基于神经网络的高压GMAW焊缝成形预测

发布时间：2018-05-02 02:40

本文选题：高压干法GMAW + 焊缝成形　；参考：《北京石油化工学院》2015年硕士论文

【摘要】：海洋石油资源的开发已经成为世界范围内的新的研究热点之一,海洋中出现越来越多的石油平台钢结构、海底管道、水下结构。由于长期浸蚀在海水中,水下焊接修复十分困难,所以研究适用于深海水下焊接技术十分关键。高压干法GMAW是具有现实意义的水下焊接方法,已成为海洋油气田开采和储运常用的水下焊接技术之一,得到广泛应用。焊缝成形是反映焊接质量的一个重要指标,通过分析高压下的焊缝成形有助于研究高压GMAW的规律,对焊缝成形进行预测能够指导高压环境下焊接工艺,有效节约高压焊接试验成本。鉴于此原因本文建立了基于BP神经网络的焊缝成形预测模型,该模型可以实现焊缝信息预测,由此优化焊接工艺方案,提高焊接质量。本文的研究内容如下：(1)首先,利用正交设计法设计了实验方案,该方案以焊接工艺参数为影响因素,每个因素取四个水平,焊缝成形尺寸为输出变量。在搭建好的高压焊接试验系统中进行高压焊接试验,采集焊缝样本数据,作为神经网络的学习训练样本和预测样本。(2)通过分析焊缝成形数据,得到高压环境下GMAW焊缝成形的规律,如高压下,焊缝成形不好,飞溅较大;环境压力的增加会引起焊缝熔宽减小,熔深增加,余高增高。(3)利用回归分析的方法建立了焊缝成形尺寸与焊接工艺参数之间的回归方程,得到了经验公式。对回归方程进行R2检验、F检验和置信区间估计,回归方程具有很高的显著性,能够用于焊缝成形预测。(4)建立了基于BP神经网络的高压GMAW焊缝成形预测模型,该模型以压力、电流、CO2比例和干伸长为输入,以熔宽、熔深和余高为输出。在VC++环境下,开发了可视化的神经网络预测系统。将该预测系统的预测结果与实际值进行比较,得到该预测系统相对误差较小,可靠度较高的结论。最后利用试验数据对建立的回归方程和BP神经网络模型的预测能力进行了对比。结果表明,BP网络的的预测能力优于回归方程,所以建立的神经网络预测模型更适用于高压GMAW焊缝成形预测。
[Abstract]:The development of marine petroleum resources has become one of the new research hotspots in the world. There are more and more steel structures, submarine pipelines and underwater structures in the ocean. It is very difficult to repair the underwater welding because of the long-term erosion in the sea water. So it is very important to study the underwater welding technology suitable for the deep sea. The high pressure dry method GMAW is the key. The underwater welding method, which has practical significance, has become one of the underwater welding techniques commonly used in the mining and storage of offshore oil and gas fields, and it has been widely used. The weld forming is an important index to reflect the quality of the welding. The analysis of weld formation under high pressure helps to study the law of high pressure GMAW, and the prediction of weld formation can be guided. Welding technology under high pressure environment can effectively save the cost of high pressure welding test. For this reason, a prediction model of weld formation based on BP neural network is established in this paper. This model can predict weld information, thus optimize the welding process and improve the quality of welding. The contents of this paper are as follows: (1) first, the orthogonal design method is used. Considering the experimental scheme, the scheme takes the welding process parameters as the influencing factor, each factor takes four levels, the weld forming size is the output variable. The high pressure welding test is carried out in the built high pressure welding test system, the sample data of the weld is collected as the learning training sample and the prediction sample of the neural network. (2) through the analysis of the weld formation. The shape data of GMAW weld formation under high pressure, such as under high pressure, the weld formation is not good, the spatter is great, the increase of environmental pressure will cause the weld weld width to decrease, the weld depth increase and the excess height increase. (3) the regression equation is established between the weld forming size and the welding process parameters by the regression analysis method, and the empirical formula is obtained. The regression equation has R2 test, F test and confidence interval estimation. The regression equation has high significance and can be used for weld forming prediction. (4) a prediction model of high pressure GMAW weld forming based on BP neural network is established. The model is input with pressure, current, CO2 ratio and dry elongation as input, and the weld width, depth and residual height are output. In VC++ ring Under the circumstances, a visual neural network prediction system is developed. The results of the prediction system are compared with the actual value, and the relative error of the prediction system is smaller and the reliability is higher. Finally, the prediction ability of the regression equation and the BP neural network model is compared with the experimental data. The results show that the BP network is used. The prediction ability of the neural network prediction model is better than that of the regression equation, so the prediction model of the neural network is more suitable for the prediction of the weld formation of high pressure GMAW.

【学位授予单位】：北京石油化工学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TG456.5

【相似文献】