Wear indicator of a padded part


Padded parts of a train set are worn over time and reach a significant wear level preventing them from providing their maximum performance rate.


Customer wished to automatically predict the padded part wear in order to alert the operator when a replacement is mandatory. There were no historical data. In order to validate the solution, there was a need to simulate the whole system to prove feasibility of building a robust predictive model, and to set up data collection parameters.


Amiral Technologies used its knowledge and expertise in industrial domains, mechanics and control theory, to build a digital twin simulating several possible real equipment behaviours including all sources of discrepancies, noise and uncertainties. The digital twin generated realistic operational data that were used to train the DiagFit predictive model and simulate its performance in predicting the wear condition.


We proved that system will generate usable data to build a robust model to predict padded parts wear and raise maintenance alerts.