Abstracts 407

198 Process Planning Knowledge Acquisition X. H. Shan, A. Y. C. Nee, A. N. Pon, pp 157-161

In this paper, a method for acquiring manufacturing planning knowledge is presented. Emphasis is placed on eliciting, analyzing and interpreting the knowledge of cutting parameter selection which a human expert uses when solving a particular problem and then transforming this knowledge into a suitable machine representation. The approach is based on repertory grid analysis and fuzzy logic operations. Production rules are automatically generated upon the successful acquisition of the relevant knowledge.

missing until now. Some problems in developing and using a decision support system, for example, designing the knowledge base and especially the acquisition of the necessary data and facts, are discussed for a particular example. A simple decision support system - called BERASYS - has been developed, implemented and tested for minimising the tool changing and adjustment time of a continuous toning line producing hollow steel sections. First experiences show a good acceptance from the operators and a remarkable reduction of the adjustment t u n e .

199 Scheduling Products with Bills of Materials C. S. Czerwinskl, P. B. Luh, pp 162- 166

A product's bill of materials specifies the hierarchical order in which parts and raw materials are to be assembled. Attempts to decompose the complex product scheduling problem into independent subproblems often ignore the bill of material constraints, and yield unnecessarily tardy schedules. This paper presents an integer programming formulation of the product scheduling problem, and a combined Lagrangian relaxation and heuristic approach. Tests using real data show that this method can obtain near-optimal schedules efficiently with quantifiable performance. Preliminary results indicate reductions in solution oscillation and work-in-process inventory. Comparisons show that integration of the bills of materials into the part scheduling can improve scheduling performance by reducing product tardiness.

203 Intelligent Systems for the Design and Manufacture of Carbon-Carbon Composites D.A. Eitman, R.B. Dirlin8 Jr., pp 182-186

Carbon-carbon has numerous advantages over any other material as an engineering material for high temperature use. It can be tailored through appropriate selection of constituents and the fibre architecture, but there is no design guide that defines these options. Also, the large number of fabrication processes provide ample oppommity for processing problems. To solve these problems, two sets of intelligent systems are being developed. A design tool employs conventional composite property-prediction methods and a set of artificial-intelligence-based directions related to fabrication process and constituent selection. This set of rules is also employed during manufacture, where sensors evaluate in-process material status, and influence the controlling parameters which affect the end- product's properties.

200 Supporting Production Management for Manufacturing S B. M. Joberto, F. W. Cruz, pp 167-171

Production and operations management (POM) is an important activity in all organizations which can make use now of protocol and networking technologies for providing more-efficient tools. This paper describes the implementation of a set of primitives suitable for supporting production management (PM) applications. The PM service is based on the MMS (Manufacturing Message Protocol), modeled differently from the MMS companion standards, and oriented for manufacturing.

201 Knowledge Based Simulation System for Manufacturing Planning J. Rlives, K. Kannel, pp 172-176

In this paper, the main principles of developing a general- purpose simulation system having a two-level (planning and operating levels) architecture is considered. To take care of the influence of production factors (system structure, production structure, equipment capacities) in the planning and operating levels, in order to achieve more flexibility and productivity of the FMS, an expert system application in both levels is very reasonable. The efficiency of the simulator depends on the skill of problem characterization, the moders construction, rule identification and result interpretation. Simulation systems having such abilities can be used to support the decisions that must be made during manufacturing planning.

202 Decision Support Systems and their Industrial Applications P. Kopacek, G. Halmschlager, V. Ambrojewitsch, pp 177-181

Decision support systems have been well known for many years, but concrete industrial applications are more or less

204 Closed-Loop Sheet Metal Forming Processes D.E. Hardt pp 187-192

Many processes could be greatly enhanced by using real- time process output control. This paper presents a framework for such process conlrol problems, then explores an application in three-dimensional sheet metal forming. The control system involves a "deformation transfer function" (DTF), defined as the ratio of the spatial frequency content of the formed part to that of the die used, calculated after each forming cycle and used in a dead-beat control algorithm to adjust the tooling. A range of part shapes have been successfully formed, requiring only two forming cycles. The signal processing necessary for system stability is presented, along with an analysis of the current forming limits of this system.

205 Hot Isnstatic Pressing: An Intelligent Processing of Materials Demonstration B.L. Henniges, pp 193-198

Intelligent processing of materials (IPM) improves the efficiency and quality of processing advanced materials such as 7-TiAI via powder metallurgy. The IPM concept, as applied to hot isostatic pressing (HIP), exploits recent advances in off-line finite element modeling (FEM), on- line prediction of material properties, process optimization, in-siva eddy current sensing, and advanced process conwol. This paper presents recent advances in HIP technologies and proposes a generic hierarchical control structure which adapts process cycles based on on-line model predictions and in-situ sensor measurements. A simple feedfurward- lag compensation satisfies a density tracking objective while maintaining mbusmess to model uncertainty and nonlinearities. Simulation results reveal substantial improvement over conventional processing methods.