Research On Feature Extraction Method Of Dynamometer And Its Application

In the process of oil exploitation,it is necessary to keep abreast of the work under the pumping unit.Traditionally,the working conditions in the underground are judged by the actual indicator diagram obtained by the experts on the well and observed by experience.The judgment results are easily affected by human factors and technical level.The dynamometer is the main basis for analyzing the working conditions of the oil well.According to the area change of the dynamometer and the geometric characteristics of the point and line angle,it can reflect the abnormal fault phenomenon.By judging the effective features of the dynamometer,the working conditions of the well can be obtained.Therefore,the use of a computer for fault identification enables faster and more accurate classification.Accurate analysis and identification of the dynamometer by computer is of great significance.In this paper,two feature extraction methods are proposed for the displacement map of the dynamometer in combination with the graphical characteristics of the dynamometer,and a secondary diagnosis model is proposed for the common fault dynamometer.Mainly complete the following.Firstly,by analyzing the line angle features of the dynamometer,a feature extraction method based on the position of the feature points,the fluctuation coefficient of the characteristic line and the variance coefficient of the special angle is proposed.The feature points,feature lines and feature angles are selected.The sample library is used to construct the classification feature vector,and the geometric feature fault determination method is proposed based on the vector.Secondly,by analyzing the area of the contour area of the dynamometer,a feature extraction method based on geometric area change is proposed.After the coordinate normalization and the point normalization are performed on the indicator diagram,and the indicator diagram is meshed,the area change characteristic value of the indicator area outside the curve is selected to establish an area filling model,and a specific network is given.A grid area calculation method is proposed,and a fault identification method based on geometric feature recognition is proposed in combination with the change of the area of six regions.Finally,a secondary diagnostic method is proposed,which combines the point-and-line feature and the area change feature.And the proposed algorithm compares the time efficiency and the recognition accuracy respectively.