【作者】 陈翔；

【导师】 邹俊；

【作者基本信息】 浙江大学 ， 工程硕士（专业学位）， 2019， 硕士

【摘要】 环形冷却机是烧结工艺中的冷却用承载,是钢铁冶金的重要设备之一。环冷机粗犷的工况使得台车车轮的轴承容易变形和磨损,导致车轮从车轴脱落,严重影响生产效率,威胁人员安全。因此研究车轮状态预警机制并实现自动化监测具有重要的意义。轴承失效的过程中,车轮在铁轨上由纯滚动变成滑动,转速降低。利用这一特征,针对台车车轮的转速测量进行研究。论文的主要工作包括:1.根据技术指标和系统需求设计总体网络架构。摄像机和传感器在线采集数据,下位机PLC协调控制,中心服务器实现图像的实时处理和报警机制。2.基于张正友法对相机进行分步标定,消除畸变并扩展对滚动车轮的采样范围;应用改进的直方图均衡和霍夫直线检测,排除光照变化和投影的影响;提出一种自适应阈值的Canny算法,并应用到霍夫圆变换中,提取车轮的旋转区域。3.提出一种基于图像特征匹配的转速测量方法。应用ORB算法对连续两帧中车轮上的标志物进行特征匹配,对筛选后的匹配点对估计仿射变换矩阵,获得滚动的瞬时转速和瞬时速度,并以此判断车轮的滚动/滑动状态。当滑动量超过阈值时触发报警。基于轮廓提取和模版法提取和识别的车轮序号,以准确地记录车轮的状态,实现精准报警。4.基于速度积分和光电开关的掉轮检测。利用光电开关检测车轮信号,根据台车的行进速度预测车轮到达时刻,通过冗余机制检测掉轮故障。基于上述算法和思想,研发基于机器视觉的环冷机台车车轮状态监测系统,实现对车轮状态的在线检测、故障预测与故障跟踪。本系统在宝钢炼铁厂4#烧结机上成功验证。瞬时转速检测耗时在300ms以内,误差小于5%,平均转速误差小于2%,综合故障检出率达99.5%。更多还原

【Abstract】 Sinter cooler is the bearing for cooling in sintering process,and it is one of the important equipment of ferrous metallurgy.The rugged working condition of the sinter cooler makes the bearing of the trolley wheel easy to deform and wear,causing the wheel to fall off from the axle,seriously affecting the production efficiency and threatening the safety of the personnel.Therefore,it is of great significance to study wheel state early warning mechanism and realize automatic monitoring.In the process of bearing failure,the wheel changes from pure rolling to sliding on the rails,and the rotational speed decreases.Using this feature,the speed measurement of the trolley wheel is studied.The main tasks of the paper include:5.Design the overall network architecture based on technical indicators and system requirements.The camera and sensor collect data online,PLC coordinate the controls and the central server to achieve real-time image processing and alarm mechanism.6.Calibrate the camera by 2 step based on the Zhang’s method to eliminate the distortion and extend the sampling range of rolling wheels.The improved histogram equalization and Hough linear detection are applied to eliminate the influence of illumination change and projection.An adaptive threshold Canny’s algorithm is proposed,which is applied to the Hough circle transform to extract the rotating region of the wheel.7.A rotational speed measurement method based on image feature matching is proposed.The ORB algorithm is used to match the markers on the wheel in two consecutive frames,and affine transformation matrix is estimated for the filtered matching point pairs to obtain a pair of instantaneous velocities--translational and rotational--of rolling,on which based the rolling/sliding state of the wheel can be judged.End systems are alerted when the sliding momentum exceeds the threshold.The sequence number is extracted and identified based on contour extraction and template method to recognize the specific wheel.8.Detect wheels falling off fault based on photoelectric sensors and speed integral method.The wheel arrival signal is detected by photoelectric switch,and the arrival time is predicted according to the travel speed of the trolley,and the failure of the wheel is detected by redundancy mechanism.Based on the above algorithms and ideas,a wheel condition monitoring system for sinter cooler based on machine vision is developed to realize on-line detection,fault prediction and fault tracking of wheel status.This system has been successfully verified on no.4 sinter cooler in Baosteel ironworks.The instantaneous speed detection takes less than 300ms,the error is less than 5%,the average speed error is less than 2%,and the comprehensive fault detection rate is up to 99.5%.更多还原