Advanced Instrumentation & Control interview questions

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

Measurement Fundamentals & Errors

ICE-301 · A pressure loop has transmitter error ±0.10%, calibrator uncertainty ±0.05% and ambient temperature effect ±0.08% of span. What is the combined uncertainty?

Root-sum-square: √(0.10² + 0.05² + 0.08²) = √0.0189 ≈ ±0.14% of span. A simple linear sum (0.23%) would overstate it, because independent errors rarely add in the same direction.

ICE-302 · Power is computed as P = V·I with V uncertain by 1% and I by 2%. What is the uncertainty in P?

For products, relative uncertainties combine by RSS: √(1² + 2²) = √5 ≈ 2.24%.

ICE-303 · A DP flow transmitter has 1% error in ΔP. What is the resulting flow error?

Q ∝ √ΔP, so δQ/Q = ½ × δΔP/ΔP = 0.5%. Square-root relationships halve relative errors, though the absolute DP error becomes a large flow error at low flow.

ICE-304 · How would you set and optimize calibration intervals for a plant's instruments?

Start from manufacturer drift data and criticality, then review as-found results. Extend the interval when as-found values are consistently within tolerance and shorten it when out-of-tolerance findings occur, with safety-critical loops treated most conservatively.

ICE-305 · What is guard banding in calibration acceptance?

Tightening acceptance limits by an amount related to calibration uncertainty, so an instrument that passes is very unlikely to actually be out of tolerance. It reduces false-accept risk at the cost of more adjustments.

ICE-306 · A temperature loop has sensor ±0.5 °C, transmitter ±0.2 °C and DCS input ±0.3 °C. What total probable error should you quote?

RSS: √(0.5² + 0.2² + 0.3²) = √0.38 ≈ ±0.62 °C. The worst case is ±1.0 °C; quote the probable error unless the application requires worst-case.

ICE-307 · A divider of two 100 kΩ resistors across 100 V is measured across the lower resistor with a 1 MΩ meter. What is the loading error?

Parallel resistance = 100k‖1M ≈ 90.9 kΩ. V = 100 × 90.9/(100 + 90.9) ≈ 47.6 V instead of 50 V, an error of about −4.8%.

ICE-308 · A thermocouple with τ = 5 s tracks a temperature rising at 2 °C/s. What is the steady-state lag error?

For a first-order sensor on a ramp, the reading lags by rate × τ = 2 × 5 = 10 °C, and the delay is 5 s.

ICE-309 · A sensor has τ = 4 s. How long after a step does it read within 1% of final value?

1 − e^(−t/τ) = 0.99 gives t = τ·ln(100) ≈ 4.6τ ≈ 18.4 s.

ICE-310 · How many repeated readings are needed to halve the random uncertainty of a mean?

The standard error of the mean is σ/√n, so halving it requires four times as many readings. Averaging does nothing for systematic error.

Sensors & Transducers

ICE-311 · You must weigh a jacketed reactor on four load cells. What design issues matter?

Use mounts that allow thermal expansion and prevent side loads, use flexible piping connections to avoid stiffness shunting load, account for wind and agitator dynamics, match or sum cells correctly, and size cells for dead load plus live load with overload margin.

ICE-312 · A full bridge has four active gauges (GF = 2), two in tension and two in compression, at 1000 µε with 10 V excitation. What is the output?

Vout = Vex × GF × ε = 10 × 2 × 0.001 = 20 mV, four times the quarter-bridge output, with inherent temperature compensation.

ICE-313 · How does a dummy gauge provide temperature compensation in a half bridge?

An identical unstrained gauge on the same material at the same temperature is placed in an adjacent arm, so thermal resistance changes cancel while strain changes appear only in the active gauge.

ICE-314 · Which vibration sensor would you choose for a large turbine with fluid-film bearings?

Non-contact eddy-current proximity probes, because they measure relative shaft displacement within the bearing clearance. Casing accelerometers or velocity sensors suit rolling-element bearings and higher frequencies.

ICE-315 · How would you reduce the response time of a temperature sensor in a pipe?

Use a smaller-diameter sensor and thermowell, a grounded-junction thermocouple, a spring-loaded tip in contact with the well bottom, heat-transfer compound, and place the tip in the high-velocity region.

ICE-316 · A charge amplifier has Rf = 1 GΩ and Cf = 1 nF. What is its low-frequency cutoff?

