End-effector tools SoH prediction: parameter identification and reliability estimation.

Estimation of remaining useful life and machines State of Health (SoH) plays a vital role in performing predictive maintenance for complex systems today. Despite the numerous researches about this topic and the huge numbers of requests from companies, it still remains a challenge. On one hand, current techniques rely on machine learning techniques not taking into account the huge amount of computational power needed and the costs of such server architectures. On the other hand, it can be very challenging to develop an edge-computing solution considering strong computational power limits. To address this issue, in this thesis work it is shown a possible edge-computing intended approach, based on up to date parameter estimations techniques such as Kalman filtering, residual error processing and global sensitivity analysis. This whole thesis work is part of the MorePRO project, owned and supervised by a collaboration between Politecnico di Torino and brain Technologies Srl. The aims of this work are to employ a multi-model approach to estimate key parameters which can be associated to CNC machine wear, to develope a way to estimate reliability of such predictions and then use this reliability index to update the nonlinear predictive models for the forecast of system states. The approach adopted in this work is to firstly study the state of the art, comparing the methods and figuring out the advantages and disadvantages. Secondly, to develop with a practical overture the previously mentioned approaches. Finally, there were made many tests to empirically demonstrate the feasability of this approach and to provide solid proofs and conclusions.