【摘要】 砂型铸造复杂铸件会出现冷隔、气孔、砂眼、缩孔等缺陷，需要对铸件质量进行预测。其中，影响铸件质量的因素包括工艺参数和产品结构，而铸件三维结构复杂，存在表征困难、形貌与铸件成形质量间关系无法确定的问题。针对上述目标和问题，课题组开展了多源异构数据驱动的砂型铸造复杂铸件质量预测研究。首先，采集了混砂造型、熔炼浇注等过程的结构化工艺参数数据和产品三维结构；其次，构建了深度卷积自编码器，实现了对复杂铸件三维结构的特征提取，获得了特征数据；最后，构建了工艺、三维异构数据驱动的砂型铸造复杂铸件质量预测模型，对铸件缺陷进行了准确预测。更多还原

【Abstract】 Sand casting complex castings will appear cold insulation, porosity, sand holes, shrinkage holes and other defects, it is necessary to predict the quality of castings. Among them, the factors affecting casting quality include process parameter data and product three-dimensional structure, but the three-dimensional structure of castings is complicated, there are difficulties in characterization, and the relationship between morphology and casting forming quality can not be determined. Aiming at the above goals and problems, this paper carried out a research on quality prediction of complex sand casting driven by heterogeneous data from multiple sources. Firstly, the structured process parameter data and product 3D structure of sand mixing molding, melting and casting were collected. Secondly, the deep convolutional autoencoder is constructed to realize the feature extraction of complex casting 3D structure and obtain the feature data. Finally, a quality prediction model of complex sand casting driven by process and three-dimensional heterogeneous data was built to accurately predict the casting defects.更多还原