(上海航天设备制造总厂有限公司，上海 200245)
摘 要: 激光熔覆修复过程中单道熔覆层形貌极大地影响修复效果，但多工艺参数对熔覆层影响的耦合作用机制尚未被研究清楚，因此，获得不同工艺参数组合与熔覆层尺寸的定量关系是亟待解决的难题。以Inconel 625合金的激光熔覆修复为背景，采用随机森林(Random Forest，RF)算法构建了激光熔覆工艺参数(激光功率、扫描速度、送粉速率)到单道熔覆层尺寸的回归模型，将模型用于特定熔覆参数组下单道尺寸的预测；同时在给定期望的单道熔覆层尺寸参数时，基于Gini不纯度选择强关联因子构建了工艺参数预测模型。结果表明，激光熔覆工艺参数预测模型的预测误差小于4%，能够准确地估计加工特定单道熔覆层截面几何形状所需的激光熔覆工艺参数。
关键字: 随机森林；激光熔覆；特征筛选；参数预测
（Shanghai Aerospace Equipments Manufacturer Co., Ltd., Shanghai 200245, China）
Abstract:The shape of the single cladding layer in the laser cladding repair process greatly affects the quality of the repair. It is necessary to control the morphology of the cladding layer to achieve high-quality repair. However, the coupling mechanism of multi-process parameters on the cladding layer was studied clearly. Therefore, obtaining the quantitative relationship between different process parameter combinations and the size of the cladding layer is an urgent problem to be solved. Based on the laser cladding repair of Inconel 625 alloy, a random forest (RF) algorithm was used to construct a regression model of laser cladding process parameter set (laser power, scanning speed, powder feeding rate) to single pass cladding size. The model was used to predict the single track size of a specific cladding parameter group. At the same time, the strong correlation factors were selected based on the Gini impurity and used to build a process parameter prediction model. The results show that the prediction error of the laser cladding process parameter prediction model is less than 4%, which can accurately estimate the laser cladding process parameters required to process the specific single-pass cladding cross-section geometry.
Key words: random forest; laser cladding; feature selection; parameter prediction