τ = RfCf = 10⁹ × 10⁻⁹ = 1 s, so fc = 1/(2πτ) ≈ 0.16 Hz. Below this, quasi-static forces are not reproduced.

ICE-317 · A 1024 PPR quadrature encoder runs at 3000 rpm. What are the resolution and channel frequency?

With 4× decoding, there are 4096 counts/rev, or about 0.088° per count. At 50 rev/s, each channel outputs 1024 × 50 = 51.2 kHz.

ICE-318 · How do you choose LVDT excitation frequency?

Make it at least about ten times the highest mechanical frequency to be measured, so the demodulator's filter can remove the carrier without distorting the motion signal, while staying within the core and coil frequency limits.

ICE-319 · Why is a variable-gap capacitive sensor non-linear, and how is it fixed?

C = εA/d varies inversely with gap, so sensitivity dC/dd = −εA/d² changes with position. A differential three-plate arrangement or measuring impedance (∝ d) makes the output linear.

ICE-320 · Compare open-loop and closed-loop Hall-effect current sensors.

Open-loop sensors amplify the Hall voltage directly; they are cheap but suffer from offset drift and core non-linearity. Closed-loop sensors drive a compensating coil to null the flux, giving better accuracy, linearity and bandwidth at higher cost and power.

Temperature & Pressure Measurement

ICE-321 · Why must thermowells be checked for wake frequency?

Flow sheds vortices at f = St·V/d. If this nears the well's natural frequency, resonance can cause fatigue failure. ASME PTC 19.3 TW gives the calculation; fixes include shorter or thicker wells, or velocity collars.

ICE-322 · A type K thermocouple reads 8.0 mV with the reference junction at 25 °C. Estimate the hot-junction temperature.

Using roughly 41 µV/°C, the reference junction is equivalent to about 1.0 mV. The total is 9.0 mV, which corresponds to about 221 °C from type K tables. Always use tables or polynomials, not a linear fit, for precise work.

ICE-323 · A 2-wire Pt100 has 5 Ω per lead. What temperature error results?

Added resistance is 10 Ω. At 0.385 Ω/°C, the error is about +26 °C, which is why 3-wire or 4-wire connections are used.

ICE-324 · Which sensor would you choose for a 1400 °C furnace?

A type R, S or B platinum–rhodium thermocouple in a high-purity ceramic sheath, or a radiation pyrometer. Type K oxidises and drifts badly at that temperature.

ICE-325 · Why should you avoid ranging a smart transmitter at a very high turndown?

Accuracy and stability specs are often stated as % of calibrated span, plus terms based on URL. At high turndown, fixed errors such as zero and temperature effects become a large fraction of the small span.

ICE-326 · What problems do impulse lines cause, and how are they mitigated?

Plugging, freezing, leaks, trapped gas in liquid lines and condensate in gas lines all cause false readings. Fixes include proper slope, heat tracing, purging, close-coupled or remote-seal mounting, and plugged-line diagnostics.

ICE-327 · A closed tank holds liquid of SG 0.9, measured 0–3 m. The LP side uses a wet leg of SG 1.0 rising 3.5 m above the HP tap. What is the calibrated range?

At 0 m: ΔP = −1.0 × 9.81 × 3.5 = −34.3 kPa. At 3 m: 0.9 × 9.81 × 3 − 34.3 = −7.8 kPa. The calibrated range is −34.3 to −7.8 kPa (elevated zero).

ICE-328 · How does ambient temperature affect remote-seal DP systems?

Fill fluid expands and its density changes, creating head and volume errors, especially with long, unequal capillaries. Use balanced seal systems, short capillaries, low-expansion fills or electronic remote sensors.

ICE-329 · An IR pyrometer reads a polished steel roll too low. Why, and what is the fix?

Polished metal has low and variable emissivity, so it emits less radiation than assumed. Set the correct emissivity, use a ratio (two-colour) pyrometer, or measure a high-emissivity target spot.

ICE-330 · A reactor exotherm must be detected within seconds. How do you design the temperature measurement?

Use a tapered or stepped thermowell with minimum wall, a small grounded-junction mineral-insulated thermocouple spring-loaded to the tip, good immersion in the high-velocity zone, and redundant sensors for safety functions.

ICE-331 · How do you protect a pressure transmitter on steam service?

Install a siphon or condensate pot so a water seal protects the sensor from live steam temperature, and correct for the water leg head in calibration.

