Automated failure classification for assembly with self-tapping threaded fastenings using artificial neural networks

Abstract

This paper presents a new strategy for the automated monitoring and classification of self-tapping threaded fastenings, based on artificial neural networks. Threaded fastenings represent one of the most common assembly methods making the automation of this task highly desirable. It has been shown that the torque versus insertion depth signature signals measured on-line can be used for monitoring threaded insertions. However, the research to date provides only a binary successful/unsuccessful type of classification. In practice when a fault occurs it is useful to know the causes leading to it. Extending earlier work by the authors, a radial basis neural network is used to classify insertion signals, differentiating successful insertions from failed insertions and categorizing different types of insertion failures. The neural network is first tested using a computer simulation study based on a mathematical model of the process. The network is then validated using experimental torque signature signals obtained from an electric screwdriver equipped with an optical shaft encoder and a rotary torque sensor. Test results are presented proving that this novel approach allows failure detection and classification in a reliable and robust way. The key advantages of the proposed method, when compared to existing methods, are improved and automated set-up procedures and its generalization capabilities in the presence of noise and component discrepancies due to tolerances.