Forecasting the Fault Detection & Condition Monitoring of Rotating Machinery by SHAP: Ex-Plain Able AI

It is well known that machine learning (ML) can help with machine diagnosis, but it is still not clear how to understand the results of the analysis. Since machine learning models are like “black boxes,” it's not clear to the user how each of the chosen traits helps with the evaluation. This work aims to find out what a Shapley additive Explanation does to find the most important faults and classify them in condition- tracking tools for spinning machines. The writers look at the case of industrially important medium-sized bearings. The Mechanical Sciences Lab at Politecnico di Torino has a test rig for industrial bearings that was used to collect shaking data for different health states. Because of the SHAP, we can understand the Support Vector Machine, or SVM, along with the k-nearest Neighbour, which diagnostics models. Using the SHAP to be a way to choose features leads to accuracy levels better than 98.60% for both models. What makes the most difference in the models' results is the skewness and shape aspect of the sound sign.