Advanced Mechatronics interview questions

125 questions with answers (MTX-301 to MTX-425). Try each one aloud before reading the answer.

Mechatronic System Design

MTX-301 · Design a line-following robot.

Use an IR reflectance sensor array to measure line position error, a microcontroller running PID on that error, a dual H-bridge driver for differential-drive DC gear motors, and a battery with regulated logic supply. Tune Kp first, add Kd to damp weaving, and limit speed on curves.

MTX-302 · Design a colour-sorting conveyor station. How do you time the diverter?

A photoelectric sensor detects the part, a colour sensor or camera classifies it, and a PLC fires a pneumatic pusher downstream. With 0.5 m between sensor and pusher at 0.25 m/s belt speed, delay = 2 s; better, count belt encoder pulses so timing stays correct when speed changes.

MTX-303 · A servo axis starts oscillating at a fixed high frequency after moving to a lighter frame. What do you do?

The new frame has a lower mechanical resonance inside the servo bandwidth. Measure the frequency response, add a notch filter at the resonance, reduce velocity loop gain if needed, and stiffen the frame or couplings as the long-term fix.

MTX-304 · Compare direct drive and geared drive for a rotary indexing table.

Direct drive (torque motor) gives zero backlash, high stiffness, low maintenance and high accuracy, but costs more and must handle full load inertia. A gearbox allows a smaller motor and better inertia matching, but adds backlash, compliance, wear and noise.

MTX-305 · Size the motor torque for a belt axis: 5 kg load, 5 m/s² acceleration, 10 N friction, 20 mm pulley radius.

Force F = m·a + friction = 5 × 5 + 10 = 35 N. Torque T = F·r = 35 × 0.02 = 0.7 N·m, plus torque to accelerate pulleys, belt and rotor. Add a safety margin of about 1.5–2 before choosing a motor.

MTX-306 · A load of 0.01 kg·m² is driven by a motor of 1 × 10⁻⁴ kg·m². What gear ratio matches inertia?

N_opt = √(J_load/J_motor) = √(0.01/0.0001) = 10. The reflected inertia becomes 0.01/100 = 1 × 10⁻⁴ kg·m², equal to the rotor inertia, giving a 1:1 inertia ratio and maximum load acceleration for given motor torque.

MTX-307 · Design a temperature control system for a lab oven.

Use a thermocouple or RTD with conditioning, a microcontroller or PID controller, and a solid-state relay switching the heater with time-proportioned PWM (period of a few seconds). Add an independent over-temperature cutout, anti-windup, sensor-break detection and slow-ramp tuning to handle the long thermal time constant.

MTX-308 · How is functional safety handled in machine design?

Assess risk, then design safety functions with a required reliability level: Performance Level (PLa to PLe) under ISO 13849-1 or SIL under IEC 62061. Achieve it through architecture (redundancy, categories), diagnostics, reliable components and validation.

MTX-309 · Design an automatic sliding door.

Use a microwave or IR presence sensor, a geared DC or BLDC motor with belt drive, an encoder for position and speed profiling, and current sensing to detect obstruction and reverse. Add end-of-travel sensing, battery backup or manual release, and safety beams at the opening.

MTX-310 · An incremental encoder on a machine with a VFD gains random counts. How do you fix it?

The cause is usually PWM noise coupling into the encoder cable. Use differential line-driver (RS-422) encoders, shielded twisted-pair cable with the shield bonded properly, separate encoder and motor cables, add filtering on the counter input and check the grounding.

MTX-311 · A system has two series components with failure rates 2 × 10⁻⁶/h and 3 × 10⁻⁶/h. What is its MTBF?

For series components with constant failure rates, λ_total = 2 × 10⁻⁶ + 3 × 10⁻⁶ = 5 × 10⁻⁶ per hour. MTBF = 1/λ = 200,000 hours.

Sensors

MTX-312 · What encoder resolution is needed for 1 µm resolution on a 5 mm-lead ball screw?

Counts per revolution = 5 mm/1 µm = 5000. With 4× quadrature decoding, the encoder needs 5000/4 = 1250 lines per revolution or more. Screw accuracy and backlash must also be within budget.

MTX-313 · How does a complementary filter fuse accelerometer and gyroscope data?

θ = α(θ + ω·dt) + (1 − α)·θ_accel. The gyro gives good short-term angle but drifts; the accelerometer gives drift-free long-term angle but is noisy. With α = τ/(τ + dt), τ = 0.5 s and dt = 0.01 s, α ≈ 0.98.

MTX-314 · A thermocouple reading is several degrees off. What are the likely causes?

Wrong thermocouple type selected, copper wire used instead of extension wire, poor cold-junction compensation, a ground loop on an earthed junction, EMI on long leads, or thermal lag and poor immersion. Check each with a reference thermometer.

