Cascade PID Control with Multiple Slaves

Process software and control books

https://en.wikipedia.org/wiki/PID_controller

http://blog.west-cs.com/blog/how-does-cascade-control-work

http://innovativecontrols.com/blog/advanced-pid-loop-tuning-methods

I've seen some people who just copy back the Master Output for the BackCal Input and BackOutput (different names on different platforms).

This doesn't give you the advantages (if really available ) of having an idea of how the process is actually responding, and allows the operator to switch modes with perhaps smaller bumps, less monetary loss.

You may wish to externally constrain under what conditions the operator can make a control configuration change, to externally build in the protection these signals give you.

The other possibility is to build an equation that sums the effect of the normal or typical ranges of the 4 variables (MV), as a model for the total DO Process Variable you are manipulating. This could be complex, but perhaps 2nd order polynomial regression for each of the 4 components will work. This model will likely will need to be developed on an operating installation. You will have to estimate the cost of developing this to the possible benefit. If you are controlling a million dollar per day reactor, with highly variable feedstock, and O2 consumption, then the complexity may be justified.

If putting the process in MANUAL control requires a few tweaks with minutes to spare before the reactor becomes unprofitable, then simplicity is what you should first shoot for.