A Learning Method for Automated Disassembly

Abstract

While joining tolerances, and therefore forces, are known in the assembly process, the determination of disassembly forces is not possible. This is caused by changes in the product properties during the product operation, which has multiple reasons such as thermal or mechanical stress on the product. Regarding the planning of disassembly tasks, disassembly times and tools cannot be planned properly. They have to be determined in the process or stay undefined, which can result in damaging of the product.

This article shows an approach to describe the necessary disassembly forces without having to investigate the complex physical influences caused by the usage of the product. To do so, a Learning Method is developed, which is sustained by a Lookup-Table for the estimation of disassembly forces based on basic input data such as hours of operation and operating characteristics. Missing values will be interpolated by using multiple linear regression. The concept will be illustrated in the example of a turbine blade connection.