Expert system for diagnosing computer numerically controlled machines: A case-study

Efficient diagnosing of a computer numerically controlled (CNC) machine is considered to be a very important component of modern manufacturing systems. It requires knowledge and expertise that are difficult or sometimes impossible to find in human experts. An expert system is proposed to overcome this problem. The diagnostic model used in this study is a hybrid reasoning approach between a deep model and a shallow model. The shallow model which represents the heuristic knowledge is provided in two different ways: a fault direct access heuristic and a heuristic for search boundary generation (for the deep model). The structure and behavior of the CNC system, or the deep knowledge, use the general deep reasoning strategy to locate the fault exactly. The hybrid reasoning approach provides an efficient and effective diagnostic model. The main objective of this paper is to propose the use of a logic interpretation of the ladder diagrams and the electrical diagrams of CNC machines for the deep model to define the fault location. Other technical knowledge of CNC machines is necessary to accomplish this process. The knowledge base of the prototype system developed in this study is production rule based; it consists of approximately 500 rules. A condition-based forward-chaining concept and a rule-ordering concept are used. These concepts reduce the knowledge base size. (C) 1997 Published by Elsevier Science B.V.