Research On Online Monitoring Method For Lift Traction Rope Surface Damage Based On Cloud And Edge-Side Collaboration

Elevator is an essential equipment in modern urban high-rise buildings,greatly facilitate people’s lives at the same time,the safety of the elevator has also been a concern,where the traction rope is one of the most important parts of the elevator,in the long use process,the surface of the traction rope will appear damage situation.If not timely detection and treatment,will bring a variety of casualties and property damage safety accidents.As the traditional elevator traction rope surface damage detection methods such as visual inspection method,acoustic emission detection method and magnetic leakage method have problems such as easy misjudgment,high cost and single detection object,it is difficult to meet the needs of the development of non-destructive testing of elevator traction rope.Therefore,the use of deep learning methods to achieve surface damage detection has become a development trend.However,when the target detection algorithm is deployed in embedded devices on the ground,the accuracy and speed will deteriorate seriously due to the low processing main frequency and small memory.In order to balance the accuracy and detection speed of traction rope surface damage detection,the YOLOv5 s model is used as a basis to design a target detection model applicable to edge-end devices by introducing effective modules on its backbone network in a targeted manner,and the cloud-side collaborative architecture is designed to transfer the identified traction rope surface damage image data to the cloud and design an elevator traction rope surface damage monitoring system locally to realize the Online monitoring of elevator traction rope surface damage,the main contents are as follows:(1)Optimize the algorithm deployed at the edge.To ensure the accuracy and speed of the detection model for embedded devices,YOLOv5 s is the primary model when constructing the edge model.Ghost Net is then employed as the backbone feature extraction network,capitalizing on its benefits of less computation and non-redundancy of feature map,thus decreasing the algorithm’s complexity and enhancing the detection speed.The attention mechanism,CA,is then implemented to seamlessly combine spatial coordinate data with the attention map,aiding the network in rapidly extracting advantageous features and augmenting the algorithm’s feature extraction capability.Replacing the original algorithm’s path aggregation network structure,the Bi-FPN structure has fused the features of multiple scales.Experiments have revealed that the YOLOv5 s model has been surpassed in detection speed by 15.4FPS,parameters by 41.5%,computation by 5.1G,and detection accuracy remains largely unchanged.(2)The edge end is tasked with gathering samples,introducing lightweight models to identify traction rope surface damage,and using Jestson AGX Xavier’s ossimport data migration tools to send data to Alibaba Cloud’s object storage(OSS)platform in real time.Additionally,a traction rope surface damage detection system is designed to be locally visible in damage images,all of which are part of the implementation of a cloud-edge collaborative architecture.(3)Design the surface damage monitoring system of elevator traction rope.In order to view the surface damage detection results of traction rope in a timely and effective manner,a surface damage monitoring system of elevator traction rope based on SpringBoot+My Batis+My SQL+CXF framework is designed,and the data source of the system is mainly collected and uploaded by the edge device Jestson AGX Xavier,and Jestson AGX Xavier uploads to the object storage(OSS)of Alibaba Cloud through the ossimport tool,and then downloads the data locally.The system is largely split into two components: the login module for the user system and the main interface display module.The user login module can perform account registration,login,retrieval and other functions,and can also switch between Chinese and English to meet the needs of different users;The main interface display module can view time information,equipment location information,user personal information and traction rope surface damage images.