Research On Target Recognition And Positioning Of Drilling Pipe Articulation Interface

As the modern drilling process continues to accelerate,the degree of automation required on drilling rigs is increasing,especially in the face of the fourth wave of industrial revolution represented by artificial intelligence and robotics,there is a greater demand for automation and intelligence of operating devices on drilling rigs.The iron driller is an important tool loading/unloading device in the drilling rig wellhead operation,and the improvement of its automation degree has a great impact on the drilling efficiency.At present,most of the iron drillers still rely on manual operation to locate the drill pipe articulation interface,so that they can move to the designated operating position,and then locate the drill pipe at a short distance through the sensor at the front of the clamp body,which cannot realize the complete automation of iron drillers.Therefore,this paper applies machine vision related technology to study the target identification and positioning of drilling rod articulation interface,establish an improved YOLOv5 drilling rod articulation interface identification model,complete the identification and positioning of drilling r od articulation interface,and enhance the automation and intelligence of drilling process.The main work of the thesis is as follows:(1)A deep neural network-based drill pipe articulation interface recognition method is proposed,and the current popular YOLOv5 network is applied to establish a model for recognizing the drill pipe articulation interface.The Dense Block module is used to replace the Focus module,the coordinate attention module is added,and the loss function is optimized to improve the feature extraction capability,feature weight enhancement,and robustness of the model.The model is improved in terms of feature extraction,feature weight enhancement and robustness.(2)Experimental validation of the above model is carried out.The basic YOLOv5 drill pipe interface recognition model and the improved YOLOv5-DCG model were trained to recognize two categories of upper and lower interfaces,and the results showed that the recognition accuracy of the YOLOv5-DCG model was improved by 2.14% compared with the improved model.Finally,by testing the recognition effect of the YOLOv5-DCG model under different brightness and deformation,it is proved that the improved model has strong robustness and can accomplish the recognition task in the actual environment.(3)The binocular matching technique is proposed to complete the research on the localization of drill pipe articulation interface.Based on the internal and external parameters,distortion coefficients and rotation translation matrix of the binocular depth camera,the block matching algorithm is used to obtain the corresponding depth information map,which is then combined with the trained recognition model to calculate and obtain the accurate coordinates of the drill pipe articulation part to achieve the accurate positioning of the drill pipe articulation interface.The experimental results show that the actual average error of positioning is 7.09 mm,which meets the engineering error requirements and has certain engineering use value.