As a rapid developing additive manufacturing (AM) technology, selective laser melting (SLM) provides a promising way for intelligent manufacturing. The SLM part quality depends largely on the process parameters in the manufacturing process. Therefore, understanding the relationships between the input process parameters and the output part performances is critical to improve the part quality. In this work, the ensemble of metamodels (EM) is adopted and an adaptive hybrid leave-one-out error-based EM (EM-AHL) is developed to predict the powder utilization rate, the energy consumption, and the tensile strength of the as-built parts. First, the Taguchi experiment design is applied to obtain the sample points and the corresponding SLM experiments are conducted to get the experimental results. Second, the correlations between the process parameters (i.e., laser power, layer thickness, scanning speed) and the three responses are fitted using the proposed EM-AHL, which is constructed by aggregating three metamodels, Kriging, Radial basis fuction (RBF), and Support vector regression (SVR), according to the local measures. Finally, K-fold cross-validation and additional experiments validation methods are adopted to evaluate the prediction accuracy of the proposed EM-AHL. Results illustrate that the proposed EM-AHL not only outperforms the stand-alone metamodels but also provides more accurate results than the EM constructed by global measures (EM-G). Among the three prediction objectives, the prediction accuracy of the proposed EM-AHL has improved by up to 20% compared to the stand-alone metamodels. Besides, the main effects and contribution rates of process parameters on the responses are analyzed. Overall, the proposed EM-AHL method exhibits the excellent capability of guiding the actual SLM manufacturing.

作为快速发展的增材制造（AM）技术，选择性激光熔化（SLM）提供了一种有前途的智能制造方式。SLM零件质量很大程度上取决于制造过程中的过程参数。因此，了解输入过程参数与输出零件性能之间的关系对于提高零件质量至关重要。在这项工作中，采用了元模型（EM）的集合，并开发了一种自适应的基于混合遗留一出错误的EM（EM-AHL），以预测粉末的利用率，能耗和拉伸强度。竣工零件。首先，使用田口实验设计获得采样点，并进行相应的SLM实验以获得实验结果。第二，工艺参数之间的相关性（即 ，激光功率，层厚，扫描速度）和这三个响应都可以使用拟议的EM-AHL拟合，该模型是通过汇总三个元模型（克里格模型，径向基函数（RBF）和支持向量回归（SVR））而构建的当地措施。最后，采用K折交叉验证和其他实验验证方法来评估所提出的EM-AHL的预测准确性。结果表明，提出的EM-AHL不仅优于独立元模型，而且比通过全局度量（EM-G）构建的EM提供更准确的结果。在三个预测目标中，与独立元模型相比，所提出的EM-AHL的预测准确性提高了20％。此外，分析了工艺参数对响应的主要影响和贡献率。