proportional band not activating when it should

[drbitboy]

Again, we don't know your process. You mentioned cycle time, so maybe this is a batch process. If it is, I wonder if PI control is even the right choice, maybe bang-bang control with a deadband would be better e.g. 100% below 978degC, 0% above 982degC with hysteresis or even linear in between. It depends how quickly the system reacts, how much thermal inertia, etc.

 

[Peter Nachtwey]

Thank you Drbitboy for all of the explanations. I guess I can play with the P and I gain ratios to unsaturate the output earlier. That is the ultimate goal afterall.

The output should be saturated when the PV is less than the proportional band. The problem I see is that your proportional band is effectively too narrow.

 

[drbitboy]

The output should be saturated when the PV is less than the proportional band. The problem I see is that your proportional band is effectively too narrow.

[get your popcorn ... ]

I don't think this is seeing the full picture of this thread; try re-reading Post #1.

I assume the quoted statement above means "... saturated when the [PV's distance from the SP (setpoint)], i.e the [error], is greater than the proportional band (PB)."

It would be more accurate to state the the P term's contribution to the CV output will be greater than 100% if the error is greater than the proportional band, and that will usually, but not always, saturate the CV output.

The CV output may also be

• not saturated when the error is greater than the PB if the I term has shifted the CV offset below the other saturation value (0%), which is unlikely but not impossible.
• saturated when the error is less than the PB if the I term has shifted the CV offset closer to saturation.

Caveat: both cases depend on how the implemented PID algorithm handles anti-windup.

The second case, which is particularly germane to this thread, is why it is the combination of P-gain (100/PB) in concert with (ratio, in this case) the I-gain (repeats/minute, or [minutes/repeat]^-1) that determines when the CV output will desaturate.

So, making the PB more, not less, narrow will cause the valve to desaturate sooner on startup transition (cf. Post #1), but so will lessening the I-gain. Of course, adjusting the tuning constants to limit the overshoot during startup transition may affect control stability and/or quality during the following (steady-state?) operation.

Sidebar: almost four decades ago, I worked in the process control and simulation department of an oil company. One of our products was operator training systems, with a DEC VAX connected to a Honeywell TDC 2000 process control console, where a process simulation in the VAX sent readings to the TDC, which TDC sent outputs back to the VAX to control the (simulated) process. At one point, we looked into saving money by emulating the TDC 2000 inside the VAX and using a custom-built keyboard and a couple of monitors, with a console desk to house them. I coded that TDC emulator, including PID controllers. In Fortran (because we're engineers).