Mechanical Feature Extraction of High Voltage Circuit Breaker based on Machine Vision

The contact displacement and speed curves are tightly linked to the performance of circuit breaker, which are key parameters of site test and widely used in the mechanical identification and prediction model of high voltage circuit breaker. In the traditional methods, acceleration or displacement sensors are installed on the insulated pull rod to measure the contact displacement. However, when the above methods are implemented at the site operation, the installation and disassembly of sensors greatly reduces the efficiency of mechanical characteristics test of the circuit breaker. In this paper, a mechanical feature extraction method for high voltage circuit breaker is proposed based on machine vision. Firstly, a high-speed camera is employed to record the video of the action process of spring operating mechanism, in which the insulated pull rods, crank arms, and many other critical parts are included. Secondly, the library function of computer vision OpenCV is used to read frame by frame and split the video into a time-lapse series of images and the optical flow method is applied to analyze the captured images to track the moving corner point and realize its feature extraction after the image preprocessing operation of filtering and noise reduction. Then, taking the point from the crank arm of operating mechanism as a corner, the angle displacement curve and angular speed curve of crank arm can be obtained. Finally, the contact displacement and other mechanical characteristics are calculated from the movement curve of crank arm, which could provide evidences for the condition assessment of high voltage circuit breaker. According to the non-contact measurement based on the machine vision, the adverse effects of mechanical sensors mounted inside circuit breaker could be avoided. Moreover, the proposed method provides a feasible way to evaluate the mechanical characteristics of a live circuit breaker in the field operating environment.