MUMP aimed to produce the prototype of a planning tool for industrial applications. The tool - called MUMP, MUlti-Method Planner - incorporates artificial intelligence techniques and automatic form feature recognition. The design of MUMP and the techniques it uses were chosen on the basis of practical experience with different process planning problems and were refined and tested in a manufacturing application.
MUMP is based on three basic components: an assumption-based truth maintenance system, a frame system and a special control mechanism. A set of methods, ranging from rule interpretation to skeletal planning, can be used for solving different aspects of a problem. MUMP is composed of modules containing the implementation of the different methods and that can be chosen in such a way to produce an optimal configuration for the type of problem to be solved.
The project developed a prototype tool (called MUMP, multi-method planner) that incorporates artificial intelligence techniques and automatic form feature recognition. The design of MUMP and the techniques it uses were chosen on the basis of practical experience with different process planning problems and were refined and tested in a manufacturing application. MUMP is based on 3 basic components: an assumption-based truth maintenance system, a frame system and a special control mechanism. A set of methods, ranging from rule interpretation to skeletal planning, can be used for solving different aspects of a problem. MUMP is composed of modules containing the implementation of the different methods and that can be chosen in such a way to produce an optimal configuration for the type of problem to be solved. The development task has been concluded, producing all the modules foreseen for MUMP as well as the user and reference manuals. The system has been used for planning the machining of medium complexity aircraft parts and for several simpler problems. For the process planning application, a knowledge engineering approach was used to identify the form features and their attributes and to analyse and reproduce the task of a human expert. The form feature recogniser completed in the projectes can recognise features such as pockets (of any form), blind through holes, and protrusions and slots, all either isolated or combines (ie, nested). The planning kernel of process plans for prismatic parts with a quality comparable to that of a human expert.
The development task has been concluded, producing all the modules foreseen for MUMP as well as the user and reference manuals. The system has been used for planning the machining of medium-complexity aircraft parts and for several simpler problems.
For the process planning application, a knowledge engineering approach was used to identify the form features and their attributes and to analyse and reproduce the task of a human expert.
The form-feature recogniser completed in the project can recognise features such as pockets (of any form), blind and through holes, and protrusions and slots, all either isolated or combined (ie nested). The planning kernel of MUMP, using a special knowledge base, can produce complete process plans for prismatic parts with a quality comparable to that of a human expert.
Exploitation
The prototype of MUMP was completed in June 1990 and the prospects for exploitation are good.
The AI group of Aeritalia has taken over internal customisation following a successful trial based on the machining of aircraft parts.
Other partners have plans for exploitation of MUMP as an additional feature or selling point of their products and systems in industrial automation and in CAD.