Wednesday, September 3, 2008

EDM controller

EDM process is very unstable, especially when working in fine regime. Often surface damage occurs by arc discharging. Unstable working is unavoidable, so the process is forced to work in liable, but effective region of working. The process is run by operator who overlook it and make a feedback control. Automation of process is the final goal of our research.

A strategy of controlling is a very important part of stable EDM action. It is complex and operator can master it after a long period of learning and collecting his own experience. Operator's knowledge and technological knowledge together with technological receipts can be added to adaptive controller which must be able to accept both of them. We are working on development of such a controller (Fig.). All basic function of controller (identification, reasoning and control) are computer made. Identification of process is well done by computer. The same is with reasoning, based on qualitative and probability assessment, which is usually reserved for human. Operator-to-computer communication is possible and controller can be improved. While the split between human way of reasoning and reasoning algorithms, a strategy of controller is developed by artificial intelligence (learning by example). Method FORS (First Order Regression System) is used. Rules IF-THEN-ELSE can be obtain by human demonstration of process leading. By this method, operator knowledge is transformed to algorithm used by computer. Rules IF-THEN-ELSE are human understandable, too.

Figure: Adaptive control system for EDM process control





No comments: