Architecture Design Of Inspection Robot Management System And Online Fault Diagnosis Method

In recent years,with the improvement of hardware computing and data transmission capabilities,intelligent inspection robots have been widely used in various industrial scenarios to replace manual execution of tedious and hazardous inspection tasks.However,during the robot inspection process,it is often out of the sight of workers,making effective monitoring and management of inspection robots a critical issue.To address the current deficiencies in realtime performance,stability,security,and inadequate monitoring of the robot’s status in the current inspection robot management system,this paper comprehensively analyzed the various requirements of the inspection robot management system.Combining with MQTT protocol,Web Socket protocol,streaming media technology,fault diagnosis,semi-supervised learning and other technologies,the robot management system is designed and implemented,which includes an online fault diagnosis subsystem based on convolutional neural network and semisupervised learning.The main research work and achievements are summarized as follows:Based on the analysis of the functional and non-functional requirements of intelligent inspection robots,a low-latency,low-bandwidth,and high-efficiency inspection robot data communication architecture suitable for heterogeneous network environments was designed.By adapting to corresponding environments and usage scenarios through multiple communication methods,a real-time communication system with high efficiency,stability,and scalability was constructed.Simultaneously,relational databases and non-relational databases are combined to achieve data persistence and efficient access.The test results demonstrated that the system could run persistently and stably,with a delay of less than 80 ms for real-time data display,less than 30 ms for real-time control,and approximately 600 ms for real-time monitoring video,with a success rate of data transmission greater than 99.6%,meeting the requirements of the management system.To address the security issue of inspection robot data transmission,considering the limited computing resources of the robot’s control system and the demand for further reducing energy consumption,this paper studied the SSL/TLS security scheme recommended by the OASIS standard for MQTT communication and summarized its drawbacks.Subsequently,several mainstream encryption algorithms were studied,and an improved MQTT encryption scheme suitable for the inspection robot management system in this paper was proposed,balancing efficiency and security.The effectiveness and security of the scheme were demonstrated through security analysis.Using multiple indicators,its superiority over the SSL/TLS scheme for this paper’s system was demonstrated.For on-line fault diagnosis of inspection robots,a convolutional neural network(CNN)structure most suitable for the system was designed,considering the system’s ability to collect data from multiple inspection robots.Then,an automatic model update strategy based on CNN and semi-supervised learning was proposed,comparing the strategy’s effectiveness using Euclidean distance and Dynamic Time Warping(DTW)metrics.Experiments show that the automatic update strategy can improve the accuracy of the original model by 2.2%,reduce the false positive rate and the false negative rate by 2.4% and 2.6%.By employing front-end and back-end separation technology and cloud servers,the system was deployed online.On this basis,the system’s interface was demonstrated and actual testing was conducted to confirm that the system could meet the requirements proposed previously.In addition,non-functional testing proved the superiority of the system presented in this paper.Finally,the online fault diagnosis capability of the system was tested and multiple indicators were used to demonstrate the reliability and real-time performance of the online fault diagnosis.