Using neural networks to identify annoying noises in vehicles

摘要

Previous papers have developed a methodology to characterise squeaks and rattles. Thus, for each noise, its origin and the means for eliminating it, are known. This paper describes the work done towards the development of a tool, based on neural networks, that determines if a squeak or rattle corresponds to any of the noises already characterised. Different types of neural networks have been evaluated. Preliminarily, it was found that for this application the best topology is a net 100-50-4. Additionally, it was found that the best training method is the gradient descent back-propagation method with a learning rate of 0.05.