MTX-315 · A load cell is rated 2 mV/V at 10 V excitation. What is the full-scale output and how do you digitise it?

Full-scale output = 2 mV/V × 10 V = 20 mV. Use ratiometric measurement (ADC reference from the excitation) with a low-noise instrumentation amplifier or a 24-bit delta-sigma bridge ADC, plus filtering, to resolve small fractions of full scale.

MTX-316 · How do you select sensors for a washdown food machine?

Use sensors with high ingress protection (IP67 or IP69K), stainless or food-grade housings, resistance to cleaning chemicals and temperature cycling, and sealed connectors. Prefer non-contact sensing and avoid crevices that trap residue.

MTX-317 · What output frequency does a 2500-PPR encoder produce at 6000 rpm?

f = PPR × rpm/60 = 2500 × 6000/60 = 250 kHz per channel. With 4× decoding, the counter must handle 1 million counts per second, so check the input and cable bandwidth.

MTX-318 · How do multi-turn absolute encoders keep turn count without power?

They use either a mechanical gear train of coded discs (geared multi-turn) or electronic counting with a battery backup or energy-harvesting Wiegand sensor. Geared and Wiegand types need no battery maintenance.

MTX-319 · How do sin/cos encoders achieve very high resolution?

They output analog sine and cosine signals per line. The controller digitises both and computes θ = atan2(sin, cos) within each period, interpolating each line into hundreds or thousands of positions. Accuracy depends on signal amplitude and phase matching.

MTX-320 · How do you compensate an ultrasonic sensor for temperature?

Speed of sound in air is approximately v = 331.3 + 0.606·T m/s (T in °C). Measure air temperature and compute v before d = v·t/2. Between 0 °C and 40 °C, speed changes by about 7%, so uncompensated error is significant.

MTX-321 · How do you calibrate a linear sensor?

Apply known reference inputs traceable to standards across the range, record outputs, and fit a line by least squares to find gain and offset. Check linearity, hysteresis (up and down runs) and repeatability, then store correction coefficients.

Actuators

MTX-322 · What limits the maximum speed of a long ball screw?

Critical speed, where the shaft's rotational speed matches its first bending natural frequency, causes whipping. It falls with the square of unsupported length and rises with root diameter and stiffer end supports. Long axes use fixed-fixed mounting, rotating nuts, belts or linear motors.

MTX-323 · Why use a peak-and-hold driver for a solenoid?

A high pull-in current gives fast actuation, then current is reduced (often by PWM) to a lower hold level once the plunger closes, since the closed magnetic circuit needs less current. This cuts heating and power and allows faster release.

MTX-324 · How much current does a piezo stack need to slew quickly?

A piezo stack is mainly a capacitor, so I = C·dV/dt. For C = 1 µF charged by 100 V in 1 ms, I = 10⁻⁶ × 100/0.001 = 0.1 A. Fast dynamic operation needs amplifiers with high current capability.

MTX-325 · Choose an actuator for 1 kN force, 100 mm stroke in 0.5 s with occasional mid-stroke stops.

Average speed is 0.2 m/s and peak power about 200–400 W. A servo-driven ball screw actuator fits: it provides force, programmable positions and holding without leaks. Pneumatics would struggle with mid-stroke stops and hydraulics would be overkill.

MTX-326 · What is an electro-hydrostatic actuator (EHA)?

A self-contained unit where a servo motor drives a reversible pump directly connected to a cylinder, with no central hydraulic supply or servo valve. It gives hydraulic force density with electric control and higher efficiency, as used in aircraft and presses.

MTX-327 · How can backlash be reduced or compensated?

Mechanically, with preloaded ball screws, anti-backlash split gears, harmonic drives or spring loading. In control, use a load-side encoder (dual-loop), or add a backlash offset to the command on direction reversal.

MTX-328 · What determines the efficiency of a lead screw?

For raising a load, η = tan λ/tan(λ + φ), where λ is the lead angle and φ = arctan μ is the friction angle. With λ = 5° and μ = 0.15 (φ ≈ 8.5°), η ≈ 0.0875/0.240 ≈ 36%. It self-locks when λ < φ.

MTX-329 · Why do relay contacts fail early on inductive DC loads and how do you fix it?

Breaking inductive current produces arcing that erodes contacts, and DC arcs do not self-extinguish at a zero crossing. Add a flyback diode or RC snubber or varistor across the load, use DC-rated contacts, or switch with a solid-state device.

MTX-330 · What force does a voice coil produce with B = 0.5 T, 10 m of wire and 2 A?

