Proceedings Article | 15 December 2003
KEYWORDS: Process control, Cameras, Neural networks, Tolerancing, Control systems, Physics, Sensors, Neurons, Manufacturing, Computer engineering
When we say that a process is "controlled" we mean that we know as much as it is possible to know about that process. When we say that a process has been "characterized", we mean that we know exactly how many critical steps there are, and that we know exactly what happens in each of those steps. Every step has a metric and a tolerance. When that metric is monitored and measured, preferably on a real-time basis, manufacturing management has the necessary data to control that process, and to reap the benefits that come with such control. Developing and installing the instrumentation needed to measure processes required the better part of the 20th Century. Today's process instrumentation can measure every conceivable physical parameter, record it, and communicate it anywhere in the world. This method does have its drawbacks: such instrumentation can be expensive, and often the parameters being measured are only indirectly related to the process. The benefits, however, of a stable, reliable process are worth the effort. The technology that supports instrumentation continually improves, as indeed, it has to if it is not to become obsolete in short order. For some time, however, the greatest improvement in the field has been in improving measurement of physical parameters based on physics, engineering and computer science. However better that focus becomes, process instrumentation development now has an alternative path. Although the end hardware, of course, is still physical, the path is based on a different model: biology, rather than physics. This new path emulates the neural network--that is, the brain. Using neural networks for process control is as old as processes themselves. Long before today's sophisticated instruments, a person would examine a process for some time, learn how it worked, and make a judgement about how well it was actually doing what it was supposed to do. Decisions about process performance were based on what someone observed, and recommendations for improvements were based on that person's experience. Instrumentation is now available that makes judgements in exactly the sme way a person would, but with the precision and tireless reliability of modern electronic equipment. The first of these instruments is called ZiCAM.