Flow & Level Measurement

ICE-332 · An orifice was sized for gas density ρd, but actual density is 1.21ρd. How is the indicated mass flow affected?

Mass flow ∝ √(ΔP·ρ). The uncompensated reading assumes ρd, so actual mass flow = indicated × √1.21 = 1.10 × indicated, meaning the meter reads 9% low.

ICE-333 · Why does a DP flow meter have limited turndown?

ΔP ∝ Q², so a 10:1 flow range is a 100:1 DP range. At low flow, the DP is tiny and its fixed error dominates. Use stacked transmitters, a different meter or a smart DP with good low-end accuracy.

ICE-334 · How is gas flow pressure- and temperature-compensated with an orifice?

Density ρ = PM/(ZRT), so mass flow ∝ √(ΔP·P/(Z·T)) using absolute pressure and absolute temperature measured near the meter. A flow computer applies the correction continuously.

ICE-335 · Which flow meter would you select for an abrasive mining slurry?

A magnetic flow meter with abrasion-resistant liner (such as polyurethane or ceramic) and suitable electrodes, because it has no obstruction and no moving parts. DP and turbine meters would erode or plug.

ICE-336 · Why does a magnetic flow meter fail on hydrocarbons?

It needs a minimum liquid conductivity, typically a few µS/cm, to develop a measurable electrode voltage. Hydrocarbons are essentially non-conductive, so use Coriolis, ultrasonic or turbine meters.

ICE-337 · What installation rules apply to a Coriolis meter on liquid service?

Keep tubes full and avoid entrained gas, orient tubes so gas cannot collect (often tubes down for liquids), support the piping to isolate vibration, and perform zeroing under full, no-flow conditions.

ICE-338 · Why are multipath ultrasonic meters used for large-pipe gas custody transfer?

Multiple acoustic paths average the velocity profile, reducing sensitivity to swirl and asymmetry. They have no pressure loss, high turndown and diagnostics such as speed-of-sound checks against gas composition.

ICE-339 · A vortex meter reads zero at low flow despite flow existing. Why?

Below a minimum Reynolds number or velocity, shedding becomes irregular, so the meter applies a low-flow cutoff. Select a smaller meter size to raise velocity, or choose a different technology.

ICE-340 · A DP transmitter spans 2 m between taps in an oil–water separator (ρw = 1000, ρo = 800 kg/m³) and reads 17.66 kPa. Where is the interface?

ΔP/g = 1800 kg/m² = 1000hw + 800(2 − hw) = 1600 + 200hw, giving hw = 1.0 m of water above the lower tap.

ICE-341 · Radar level on an agitated, foaming tank gives erratic readings. What would you do?

Apply false-echo mapping for the agitator, move the antenna away from the agitator, or use guided wave radar or a stilling well. Choose a frequency and antenna suited to foam, or use DP measurement if density is stable.

ICE-342 · An ultrasonic level sensor is compensated for 20 °C air but the vapour space is at 40 °C. What is the error?

c ≈ 331 + 0.6T gives 343 m/s at 20 °C and 355 m/s at 40 °C. The computed distance is about 3.4% short, so level reads high. Use built-in temperature compensation.

Signal Conditioning & Data Acquisition

ICE-343 · A 12-bit system samples at 1 kHz a signal up to 100 Hz. What anti-aliasing filter order is needed?

Attenuation of about 72 dB is needed at 500 Hz, which is 0.7 decade above 100 Hz. A Butterworth filter gives 20n dB/decade, so 14n ≥ 72 and n ≈ 6. Alternatively, oversample and filter digitally.

ICE-344 · How do you reject 50/60 Hz mains noise on thermocouple inputs?

Use differential inputs with high CMRR, twisted shielded cable, and integrating or sigma-delta ADCs whose integration time is a multiple of the line period, placing notches at 50 and 60 Hz.

ICE-345 · Several 4–20 mA loops from equipment at different ground potentials feed one input card with common return. What is the risk and fix?

Ground loops cause circulating currents and offset errors between channels. Use channel-to-channel isolated input cards or loop-powered signal isolators.

ICE-346 · Compare zener barriers and galvanic isolators for intrinsic safety.

Zener barriers are simple and cheap but need a dedicated high-integrity earth and add series resistance. Galvanic isolators need no IS earth, break ground loops and add less loop burden, at higher cost.

