Research On Key Technologies Of Mobile Robots For Power Plant Inspection

At present,power plant safety inspections mainly rely on manual inspections and manual transcripts of instrument data.There are problems such as high labor costs,scattered data and information,and strong subjectivity,and it is difficult to ensure that hidden dangers are dealt with in time.This article has designed an autonomous navigation and dynamic obstacle avoidance,Infrared thermal imaging temperature warning,environmental data collection,dial recognition,and web-side monitoring of power plant inspection mobile robots.Through the kinematics analysis of the robot,the establishment of the model of an omnidirectionally movable chassis based on the mecanum wheel is completed.In view of the large blind area of the single lidar and insufficient point cloud data characteristics,the PCL-based front-end and back-end lidar data source splicing and fusion were completed,thereby improving the robot’s positioning accuracy and dynamic obstacle avoidance performance.The state space model of the cooperative movement of the electric push rod and the pan/tilt is established,and the dynamic adjustment control of the attitude under different dial positions is realized.In view of the need to detect the concentration of some dangerous gases in the power plant environment,the discrete data collection of carbon monoxide gas concentration,methane gas concentration and smoke has been completed.The infrared thermal imaging camera is used to realize remote display of real-time images,temperature monitoring and abnormal conditions alarm.Aiming at the problem of large-scale image construction and odometer dependence of the particle filter-based Gmapping SLAM algorithm,the Cartographer solution based on graph optimization SLAM was deployed in the robot and the establishment of a twodimensional grid map was completed.And successfully applied to navigation.Due to the large limitation of the single-line laser detection range,the realization of the visual SLAM solution based on depth camera was completed.Taking ORB-SLAM2 as an example,a sparse point cloud 3D map was established.Taking RTAB-Map as an example,a dense point cloud three-dimensional map was established,and it was successfully converted into a 2D raster map for navigation,which solved the problem of collisions caused by potential obstacles that cannot be detected.The path derived from the A star algorithm has problems such as poor smoothness and large turning range,which are often difficult to directly apply in actual engineering.This thesis designs a static path search algorithm based on minimizing snap optimization A*.The trajectory is described by a piecewise polynomial,and the robot dynamics constraints are combined to improve the smoothness of the path.The dynamic sliding window algorithm is selected to realize the robot’s dynamic obstacle avoidance function,and two inspection schemes of fixed point and custom point are designed under the ROS framework.Aiming at the problem of difficult high-precision and intelligent identification of dial data,and the identification of industrial dials in the video stream was realized by deploying darknet10.1 and YOLOv4-tiny algorithms.Completed the static web page design based on HTML and CSS,and empowered it with a Java Script-based trigger function.Based on Web Socket,it realized the functions of web page to ROS end instructions,real-time video display,and related data display.In this thesis,the self-developed mobile robot for power plant inspection operations has completed the main functional modules and algorithm performance tests in the laboratory environment,which meet the needs of normal inspection tasks,and have many advantages such as high inspection efficiency,stable and reliable operation,and labor cost saving.