What is unsupervised learning? Unsupervised Learning (see wikipedia) is used when no historical data is available that involves past faulty behaviour and/or ageing progression. Therefore,
Car parts wear
The analysed device was made of a moving part with a rubber element cleaning a fixed part. The rubber element supports the whole friction and is subject to wear after each move, the quality of cleaning being altered as wear level increases.
Customer wished to automatically measure the wear of the rubber element in order to display its wear level to the driver and help planning its replacement.
DiagFit was used to build a wear indicator based on acceleration data captured from the moving part and per type of rubber. Wear indicator was obtained using data from devices with a new rubber piece, learning the normal behaviour and then detecting devices with wear conditions. Wear indicator can be further refined using historical data for both new and old rubber pieces.
Using available data including acceleration data for different parts and rubber types, Diagfit was able to classify worn parts with a 99.3% true positive rate and 2.8% of false positive rate. 50% of data was used to train the model.