F = B·I·L = 0.5 × 2 × 10 = 10 N, assuming all wire is perpendicular to the field. Force is linear in current, so a current amplifier gives direct force control.

MTX-331 · How does actuator saturation affect a control design?

The actuator cannot deliver demanded effort beyond its limits, so the loop effectively opens, causing integrator windup, overshoot and slower response than the linear design predicts. Design motion profiles within limits, add anti-windup and check performance in nonlinear simulation.

Electric Motors & Drives

MTX-332 · What is the transfer function of an armature-controlled DC motor from voltage to position?

θ(s)/V(s) = K/[s((Js + b)(Ls + R) + K²)], with K = Kt = Ke. The free s is the integration from speed to position. Ignoring inductance gives a first-order speed model with a mechanical time constant.

MTX-333 · What is field-oriented control?

Vector control that transforms three-phase currents into a rotating d-q frame aligned with the rotor flux (Clarke and Park transforms). Id controls flux and Iq controls torque independently with PI loops, giving DC-motor-like fast, smooth torque control of AC motors.

MTX-334 · A 24 V DC motor has R = 2 Ω and Ke = Kt = 0.05 V·s/rad. Find stall current, stall torque and no-load speed.

Stall current = 24/2 = 12 A. Stall torque = 0.05 × 12 = 0.6 N·m. No-load speed (ignoring friction) = 24/0.05 = 480 rad/s ≈ 4584 rpm.

MTX-335 · Calculate the mechanical time constant of a DC motor with J = 1 × 10⁻⁴ kg·m², R = 2 Ω and K = 0.05.

τm = J·R/(Kt·Ke) = (1 × 10⁻⁴ × 2)/(0.05 × 0.05) = 0.08 s. Speed reaches about 63% of its final value 80 ms after a voltage step.

MTX-336 · Why should the electrical time constant be much smaller than the mechanical one?

τe = L/R (for example 2 mH/2 Ω = 1 ms) sets how fast current and torque change. When τe ≪ τm, the motor model simplifies to first order, and the current loop can be tuned much faster than the speed loop, allowing cascade control.

MTX-337 · Why is space vector PWM preferred over sine-triangle PWM?

SVPWM uses the DC bus better: it allows about 15% more output voltage (a factor of 2/√3) in the linear range, and it reduces harmonic distortion and switching losses. It is the standard modulation in FOC drives.

MTX-338 · How does sensorless BLDC control work and what is its limitation?

The controller detects back-EMF zero crossings on the floating phase in six-step drive, or uses an observer of flux and back-EMF in FOC, to estimate rotor position. Back-EMF is near zero at standstill and low speed, so start-up uses open-loop ramping or signal-injection methods.

MTX-339 · A 4-pole, 50 Hz induction motor runs at 4% slip. What is its speed?

Ns = 120 × 50/4 = 1500 rpm. N = Ns(1 − s) = 1500 × 0.96 = 1440 rpm. Rotor current frequency = s·f = 2 Hz.

MTX-340 · How do you deal with stepper motor resonance?

Steppers resonate at low speeds and can lose torque in mid-band ranges. Use microstepping, avoid dwelling at resonant speeds with acceleration ramps, add mechanical damping or a damper on the shaft, use drivers with anti-resonance algorithms, or go to closed-loop stepper control.

MTX-341 · A motor delivers 2 N·m for 0.2 s and 0.5 N·m for 0.8 s every cycle. What is the RMS torque?

T_rms = √((2² × 0.2 + 0.5² × 0.8)/1.0) = √(0.8 + 0.2) = 1.0 N·m. The motor's continuous torque rating must be at least 1.0 N·m, and its peak rating at least 2 N·m.

MTX-342 · How do you size a braking resistor for a decelerating flywheel?

Kinetic energy E = ½Jω². For J = 0.01 kg·m² at 300 rad/s, E = 450 J. Decelerating in 0.5 s gives 900 W average (about 1.8 kW peak at constant torque); the resistor's average rating depends on cycle frequency, and its value must keep bus voltage below the drive's limit.

Microcontrollers & Interfacing

MTX-343 · How would you structure firmware for a motor controller?

Trigger the ADC from the PWM timer at the centre of the PWM period, run the current loop in the ADC-complete ISR (for example at 10–20 kHz), the speed and position loops in a slower timer ISR, and communications, diagnostics and the user interface in the background or RTOS tasks.

MTX-344 · When would you use fixed-point arithmetic and what is Q15 format?

On MCUs without an FPU, fixed-point is much faster than software floating point. Q15 represents values in [−1, 1) as signed 16-bit integers scaled by 2¹⁵. Watch for overflow and scaling, and use saturating arithmetic.

MTX-345 · What is priority inversion and how is it prevented?

