Automated Maintenance Plan Generation Based On CAD Model Feature Recognition

Maintenance takes care of the health of equipment and/or mechanical system. Currently, it is planned and executed by humans, since disassembly/assembly procedures and its level of difficulty, makes it very difficult to automate the process. The first step to move towards automated maintenance is to create a system that is able to provide a maintenance procedure automatically by recognizing 3D CAD models. The proposed framework identifies products at the abstract level. This is done by computer-developed algorithm. The method recognizes the selected product from the CAD model (using SolidWorks) by extracting geometrical information and then associating it to a knowledge base to identify the product by looking for User-Defined Features (UDF). This method also supports the process to acquire and store new geometrical information about products that are new to the knowledge base system. The system will link the product definition, maintenance procedures, and disassembly rules and it will build a case-base for the identified model. If knowledge case acquired exists then a previously built maintenance plan can be reviewed and used. Otherwise, a new case will be built based on the implemented rules in the knowledge base system. Moreover, available maintenance plans can be used for similar products. Also, this plans can be modified to accommodate different types of products to use the same or a modified maintenance plan. Finally, a special algorithm has been developed to recognize products created in other CAD systems. The fact that a product can be identified automatically and be linked to a knowledge base system without human intervention makes it an important progress towards automated maintenance. Further research is underway to provide maintenance procedures at the component level, where an optimal sequence for disassembly and maintenance plans will be generated for different levels of the product structure hierarchy. (C) 2018 The Authors. Published by Elsevier B.V.