Optimization of Running-in surface morphology parameters based on Auto-ML model

Running-in is an important and relatively complicated process. Surface morphology prior to run-ning-in affects the surface morphology following the running-in process, which in turn influences friction and wear characteristics of the workpiece. Therefore, it is important to establish a running-in model based on surface morphology to investigate the process and optimize the surface design. According to the existing research, the application of machine learning in running-in wear modeling is not sufficiently investigated. The black-box model has a robust complex object simulation per-formance. In this paper, five machine learning methods are applied for running-in modeling per-formance based on surface morphology parameters. It is observed that the support vector machine method has the best performance. Then, the support vector machine model is carried out and tested. Finally, the change law of surface morphology parameters is obtained and optimization direction of surface morphology design is proposed. When the surface is required to have good oil storage per-formance, the root mean square deviation (Sq), the root mean square slope (Sdq), and the height of surface core roughness (Sk) should be increased, and the height difference of surface morphology (Sdc) and the surface area ratio (Sdr) should be moderately high. When surface improved surface supporting performance is desired, larger values of parameters the height difference of surface morphology (Sdc), the root mean square slope (Sdq), and the surface area ratio (Sdr) should be used.

磨合（running-in）是一项重要且相对复杂的工艺过程。磨合前的表面形貌会影响磨合后的表面形貌，而后者又会对工件的摩擦磨损特性产生影响。因此，基于表面形貌建立磨合模型，以探究该工艺过程并优化表面设计，具备重要的研究价值。现有研究表明，机器学习在磨合磨损建模领域的应用尚未得到充分探索。黑箱模型（black-box model）对复杂对象具备优异的仿真性能。本文基于表面形貌参数，采用五种机器学习方法开展磨合建模性能研究，结果显示支持向量机（support vector machine）方法的建模性能最优。随后，对该支持向量机模型进行了构建与测试。最终，本研究得到了表面形貌参数的变化规律，并提出了表面形貌设计的优化方向：当要求表面具备良好的储油性能时，应增大均方根偏差（root mean square deviation, Sq）、均方根斜率（root mean square slope, Sdq）以及表面核心粗糙度高度（surface core roughness, Sk），同时适度提升表面形貌高度差（surface morphology height difference, Sdc）与表面面积比（surface area ratio, Sdr）；当需要提升表面的承载性能时，则应选取更大的表面形貌高度差（Sdc）、均方根斜率（Sdq）以及表面面积比（Sdr）。