Overview
I developed the algorithm and graphical interface for Automotive Aeroacoustic Noise Prediction Software (WNPS) at SAIC Volkswagen. The software predicts the wind noise propagation paths of the holes & gaps and provides guidance for the development of wind noise performance on the early stage.
Physics-informed Neural Network (PINN) and data science techniques are adopted to analyze experimental data and predict.
Prediction task
The software reads vehicle information and experimental configurations and predicts an Articulation Index, a measure of speech intelligibility in noise. Engineers can compare the output across test states to study how changes to sealing strips affect wind-noise performance.
The interface brings vehicle data, model inputs, and predicted results into one workflow. Its comparison plots make it easier to inspect the effect of different experimental configurations.
The prediction interface, with annotated input and output panels.
My contribution
I was the primary developer for the prediction algorithm, deployment, and GUI, and managed a team of three. The implementation used Python, PyTorch, and PyQt.