One of the most common causes of unacceptable control performance is the difficulty of identifying and compensating for serious controller interaction in a multiloop environment.

By systematically isolating each of the loops in question and tuning them one at a time (where each interacting loop is frozen on manual allowing this) . Even with a better method how is it possible to know when the system to be tuned is at the optimum point of running? With any tuning method the actual variables chosen is still subjective to the knowledge the tuner has with the process in question. If this fact were not true then the variables used would be same for all tuners.

I'm not a big process guy... Automotive robot/tooling assembly is my game.
However, All the CPU's "loops" of code, must be resolved deterministicaly in 1 scan.
The update rate is the "scan time + I/O update" even if STI or scan interrupt
instructions are used in the application code created, basically it all boils down
to how fast the RTL app+hardware can service the "total loop time" required !

THIS CANNOT BE TESTED ONE AT A TIME.... Validation is by system update rate !
[ it is ok to start/limp up one "loop" at a time...] but, you don't run this way !!!!
so...why test this way ? Auto tooling doesn't Validate like DCS, but we do BUY-OFF !
Generally we want 20hrs continous ...with 4hrs contigous... AT RATE = Cyc time.
WITH EVERYTHING MOVING AS IT WOULD IN PROCESS AT THE OEM's SITE.

MY DCS friends tell me... "SYS VAL" MUST RUN ALL, AT THE SAME TIME !!!
or your just asking for trouble when you get back to your home plant for buy-off.

regards,
Ed Driscoll 

In tuning the PID gains of one of the controllers the effect of the other loops is treated as a disturbance. Validation comes in when these disturbances are introduced and performance evaluated. If the performance/stability are not acceptable, modeling and MIMO design techniques become attractive.

There is no short cuts to the actual knowledge necessary to address loop interactions to any process. The tuner has to be aware of the basic laws of physics, piping layouts, pump curves and actual expectations with presently used components. Sometimes the smallest change in how what is being done does more than all the best tunings. For a much lower price and bigger payback.

It depends and may not be a set of single tuning parameters will suffice the performance of the main controlled variable. Especially if the disturbances have got different types of nature. The most important is the band width of the main controller and its ability to address all the disturbances under one time response.Only optimization is possible identifying clearly which one is a disturbance and which one is an required interaction. Adaption of tuning parameters are possible in multivariable domain where the parameters may be made sensitive to the required interaction and insensitive to the disturbances. Its really a tricky tuning issue and one needs to identify very clearly the variables and its effect on the main controlled variable.

Interesting comments. Tomorrow, in a webcast, we will present how we do it. I have been doing this for 35 years and I agree with your comments. The difference now is that we are using a tool modeling not only loops but intercations as well. You can select tuning parameters (or modify the control strategies) based on operation and process goals but taking in account interactions. Simulation will show how each loop responds considering inteactions and other loops tuning parameters.