
In industrial large objects a good tuning can take many hours even days. You are a technician or field engineer that has to adjust the PID factors and go to another job. It typically is used for a starting point. Fo one object the tune can work great for other - a lot worse, for some may not work at all. It depends on many factors including the type of the object. And at the end of the day, that is what lets us process control professionals sleep at night.Ī good autotune is not easy at all if at all possible to do it good for any object you want to control. Robust tuning will keep your loop stable over a wide range of process conditions. When you tune a loop, the most important thing you can hope for is 'robustness'. And this is exactly what the men in white coats at Control Engineering Magazine found with 80% of the vendors that they tested. So all autotune loops that I have ever seen are forced into unsatisfactory closed loop tuning, which may work for the process conditions at that exact moment in time, but if the process changes ever so slightly, there is a strong chance that the tuning will fall over. That would really give the operators high blood pressure! No-one wants a controller that every-now-and-then decides to go into open loop (manual) mode for a spot of tuning. This is because, with the loop closed, the feedback will mask all sorts of process characteristics that are essential to know for proper tuning. Why do vendors struggle to provide good autotuning? Well, all autotune algorithms have to work in closed loop (auto).īut the only way to properly tune a PID loop is to use open loop tuning.

Unfortunately they don't have the balls to name the vendors, so if you are planning on using autotune in the near future, a few prayers to the Gods of process control may be in order, to try and beat the 4 out of 5 chance that the tuning will end up as a dog's breakfast. To cut a long story short, they looked at 5 different PID Controller vendors who promised Autotune functions and found that only 1 out of 5 of the Autotune algorithms actually worked! There is a great article in the Jan '09 Edition Of Control Engineering magazine: "Is Automated PID Tuning Dependable?".
