ADIOSS Automatic Diagnostic Of System Simulations
This addresses the need to reduce costs and lead times in automotive engineering by automating fault detection in simulations, though it appears incremental as it builds on established techniques.
The paper tackles the problem of detecting faulty modules in automotive system-level simulation workflows after model updates, proposing a method that requires only a limited number of simulations and integrates easily into existing processes.
Automotive engineering makes extensive use of numerical simulation throughout the design process. The development of numerical models, their validation against experimental tests, and their updating during vehicle and engine projects constitute a core engineering activity. However, this activity must continuously evolve to reduce costs and lead times. In this context, we propose a method for detecting faulty modules within a system-level simulation workflow, represented as a graph of 0D models, following model updates. The proposed method requires a very limited number of system simulations and can therefore be easily integrated into existing engineering processes. It is designed as a toolbox based on well established and widely validated techniques, including Dynamic Mode Decomposition commonly used for 3D model reduction, linear programming, and autoencoders.