Process Control Interview Questions

PID controller is the most common feedback control algorithm combining three modes: Proportional (P) produces output proportional to current error, Integral (I) accumulates past errors to eliminate steady-state offset, and Derivative (D) predicts future error based on rate of change. Together, these provide responsive, accurate control. The controller output is: Output = Kp*e + Ki*integral(e) + Kd*de/dt, where e is the error between setpoint and process variable.

Open-loop control applies a predetermined action without feedback - output is not affected by process response (e.g., timed irrigation). Closed-loop control continuously measures the process variable, compares to setpoint, and adjusts output to minimize error (e.g., thermostat). Closed-loop provides automatic correction for disturbances and process variations, making it essential for precise process control. Most industrial control systems use closed-loop (feedback) control.

Setpoint (SP) is the desired value of the controlled variable set by the operator. Process Variable (PV) is the actual measured value of the controlled parameter. Error is the difference between SP and PV (Error = SP - PV for direct acting, PV - SP for reverse acting). The controller's job is to minimize error by adjusting the manipulated variable (controller output) that affects the process.

Proportional band (PB) is the percentage of input range that causes 100% change in output. Controller gain (Kp) is inversely related: Kp = 100/PB. A 50% PB means Kp = 2 (50% input change causes 100% output change). Narrow PB (high gain) gives more aggressive response but may cause instability; wide PB (low gain) gives sluggish but stable response. Modern controllers typically use gain directly; traditional pneumatic controllers used proportional band.

Direct acting controller increases output when process variable increases (e.g., cooling system - temperature rises, increase cooling). Reverse acting controller increases output when process variable decreases (e.g., heating system - temperature falls, increase heating). Selection depends on the process and final control element: consider if the valve fails open or closed and the effect on the process. The combination of controller and valve action must provide negative feedback for stable control.

Integral action eliminates steady-state error (offset) that exists with proportional-only control. It continuously sums the error over time and adjusts output until error becomes zero. The integral term (Ki * integral of error) increases as long as any error exists, eventually driving the process variable to exactly match the setpoint. Integral time (Ti) or reset rate determines how fast this correction occurs. Too much integral can cause overshoot and oscillation.

Cascade control uses two controllers in series - a primary (master) controller sets the setpoint for a secondary (slave) controller. The secondary loop controls a faster, inner variable that affects the primary variable. Example: temperature control of a reactor where primary controller outputs setpoint to steam flow controller. Cascade control improves response to disturbances in the secondary loop before they affect the primary variable, providing faster correction and better control quality.

Derivative action responds to the rate of change of error, providing anticipatory control. It applies corrective action based on how fast the error is changing, not just its magnitude. Derivative is useful for slow processes where early correction prevents overshoot. However, derivative amplifies noise, so it is often used with filtering or omitted entirely (PI control) for noisy processes. Derivative time (Td) determines the strength of this predictive response.

In automatic mode, the controller calculates output based on PID algorithm to minimize error between setpoint and process variable. In manual mode, the operator directly sets the controller output, bypassing automatic calculations. Manual mode is used during startup, shutdown, maintenance, or troubleshooting. Bumpless transfer ensures smooth transition between modes by matching internal states before switching, preventing sudden output changes that could upset the process.

A feedback control loop consists of: Process (system being controlled), Sensor/Transmitter (measures process variable), Controller (compares PV to SP and calculates output), Final Control Element (valve, VFD - implements control action), and Signal path (4-20 mA, fieldbus connecting elements). The loop creates a circular path: controller output affects process, sensor measures result, controller adjusts based on error. Each element contributes to overall loop performance and stability.

Ratio control maintains a fixed ratio between two process variables by having one variable (wild) set the setpoint for another (controlled) through a ratio station. Example: maintaining air-to-fuel ratio in a combustion system - fuel flow is wild flow, air flow is controlled to maintain stoichiometric ratio. Another example is blending two streams in fixed proportion. Ratio station multiplies wild flow by ratio factor to generate controlled flow setpoint.

Overshoot occurs when process variable exceeds setpoint before settling. Common causes: too high proportional gain (aggressive response), excessive integral action (windup), insufficient derivative action, or fast setpoint changes. Reduction methods: reduce proportional gain, increase derivative time, use setpoint ramping/filtering, enable anti-windup protection, or implement two-degrees-of-freedom control. Some overshoot may be acceptable for faster response; the balance depends on process requirements.

Feedforward control measures disturbances before they affect the process and applies corrective action preemptively. Unlike feedback control that reacts to error after it occurs, feedforward anticipates disturbances using a process model. Example: measuring feed rate change to a heat exchanger and adjusting steam flow before outlet temperature changes. Feedforward provides faster disturbance rejection but requires good disturbance measurement and accurate process model. Often combined with feedback control.