ICE-347 · A 24 V loop has a transmitter needing 12 V minimum, a 300 Ω barrier, 50 Ω of cable and a 250 Ω receiver. Will it work at 20 mA?

Drop = 0.020 × 600 = 12 V, leaving exactly 12 V, so there is no margin. Over-range current (such as 21 mA alarm) will fail. Raise the supply or reduce the resistance.

ICE-348 · An ADC has measured SINAD of 74 dB. What is its effective number of bits?

ENOB = (SINAD − 1.76)/6.02 = (74 − 1.76)/6.02 ≈ 12.0 bits.

ICE-349 · How fast should a digital control loop sample?

Sample well above the closed-loop bandwidth; common rules are 5–10 samples per closed-loop rise time, or a sample time of no more than about one-tenth of the dominant process time constant. Slower sampling adds effective dead time.

ICE-350 · How is thermocouple burnout detected, and how do you choose its direction?

A small current through the sensor drives the input to an extreme when the circuit opens. Choose upscale or downscale so the failure leads the loop or SIS to the safe action.

ICE-351 · What are the NAMUR NE 43 signal levels?

Measurement range is about 3.8–20.5 mA. Signals at or below 3.6 mA or at or above 21 mA indicate transmitter failure, allowing the system to distinguish over-range from a fault.

ICE-352 · What trade-off does a first-order digital filter on PV introduce?

It reduces noise but adds lag equal to the filter time constant, reducing phase margin. Keep the filter time constant small compared with the loop's dominant time constant, and retune after adding it.

Control System Theory

ICE-353 · Compare feedforward and feedback control.

Feedforward measures a disturbance and acts before it affects the output, but it needs a model and cannot correct unmeasured disturbances. Feedback corrects any error after it appears. They are usually combined, with feedback trimming feedforward.

ICE-354 · Specify ζ and ωn for 10% overshoot and a 2 s (2%) settling time.

ζ = −ln(0.1)/√(π² + ln²0.1) = 2.303/3.895 ≈ 0.59. ωn = 4/(ζTs) = 4/(0.59 × 2) ≈ 3.4 rad/s.

ICE-355 · For G(s) = K/(s(s + 5)) with unity feedback, find K for a ramp error of 0.02, and comment.

Kv = K/5 = 1/0.02 = 50, so K = 250. Then ωn = √250 ≈ 15.8 and ζ = 5/(2 × 15.8) ≈ 0.16, giving about 60% overshoot, so a lead-lag compensator is needed rather than gain alone.

ICE-356 · How would you design ratio control for blending two streams?

Measure the uncontrolled (wild) flow, multiply by the desired ratio and use it as the setpoint of the controlled flow loop. Characterize both flows linearly and allow a quality analyser to trim the ratio.

ICE-357 · What is split-range control? Give an example.

One controller output drives two or more final elements over different signal ranges. For example, a reactor temperature controller opens the cooling valve at 0–50% output and the heating valve at 50–100%, with care at the transition gap or overlap.

ICE-358 · What is override (selector) control?

Several controllers compute outputs and a high or low selector passes the most limiting one to the final element. For example, a compressor flow controller is overridden by a discharge pressure limit. The non-selected controllers need anti-windup via external reset feedback.

ICE-359 · How does a Smith predictor handle long dead time?

It uses a process model without the delay to give the controller immediate feedback, and adds the model mismatch as a correction. This allows tighter tuning, but it is sensitive to errors in the dead-time estimate.

ICE-360 · When would you use model predictive control instead of PID?

For multivariable processes with strong interactions, constraints and dead times, such as distillation columns, where MPC optimizes future moves over a horizon while honouring limits on inputs and outputs.

ICE-361 · How does the relative gain array guide loop pairing?

λij is the ratio of open-loop gain to gain with other loops closed. Pair variables with λ close to 1, avoid negative λ (it can cause instability), and expect strong interaction for λ much greater than 1.

ICE-362 · For ẋ = [0 1; 0 0]x + [0; 1]u, find state-feedback gains for poles at −2 ± j2.

The desired polynomial is s² + 4s + 8. With u = −[k₁ k₂]x, the characteristic equation is s² + k₂s + k₁, so k₁ = 8 and k₂ = 4.

PID Control & Tuning

ICE-363 · What is integral windup and how do you prevent it?

While the actuator is saturated, the integral term keeps accumulating error, causing large overshoot when the error reverses. Prevent it with anti-windup: clamp or stop integration at limits, back-calculation, or external reset feedback.

