Turn a manual step test into stable PID gains, and avoid the overshoot traps that naive tuning creates.
What the Open-Loop Method Really Gives You
The open-loop test measures the process, not the controller. You step the output in manual mode and watch the PV. First, you read dead time L from the flat start. Second, you read the reaction rate R after the PV moves.
The method fits a first-order plus dead-time model. Therefore, you can predict proportional gain Kc, integral time Ti, and derivative time Td. However, the classic rules assume a self-regulating lag. They break on integrating and pure-dead-time loops.
The source example shows a 10% step on a generic process. It found R near 1.33 %/min and L near 2 min. Those two numbers drive the whole tuning.
The Process Reaction Curve Math
Draw a tangent at the steepest PV slope. Its slope is R, the reaction rate. The intercept back to the step time is L, the dead time. The step size is Δm, your output change.
Use the Ziegler-Nichols open-loop rules:
- Kc = 1.2 / (Δm · R · L)
- Ti = 2 · L
- Td = 0.5 · L
These gains aim for a quarter-amplitude decay, which often overshoots in real plants. We usually cut the calculated Kc by 20% to 30%. The loop stays calmer and the operator stops calling the panel.
Running the Test on Yokogawa CENTUM VP
CENTUM VP stores PID settings in a function block. Use this sequence to capture a clean curve:
- Put the loop in manual (M) and wait for a steady PV at the normal load.
- Step the output by 10% and record the time and the new MV exactly.
- Let the PV move for at least 5·L and export the trend to a CSV.
- Fit L and R in Excel, then load Kc, Ti, and Td into the PID block.
For related CENTUM VP hardware, see the Yokogawa CENTUM VP analog input module and CENTUM VP dual-redundant module.
Finally, switch to auto with a small setpoint bump. Confirm the gains before the plant notices.
Tuning the Same Loop in ABB 800xA
ABB 800xA uses Control Builder and the AC 800M controller. First, open the PID block and set the scaling to engineering units. Second, repeat the same manual step through OPC UA or the compact HMI.
800xA trend servers make the curve easy to read. However, watch the sample time. A slow archive rate hides the true dead time. Trend at one second during the test, not at the ten-second history rate.
For 800xA integration work, related options include the ABB 800xA INSUM communication interface and 800xA PROFIBUS interface kit.
Push the tuned values to the controller only after the bump test passes. A good rule beats a fast rule every time.
Traps That Ruin Naive Tuning
- Integrating process: Level or batch tanks can wind up. Use a slower integral.
- Large dead time: The rule can produce excessive gain. Cut Kc and consider dead-time compensation.
- Non-linear valve: A sticky stem can distort the curve. Calibrate the positioner first.
- Noise: Derivative on a noisy PV amplifies spikes. Add a light filter or drop Td.
Document the as-found and as-left gains. Keep the step-test CSV with the loop folder.
Conclusion and Action Advice
Ziegler-Nichols open-loop tuning is still a fast route to a working PID. Measure L and R on a real step, then soften the gains for plant service. Apply the same method on Yokogawa CENTUM VP and ABB 800xA with OPC UA data. Your loops settle faster, and your operators can trust the board again.
Author: Zhang Ming is an industrial automation engineer with over 10 years of experience in PLC, DCS, and control systems.