Quality prediction of disturbed ultrasonic metal welds

Abstract

Ultrasonic Metal Welding is an industrially established welding process. Applications such as wiring harness assemblies, battery cell production and joints in power electronics show its relevance for current production and future technologies, but also allow zero failures. Still the mechanisms and influences involved in the joint formation process are not fully understood yet, and non-destructive testing methods to validate the joint quality are not available. Strict quality control of unprocessed material and destructive testing of welded sample parts are established methods to ensure statistical validation of sufficient joint quality.This study evaluates the influences of selected, pre-defined disturbances on the welding process of a statistically relevant number of welds. These disturbances include changes of material hardness, thickness, surface quality, rolling direction and welding position. The joint quality of these disturbed welds as well as a reference data set was determined by destructive shear testing. During welding, measurements of process variables like horn and anvil vibrations and the penetration depth of the horn were taken at high frequencies. Internal, statistical data provided by the welding machine is labelled with its shear strength. Subsequently, process characteristics are derived from this dataset, allowing for the estimation of the achieved weld quality. In this paper, the underlying robust estimation method to predict the weld quality based on these measurements will be presented and compared to current limit values and linear regression used in the industry to monitor the welding process.