Real-time abnormal monitoring of electro-hydraulic servo valve characteristic curve based on sliding window
Aiming at the problems of difficulty in automatically marking of feature parameters,high calculation cost and low recognition accuracy in the real-time abnormal monitoring of electro-hydraulic servo valves,a real-time abnormal monitoring method for valves facing sliding window servo valves is proposed.First of all,the pulse width modulation signal is used to realize the digital drive of the valve and reduce the transmission noise interference of the analog current.According to the sensor feedback data,the valve characteristic curve is plotted,the identification characteristics of the inflection points are analyzed,and the cost function under the linear piecewise regression model is constructed based on its characteristics.The sliding window is selected as the discretization optimization algorithm,and the fixed number of 4 inflection points on the feature curve are quickly found.The valve data such as valve stroke difference and dead zone are calculated according to their numerical values,and the valve work is evaluated by comparing with the standard values.The window width is set to 2 500,and the speed and accuracy of abnormal discovery are balanced.Finally,the proposed method is verified in a real power plant environment.The test results show that the proposed method can quickly and accurately extract the inflection point of the valve characteristic curve under high noise environment,and then identify the abnormal performance of the valve on this basis,the proposed method can retain the local fluctuation characteristics of the curve and reduce the misjudgment rate of the inflection points at the same time compared with the traditional curve smoothing method.
NETL
NSTL
万方数据
