The aim of this work is to propose a methodology for rapidly predicting suitable process parameters for additive manufacturing of a 316L-Cu multi-material part with a compositional gradient by using machine learning. Specifically, an algorithm based on a multivariate Gaussian process is developed to predict part density and surface roughness for a given set of laser power, velocity, and hatch spacing values. The training data for the algorithm is collected using a high-throughput experimentation method that allows for rapid measurement of part density, and surface roughness. After the model is validated using leave-one-out cross validation method, process parameter maps are generated for 316L-Cu parts manufactured using selective laser melting with premixed powder at mass fractions of 0.25, 0.50, and 0.75. A set of suitable process parameters are predicted using the process maps. It is shown that process parameters are a nonlinear function of gradient composition and neither process parameters of 316L or Cu are suitable for the graded region of a 316L-Cu multi-material part. Generated process maps provide a firsthand knowledge of process-property relationships for regions of compositional grading in 316L-Cu parts.

这项工作的目的是提出一种方法，用于通过使用机器学习技术快速预测适合的工艺参数，以进行具有成分梯度的316L-Cu多材料零件的增材制造。具体而言，开发了一种基于多元高斯过程的算法，以针对给定的一组激光功率，速度和剖面线间距值预测零件密度和表面粗糙度。使用高通量实验方法收集算法的训练数据，该方法可以快速测量零件密度和表面粗糙度。使用留一法交叉验证方法对模型进行验证后，将生成质量参数为0.25、0.50和0.75的预混合粉末的选择性激光熔融生产的316L-Cu零件的工艺参数图。使用过程图可预测一组合适的过程参数。结果表明，工艺参数是梯度成分的非线性函数，而316L或Cu的工艺参数均不适用于316L-Cu多材料零件的渐变区域。生成的过程图为316L-Cu零件的成分分级区域提供了过程属性关系的第一手知识。