Split-range control uses a single controller output to manipulate two or more final control elements sequentially over different portions of the output range. Example: pressure control using a vent valve (0-50% output) and supply valve (50-100% output) - as controller output increases, first vent closes then supply opens. Applications include temperature control with heating and cooling, pressure control with supply and vent, and pH control with acid and base addition.

Standard controller output is 4-20 mA (analog) or 0-100% (digital), representing the commanded position or action of the final control element. 4 mA/0% typically corresponds to valve closed (or minimum action), and 20 mA/100% to valve fully open (or maximum action). This range allows detection of signal failures (below 4 mA indicates fault). The output is scaled to the final element - for a valve, it represents demanded position; for a VFD, speed setpoint.

Ziegler-Nichols provides two methods: (1) Ultimate gain method - with I and D disabled, increase P gain until sustained oscillation occurs (ultimate gain Ku) at ultimate period Pu. PID settings: Kp = 0.6Ku, Ti = Pu/2, Td = Pu/8. (2) Reaction curve method - apply step change in manual, measure process response to get delay time L and time constant T. PID settings: Kp = 1.2T/(K*L), Ti = 2L, Td = 0.5L. These are starting points requiring refinement; often produce aggressive tuning.

Integral windup occurs when the controller output saturates (at 0% or 100%) but the integral term continues accumulating error, causing excessive overshoot when the output returns to normal range. Prevention methods: external reset (integral tracks actual valve position), conditional integration (disable integral when output saturated), back-calculation (reduce integral based on saturation), and velocity form algorithm. Modern controllers include anti-windup protection that must be properly configured for cascade, override, and split-range applications.

Cascade design requirements: secondary loop must be faster than primary (typically 3-5 times faster), secondary measurement must indicate disturbances affecting the primary, and secondary variable should respond to the primary manipulation. Tuning procedure: first tune secondary loop with primary in manual (fast, tight tuning), then tune primary loop with secondary in cascade (slower, conservative tuning). Secondary should reject disturbances before affecting primary. External reset from primary to secondary prevents windup during mode changes.

Process characterization identifies key parameters: process gain (Kp - change in PV per unit change in output), dead time (delay before response begins), and time constant (time to reach 63.2% of final value). Methods: step test (apply output step in manual, record response), pulse test, or relay auto-tune. Calculate process parameters using tangent method or software analysis. These parameters determine whether process is self-regulating or integrating, fast or slow, and guide tuning method selection. Document at multiple operating points for nonlinear processes.

Override control uses high or low selectors to choose between multiple controller outputs, allowing one control objective to take priority over another under certain conditions. Example: compressor control where normal flow control is overridden by low suction pressure protection - selector chooses minimum output. Override configurations: high selector for protective limits on maximum, low selector for protective limits on minimum. Anti-windup (external reset) must be configured for the controller not selected to prevent windup.

Valve characteristics describe the relationship between valve position and flow. Linear valves provide proportional relationship, equal percentage valves give same percentage flow change per unit position change (small flow at low opening, large at high), and quick opening provides large flow change at low opening. Process gain varies with valve position for non-linear valves. Equal percentage valves compensate for process non-linearities, providing more consistent loop gain across operating range. Proper matching of valve to process characteristics ensures stable control throughout the operating range.

Dead time (transport delay) makes control difficult because the controller responds to errors that reflect past, not current, conditions. Compensation strategies: reduce controller gain and increase integral time (detuned PI), use Smith Predictor (model-based compensation predicting future PV), implement model predictive control for complex processes. Design considerations: minimize physical dead time where possible, ensure accurate dead time estimation (often varies with flow rate), and consider feedforward control to reduce disturbance effects before dead time delay.

PID controller forms differ in how P, I, D interact: Ideal (ISA) form - Kp multiplies all terms: Kp*(e + 1/Ti*integral(e) + Td*de/dt). Series (interacting) form - terms multiply: Kp*(1 + 1/Ti*s)*(1 + Td*s). Parallel (independent) form - each term independent: Kp*e + Ki*integral(e) + Kd*de/dt. Conversions between forms required when moving tuning parameters between different controller types. Series form was common in pneumatic controllers; ideal form is standard in DCS/PLC.

