Extended friction energy model for milled surface wear prediction

Abstract

Milling, as a core process for manufacturing key components of aero-engines, forms a gradient metamorphic layer (residual stress and work hardening) on the workpiece surface, which significantly affects its fretting wear behavior. However, most existing research focuses on point contact conditions, and studies on wear under face-to-face contact conditions are still seriously lacking. In addition, constructing a prediction model for such a complex surface faces a double challenge: traditional machine learning methods are limited by small sample data and insufficient physical interpretability; finite element simulation requires repeated experimental calibration of tribological parameters. Therefore, this study, through systematic experiments, for the first time revealed the four-dimensional nonlinear coupling wear mechanism (normal pressure, displacement amplitude, cycles, frequency) of the milled surface under face-to-face contact conditions. Based on this, an extended weighted friction energy model was innovatively proposed. By introducing a bias term, it breaks through the limitations of the traditional model's assumption of homogeneous materials and significantly enhances its adaptability to complex working conditions. Further combined with ALE technology, a fretting wear simulation model that does not require parameter calibration was developed. Experimental verification shows that only 12 training samples are needed to achieve high-precision prediction of the wear rate of the milled surface, reducing the error by 27.6% compared with the traditional model; the wear depth prediction error of the developed simulation model is 7.98% under low cycle numbers. This method, by integrating the advantages of physics-driven and data-driven approaches, provides new theoretical support for the prediction of complex surface wear.