Process curve analysis with machine learning on the example of screw fastening and press-in processes

Abstract

With the increasing degree of digitalization in manufacturing industry, advanced data analysis techniques such as machine learning (ML) are moving into focus. In several production processes like machining or welding, ML applications already show high potential for process monitoring, optimization, or control. Although screw fastening and press-in processes are frequently employed in modern assembly lines, there are only few research approaches addressing the application of ML so far. Therefore, this paper starts with an overview of conventional and first ML-based approaches to process curve monitoring. Following this, it is shown how ML can be used to detect and classify errors in screw fastening by analyzing the resulting torque curve. In addition, an assembly line consisting of several press-in processes is used to demonstrate that ML also allows the analysis of cross-process interactions by evaluating several process curves at once. In this way, defective parts can be detected and rejected at an early stage, eliminating further processing and testing steps while simultaneously reducing costs.