It is shown how an AI-assisted workflow accelerates the development and verification of vehicles’ noise, vibration, and harshness (NVH) behavior. Typically, many design parameters must be controlled in CAE-software models to accomplish multiple goals. By using the AI workflow, the NVH characteristics of components are analyzed and optimized to the predefined target value. In the digital phase, this enables a target-oriented and faster development of vehicle systems under NVH aspects. We demonstrate that AI-metamodels can be trained to accurately predict NVH-simulation results for new design parameters, when the training data is provided by several reference simulations. This is the key of the massive acceleration given by the NVH-AI-workflow, which is illustrated by two case studies. First, a regression problem consisting of 27 parameters (stiffness and inertia) is defined for a multi-body simulation (MBS) crankshaft model. The crankshaft vibration angle is minimized by exploiting the trained AI metamodel in an optimization loop, such that the NVH target is ensured. Second, the demonstrator “CUBIC AI” is presented, in which the user can literally draw the desired NVH-behavior, and then, without iterative process, the best design parameters are identified. This approach is illustrated using a simplified model of a vehicles’ vertical dynamical behavior.
Session:
Poster
|