Research On Automatic Analysis And Positioning Of S700k Switch Machine Control Circuit Fault

With the railway developing trend turns to high speed and overload, there’s no doubt thatmore stringent requirements should put forward on fault diagnosis technology of S700kswitch. However, the fault detecting method in control circuit existed still stays in the originaltechnical level, which is mainly determined by operators experience and simple apparatus. Itseems that the ability to process faults not equals to the requirements, such as handling failurefor a long time, inaccuracy in fault location. Thus, the timeliness and accuracy of faultdetection become urgent.In order to meet the needs of the development of railway, it is necessary to improve theoriginal technology or innovation by using existing mature technologies. A convenient waywas provided for switch control circuit since the development and gradually perfection inintelligence technology field. Especially, an experimental platform for fault diagnosis wasprovided by combining the gradually mature artificial neural network.In this paper, firstly, according to the characteristics of the control circuit, the followingwas introduced divided into expression circuit, starting circuit: the working principle of thecircuit, the fault diagnosis process, the troubleshooting process. Then, expounded thenecessity of using the neural networks. Then the basic principle of BP neural network,algorithm and process were introduced, and L-M numerical optimization algorithm isproposed. Secondly, the failure mode of expression circuit and startup circuit was carried outrespectively. After completing the whole model, we should obtain the corresponding trainingsamples, which were obtained from the microcomputer monitoring and from the locale, asmere microcomputer monitoring data can’t give an exact fault position. At the same time, thedata trade-offs, supplement and normalized processing were included. Finally, neural networkwas trained. Due to the expression circuit fault is divided into normal-position circuit faultand the anti-position circuit fault, and start-up circuit fault can be divided intonormal-positioning-anti-position circuit fault and anti-position-normal-positioning circuitfault. So an neural network mode should be constructed for each type of fault. Eventually, putthe models together to constitute a parallel neural network system, making use of thecharacteristics of the same branch in different circuits to locate accurate positioning of fault.Test results showed the effectiveness and accuracy of L-M numerical optimizationalgorithm, solving the timeliness and accuracy of fault detection, and reach the ideal result,thus this method can be used for detecting malfunction of S700K switch machine controlcircuit automatically.