Flexible planning logic for technology planning

Abstract

The ever-rising expectations of customers regarding quality and diversity of products require new products to be launched with a minimum of upheaval and consequently rapid, defect-free and economical processes. These goals can be attained and guaranteed in the planning departments of businesses through the sharing of secure and up-to-date planning data, leading to a high degree of flexibility and a minimum of interruptions in the production processes. In terms of technology planning, this results in a demand for economic methods of planning and NC programming, leading in turn to continuous, systematic acquisition of specialised in-house technological knowledge. However, most of the applications on the market are unsatisfactory in terms of acquisition of specialised knowledge of technology, only allowing specialised planning knowledge to be gathered or entered through costly procedures and for it to be organised statically. For rational processing and evaluation of planning knowledge, it is therefore sensible to systematically record technological discoveries made in the course of the planning and running-in process and to make them available for future planning processes. This processing of empirical knowledge along with rational and safe technology planning is supported by the use of DP systems, thus enabling planning knowledge to be kept constantly up-to-date, growing along with the data specific to the enterprise. For this purpose, a suitable information model must be devised that represents the integration point within the NC process chain. Using a feature-based product model introduced in the design departments as a starting point, the model is extended to take in a production-orientated view. The model should be designed in such a way that, above all, the technological interconnections are represented and that the technology planning, NC programming and running-in process are integrated through the model. In order to evaluate and process empirical knowledge, methods are applied from machine learning from Bayesian belief networks. This enables a costly human acquisition of the empirical knowledge of technology to be avoided and a rapid and rational decision-support system tailored to the particular problem to be developed. Also, by using suitable Bayesian belief networks, planning knowledge can be condensed, the numerous different planning conditions can be automatically diverted and the decision-base kept extremely up-to-date.