A high-priority task waits for a resource held by a low-priority task, which is pre-empted by a medium-priority task, so the high-priority task waits indefinitely. Priority inheritance or priority ceiling protocols in the RTOS mutex prevent it.

MTX-346 · Why use a hardware quadrature encoder interface rather than interrupts?

A QEI or timer in encoder mode decodes A/B signals and counts in hardware, with no CPU load and no missed counts at high rates. Interrupt-based counting at hundreds of kHz would overload the CPU and lose counts.

MTX-347 · An I²C bus locks up with SDA held low after a reset. How do you recover it?

A slave was reset mid-transfer and is still driving SDA. The master toggles SCL as GPIO up to nine times until the slave releases SDA, then generates a STOP condition and reinitialises the peripheral. Add timeouts in the driver.

MTX-348 · How does CAN bit rate limit bus length?

All nodes must see a bit within one bit time to arbitrate, so propagation delay limits length. Classic CAN supports about 40 m at 1 Mbit/s, and lower bit rates allow longer buses, such as roughly 500 m at 125 kbit/s.

MTX-349 · Calculate the UART baud error for 115200 baud from a 16 MHz clock with 16× oversampling.

Divisor = 16 × 10⁶/(16 × 115200) − 1 ≈ 7.68, rounded to 8. Actual baud = 16 × 10⁶/(16 × 9) ≈ 111,111, an error of −3.5%. This exceeds the roughly ±2% tolerance, so use 8× oversampling or a baud-friendly crystal like 14.7456 MHz.

MTX-350 · How do you implement safe firmware updates in the field?

Use a bootloader with dual flash banks (A/B) or a backup image, verify the new image with a CRC or cryptographic signature before switching, and roll back automatically if the new firmware fails to start or refresh the watchdog.

MTX-351 · How do you reduce power consumption in a battery-powered sensor node?

Spend most time in deep sleep with wake-ups from an RTC or interrupt, turn off unused peripherals and sensor power, lower the clock, batch radio transmissions, and use DMA. Average current = Σ(I × duty), which is usually dominated by sleep current and radio time.

MTX-352 · A digital control loop shows jitter in its sample time. Why is that a problem and how do you fix it?

Jitter makes the effective sample time vary, adding phase noise and degrading or destabilising the loop, especially derivative terms. Run the loop from a hardware timer interrupt at high priority, avoid long critical sections and keep variable-time tasks out of the ISR.

Signal Conditioning & Data Acquisition

MTX-353 · What is a Kalman filter?

An optimal recursive estimator for linear systems with Gaussian noise. Each step predicts the state from a model, then corrects it with the measurement weighted by the Kalman gain, which balances process noise against measurement noise. It is used for sensor fusion and state estimation.

MTX-354 · What ADC resolution is needed to resolve 1 mV over 0–10 V?

Levels needed = 10 V/1 mV = 10,000. Since 2¹³ = 8192 is too few and 2¹⁴ = 16,384 is enough, at least a 14-bit ADC is required; a 16-bit device gives margin for noise.

MTX-355 · What is ENOB and how is it calculated?

Effective Number of Bits reflects real ADC performance including noise and distortion: ENOB = (SINAD − 1.76)/6.02. An ADC with SINAD of 74 dB has ENOB ≈ 12 bits, even if it outputs 16 bits.

MTX-356 · How much attenuation at 5 kHz does a 4th-order Butterworth filter with 1 kHz cutoff give?

Far above cutoff, the roll-off is 20n = 80 dB/decade. 5 kHz is log₁₀(5) ≈ 0.7 decades above, giving about 56 dB attenuation. For fs = 10 kHz, the Nyquist frequency of 5 kHz is attenuated by about 56 dB.

MTX-357 · A sensor reading shows 50 Hz hum only when the PC is connected. What is happening?

A ground loop: the PC and the sensor are earthed at different points, so mains-frequency current flows through the signal ground. Use differential inputs, isolate the DAQ or USB link, or connect grounds at one point only.

MTX-358 · How does oversampling improve ADC resolution?

Averaging N samples reduces uncorrelated noise by √N. Each 4× oversampling with decimation gives about one extra bit, if there is enough noise or dither to spread the signal across codes. Gaining 2 bits needs 16× oversampling.

MTX-359 · Design the signal chain for a K-type thermocouple measuring 0–500 °C.

Output is about 41 µV/°C, about 20.6 mV at 500 °C. Use an RC input filter, an instrumentation amplifier or dedicated thermocouple IC with gain of about 100 to reach a 2 V range, cold-junction compensation, open-circuit detection and linearisation from NIST tables.

MTX-360 · How should shielded cables be grounded for low-level analog signals?

