Computer Vision Based Surface Damage Diagnosis For Steel Wire Ropes

With the socio-economic and industrial development,steel wire ropes(SWRs)have more and more extensive applications in many important fields.However,during the use of the SWR,it is inevitable that the SWR will cause a certain degree of damage.In order to achieve the detection,the warning and the maintenance of SWR damages as early as possible,it is necessary to carry out non-destructive testing research on SWRs.However,most computer vision-based surface SWR damage diagnosis methods are using traditional machine learning methods,instead of deep learning method methods.In recent years,the surface damage diagnosis method based on deep learning has made some achievements in many fields.Compared with the traditional machine learning method,the deep learning method is more effective.Therefore,in this paper,the convolutional neural network method is introduced to computer vision-based surface SWR damage diagnosis.The main research contents of this paper can be summarized as follows:(1)In order to investigate the computer vision-based SWR surface damage diagnosis,this paper designs an experiment rig that can be used to collect SWR surface images.The design of the carriage rail device allows easy image data acquisition of different sizes of SWRs.The design of a strip light source at an angle of 60 degrees from the camera improves diffuse reflection and reduces the influence of the reflection of SWR surface on the damage detection.The addition of a sleeve device stabilizes the wire rope and improves imaging quality.At the same time,this article also analyzes three factors that affect the quality of image acquisition,which provides some references for the future design of computer vision-based surface damage image data acquisition devices.(2)For the pattern recognition of SWR surface damage,traditional machine learning methods based on artificial feature extraction have problems that rely on a large amount of prior knowledge and poor model adaptability.Therefore,this paper proposes a convolution neural network-based damage detection method for SWR surface,which can well identify normal SWR type,broken wire SWR type,and worn SWR type.A comparison with machine learning methods based on manual feature extraction reveals that the convolutional neural network method proposed in this paper is more suitable for SWR surface damage detection.This paper also analyzes the effect of different convolutional,pooled and full-connected layer depths on the final classification performance in the design of the convolutional neural network structure.This analysis can bring some implications for the future design of a deep learning network for SWR surface damage recognition.(3)For the target detection of SWR surface damage,based on the existing literature research,no scholars have applied the wire rope target detection method to the wire rope surface damage diagnosis.Therefore,this paper proposes a target detection method based on improved YOLOv3 network.By introducing the idea of deep separable convolution to YOLOv3,it can reduce the model scale and increase the speed of model training and inference.Comparisons show that the network designed in this paper is smaller,more convenient for production deployment,and better target detection than YOLOv3 network model.In order to further satisfy the needs of actual production deployment,this paper uses OpenVINO toolbox developed by Intel Company to accelerate model reasoning and improve the speed of model training on CPU.