Journal of Intelligent Manufacturing (2000) 11, 559±572 A three-level knowledge-based system for the generation of live and safe Petri nets for manufacturing systems L. CASTILLO, J. FDEZ-OLIVARES and A. GONZA Â LEZ Departamento de Ciencias de la Computacio Ân e Inteligencia Arti®cial E.T.S. IngenierõÂa Informa Âtica, Universidad de Granada 18071 Granada, Spain E-mail: L.Castillo@decsai.ugr.es, Faro@decsai.ugr.es, A.Gonzalez@decsai.ugr.es New generation manufacturing systems are involved in a transformation process which aims for more reliable production processes and with a lower response time to the demand of the market. This work presents an application of arti®cial intelligence planning techniques for the automatic generation of the control program for a manufacturing system expressed as a safe and live Petri net. The advantage of the system presented here is straightforward: it allows for a fast generation of sound results free of human errors, reducing the cost and duration of the development phase of control programs. Keywords: Intelligent manufacturing systems, arti®cial intelligence planning, sequential control programs, Petri nets 1. Introduction New generation manufacturing systems are involved in a deep transformation process with the aim of obtaining more reliable manufacturing processes and with a greater response capability to the demands of a market in continuous change. The achievement of this goal requires the development and use of new technologies, specially from the ®eld of arti®cial intelligence. At present, there are several initiatives that follow this technological development such as international research and development programs (IMS, 1995) or special interest groups (AAAI, 1996; PLANET, 1998). Amongst all of the problems that are faced by these initiatives, one of the most challenging ones consists of automating the design process of the control program that drives the operation of a manufacturing system by using arti®cial intelligence planning techniques (Allen et al., 1990; Nilsson, 1980; Russell and Norvig, 1995; Aylett et al., 1998; Castillo et al., 1998; Klein, 1994). The reasons for this interest are that this is a slow and expensive process, always subject to possible human errors and hardly adaptable to the life-cycle of manufacturing systems, having direct and negative effects on aforementioned reliability and speed goals. The use of these arti®cial intelligence techniques in the automation of the design of control programs would provide solutions to these negative effects because they can obtain sound results and very fast (thanks to the growing speed of modern computers) and they are easily adaptable to new environments. In this sense, this work presents a three-level system, based on arti®cial intelligence planning techniques which, starting from: * a high level description of the manufacturing system layout, * a description of raw products, and 0956-5515 # 2000 Kluwer Academic Publishers