Research On High Precision Automatic Detection System Of Oil-water Interface

Crude oil is an indispensable and important energy source for the development of China’s national economy.From newly mined crude oil to real-life refined oil,it needs to go through such processes as solid impurities separation,gas impurities separation,crude oil,and sewage separation,among which the oil-water separation process is still a difficult problem for researchers at home and abroad.Although researchers at home and abroad have made some achievements in oil-water interface detection,the interface detection and oil-water separation control system between crude oil and sewage still need to be further improved.Aiming at the deficiency of the existing oil-water interface detection accuracy,this paper adopts the improved K-means image segmentation algorithm to obtain the interface between crude oil and sewage.Aiming at the problem that the traditional PID controller as the controller of the oil-water separation control system can not meet the actual oil-water separation control effect,this paper adopts the BP neural network learning algorithm to automatically adjust the three parameters of the incremental PID controller,,₄)and₍（9）,to realiee the reasonable control of the drainage valve opening siee of the oil-water separation tank.The main contributions of this paper are reflected in the following aspects:（1）The template matching method is used to locate the oil-water interface and observe the distribution of emulsion between the oil-water interfaces;the improved Canny edge detection operator is used to detect the edge information of the oil-water interface;the seed filling region distance calculation algorithm is used to measure the distance between the oil-water equivalent interface and the upper boundary of the image.（2）The CNCS-Kmeans hard clustering algorithm was proposed and applied in the oil-water interface detection of image segmentation experiment.Using Python software Open CV computer vision library,CNCS-Kmeans hard clustering algorithm,classical K-means hard clustering algorithm,and Gaussian mixture model EM algorithm optimieed soft clustering algorithm（GMM-EM）were applied to the oil-water interface detection image segmentation process.The results show that the CNCS-Kmeans hard clustering algorithm adopted in this paper can reduce the number of training iterations and running time of the algorithm when it is applied to oil-water interface image segmentation and has a better global segmentation effect.（3）In the process of oil-water separation,BP neural network PID is used to control the opening of the drain valve,which can effectively track the position and height of the oil-water interface.Through the modeling and Simulation of the model in the Simulink module,and the construction of the actual oil-water interface experimental environment and platform.The simulation and experimental results show that the BP neural network PID controller proposed in this paper has a good tracking effect on the oil-water interface,and improves the reliability and accuracy of the system,which is suitable for the control system of oil-water separation.