ICE-364 · Lambda-tune a PI controller for K = 2, τ = 10 min, θ = 2 min with λ = 10 min.

Kc = τ/(K(λ + θ)) = 10/(2 × 12) ≈ 0.42, and Ti = τ = 10 min. The closed loop then responds like a first-order lag of about 10 min plus the 2 min dead time.

ICE-365 · A loop oscillates at Ku = 4 with Pu = 2 min. Give Ziegler–Nichols PI and PID settings.

PI: Kc = 0.45Ku = 1.8, Ti = Pu/1.2 ≈ 1.67 min. PID: Kc = 0.6Ku = 2.4, Ti = Pu/2 = 1 min, Td = Pu/8 = 0.25 min. Expect about quarter-amplitude decay and consider detuning.

ICE-366 · What are the SIMC tuning rules?

For FOPDT, PI settings are Kc = τ/(K(τc + θ)) and Ti = min(τ, 4(τc + θ)). Choosing τc = θ gives a good balance between speed and robustness, and the 4(τc + θ) limit improves disturbance rejection for lag-dominant processes.

ICE-367 · How should you tune a surge-tank level loop?

Use averaging (loose) control: low gain and long integral time so the outflow changes smoothly and the tank absorbs flow upsets, while keeping level within limits. Because level is integrating, too much integral action causes slow oscillations.

ICE-368 · What is the correct procedure for tuning a cascade loop?

Put the primary loop in manual, tune the secondary loop first for fast, tight response, then switch it to cascade and tune the primary. The inner loop should be several times faster (commonly 3–5×) than the outer.

ICE-369 · A loop is oscillating. How do you diagnose the cause?

Put the controller in manual. If oscillation stops, it is loop-induced (tuning or valve problem); if it persists, it is an external disturbance or interaction. Then check valve stiction, deadband and saturation before retuning.

ICE-370 · How does valve stiction show up in loop data, and what is the fix?

It causes a limit cycle with a sawtooth controller output and square-wave-like PV in flow loops. Retuning only changes the amplitude and period; the real fix is valve maintenance or replacement.

ICE-371 · When is gain scheduling needed?

When process gain or dynamics vary strongly with operating point, as in pH neutralisation, heat exchangers at varying load or equal-percentage valves. Controller settings are interpolated from a scheduling variable such as flow.

ICE-372 · Compare position and velocity forms of discrete PID.

The position form computes the absolute output each sample. The velocity form computes Δu and adds it to the previous output, giving inherent anti-windup and bumpless transfer, but it needs care with P-only control.

ICE-373 · Design a static feedforward for process gain Kp = 2 and disturbance gain Kd = 0.5.

Kff = −Kd/Kp = −0.25. For dynamic compensation, add a lead-lag (τp·s + 1)/(τd·s + 1) and any dead-time difference, and combine it with feedback.

Stability Analysis (Root Locus, Bode, Nyquist)

ICE-374 · What are gain margin and phase margin?

Gain margin is the factor by which loop gain can increase before instability, measured at the −180° phase crossover. Phase margin is the extra phase lag tolerable at the 0 dB crossover. Both are read from Bode plots; typical targets are GM > 6 dB and PM of 30°–60°.

ICE-375 · What does the Routh–Hurwitz criterion tell you, and how are special cases handled?

It gives the number of right-half-plane roots without solving the polynomial. A zero first-column element is replaced by a small ε. A full row of zeros signals symmetric roots; use the derivative of the auxiliary polynomial.

ICE-376 · Find the phase margin of L(s) = 10/(s(s + 1)).

|L| = 1 at ω⁴ + ω² − 100 = 0, giving ω ≈ 3.08 rad/s. Phase = −90° − arctan(3.08) ≈ −162°, so PM ≈ 18°. GM is infinite because phase never reaches −180°.

ICE-377 · For L(s) = K·e^(−s)/(s + 1), what is the maximum stable K?

Phase crossover: ω + arctan ω = π gives ω ≈ 2.03 rad/s. |L| = K/√(1 + ω²) = K/2.26, so K must be less than about 2.26.

ICE-378 · Sketch the key features of the root locus of K/(s(s + 2)(s + 4)).

Asymptotes are at 60°, 180° and 300°, centred at −2. Breakaway is at s ≈ −0.85 (from 3s² + 12s + 8 = 0). The jω crossing is at K = 48, ω = √8 ≈ 2.83 rad/s, so the system is stable for 0 < K < 48.

