CNets has sponsored a master thesis analyzing a new method to create a health indicator for machines. The master thesis was handed in and evaluated by KTHs School of Electrical Engineering and Computer Sciences (EECS).
The work is based on distance measurements transformed into a vector space through a feed-forward neural network. The neural network is trained using a multi-objective optimization algorithm to optimize criteria that are desired in a health indicator. The constructed health indicator is used as input to a gated recurrent unit (a neural network that handles sequential data) to predict the remaining life of a system in question.
The Master thesis Machinery Health Indicator Construction using Multi-objective Genetic Algorithm Optimization of a Feed-forward Neural Network based on Distance was carried out by Jacob Nyman in the field of machine learning:
