Research Experiences for Undergraduates in Automotive Technologies

Proposed Project:
Independent Component Analysis (ICA)

Independent component analysis (ICA) is a new, non-linear adaptive signal processing technique used to unmix vibration and acoustical measurements into independent source channels. In order to perform in-situ health monitoring of gearboxes and bearing assemblies, uncorrupted signatures of these machinery components are required. To obtain these uncorrupted measurements, costly modal analysis techniques are currently used to identify optimal accelerometer placement. ICA, by contrast, produces these unmixed signatures from mixed measurements based on the principal of information maximization. This project will gather data from a roller bearing machine, where several machinery components signatures are simultaneously measured. The measurements contain several machinery component signatures that are mixed in an unknown way. This data will then serve as input to the ICA algorithms to test the separability of the mixed machinery components: including faulty bearing assemblies. This would eliminate the costly modal analysis that accompanies most health monitoring implementations for sensor placement.

Back to PROJECT DESCRIPTIONS page...