ICE-379 · Apply the Nyquist criterion to L(s) = K/(s − 1) with unity feedback.

P = 1 open-loop RHP pole, so stability needs one counter-clockwise encirclement of −1, which happens when K > 1. This is confirmed by the closed-loop pole at s = 1 − K.

ICE-380 · What maximum phase lead does C(s) = (Ts + 1)/(αTs + 1) give with α = 0.1?

φmax = arcsin((1 − α)/(1 + α)) = arcsin(0.818) ≈ 55°, occurring at ωm = 1/(T√α). Place ωm at the new gain crossover.

ICE-381 · When would you use a lag compensator?

To raise low-frequency gain and reduce steady-state error without much change to crossover and phase margin. Place its zero and pole well below crossover, typically a decade below, so the added phase lag there is small.

ICE-382 · What does a peak sensitivity Ms = 2 guarantee?

1/Ms is the minimum distance from L(jω) to −1. Ms = 2 guarantees GM ≥ Ms/(Ms − 1) = 2 (6 dB) and PM ≥ 2·arcsin(1/(2Ms)) ≈ 29°. Typical robust designs use Ms of about 1.2–2.0.

ICE-383 · A loop has PM = 45° at ωgc = 0.5 rad/s. How much extra dead time can it tolerate?

Delay margin = PM (rad)/ωgc = 0.785/0.5 ≈ 1.57 s.

ICE-384 · What is conditional stability, and why is it dangerous?

The system is stable only within a range of gain and becomes unstable if gain falls, as well as if it rises. Actuator saturation reduces effective gain and can trigger large oscillations, which the Nyquist plot reveals.

PLC Programming

ICE-385 · Outline a PLC program for a sump with two pumps, level switches and duty alternation.

Use start-on-high and stop-on-low with a seal-in; alternate the lead pump each cycle using a toggle bit; start the lag pump on high-high; include overload and run-feedback faults with automatic standby takeover, and alarm on high-high and on pump failure.

ICE-386 · A 5 ms sensor pulse is read by a PLC with a 20 ms scan. What happens and how do you fix it?

The pulse can occur entirely between input reads and be missed. Use a high-speed counter, an event or interrupt input, a pulse-stretching input, or a faster periodic task with shorter input filters.

ICE-387 · When would you choose Structured Text over Ladder Diagram?

ST suits mathematics, loops, arrays, string handling and complex state machines. LD suits interlocks and discrete logic that maintenance staff must troubleshoot online.

ICE-388 · Why should a PLC PID run in a periodic task?

The integral and derivative terms assume a constant sample time. Running in the continuous scan with variable timing causes inconsistent tuning, so use a timed task or pass the actual elapsed time to the block.

ICE-389 · How do you debounce a mechanical limit switch in a PLC?

Use the input module's filter time or require the input to be stable for a short time, typically tens of milliseconds, with a TON before acting on it.

ICE-390 · How does a hot-standby redundant PLC work?

A primary CPU executes and continuously synchronizes data to the secondary. On a primary fault, the secondary takes over within a bounded switchover time, ideally bumpless with outputs held during transfer.

ICE-391 · How would you implement a robust sequence in Structured Text?

Use an integer step variable with a CASE statement, where each step defines actions and transition conditions, with timeouts per step, a fault step, and hold or abort handling so only one step is ever active.

ICE-392 · How should a PLC handle a failed analog transmitter?

Detect out-of-range signals (for example below 3.6 mA or above 21 mA), set a bad-quality flag, alarm, and either hold the last good value, switch to a backup, or drive dependent logic to a safe state, never using the raw value blindly.

ICE-393 · What are the risks of online edits to a running PLC, and how are they managed?

Edits can cause unexpected output changes, lost data or scan overruns. Manage them with management of change, offline testing or simulation, version control, a backup before download, and coordination with operations.

ICE-394 · How do you harden a PLC against cyber attacks?

Put the CPU key switch in RUN, disable unused services and ports, use password protection and role-based access, segment networks with firewalls per IEC 62443 zones and conduits, and monitor for unauthorized downloads.

DCS & SCADA

ICE-395 · Compare PLC, DCS and SCADA.

