An embedded fuzzy computing unit for highspeed industrial process control

A new hardware architecture for controlling an electrodischarge machining (EDM) process is presented. In EDMachining sparks are discharged between tooland workpiece-electrodes thus producing very accurate forms with a well defined surface regardless of the mechanical qualities of the workpiece. The control of EDMachining is aimed at a stable process conductance, with maximum removal rate, high quality surfaces or any combination of these two demands. The EDM removal process is known for its high stochastic and non-linear nature and is therefore extremely difficult to handle. The absence of a complete mathematical model for the removal process and the large amount of operator's knowledge have led to the application of fuzzy logic methodologies. Today most of the fuzzy control systems are software solutions which show a significant lack of data processing speed or possible complexity of fuzzy algorithms. This paper works out a 'fuzzy in hardware solution' which was designed to satisfy the hard timing constraints of EDM control and is closely linked to industrial standards. It is pointed out how embedded fuzzy computing components can be integrated into a high speed multiprocessing system, showing some aspects of communication mechanisms, data flow strategies and multiprocessing management.<>