For low-frequency signals, ground the shield at one end only, usually the receiver end, to avoid ground-loop current in the shield. Use twisted pairs for the signal. For high-frequency noise, both-end grounding or one end through a capacitor works better.

MTX-361 · How do FIR and IIR digital filters compare?

FIR filters are always stable and can have exactly linear phase, but need many coefficients for sharp responses. IIR filters achieve sharp responses with few coefficients and low delay, but have nonlinear phase and can become unstable or suffer quantization problems.

MTX-362 · When do you need simultaneous sampling instead of a multiplexed ADC?

When phase relationships between channels matter, such as three-phase current measurement, vibration analysis or power calculations. A multiplexed ADC samples channels one after another, adding time skew that becomes phase error at high frequencies.

MTX-363 · Calculate the thermal noise of a 1 kΩ resistor at 300 K over 10 kHz bandwidth.

vn = √(4kTRB) = √(4 × 1.38 × 10⁻²³ × 300 × 1000 × 10⁴) ≈ 0.41 µV rms. Lower resistance and narrower bandwidth reduce it.

Control Systems

MTX-364 · A position loop overshoots badly and rings. How do you tune it?

First tune the inner current and velocity loops. Then reduce position proportional gain or add velocity-loop damping, reduce integral gain, add velocity feedforward to cut following error, and check for mechanical resonance or backlash. Use step and frequency response measurements.

MTX-365 · What is state feedback with pole placement?

Control u = −Kx + r places closed-loop poles at chosen locations, the eigenvalues of (A − BK). This is possible for any pole set if the system is controllable (rank of [B AB … Aⁿ⁻¹B] = n).

MTX-366 · What is a Luenberger observer?

A model-based estimator that reconstructs unmeasured states: x̂' = Ax̂ + Bu + L(y − Cx̂). The gain L sets how fast the estimation error decays. It requires observability and is used with state feedback when not all states are measured.

MTX-367 · What is LQR control?

Linear Quadratic Regulator: state feedback that minimises J = ∫(xᵀQx + uᵀRu)dt. Q weights state errors and R weights control effort. It gives guaranteed good robustness margins for full state feedback.

MTX-368 · How do you choose the sample rate for a digital controller?

Sample at roughly 10–30 times the closed-loop bandwidth. Slower sampling adds phase lag of about ωTs/2 from the zero-order hold and computation delay, eroding phase margin.

MTX-369 · A loop has a 100 rad/s crossover and a 2 ms delay. How much phase margin is lost?

A pure delay adds phase lag ω·Td = 100 × 0.002 = 0.2 rad ≈ 11.5°. If the original phase margin was 45°, it falls to about 33.5°.

MTX-370 · How does feedforward improve a servo's tracking?

Velocity feedforward adds a command proportional to reference velocity, and acceleration feedforward adds torque proportional to reference acceleration × inertia. They supply most of the needed effort directly, so the feedback only corrects errors, greatly reducing following error without raising feedback gains.

MTX-371 · What is Model Predictive Control (MPC)?

At each sample, MPC uses a model to predict future outputs over a horizon, solves an optimisation that minimises tracking error and effort subject to constraints, applies the first control move, and repeats. It handles constraints and multivariable coupling explicitly.

MTX-372 · What is gain scheduling?

Using different controller gains for different operating points, selected or interpolated by a measured scheduling variable such as speed, load or temperature. It handles nonlinear plants with a set of linear controllers.

MTX-373 · How is a cart-pole inverted pendulum stabilised?

Linearise about the upright position, giving an unstable pole. Use state feedback (pole placement or LQR) on cart position and velocity and pendulum angle and rate, using an observer or filtered derivatives for the rates. A swing-up controller (energy-based) brings it upright first.

MTX-374 · How is a notch filter used in a servo loop?

A notch filter in series with the velocity or torque loop attenuates a narrow band around a mechanical resonance, so loop gain can be raised without exciting it. Its centre frequency, width and depth must be tuned, and it adds phase lag below the notch.

PLC & Industrial Control

MTX-375 · Design a bottle-filling sequence in a PLC.

Use SFC or a state machine: conveyor runs until a sensor detects a bottle, stop conveyor, lower nozzle, open fill valve until a level sensor or timer, close valve, raise nozzle, restart conveyor. Add timeouts, interlocks for missing bottles, an E-stop and a fault-reset step.

MTX-376 · A PLC misses pulses from a flow meter. How do you fix it?

Pulses shorter than the scan time are missed. Wire the meter to a high-speed counter input or module, which counts in hardware, or use an interrupt input. Also check input filter time, which may suppress short pulses.

MTX-377 · How is PLC redundancy implemented in critical processes?