A PLC provides fast, deterministic control of machines and discrete processes. A DCS provides integrated plant-wide continuous process control with a unified database. SCADA provides supervisory monitoring and control over wide geographic areas through RTUs or PLCs and communications links.

ICE-396 · Outline a SCADA design for a 200 km pipeline.

Use RTUs or PLCs at pump and valve stations with local logic, fibre with radio or satellite backup, an event-timestamping protocol such as DNP3, redundant master servers at a main and backup control centre, leak detection, and a historian.

ICE-397 · How do you assign controller scan rates and loading in a DCS?

Fast loops such as flow and pressure run at short execution periods, while slow loops such as temperature and analysers run slower. Spread modules across controllers to stay below the vendor's CPU load limit with margin for future expansion.

ICE-398 · How would you migrate a legacy DCS with minimal downtime?

Audit the I/O and logic, convert and test the configuration on a simulator, plan a hot cutover loop-by-loop where possible or during a turnaround, run loop checks, train operators on the new graphics, and keep fallback arrangements.

ICE-399 · How do you rationalize an alarm?

For each alarm, document cause, consequence, operator action and allowable response time; delete those needing no action. Set priority by severity and urgency, and set limits, deadbands and delays to avoid nuisance alarms.

ICE-400 · How should a control network be segmented?

Follow the Purdue levels with firewalls between zones, a DMZ between plant (level 3) and enterprise (level 4) networks, no direct connections from the business network to control, and redundant ring or star topologies for the control layer.

ICE-401 · What is the trade-off in historian data compression?

Exception deadbands and swinging-door compression save storage by discarding points within a tolerance. Settings that are too loose hide fast transients needed for incident analysis, so tune per tag type.

ICE-402 · How should a SCADA system behave on loss of communication to an RTU?

The RTU should keep running local control and safe-state logic, buffer timestamped events, and resume reporting on reconnection. The master marks the data stale, alarms, and blocks remote commands until communication is restored.

ICE-403 · Fifty RTUs each return 500 bytes per poll, every 2 s. What bandwidth is needed?

50 × 500 × 8 / 2 = 100 kbit/s of payload, plus protocol overhead and retries. Report-by-exception could greatly reduce this.

ICE-404 · Why is OPC UA preferred over OPC DA across firewalls?

OPC DA uses DCOM, which needs wide dynamic port ranges and Windows security. OPC UA uses a single configurable TCP port, built-in encryption, authentication and a platform-independent information model.

Industrial Communication (HART, Modbus, Profibus, Fieldbus)

ICE-405 · Why does a HART loop need a minimum loop resistance?

The HART modem signals by modulating current, which needs a load resistance (commonly 250 Ω) to develop a readable voltage. A low-impedance input such as a direct short or some isolators will block communication.

ICE-406 · An FF H1 segment has 12 devices at 20 mA each on 1 km of type A cable (44 Ω/km loop). Is the voltage adequate with a 24 V supply?

Drop = 0.24 A × 44 Ω ≈ 10.6 V, leaving about 13.4 V at the far end, above the typical 9 V device minimum. Add terminators at both trunk ends and check spur lengths.

ICE-407 · A Modbus device returns nonsense floating-point values. What would you check?

Check 32-bit word order (register swap), byte order, off-by-one addressing (40001 notation versus zero-based register 0), the register type (holding versus input) and the data type (float versus scaled integer).

ICE-408 · Estimate the Modbus RTU transaction time to read 10 holding registers at 9600 baud.

Each character is 11 bits, about 1.15 ms. The request is 8 bytes and the response 25 bytes, giving 33 characters ≈ 38 ms, plus 3.5-character gaps (about 4 ms each) and slave processing time.

ICE-409 · How does Profibus DP cable length vary with baud rate?

For RS-485 type A cable: about 1200 m at 93.75 kbit/s, 1000 m at 187.5 kbit/s, 400 m at 500 kbit/s, 200 m at 1.5 Mbit/s and 100 m at 12 Mbit/s per segment, with repeaters to extend.

ICE-410 · How would you choose between HART, Foundation Fieldbus and Profibus PA for a new plant?

Consider the existing DCS support, need for control in the field, diagnostic depth, wiring savings, hazardous area requirements, staff skills and vendor interoperability. HART is simplest and widely supported, while FF and PA offer fully digital multidrop networks.

ICE-411 · How does EtherCAT achieve high determinism?

A single frame passes through all slaves, and each reads and inserts its data on the fly in hardware, giving very short cycle times. Distributed clocks synchronize devices precisely for motion control.