Stability assessment methods: time-domain analysis (observe oscillation, damping ratio - quarter decay ratio target, overshoot percentage), frequency-domain analysis (Bode plot showing gain and phase margins - typically >6 dB gain margin, >45 degree phase margin), and empirical indicators (decreasing oscillation amplitude, settling within acceptable time). Unstable loops show increasing oscillation amplitude or runaway response. Loop stability depends on total loop gain - product of controller, process, valve, and sensor gains. Reduce gain or add filtering if marginally stable.

Lambda tuning (Internal Model Control) provides user-specified closed-loop time constant (lambda), allowing tradeoff between response speed and robustness. For first-order plus dead time process: Kc = T/(K*(lambda + L)), Ti = T, Td = 0 (PI controller). Lambda is typically set to 1-3 times the dead time. Lambda tuning is preferred when: conservative, non-oscillatory response is needed, process has variable characteristics, or safety requires robust stability. It produces less aggressive tuning than Ziegler-Nichols with better robustness to process variations.

Batch control involves discrete, time-based operations with defined start and end points, following recipes and sequences through phases (charging, heating, reacting, discharging). ISA-88 standard defines equipment hierarchy (unit, equipment module, control module) and recipe structure. Batch requires: sequential function charts for phase logic, flexible setpoint profiles, transition conditions between phases, and recipe management. Unlike continuous control with steady-state operation, batch involves frequent startup/shutdown and product changeovers requiring different tuning and control strategies.

Oscillation diagnosis: determine oscillation type - continuous sustained (controller at stability limit), decaying (acceptable if reasonable), increasing (unstable, immediate action needed), or irregular (external disturbance, valve sticking). Diagnostic steps: put controller in manual to check if oscillation stops (controller-induced vs external), check for valve stiction/hysteresis (monitor valve position vs output), examine interaction with other loops, and verify sensor signal quality. Fixes: reduce controller gain, increase derivative for slow oscillation, fix mechanical issues (valve, actuator), or retune using appropriate method.

Combined FF-FB system: feedforward handles measured disturbances quickly; feedback corrects for modeling errors and unmeasured disturbances. Design steps: identify major disturbances that can be measured, develop feedforward model (gain and dynamics matching process response), implement feedforward signal summed with feedback output, tune feedback for unmeasured disturbances (can be less aggressive since FF handles primary disturbances). Lead-lag compensation in FF path may be needed to match process dynamics. FF improves disturbance rejection; FB provides accuracy.

Auto-tuning methods automatically determine PID parameters through process identification. Common approaches: relay feedback (induces controlled oscillation using relay, measures ultimate gain and period), step response analysis (applies step change, extracts process model), and pattern recognition (adjusts parameters based on closed-loop response patterns). Auto-tuning provides initial parameters that may require fine-tuning. Considerations: ensure process can tolerate test disturbances, run at normal operating conditions, and verify results with manual analysis. Many DCS platforms include built-in auto-tune functions.

Regulatory control focuses on rejecting disturbances while maintaining setpoint (disturbance rejection) - most common in process control where setpoints change rarely. Servo control focuses on tracking setpoint changes (setpoint tracking) - important when setpoints change frequently. Tuning priorities differ: regulatory tuning emphasizes disturbance rejection with minimal overshoot; servo tuning emphasizes fast setpoint tracking with acceptable overshoot. Two-degrees-of-freedom control allows independent optimization of both by separating setpoint and PV paths through the controller.

Process gain nonlinearity (gain varies with operating point) causes inconsistent control performance. Solutions: equal percentage valves (compensate for process gain increase with flow), gain scheduling (change controller parameters based on operating point or throughput), adaptive control (automatically adjusts tuning), characterization (linearize valve signal with function block), or model predictive control. Identification: run step tests at multiple operating points, compare responses. For severe nonlinearity, single tuning provides acceptable performance only over limited range.

Loop interaction occurs when control action in one loop affects the process variable of another loop (coupling). Diagnosis: Relative Gain Array (RGA) analysis quantifies interaction - diagonal elements near 1 indicate low interaction. Management: proper variable pairing (minimize RGA off-diagonal elements), decoupling control (feedforward from interacting outputs), sequential loop tuning (tune tighter loop first, detune interacting loops), detuning (reduce gains on both loops), or multivariable control (MPC handles interactions explicitly). Severe interaction may require process redesign.

Proper valve sizing is critical for control performance. Oversized valves: operate at small opening where small position changes cause large flow changes (high gain, sensitive to stiction), poor rangeability, and control valve hunting. Undersized valves: limit maximum flow capacity, operate near wide-open with little control authority. Target: valve operates at 60-80% open at normal conditions with capacity for upsets. Cv calculation must account for pressure drop variation with flow. Installed characteristic differs from inherent due to system pressure drop distribution.