A hot-standby pair of CPUs runs in sync, with the backup taking over within milliseconds if the primary fails, without bumping outputs. Redundant power supplies, networks (ring topologies) and I/O can be added.

MTX-378 · How do you implement PID in a PLC correctly?

Scale the process variable to engineering units, call the PID block at a fixed cyclic interrupt rate, clamp output with anti-windup, use bumpless transfer between manual and auto, and handle sensor faults by switching to manual or a safe output.

MTX-379 · What are PLCopen motion function blocks?

A standard set of IEC 61131-3 function blocks for motion control, such as MC_Power, MC_Home, MC_MoveAbsolute and MC_MoveVelocity, with a defined axis state machine. They make motion programs more portable across PLC vendors.

MTX-380 · What is Safe Torque Off (STO)?

A drive safety function that removes power to the motor's switching devices through a safety-rated circuit, so the motor cannot produce torque. It replaces a main contactor for stopping and allows quick restart. A vertical axis still needs a brake.

MTX-381 · A PLC has a 10 ms scan. What is the fastest input signal it can reliably read?

An input must be stable for at least one full scan (plus input filter delay) to be seen. Pulses shorter than about 10 ms may be missed, so reliable pulse rates are well below 50 Hz. Faster signals need high-speed counters.

MTX-382 · What is the Purdue model?

A hierarchical reference for industrial networks: Level 0 (field devices), 1 (PLCs and controllers), 2 (supervisory HMI and SCADA), 3 (operations and MES), 4–5 (enterprise IT). It guides network segmentation and security zones.

MTX-383 · How do you troubleshoot an intermittent PLC input?

Monitor the input online and with a trend, check the input LED, measure voltage at the terminal while the fault occurs, inspect wiring, connectors and sensor alignment, and check for noise from nearby power cables. Use event logging to capture the timing.

MTX-384 · What is ISA-88 and why is it used?

ISA-88 is a standard for batch control that separates recipes from equipment control and defines a physical model (cell, unit, equipment module) and procedural model (procedure, unit procedure, operation, phase). It makes batch systems modular and recipe-driven.

Hydraulics & Pneumatics

MTX-385 · Size a hydraulic cylinder to produce 50 kN at 160 bar.

A = F/p = 50,000/(16 × 10⁶) = 3.125 × 10⁻³ m². D = √(4A/π) ≈ 63 mm. A standard 63 mm bore gives only about 50 kN with no margin, so choose the next size up, such as 80 mm, to allow for losses and margin.

MTX-386 · What pump flow and motor power are needed to extend a 63 mm cylinder at 0.1 m/s at 160 bar?

Area ≈ 3.12 × 10⁻³ m², so Q = A·v = 3.12 × 10⁻⁴ m³/s ≈ 18.7 L/min. Fluid power = p·Q = 16 × 10⁶ × 3.12 × 10⁻⁴ ≈ 5 kW. With 85% pump efficiency, the electric motor needs about 5.9 kW.

MTX-387 · What sets the natural frequency of a hydraulic servo cylinder?

The oil column acts as a spring. For a mid-stroke symmetric cylinder, ωh = √(4βA²/(V·m)), where β is bulk modulus, A piston area, V total trapped volume and m moving mass. Short hoses, larger area and air-free oil raise ωh and allow higher bandwidth.

MTX-388 · What are the challenges of servo-pneumatic positioning?

Air compressibility gives low stiffness and a low natural frequency; valve nonlinearity, friction and stick-slip cause limit cycles; and dynamics change with position and load. Solutions are proportional valves, pressure sensors, model-based or sliding-mode control and friction compensation.

MTX-389 · How much heat is generated when a 30 L/min pump dumps over a 150 bar relief valve?

All throttled power becomes heat: P = Δp·Q = 15 × 10⁶ × 5 × 10⁻⁴ = 7.5 kW. This is why pressure-compensated pumps, unloading valves or load sensing are used rather than continuous relief flow.

MTX-390 · Estimate air consumption of a 50 mm, 200 mm stroke double-acting cylinder at 6 bar gauge, 10 cycles/min.

Swept volume per stroke = 1.963 × 10⁻³ × 0.2 ≈ 0.39 L, so about 0.79 L per cycle ignoring the rod. Compression ratio = (6 + 1.013)/1.013 ≈ 6.9, so free air ≈ 5.4 L per cycle, or about 54 L/min.

MTX-391 · What is a load-sensing hydraulic system?

The pump senses the highest load pressure through a signal line and supplies only the flow needed at a pressure just above it (a fixed margin). This reduces throttling losses and heat compared with fixed pressure systems, as in mobile machinery.

MTX-392 · Why is contamination control critical in hydraulic servo systems?