ICE-412 · How would you troubleshoot an intermittent RS-485 network?

Check termination at both ends only, fail-safe biasing, daisy-chain topology without long stubs, shield grounding, common reference, baud and parity settings, and duplicate addresses. Use an oscilloscope to inspect signal quality.

ICE-413 · Is WirelessHART suitable for closed-loop control?

It is mainly used for monitoring and slow loops. Update rates of seconds, mesh latency and battery life trade-offs make it unsuitable for fast critical control, though it is acceptable for slow, non-critical loops with care.

ICE-414 · Modbus TCP has no security. How do you protect it?

Place devices in a segmented zone, use firewalls that inspect Modbus function codes (for example blocking writes from untrusted hosts), use VPNs for remote access, or adopt Modbus/TCP Security with TLS where devices support it.

Control Valves & Safety Instrumented Systems

ICE-415 · Size a valve for 200 gpm of water at 16 psi drop, and state how you would select the body.

Required Cv = 200/√16 = 50. Select a valve whose rated Cv places maximum flow near 70–80% open and minimum flow above about 10–20% open, then check for cavitation, noise and actuator force.

ICE-416 · Calculate the choked liquid ΔP for FL = 0.9, P1 = 100 psia, Pv = 10 psia and FF = 0.94.

ΔPchoked = FL²(P1 − FF·Pv) = 0.81 × (100 − 9.4) ≈ 73 psi. Above this, flow does not increase with ΔP and severe cavitation or flashing occurs.

ICE-417 · Why does an equal-percentage valve often give a linear installed response?

When the valve takes a small share of the system pressure drop, its ΔP falls as flow rises. The equal-percentage curve compensates, so the installed gain becomes roughly constant.

ICE-418 · A single-channel SIF has λDU = 2 × 10⁻⁶ /h and an annual proof test. What SIL does it meet?

PFDavg ≈ λDU × TI/2 = 2 × 10⁻⁶ × 8760/2 ≈ 8.8 × 10⁻³, which lies in the SIL 2 range (10⁻³ to 10⁻²), subject also to architectural constraints.

ICE-419 · For the same sensors in 1oo2 with a 5% common-cause factor, estimate PFDavg.

Independent part ≈ (λDU·TI)²/3 = (0.0175)²/3 ≈ 1.0 × 10⁻⁴. Common cause ≈ β·λDU·TI/2 = 0.05 × 8.8 × 10⁻³ ≈ 4.4 × 10⁻⁴. The total is about 5.4 × 10⁻⁴, in the SIL 3 range, but dominated by common cause.

ICE-420 · LOPA: initiating event 0.2 /yr, relief valve PFD 0.01, target 10⁻⁵ /yr. What SIL is required?

Required PFD = 10⁻⁵/(0.2 × 0.01) = 5 × 10⁻³, a risk reduction factor of 200, so SIL 2 is required.

ICE-421 · What does partial stroke testing achieve, and what are its limits?

Moving an ESD valve a small amount online detects stuck stems and actuator faults, improving PFDavg between full proof tests. It cannot verify full closure or seat tightness, so full stroke tests are still needed.

ICE-422 · Why must the SIS be independent of the BPCS?

A shared sensor, logic solver or valve creates a common-cause failure, so a BPCS fault could both start the hazard and disable protection. IEC 61511 restricts the credit given to BPCS layers and requires independence for claimed protection layers.

ICE-423 · Why is 2oo3 voting often chosen over 1oo2 for sensors?

1oo2 is very safe but any single false signal trips the plant. 2oo3 tolerates one dangerous failure and one spurious failure, giving high safety with fewer spurious trips, and it allows online maintenance of one channel.

ICE-424 · Estimate the actuator force for a globe valve with a 3 in seat and 200 psi shutoff ΔP.

Seat area = π × 1.5² ≈ 7.07 in², so unbalanced force ≈ 200 × 7.07 ≈ 1414 lbf. Add packing friction, seat-loading force for the required shutoff class, and a safety margin.

ICE-425 · How do you choose fail-safe positions for valves in a fired heater?

Choose the position that leads to a safe state on air or signal loss. Fuel gas valves are fail-closed, process flow through the tubes is fail-open to protect tubes from overheating, and stack dampers usually fail open.

Other levels: Basic · Intermediate · Tricky · All Instrumentation & Control questions