By Russ Kinner, Contributing Writer
An engineer learns that sometimes introducing an error into a process can correct a problem
At a previous employer, I was called in to determine why the moisture level of a food product was fluctuating, causing the plant to overpack this product to meet minimum packaging weight requirements. The oven that was feeding the food process was moved from another plant in fine working order, and after a number of specialists reviewed the system it was declared to be working as designed. Still, the moisture content went from under 6% to as high as 15% and there didn’t seem to be a pattern to the variability.
The problem quickly came to a head as the former plant that housed the equipment corrected for weight in every package so a moisture variance was not a big deal. The new plant packaged its product by slicing and chopping a slab of material meaning that corrections could not be made easily on the fly.
I reviewed the control system and everything seemed to be operating normally. The PLC output fed a motorized air control valve for the burner. I set up a trend to monitor the loop and discovered that the output was oscillating by approximately 5%. No amount of tuning seemed to reduce that oscillation. I eventually climbed up on the oven and looked at the bi-directional motor. It moved up, then down every few minutes by several percent, correcting for the output changes. There was the clue; it moved roughly 5% each time.
After reviewing the documentation of the valve, I learned that it had a built-in hysteresis to prevent the valve from continuing hunting for the “correct” position. While that saved wear and tear on the valve, it allowed several percent of error to accumulate before moving to correct the error. This caused the oscillation in the controller, not the setting of the tuning within the PID (a generic control loop feedback mechanism.)
I added a timer (30 seconds seemed to do the trick) to the PLC to cause a momentary error to the PID output of about 10%, which drove the valve further open and then corrected back to 0% error after returning to the normal output to the PID. It corrected the problem and allowed the plant to reduce the over pack. This saved about $10,000 per line each week.
Needless to say, they were happy with the results. In fact, I received a large box by truck shipment at my office containing cases of all the products the plant produced - I was one popular guy at the office as I handed out candy bars and other snacks for the next week!
Russ Kinner is a Senior Controls System Engineer based in Phoenix, AZ. He is responsible for guiding electrical/controls projects and also performs this function within multi-discipline projects.