Predictive Maintenance of an Electro-Injector through Machine Learning Algorithms

This work aims to define a system for measuring the "lift" of the anchor (the final part of the shutter) present inside the injector based on the use of Machine Learning classification algorithms. The measurement method determined is a non-invasive method, which guarantees that the internal organs of the injection system are not damaged to carry out the measurement and that it can be performed after welding the injector to prevent the "lift" from changing later. This measurement method provides for the classification of the currents circulating inside the solenoid, each of which can be associated with a specific value of the "injector lift. This approach is part of predictive maintenance techniques, a type of maintenance that tries to predict incorrect behavior of the system avoiding that critical operating conditions are reached. Finally, an analysis of the possible techniques for measuring the injector "lift" is carried out through the use of Machine Learning algorithms.