Most hydraulic failures come from contamination. Particles cause spool sticking, wear and leakage, especially in servo valves with micrometre clearances. Fine filtration, cleanliness targets under ISO 4406, breathers and oil analysis are essential.

MTX-393 · A vertical cylinder drifts down when the valve is centred. What is the cause and fix?

Leakage across the spool valve or piston seals lets the load creep down. Fit a pilot-operated check valve or counterbalance valve close to the cylinder port to hold the load, and inspect piston seals if drift continues.

MTX-394 · What should you consider when replacing hydraulic cylinders with electromechanical actuators?

Compare peak and continuous force, speed, duty cycle and shock loads, energy efficiency, holding needs (brakes), environment, maintenance, leaks and total cost of ownership. Electric actuators give precision and cleanliness but may struggle with very high force or shock.

System Modeling & Simulation

MTX-395 · What is hardware-in-the-loop (HIL) testing?

The real controller hardware and software run against a plant model simulated in real time on a dedicated computer, connected through real I/O signals. It tests faults, edge cases and dangerous scenarios safely before the real machine exists.

MTX-396 · What is the resonant frequency of a motor connected to a load through a flexible shaft?

For two inertias J₁ and J₂ linked by torsional stiffness k, the anti-resonance is √(k/J₂) and the resonance is ωr = √(k(J₁ + J₂)/(J₁·J₂)). It sets the upper limit on servo bandwidth.

MTX-397 · What is acausal (equation-based) modelling?

Components are described by equations connecting physical ports, without fixing which variables are inputs or outputs, as in Modelica or Simscape. The tool sorts the equations, so multi-domain physical models are built by connecting parts like a schematic.

MTX-398 · What constraints apply to real-time simulation for HIL?

Each model step must finish within its fixed time step, so solvers are fixed-step, often explicit, and the model may need simplification. The step must be small enough for the fastest dynamics and for PWM or I/O timing, often 1–100 µs for power electronics.

MTX-399 · A motor's thermal model has R = 0.5 K/W and C = 200 J/K. What happens with 20 W of losses?

Steady-state rise ΔT = P·R = 20 × 0.5 = 10 K. Time constant τ = R·C = 100 s, so the temperature reaches about 63% of the rise (6.3 K) after 100 s and is nearly settled after 400 s.

MTX-400 · What is model order reduction?

Approximating a high-order model with a lower-order one that keeps the important dynamics, for example by removing fast poles or using balanced truncation. It speeds up simulation and controller design, especially for real-time or embedded use.

MTX-401 · When do you use averaged versus switching models of a PWM motor drive?

Averaged models replace switching with the mean voltage, run quickly and suit control design and long simulations. Switching models capture ripple, losses, dead-time and EMI but need very small time steps, so they are used for power-stage design and verification.

MTX-402 · A second-order step response shows 25% overshoot. What is the damping ratio?

ζ = −ln(Mp)/√(π² + ln²(Mp)). With Mp = 0.25, ln 0.25 = −1.386, so ζ = 1.386/√(9.87 + 1.92) ≈ 0.40. The peak time then gives ωn.

MTX-403 · What is Monte Carlo simulation used for in mechatronic design?

Running a model many times with parameters such as tolerances, friction and sensor errors sampled from their distributions. It shows the statistical spread of performance and the probability of failing a specification.

MTX-404 · What is FMI co-simulation?

The Functional Mock-up Interface is a standard for exchanging models as Functional Mock-up Units (FMUs). Models from different tools can be combined and run together, either with the solver included or with a master algorithm coordinating time steps.

MTX-405 · How do Tustin and zero-order-hold discretisation differ?

ZOH discretisation exactly matches the continuous plant's response to piecewise-constant inputs and suits plant models. Tustin (bilinear) maps s = (2/Ts)(z − 1)/(z + 1), preserving stability and approximating frequency response, and is used to convert continuous controllers; it warps frequency near Nyquist.

Machine Vision

MTX-406 · Design a vision inspection for 600 parts per minute.

That is 10 parts/s, so the total budget is 100 ms per part for trigger, exposure, transfer, processing and output. Use a photo-eye trigger, strobed lighting with short exposure, a camera with enough resolution, processing well under 100 ms, and a timed or encoder-tracked reject.

MTX-407 · A conveyor moves at 0.5 m/s and the pixel size on the part is 0.05 mm. What exposure keeps blur below one pixel?

Exposure < pixel size/speed = 0.05 × 10⁻³/0.5 = 100 µs. Use a strobe or high-intensity light to get enough signal in that time.

MTX-408 · What does camera calibration determine?

Intrinsic parameters (focal length, principal point, lens distortion coefficients) and extrinsic parameters (rotation and translation of the camera relative to the world). It is done by imaging a known target such as a checkerboard from several poses.

