RoboUcs & Computer-Integrated Manufacturing, Vol 4, No 3/4, pp 457-464, 1988 0736-5845/8853 00 + 0 00 Printed m Great Bntam © 1988 Pergamon Press plc • Paper NEW TRENDS IN MACHINE TOOL MONITORING AND DIAGNOSTICS LASZLO MONOSTORI Computer and Automation Institute, Hungarian Academy of Sciences, Kende u. 13-17, Budapest, H-1502 Hungary The line of evolution of manufacturing systems indicates rapidly increasing complexity at every system level, which necessitates enhanced requirements for the monitoring and diagnostic subsystems applied in these manufacturing complexes. This means they must correspondmin performance, complexity and intelligencemto the entire material and data-processing system. This paper summarizes the fundamental requirements for a new family of monitoring and diagnostic equipment and describes two multipurpose, flexible machine tool monitoring systems, which can be regarded as first attempts in this direction. Special emphasis is placed on the generation of reference data for such complex monitoring equipment. Process modelling and teaching approaches are discussed. Pattern recognition methods during learning and decision-making are suggested. Among the most important research and development trends, the use of AI techniques in monitoring and diagnostics is also investigated. 1. INTRODUCTION Four main phases of evolution of manufacturing systems can be distinguished: 5 • direct numerical control (DNC) of groups of machine tools (in the late sixties); n flexible manufacturing systems (FMS), where machines are equipped with automatic workpiece and tool exchangers, and some kind of on-line computer based scheduling is incorporated (start- ing in the seventies, and still spreading); • computer integrated manufacturing (CIM~5), which can be characterized by the synthesis of CAD, CAPP and CAM techniques (imminent breakthrough from experiment into industrial practice); n intelligent manufacturing systems (IMS, tenta- tively forecast by Hatvany and NemesT), capable of "solving, within certain limits, unprecedented, unforeseen" situations, on the basis "even of incomplete and inprecise information" (this case seems to be substantiated by some research and development projects launched in some highly industrialized states). Having surveyed the above line of evolution, rapidly increasing complexity can be observed at each stage and at every system level, such as machines, controllers and monitoring subsystems with the related sensory components, workshop scheduling, system supervision, etc. As a direct con- sequence, new requirements arise from the system design phase through operation and maintenance as well as the reconfiguration ability of the whole sys- tem. One of the most important conditions for highly integrated manufacturing systems is the ability to automatically recognize different--incipient, partial, total and catastrophic--failure situations. This abil- ity enables the system to determine and implement remedial actions, such as, in the case of partial failure situations, the optimal operating mode. 6 Obviously this necessitates enhanced requirements for the monitoring and diagnosticsubsystems applied Acknowledgements--I wish to thank Professor Manfred Weck, head of the Machine Tool Faculty, and Dr. Lutz Kiihne, head of the Machine Tool Monitoring and Diagos- tic Group, for their hospitality, encouragement and their theoretical and practical assistance. Their support allowed me to conduct my research and implement the digital signal processing and pattern recogmtion procedures in the multiprocessor based monitoring system developed during my 12-month scholarship in 1983/84 at the Tech- nical University of Aachen. A part of this investigation was supported by the State Office for Technical Development and the VILATI CNC manufacturing company. Special thanks are due to Mr. Ott6 Bfinhegyland his colleagues for their help and coop- eration. 457