Process Control, Instrumentation and Digitalisation
Scope
Control questions test loop thinking: what is measured, what is manipulated, what disturbs, and what happens on failure. Instrumentation and digitalisation extend this to sensor selection, alarm management, safety instrumented systems, and data integrity.
Core rules
- A feedback loop has a sensor, controller, final element, and process. Tuning depends on process gain, time constant, and dead time. Dead time is the main enemy of stability.
- PID: proportional acts on current error, integral removes offset, derivative anticipates. Derivative amplifies noise; integral windup must be prevented.
- For a first-order-plus-dead-time model, a Ziegler-Nichols PI starting point from the reaction curve is Kc = 0.9 T / (K tau) and Ti = 3.3 tau. Use it as a starting point, then verify stability and robustness.
- Control valve failure position is a safety decision, not an instrumentation default. Fail-closed, fail-open, and fail-last must match the HAZOP consequence.
- Safety instrumented functions are managed by SIL targets. Independence, proof-test interval, and systematic capability all matter.
- Digitalisation adds cybersecurity, alarm rationalisation, data validation, and management of change for software. A dashboard is not automatically a control.
Worked example
A process has gain K = 2, time constant tau = 2 min, and dead time theta = 0.5 min. Using the reaction-curve PI correlation, Kc = 0.9 x 2 / (2 x 2) = 0.45 and Ti = 3.3 x 2 = 6.6 min. The small dead-time-to-time-constant ratio of 0.25 suggests a reasonably controllable loop. If theta rose to 4 min, the same settings would be too aggressive; detune or add feedforward. Always check the final element's range and authority before blaming the controller.
Traps
- Tuning a loop before checking sensor calibration and valve stroking.
- Using derivative action on a noisy flow signal.
- Treating an alarm as a control action. Alarm flood is a design defect.
- Claiming SIL without proof-test coverage, independent sensors, and documented functional testing.
Instrumentation is only as good as its installation. Impulse lines, thermowells, and grounding faults cause more problems than controller algorithms.