铁路货车智能铆接系统的研究与设计

Research and Design of Intelligent Riveting System for Railway Freight Cars

为保证货车运行的可靠性,在转向架上采用了铆接技术。短尾拉铆销在铆接前后从外形上看没有明显差异,操作人员在安装时还需进行其他部件的装配工作,常造成铆接不到位、漏铆等操作问题,将会给实际运用中带来严重的安全隐患。针对铁路货车转向架基础制动梁组装工序中出现的拉铆钉的漏铆、铆接不到位等问题,文章研究了一种检测铆接质量和铆接位置的方法。在铆接过程中采集流量传感器和压力传感器信号,间接得到铆接力与铆接位移,绘制铆接力与位移关系曲线。利用经验阈值法对铆接曲线进行分析,判断铆接结果的质量。同时,采集每个铆接位置的背景图像并提取像素点特征值,利用主成分分析法(PCA)对特征值进行降维处理。最后,利用卷积神经网络(CNN)进行训练和识别。试验结果表明,此方法能够根据合格曲线正确完成铆接结果的判断,并能准确地判断出铆接位置,可为铁路货车的行车安全提供一种有效的技术保障。

In order to ensure the reliability of freight car operation,riveting technology is adopted on the bogie.There is no obvious difference in the appearance of short-tail rivets before and after riveting.Operators in the installation also need to assemble other parts,which often causes riveting not in place,riveting omission and other operational problems,bringing serious safety hazards in practical application.In response to the problems of rivet omission and incomplete riveting during the assembly of the brake beam on the bogie,this article proposes a method for detecting the quality and position of the riveting.Flow sensor and pressure sensor signals are collected during the riveting process to indirectly obtain the riveting force and riveting displacement,and the relation curve of riveting force and displacement is drawed.The riveting curve is analyzed using the empirical threshold method to judge the quality of the riveting effect.At the same time,the background image of each riveted position is collected and the pixel point eigenvalues are extracted,and the eigenvalues are dimensionally reduced using principal component analysis(PCA).Finally,convolutional neural network(CNN)is used for training and recognition.The experimental results show that this method can correctly determine the riveting results according to the qualified curve,and accurately identify the riveting position,which can provide effective technical support for the driving safety of railway freight cars.