摘 要 随着厚板在结构件中应用的越来越多,窄间隙焊接具有比较广泛的应用。由于焊缝截面形状参数能直接影响接头的性能,同时窄间隙焊接坡口形状特殊,因此需要建立一种能够快速、简单预测窄间隙焊缝截面形状参数的方法,通过研究焊缝截面形状参数与窄间隙焊接工艺参数之间的关系,提前预知焊缝截面形状,从而为优化焊接工艺参数、节约焊接成本以及减少不必要的浪费提供一种技术手段。 基于以上情况,本课题采用 BP 神经网络算法,以 MATLAB为开发平台,建立了一个窄间隙焊缝截面形状参数的预测模型。 通过MATLAB内部的 GUI设计模块,设计了人机交互的按钮式操作界面。所设计的预测软件主要包括“学习训练”、“预测应用”两大功能块,其中“学习训练”功能块能够建立焊缝截面形状参数与窄间隙焊接工艺参数之间的内在关系, “预测应用”功能块是在“学习训练”这个功能块的基础上,根据之前学习的焊缝截面形状参数与窄间隙焊接工艺参数之间的内在关系对新的工艺参数进行预测。 试验结果表明:针对窄间隙摇动电弧焊接工艺建立的预测模型,实现了对预测结果的可视化; 当摇动频率在0.5~3.5Hz、摇动幅值75~115° 、侧壁停留时间40~100ms时,所预测的焊缝表面弯曲度和侧壁平均熔深误差范围在 10%以内。 58972 毕业论文关键词:窄间隙焊接;摇动电弧;焊缝成形;BP 神经网络;
MATLAB Abstract With more and more application of thick plate used in structure, narrow gap welding is widely used. The weld shape parameters affect the quality of the joint directly. Meanwhile, the narrow gap welding groove shape is special. Therefore, it is necessary to establish a kind of software which can rapid and simple predict the narrow gap weld shape. By studying the relationship between process parameters and the narrow gap weld shape welding, it can predict of the weld shape in advance. Then, it provides a new method to optimize the welding parameters, save welding cost and reduce unnecessary waste to. Based on the above situation, this project uses the BP neural network algorithm, by using MATLAB as the development platform, and establishes a model for the prediction of the shape parameters of the narrow gap weld section. Through the GUI internal, it designs of the man-machine interactive button operation interface. The prediction software mainly consists of "learning and training", "blocks of prediction application" two big functions, in which "learning training function block” can be used to establish the weld cross-section shape parameters and narrow gap welding the intrinsic relationship between process parameters. The "prediction application function” block is based on the "learning habits the function block”. According to the previous study of the relationship between the weld shape and narrow gap welding parameters, it can predict on the new process parameters. Experimental results show that the prediction model for narrow gap arc welding process, the realization of the visualization of the prediction results. When the shaking frequency in the 0.5~3.5Hz, shaking amplitude 75~115 degrees, the side wall of the residence time 40~100ms, error range of the predicted weld surface curvature and average penetration depth of side wall is within 10%.
Keywords: Narrow gap welding; Flux cored wire; Weld shape; BP algorithm; MATLAB