By Dovydas Kersys · Chapter 42 of 58 in Machine Commissioning.
Before you begin: Follow the machine documentation, risk assessment, site procedures and lockout/tagout requirements. Commissioning can expose personnel to electrical energy, stored pressure and unexpected motion. Never defeat a safety function merely to make a test easier. Examples and diagrams are generic and must be checked against the actual machine.
Closed-loop control should be commissioned in a strict order: sensor direction, scaling, actuator direction, manual response, limits, then tuning. Tuning cannot repair reversed action or a sticky valve.

Commission the physics before tuning the mathematics.
Why this matters during commissioning
An incorrectly signed control loop can drive immediately to an unsafe limit.
Commissioning procedure
- Verify process variable increases/decreases correctly with the real process.
- Place output in manual and make small changes to confirm actuator direction.
- Confirm output limits and safe manual values.
- Check controller action (direct/reverse) against process response.
- Begin conservative tuning and observe step response.
- Test setpoint changes, disturbances, mode transfers and alarm limits.
Common failure patterns
- Controller action reversed.
- Integral windup caused by actuator saturation.
- No bumpless transfer between manual and automatic.
- Tuning hides a slow or sticking actuator.
Evidence to record
- Initial tuning values
- Process response
- Output limits
- Final tuning values
- Overshoot/settling observations
- Mode-transfer result
FIELD RULE: If the loop behaves strangely in manual, do not tune it in automatic.
PART VIII
Startup and Production Proving

Startup and Production Proving
Continue the series
- Previous chapter: Control Valve Commissioning
- Next chapter: Dry Runs and Single-Cycle Tests: Commissioning Guide
- Browse all Machine Commissioning chapters
Adapted from Machine Commissioning for Maintenance and Automation Technicians, by Dovydas Kersys. © 2026 Dovydas Kersys. Featured cover image is an AI-generated editorial illustration.