MTX-409 · How do you convert pixel coordinates to millimetres for an XY table?

Image a calibration target or move the table to known positions, then fit a transform (scale, rotation, offset, and distortion correction if needed) from pixels to machine coordinates. Recheck calibration after any change of lens, focus or camera mount.

MTX-410 · What is sub-pixel edge detection?

Estimating edge location to a fraction of a pixel by fitting a curve to the intensity profile or gradient across the edge. With good contrast and low noise, repeatability of around 1/10 pixel or better is achievable.

MTX-411 · When would you choose deep learning over rule-based vision?

Deep learning suits variable natural defects, textures and appearances that are hard to define with rules, such as scratches on wood. Rule-based tools suit precise measurement, gauging and well-defined features. Deep learning needs labelled data and validation and is less transparent.

MTX-412 · A stereo rig has f = 1000 px and baseline 0.1 m. What depth corresponds to 20 px disparity?

Z = f·B/d = 1000 × 0.1/20 = 5 m. Depth resolution worsens with distance, since ΔZ ≈ Z²·Δd/(f·B).

MTX-413 · How do you light highly reflective metal parts?

Use diffuse dome or coaxial lighting to avoid specular hot spots, or dark-field low-angle lighting to highlight scratches and engravings. Polarisers on light and lens can also suppress glare.

MTX-414 · What line rate does a line scan camera need for 0.1 mm pixels on a web moving at 1 m/s?

Line rate = speed/pixel length = 1/(0.1 × 10⁻³) = 10 kHz. Use an encoder to trigger lines so pixel spacing stays square if web speed changes.

MTX-415 · How do you verify that a vision gauging system is capable?

Perform a measurement system analysis (gauge R&R) with several parts and repeated measurements, and check accuracy against a reference. A common rule is that measurement resolution should be about one-tenth of the tolerance.

Industry 4.0, IoT & Digital Twins

MTX-416 · Design an IIoT predictive maintenance system for a fleet of motors.

Fit vibration, temperature and current sensors; process at the edge (FFT, RMS, envelope features) and send features over MQTT or OPC UA to a platform. Store time series, build baselines and anomaly or remaining-life models, raise alerts to maintenance, and secure all links.

MTX-417 · How do you build a digital twin of a production line?

Combine a model of machines and flows (physics or discrete event), a data link to PLCs and sensors through OPC UA or MQTT, state synchronisation, and analytics. Validate it against real data, then use it for monitoring, what-if planning and virtual commissioning.

MTX-418 · Estimate raw data rate from 100 sensors sampled at 1 kHz with 2 bytes per sample.

100 × 1000 × 2 = 200 kB/s, about 1.6 Mbit/s, or around 17 GB/day before overhead. This is why edge processing sends features or events instead of raw data.

MTX-419 · How would you secure a legacy PLC network being connected to IT?

Segment it into zones with firewalls and conduits per IEC 62443, add an industrial DMZ between IT and OT, restrict and monitor traffic, use secure remote-access gateways, harden and patch where possible, and use OPC UA with encryption for new connections.

MTX-420 · When would you choose OPC UA over MQTT, or the other way round?

OPC UA suits machine-to-machine and plant-level integration, with rich information models, security and client-server or PubSub. MQTT suits lightweight, scalable telemetry to cloud platforms. They are often combined, such as OPC UA over MQTT or Sparkplug B payloads.

MTX-421 · How do you retrofit an old machine for IIoT monitoring?

Add non-invasive sensors (current clamps, vibration, temperature) and read existing PLC data through its protocol or a gateway, without changing safety-critical control. Send data through an edge device to a dashboard, then add OEE and alerts.

MTX-422 · What are the challenges of deploying machine learning at the edge?

Limited CPU, memory and power require compressed or quantised models. Models must be monitored for data drift and updated securely. Results must be deterministic enough for production, and training data must reflect real operating conditions.

MTX-423 · Why is semantic interoperability important in Industry 4.0?

Data is useful only when systems understand what it means. Standard information models, such as OPC UA companion specifications and Asset Administration Shell submodels, give units, context and structure so different vendors' equipment can be integrated without custom mapping.

MTX-424 · Can cloud computing close a motion control loop?

No. Internet latency is tens of milliseconds and variable, while motion loops need deterministic microsecond-to-millisecond timing. Fast control stays in drives and PLCs; the cloud handles supervisory optimisation, analytics and fleet management.

MTX-425 · Line OEE improves from 60% to 70%. What is the effect on output?

Output scales with OEE at the same planned time, so good output rises by 70/60 − 1 ≈ 16.7%. This often justifies IIoT investment without buying new machines.

Other levels: Basic · Intermediate · Tricky · All Mechatronics questions