基于神经网络数据处理的轮缘踏面检测研究  

Wheel Profile Measurement Based on Neural Networks Digital Processing

作　　者：杨志强[1] 翁孟超[1] 张济民[2] 王晓燕[1] 

机构地区：[1]浙江师范大学交通学院,金华副教授321019 [2]同济大学铁道与城市轨道交通研究院,上海200331

出　　处：《城市轨道交通研究》2007年第3期42-45,共4页Urban Mass Transit

摘　　要：介绍了一种基于神经网络数据处理技术的轮缘踏面自动检测方法。该方法能自动判断被测轮对轮缘踏面是否因磨耗过限而需要镟修。它通过数码相机采集被测轮对的原始图像,输入计算机,通过计算和处理,并与基于BP神经网络建立的模型相比较,从而自动作出判断。这种非接触式自动测量方法与多年来采用的手工测量方法相比较有很大的提高,可判断自动检测的精度在0.2mm范围之内,而工程测量的精度要求为0.5mm。This paper introduces a method which is based on the neural networks digital processing technology for wheel profile automatic measuring. This method can automatically judge whether the wheel profile needs to repair because of wears and tears. It captures the picture of a digital camera, inputs it into the computer for certain computation and processing, compares the result with the BP neural network establishment model, thus makes the judgement automatically. The authors argue that this method is of great importance to wheel measurement, and will have a wide foreground of application.

关 键 词：轮对踏面 外形检测 数据处理 神经网络 

分 类 号：U260.331.1[机械工程—车辆工程]