Peter Nachtwey
Member
Quickly make trial temperature changes.
Try and iterative technique where you make trial temperature changes.
A function that computes the undesirable index is still required. Instend of minimizing using advanced techniques one just evaluates U in a seres of nest loops making trial temperature changes.
Start each iteration with U set to a big number.
Now evalute U with all the combinations of temperatures of each zone using the range of Temp(z)-1 to Temp(z)+1. The number of calcuations can be large if there are a lot of zones.
However, in the case of controlling temperature when one has minutes to make many guesses ( trial temperature changes ) and check to see if a new minimum is found, this may a better way. One could continually make trial changes to the temperatures in each zone. If the new undesirable index is smaller than the last then the trial temperature for that zone that is new temperature set point. This should be fast enough to adjust for changing of weighting between zones or changes in the ambient heat. With ten zones that could be done in a MicroLogix with floating point and the computer function block. With 10 zones the U function would have to be calcuated 3^10 or 59000 times for each iteration. At one millisecond per U function calcuation that would take about a mimute per update.
At least this trial and error techinique could be understood by most. I still wouldn't use Fuzzy Logic. One must have an a function that must be evaulated and minimized. In this minimization is done by case by trial and error in stead of using calculus and arrays.
Try and iterative technique where you make trial temperature changes.
A function that computes the undesirable index is still required. Instend of minimizing using advanced techniques one just evaluates U in a seres of nest loops making trial temperature changes.
Start each iteration with U set to a big number.
Now evalute U with all the combinations of temperatures of each zone using the range of Temp(z)-1 to Temp(z)+1. The number of calcuations can be large if there are a lot of zones.
However, in the case of controlling temperature when one has minutes to make many guesses ( trial temperature changes ) and check to see if a new minimum is found, this may a better way. One could continually make trial changes to the temperatures in each zone. If the new undesirable index is smaller than the last then the trial temperature for that zone that is new temperature set point. This should be fast enough to adjust for changing of weighting between zones or changes in the ambient heat. With ten zones that could be done in a MicroLogix with floating point and the computer function block. With 10 zones the U function would have to be calcuated 3^10 or 59000 times for each iteration. At one millisecond per U function calcuation that would take about a mimute per update.
At least this trial and error techinique could be understood by most. I still wouldn't use Fuzzy Logic. One must have an a function that must be evaulated and minimized. In this minimization is done by case by trial and error in stead of using calculus and arrays.