基于XGBoost算法的焊接接头固有变形数据库构建

Establishment of Inherent Deformation Database of Welded Joints Based on XGBoost Algorithm

基于固有变形的大型结构焊接变形预测中,焊接接头的固有变形数据库必不可少。针对焊接接头的特点,研究了基于XGBoost算法的固有变形数据库的构建方法:在梳理焊接接头固有变形影响因素的基础上设计了固有变形数据库的结构,采取网格搜索方法自动确定模型的最优超参数,建立了固有变形预测模型。并通过简单实例验证了提出的数据库构建和完善方法在工程实际中应用的可行性。研究成果对于挖掘生产现场的焊接变形大数据,实现焊接结构变形精准预测有重要的意义。

In the welding deformation prediction of large-scale structures based on inherent deformation, the inherent deformation database of welded joints is indispensable. According to the characteristics of welded joints, the establishment method of the inherent deformation database based on XGBoost algorithm was studied. The structure of the inherent deformation database was designed on the basis of the inherent deformation factors of welded joints;the grid search method was adopted to determine the optimal hyper parameters and the inherent deformation prediction model was built. A simple example was used to verify the application of the proposed database establishment and improvement method in engineering practice. The research results are of great significance to mining large data of welding deformation at the production site and precisely predicting the deformation of welded structure.