Design Of Drilling Cuttings Detection And Automatic Feeding System Based On Machion Vision

At present,among the basic energy consumed by human beings,petroleum energy occupies a dominant position and has become an important strategic resource in China and even the world.Oil drilling engineering is an important part of oil exploitation,accounting for more than half of the oil exploitation cycle.If a drilling accident occurs,it will greatly affect the efficiency of drilling construction.When drilling,the downhole situation can be analyzed according to the cuttings brought by mud to prevent drilling accidents,but the traditional contact cuttings detection equipment based on weighing is easy to be corroded by cuttings,which leads to poor reliability of cuttings detection.In addition,when drilling,drilling fluid needs to be added to help the construction.Usually,raw materials need to be manually transported to the drilling fluid preparation site,which is extremely labor-intensive,and it is difficult to meet the requirements of rapid feeding in an emergency.The palletizing robot can carry the material bag,but it still needs to manually place the material bag in the preset position in advance,which is inefficient.In order to solve the above problems,this paper first studies and designs a non-contact online volume detection method for drilling cuttings based on binocular camera,calculates the cuttings volume in a certain area by using binocular stereo imaging principle,and realizes continuous cumulative detection of cuttings volume by using image mosaic method.On this basis,a set of drilling cuttings volume detection device including binocular camera,embedded GPU computer and 4G communication module is designed and developed,which can be hoisted at the drilling site,automatically and continuously detect cuttings volume,and upload the detection results to the web server for storage through 4G network,and display them on the web page.The test results show that the relative detection error of drilling cuttings volume detection system is less than 13%,which can meet the real-time monitoring requirements of drilling site.Then,this paper develops a set of multi-objective bag detection algorithm based on the improved CenterNet network,which improves the circular two-dimensional normal distribution heat map of CenterNet network into an elliptical two-dimensional normal distribution heat map with an angle,and obtains the center position and attitude angle of the bag through ellipse fitting,which effectively solves the problem of non-monotonic angle change in bag attitude detection.On this basis,the bag detection of the global camera at the far-view end and the near-view camera at the end of the manipulator is further studied.According to the different working conditions of the two cameras,the network structure is improved,and the far-view bag detection algorithm and the near-view bag detection algorithm are obtained,which lays the foundation for the robot to automatically grab.Finally,the bag detection algorithm,robot control algorithm and robot grasping strategy are combined to control the robot to carry out automatic and autonomous grasping test.The test results show that the final detection accuracy of the material bag can reach more than 97.6%,which initially meets the control requirements of the robot’s autonomous grasping.