight line between the end points, so the area is the width times the average height, .
- Here , so and .
- With one panel from to : , which is option (C).
- Test against the true value. A fine Simpson rule with 20000 panels gives , which agrees with . The options are (1.0000), (0.3679), (0.6839) and (1.8591). Their distances from are 0.253, 0.379, 0.063 and 1.112, so (C) is the closest.
Trap: option (D), , is what you get from the average of and , forgetting the minus sign in the exponent; it is larger than , which is impossible because the integrand never exceeds on . Check on the rule: more panels move the estimate toward the true value (1 panel 0.6839, 2 panels 0.7314, 4 panels 0.7430; Simpson with 2 panels 0.7472).
Process and sensor data are usually samples at discrete instants, so areas under them, such as energy used by a machine or total heat input, are estimated with the trapezoidal rule exactly as here.