Integrating processes have no natural equilibrium - output continues changing with any non-zero input (level rises/falls indefinitely). Control characteristics: process gain is ramp rate per unit input, no self-regulation. Tuning approach: integrating processes require tight P-only or PI control; derivative generally not helpful. For level control: determine required tightness based on downstream needs (tight control for feed to sensitive process, loose control for surge tanks). Accumulator sizing affects control difficulty. Lambda tuning for integrators: Kc = 2T/(K*lambda^2), Ti = 2*lambda.

Pure derivative amplifies high-frequency noise (derivative of noise is larger noise). Derivative filter (first-order lag) limits high-frequency gain: filtered derivative = Td*s/(alpha*Td*s + 1). Alpha (typically 0.1) sets filter time constant as fraction of Td. Smaller alpha = more filtering but less true derivative action; larger alpha = less filtering but more noise sensitivity. Some controllers use derivative on PV only (not error) to avoid derivative kick on setpoint changes. Filter configuration is critical for noisy processes like flow measurement.

MPC implementation involves: process modeling (step response or transfer function models through testing), variable selection (controlled variables, manipulated variables, disturbance variables), constraint definition (valve limits, rate limits, process constraints), tuning (prediction horizon, control horizon, move suppression weights, CV/MV priorities), and real-time optimization integration. Design steps: steady-state analysis, dynamic matrix identification through plant tests, controller configuration, simulation validation, commissioning with constraints active, and performance monitoring. MPC handles multivariable interactions, constraints, and optimization simultaneously, ideal for complex units like distillation columns and reactors.

Robustness analysis ensures controller performs adequately despite process variations. Methods: sensitivity analysis (vary process parameters +/-20%, check stability margins), structured singular value (mu) analysis for multivariable systems, Monte Carlo simulation with parameter distributions, and H-infinity design for worst-case optimization. Key metrics: maximum sensitivity Ms (peak of sensitivity function, target <2), complementary sensitivity Mt, gain and phase margins vs parameter variations. Document acceptable parameter ranges for maintained stability. Lambda tuning inherently provides robustness by setting closed-loop bandwidth below process uncertainty.

Decoupling compensates for loop interactions using feedforward elements between loops. Design: calculate decoupler transfer functions from process model (ideal decoupler inverts off-diagonal elements), implement realizable approximations (may require lead-lag elements), and tune decoupled loops independently. Static decoupling uses steady-state gains only; dynamic decoupling matches time response. Partial decoupling may be practical when full decoupling is complex. Alternative: use MPC which handles interactions inherently. RGA analysis guides decoupler design - large off-diagonal elements indicate significant interaction requiring decoupling.

Inferential control uses calculated (soft sensor) values when direct measurement is unavailable, expensive, or slow (e.g., product quality). Design: identify correlating measurements, develop model (empirical regression, first-principles, neural network), implement online calculation, validate against lab or analyzer data, and design bias update from intermittent measurements. Key considerations: model accuracy vs complexity, update frequency, input validation, and bias correction scheme. Applications: distillation composition from temperature profile, reactor conversion from heat balance, crude assay from density and temperature. Implement confidence indication and fallback strategy.

Comprehensive loop assessment methodology: data collection (time series of PV, SP, OP, status flags), statistical analysis (mean, variance, standard deviation), oscillation detection (autocorrelation, spectral analysis), valve performance (backlash, stiction from MV-OP analysis), controller utilization (time in auto, saturation frequency), economic impact analysis (variability cost, constraint violations). Key metrics: IAE (Integral Absolute Error), Harris index (actual vs minimum variance), valve travel, oscillation period and decay ratio. Use automated tools for plant-wide assessment, prioritize loops by economic impact, and track improvements through KPIs.

Anti-surge control prevents compressor operation beyond surge line (minimum flow limit). Design elements: surge detection (distance to surge from operating point), surge control line (safety margin below surge line, typically 10%), recycle valve sizing (pass surge flow at trip conditions), fast-acting control loop (high gain, minimal dead time), hot bypass or cold recycle selection, and integration with load/performance control. Implementation: PID with derivative for fast response, specialized anti-surge controllers with adaptive algorithms, measurement validation (suction/discharge pressure, temperature, flow). Include post-surge recovery logic and coordinate with ESD system.

Sampling time affects digital controller performance through: discretization (conversion from continuous to discrete), aliasing (frequencies above Nyquist appear as low frequency noise), phase lag (sample-and-hold introduces delay equivalent to T/2), and computational delay. Guidelines: sample 10-20 times per closed-loop time constant, minimum 10x faster than process dominant time constant. Fast sampling wastes resources; slow sampling degrades performance. For cascade loops, secondary should sample faster. Anti-aliasing filter required before sampling (cutoff at 1/5 to 1/10 sample rate). Consider computational delay in tuning.

