📖Program Curriculum
Project details
This project aims to develop an automatic acoustic fault diagnosis system that can mimic the industrial engineers’ “listening and diagnostic” capability.
As demonstrated from literature, human auditory system possesses exceptional capability to extract and simplify characteristic information of audio signals. To model this functionality, the proposed “machine hearing” methodology will include two main stages. Firstly, the information extraction method of human ear, especially the cochlea, based on nonlinear processing mechanism will be analysed, to achieve robust extraction of the acoustic features. Then, in order to simulate human auditory central nervous system to effectively process the time-frequency information output from the “artificial cochlea”, a novel deep neural network model will be developed based on the real auditory signal processing procedure, as well by including memory elements in the network.
The developed methodology will be validated using acoustic data collected from microphones and data from acoustic emission sensors for damage detection of composite structures (e.g., representing aeronautical or automotive components), fault diagnosis of electric motor, fuel cells and/or batteries.
The outcome of this research can be widely applicable to various industrial sectors, e.g., for transportation, energy, manufacturing, and healthcare applications.
Show less
