CONTROL & INSTRUMENTATION AUTOMATION

Control Is Only Ever as Good as the Measurement

A control system cannot be better than what it is told. A PID loop fed by a badly installed sensor will hold a wrong number very precisely, and everyone will trust it because it appears on a screen. Most “the automation is not working” complaints turn out, on inspection, to be measurement problems.

• A drifting sensor nobody has calibrated in three years, quietly moving your process off target
• A thermowell inserted too shallow, reading the pipe wall rather than the product
• A flow meter installed with no straight length upstream, reading turbulence
• A signal cable run alongside a VFD output, picking up noise that looks like process variation
• An instrument specified for the wrong range, so the operating point sits at 3% of span with no useful resolution

Choosing the Instrument for the Duty

The cheapest instrument that meets the duty is the right one. The cheapest instrument that does not is the most expensive thing on the P&ID, because everything downstream of it is wrong.

Where It Is Installed Matters as Much as What It Is

Instrument selection gets attention in the specification; installation detail rarely does. It is where accuracy is actually lost.

• Straight lengths upstream and downstream of flow meters, per the manufacturer — not per what fits in the available space
• Thermowell insertion depth sufficient to read the product, oriented into the flow
• No dead legs. In a hygienic plant a dead leg is both a measurement error and a contamination risk
• Drainable and CIP-able — an instrument that cannot be cleaned in place will eventually be bypassed
• Accessible for calibration without scaffolding, or it will not be calibrated
• Isolation valves so an instrument can be replaced without draining the line

Hygienic Design for Dairy and Food

• Product-contact materials and surface finish appropriate to food contact
• Hygienic process connections rather than threaded fittings in product lines
• Crevice-free, self-draining installation — no horizontal pockets holding product through a wash
• Withstands CIP chemistry and temperature, cycle after cycle, not once
• Documented, because an auditor will ask what is in contact with product and expect an answer with certificates

For the wash regime these instruments must survive, see dairy plant automation.

Calibration, Drift and Records

An uncalibrated instrument is an opinion with a display. Drift is slow and invisible, which is exactly what makes it dangerous — the plant moves off target gradually and everyone adapts to the wrong number.

• A calibration schedule by criticality, not one blanket interval for everything
• Records with traceability — as-found and as-left values, so drift is visible over time rather than silently corrected
• Tight discipline on legally significant measurements — pasteurisation temperature, reception quantity
• Verification built into the control system where possible: two independent measurements that should agree, alarming when they diverge
• Calibration status visible, so an operator knows whether to trust a reading

As-found readings are the valuable ones. If an instrument is routinely found well out of tolerance, the interval is wrong or the installation is.

Signal Integrity — Where Good Instruments Go Wrong

• Signal cable segregated from power, and never run alongside VFD output cabling. Drives are the single most common source of instrument noise in a process plant
• Screened cable, earthed at one end only — earthed at both ends creates the loop it was meant to prevent
• A single-point earthing philosophy for the instrument system, decided and documented
• 4–20 mA remains robust and diagnosable; a reading of 0 mA is a broken loop, not a zero measurement, and that distinction matters
• Digital protocols where the diagnostics justify them, not because they are newer
• Loop-checked at commissioning, every point, field to screen, and recorded

Loop Tuning

A loop that oscillates gets put into manual, and once in manual it stays there. The plant then runs on operator attention instead of control, which works until the shift changes.

• Verify the measurement first — see above. Tuning around a bad sensor is wasted effort
• Check the final control element — a sticking valve or a worn positioner cannot be tuned out
• Tune for the actual operating range, not one convenient setpoint
• Handle the practicalities — bumpless transfer, sensible output limits, anti-windup on integral action
• Record the parameters so the next engineer inherits a starting point rather than a mystery
• Review loops in manual periodically. Every loop sitting in manual is a control system not doing the job it was paid for

What to Instrument — and What Not To

More instruments is not better. Every one is capital, calibration, spares and another thing to fail.

• Instrument what you control — the loop needs it
• Instrument what you must prove — auditors, customers, payment
• Instrument what you cannot afford to discover late — the excursion that spoils a batch
• Instrument what pays for itself — energy per unit, yield, the numbers that fund the next project
• Do not instrument what nobody will look at, act on, or maintain

On retrofits the answer is often that fewer, better-placed instruments beat more of them. Where the aim is visibility rather than control, see plant monitoring and dashboards.