Erosion wear of biomimetic valve plug prediction with random forest regression based on Mahalanobis distance: an experimental study

Erosion wear is not entirely avoidance. Significant erosion occurs due to solid particle impacting, especially in applications involving sand production. The repair of hydraulic equipment in appropriate period can avoid catastrophic disasters. Therefore, the erosion prediction is an essential research for the hydraulic system. After billions of years of evolution, several biology exhibit unique erosion wear resistance performance. The biomimetic design can be a successful approach to overcome the limitations of typical approaches. Thus, we first designed a biomimetic valve plug inspired by biology to improve the erosion wear resistance performance. Then experiments have been conducted under different valve opening, biomimetic valve structure and the results were selected for analysis and prediction. Predicted erosion wear through various machine learning techniques were compared, and the random forest regression (RFR) model was chosen as a practical approach for erosion prediction. Finally, we used Mahalanobis distance (MD) as a structural feature to analyze erosion wear prediction. The results indicate that biomimetic structure with MD structure feature is more accurate. This innovative approach explores the positive role of biomimetic structures in the prediction of erosion wear.