Nonlinear control strategies: gain scheduling (tune at multiple operating points, interpolate parameters), feedback linearization (transform nonlinear system to linear), variable structure/sliding mode control (robust to parameter variations), neural network control (learn nonlinear relationships), nonlinear MPC (use nonlinear models in optimization), and fuzzy logic (rule-based expert system approach). Implementation considerations: model accuracy requirements, computational demands, commissioning complexity, and maintenance capability. pH control exemplifies extreme nonlinearity - titration curve gain varies 1000:1 requiring special techniques like adaptive or reagent-demand control.

State estimators reconstruct unmeasured states from available measurements for control and monitoring. Kalman filter provides optimal estimation for linear systems with known noise statistics. Extended Kalman Filter (EKF) handles nonlinear systems. Moving Horizon Estimation (MHE) handles constraints. Applications: fault detection (estimate vs measurement deviation), soft sensors (estimate quality variables), unmeasured disturbance estimation, and sensor validation. Implementation: develop state-space process model, tune estimator gains (noise covariance matrices), validate estimation accuracy, and integrate with control strategy. Consider computational requirements for real-time implementation.

Degradation identification: trend key metrics over time (variance, valve travel, controller mode time, oscillation frequency), detect drift in process model parameters, monitor maintenance records (calibration results, valve overhaul). Common causes: valve wear/stiction (increasing hysteresis), sensor drift, fouling changing process dynamics, changed operating conditions. Address through: regular loop audits (annual assessment), automated monitoring software, proactive maintenance (valve overhaul, sensor calibration), model update for APC systems. Implement benchmark testing after maintenance to verify restoration. Economic justification through variability reduction and constraint optimization.

RTO integration hierarchy: RTO (steady-state economic optimization, minutes to hours cycle) provides setpoints to APC layer, which provides setpoints to regulatory (DCS/PLC) layer. Design considerations: execution frequency matching, constraint handling consistency, model consistency between layers, steady-state detection before RTO execution, feedback from actual to update RTO models. Interface: RTO writes setpoint targets to MPC CVs or directly to PID setpoints with feasibility checking. Implement range limits, rate limits, and operator override capability. Validate economic benefits through production accounting.

Distillation control strategy design: identify degrees of freedom (typically 5: reflux, distillate, bottoms, reboiler duty, condenser duty), select product quality control points (direct composition or inferential from temperature), choose inventory controls (level in accumulator and column base, pressure), and address constraint handling. Common configurations: LV (reflux flow, boilup), DV (distillate flow, boilup), L/D-V/B (ratio control). Use RGA for variable pairing. Implement composition/temperature cascade, constraint control for flooding/weeping limits, feed-forward from feed rate/composition, and consider dual composition control for high-purity columns. MPC effective for multivariable optimization.

Online valve diagnostics identify problems before failure using data from smart positioners and control signals. Diagnostic parameters: friction/stiction (hysteresis, dead band from valve signature), response time (step response test), seat leakage (minimum controllable flow), actuator performance (travel time, air consumption), and positioner tracking. Implementation: configure HART/Fieldbus polling for diagnostic data, establish baseline signatures during commissioning, trend key parameters, set alarm thresholds for degradation. Integrate with asset management software for predictive maintenance scheduling. Combined with control loop analysis identifies whether loop issues are valve-related.

Adaptive control automatically adjusts parameters for changing process dynamics. Approaches: self-tuning regulators (continuous model identification + tuning recalculation), model reference adaptive control (adjust parameters to match reference model response), gain scheduling (pre-computed tuning vs measured operating condition), and extremum seeking (optimize performance metric continuously). Implementation considerations: identification excitation requirements, stability during adaptation, parameter bounds to prevent unrealistic values, and when to trigger adaptation. Applications: processes with varying composition, throughput-dependent dynamics, or seasonal variations. Test thoroughly before deploying to prevent instability.

Coordinated control balances energy demand with boiler capability for stable power generation. Control modes: boiler-following (turbine controls MW, boiler follows), turbine-following (boiler controls pressure, turbine follows), and coordinated (both respond to load demand simultaneously). Design elements: MW demand controller, pressure controller, fuel-air ratio control with O2 trim, feedwater control (three-element), and steam temperature control (desuperheater spray). Cross-limiting prevents fuel-rich or air-rich combustion. Runback logic handles constraints (mills, fans). Coordinate with automatic generation control (AGC) for grid frequency response. Implement extensive simulation before commissioning.