Applications of GMDH-type modeling in manufacturing

This paper introduces the group method of data handling (GMDH) algorithm which addresses the problem of modeling complex systems. It is a heuristic self-organizing modeling approach for modeling different processes/phenomena on a metalworking shop floor. These models which can be used in monitoring/prediction of the processes also present information in a concise and accurate form to the planning functions, such as process planning and scheduling. This helps in making these functions more dynamic and the decisions more realistic. Most shop floor processes are difficult to model because of the large number of influencing factors and the complex interactions between them. The GMDH is particularly well suited for the modeling and control of manufacturing processes and shop floor phenomena. In this paper the GMDH algorithm is described and its application in three problems, i.e., modeling of positioning error of a machine under varying thermal conditions, modeling of tool wear and tool wear rate, and modeling of cutting forces is presented. These initial applications of the method suggest that the GMDH algorithm has good potential for use in the abstraction and computer control of such processes.