Real-time prediction and classification of erosion crater characteristics in pulsating water jet machining of different materials with machine learning models

Erosion caused by water droplets is constantly in flux for practical and fundamental reasons. Due to the high accumulation of knowledge in this area, it is already possible to predict erosion development in practical scenarios. Therefore, the purpose of this study is to use machine learning models to predict the erosion action caused by the multiple impacts of water droplets on ductile materials. The droplets were generated by using an ultrasonically excited pulsating water jet at pressures of 20 and 30 MPa for individual erosion time intervals from 1 to 20 s. The study was performed on two materials, i.e. AW-6060 aluminium alloy and AISI 304 stainless steel, to understand the role of different materials in droplet erosion. Erosion depth, width and volume removal were considered as responses with which to characterise the erosion evolution. The actual experimental response data were measured using a non-contact optical method, which was then used to train the prediction models. A high prediction accuracy between the predicted and observed data was obtained. With this approach, the erosion resistance of the material can be predicted, and, furthermore, the prediction of the progress from the incubation erosion stage to the terminal erosion stage can also be obtained.