基于DeepLabV3s的曳引轮磨损测量研究

Traction sheave wear measurement based on DeepLabV3s

在不同的光照环境下,对电梯曳引轮的磨损量进行非接触式测量时存在误差较大这一问题,为此,提出了一种基于改进DeepLabV3的曳引轮磨损自动测量算法。首先,构建了曳引轮绳槽的物理模型,基于采集到的曳引轮绳槽图片,建立了曳引轮绳槽数据集;然后,采用融合SEnet和ECAnet双注意力机制的DeepLabV3s模型,对数据集进行了训练,实现了钢丝绳与曳引轮的分类目标;提出了一种融合曳引轮图像特征的图像处理算法,用相关匹配法识别并截取了目标区域,定位到磨损点,并计算了其磨损量;最后,为了对上述算法的性能进行验证,搭建了测量实验平台,进行了算法的鲁棒性验证和误差分析实验。实验结果表明:采用该算法进行测量所得绝对误差小于0.049 mm,均方根误差小于0.044 mm,且算法运行时间小于2.50 s。研究结果表明:与传统测量方法相比,该自动测量方法具有高精度、自动化、非接触的特点,能适应不同光照环境,快速、准确地测量曳引轮绳槽的磨损量,解决了不同光照环境下的曳引轮磨损非接触式测量问题。

Aiming at solving the problem of large error in the non-contact measurement of traction sheave wear in different lighting conditions,an automatic measurement algorithm of traction sheave wear based on improved DeepLabV3 was proposed.First,the physical model of the traction sheave rope groove was built,a data set of traction sheave rope groove was built based on the collected traction sheave rope groove images.Then,the DeepLabV3s model with squeeze-and-excitation networks(SEnet)and efficient channel attention networks(ECAnet)dual attention mechanism was used to train the data set to classify wire rope and traction sheave.An image processing algorithm integrating the image features of the traction sheave,which using correlation matching algorithm identified and intercepted the target area to locate the wear point and calculate it,was proposed.Finally,in order to verify the performance of the above algorithm,a measurement experiment platform was built,and the robustness verification and error analysis experiments of the algorithm were carried out.The experimental results show that,through algorithm measurement,the absolute error is less than 0.049 mm,the root mean square error is less than 0.044 mm,and the operating time is less than 2.50 s.The results indicate that,comparing with the traditional measurement,it has the characteristics of high precision,automation and non-contact,and can quickly and accurately measure the wear of the traction sheave rope groove in different lighting conditions,which solves the problem of non-contact measurement of traction wheel wear under different lighting environments.