Knowledge-based modelling of a TV-tube manufacturing process

An alternative approach for modelling a complex manufacturing process is proposed. This work has been done in the framework of it European ESPRIT project. From a global point of view, the project has been developed according to the waterfall model of a software lifecycle (Royce, 1970). At a lower level, the process modelling is evaluated through the proposals of Cohen and Howe (1989) which have defined five different types of stages to represent an idealized cyclic model of empirical Artificial Intelligence (AI) research. First, the global architecture for supervision is presented, including the process simulation system. This paper will focus on the modelling phase of the manufacturing process for simulation. Then the basic structure of specification (per workstation) is presented. This structure involves both quantitative and qualitative knowledge. It is shown why and how qualitative relations are used, particularly in conjunction with the quantitative one, to model and to predict some results of the manufacturing process. As illustration, a simulation of a TV-tube manufacturing process by means of the G2 Expert System is presented. Finally, some relevant results from this simulation are analyzed, using the criteria proposed by Cohen and Howe (1989) in their analysis of experimental results