Dynamic conformity assessment for joining force monitoring using Bayes filters

Monitoring force-displacement or force-time curves is a widely used quality control technique in the field of mechanical joining. For online monitoring of self-piercing riveting, envelope curves are often used to define a tolerance zone for the measured setting force. However, the measurement uncertainty is typically not considered and the force curve of a joint can be wrongly rated as non-conform due to measurement errors and noise. In this article, we present a method for dynamical online filtering and uncertainty determination for noisy force curves using two types of Bayesian filters. The methodology is based on a Bayesian probability framework using a priori information for the process curve and sensor noise. To investigate the general feasibility of the method, force measurements with different noise levels are simulated and processed. The conformity is further assessed taking the uncertainty of the filtered signal into account. The results show that the Bayes filter technique is principally able to reduce noise for well-known characteristics of the process curve and sensor noise. Advantages over common filtering techniques, especially for experimental conditions with less known characteristics, are still to be verified. The methodology could be used in future for closed-loop controls to adapt process parameters dynamically. </jats:p>