AI-Based Diagnostic Tool for Offline Evaluation of Measurement Data on Test Benches

Abstract

Test benches are becoming increasingly important in the development of modern vehicles. It does not matter whether the vehicle has a conventional, hybridized or fully electric drive. This trend is further strengthened by shorter development times, cost effectiveness and measures such as Road-to-Rig. In order to generate long running times and thus operate the test bench as effectively as possible, downtimes must be reduced to a minimum. In addition to the interruptions for setup and commissioning work, the downtimes primarily include the time for measuring data analysis in the event of an error. The procedure in the event of an error related shutdown is first of all to isolate the affected components and convert the required measurement data. These are then manually evaluated, categorized and logged by the test bench operator. The approach of the diagnostic tool developed here is the automated pre-evaluation of measurement data in the event of an error, before the test bench operator arrives at the testbench. This offers the possibility of efficient error analysis and support for the test bench operator. Thanks to the AI-based approach, the diagnostic tool learns independently and without intervention from the test bench operator, possible wear-related changes to the components over the course of time.