Research On Railway Multi-sensor Track Section Occupation Detection Method

Railway signal refers to signal display, interlocking and block equipment. The operating conditions of rolling stock, the status of train operation equipment, and the instructions and orders can be conveyed to railway staff to ensure driving safety. It is very important to detect whether there is a vehicle on the line in railway signal system, which refers to inter-station line or station section line.This thesis researches on the occupancy detection method of railway station track section. A new detection method of axis counter equipment is proposed based on the typical track occupancy detection. After summerizing the development history, current situation and the existing problems of the detection method, this thesis reviews the related theories, to solve the critical problem of the detection method, and finally verifies the method through simulation and actual test.The main research content and innovations of this thesis are as follows.1. Many existing axis counting sensors provide single data without fault tolerant, while the proposed method applies infrared axis counting sensor, proximity axis counting sensor and train existing sensor to count train's axis. It includes the installation approach of the sensor, the approach of data acquisition, the approach of the encoding sequence, and the feature extraction of the data, which are the basis of further processing axis counting data and detecting the occupancy of rail sections. All the output data generated by the sensors are binary data, so the formed code data is accurate and easy to process and transmit, which make our system have strong anti-interference performance.2. In order to remove interference and improve the accuracy of the axis counting, this thesis proposes a KMP meter axis matching detection algorithm by axis counting coding and combining the KMP matching algorithm with the coding data after the installation, collection, coding of infrared axis counting sensor, proximity axis counting sensor and train existing sensor based on axis counting code. It judges whether the output value is an axis of data of the train according to the output value of axis counting matching algorithm, which is getting from the extraction of the characteristics of integrity, orderliness and similarity of data code for counting axis. The algorithm can be easily implemented. It is fast and suitable for the less interference section because of its'low interference ability. The related results are shown by the simulation.3. The time element is added to the counting encoding sequence in order to improve the identification ability of interference in complex environment. The author proposed the time-series change rate matching (TSCRM), which takes the change speed and acceleration of the time series as a means of matching series to implement similarity match between the standard axis counting coding data and the coding data to be detected. This method has simple principle, fast processing speed and strong anti-interference ability, and can obtain accurate data through simulating and analyzing field test data.4. Since track section occupancy is more complicated, in order to improve the judging accuracy of track section occupancy, this paper proposes a method based on D-S evidence theory and information fusion by chaotic neural network to solve the problem of detection occupation of track sections. The fusion information contains six aspects, including the existence detection of any object, track section occupancy, speed of the train, acceleration of the train, movement direction of the train, area of the train. The method determines the basic probability assignment, which has three cases:car passing, no car passing, and uncertainty by statistical evidence and maximizing deviation method. Then it applies track occupancy detection model of Multi-sensor temporal recursive to determine whether there is a car, calculates the weight data of sensors through historical data by applying neural network technology to avoid the limitations of D-S evidence theory model where the neural network is easy to fall into local optima into account. This kind of chaotic neural network can effectively prevent the neural network from falling into local minimum in training process by introducing chaos mechanism, and recognize the tiny difference model better. This paper makes judgments more accurate based on the two fusion algorithms.5. In order to validate the occupancy detection method of track section based on the axis counting matching and information fusion, the system designs a specific test program by using CAN bus structure to connect the equipment near the track and the indoor equipment to the bus and completing data collection and information transmission, including the reliability of transmission and reliability of controlling. At the same time, the relay control unit is designed by detection reliability controlling to complete the safety protection of track section occupancy detection in the way of sub-controlling parallel with the original track circuit, which not only improves the occupancy detection accuracy but also can detect whether track circuit detection is cracked or not by the principle of "fault oriented to security". After continuous actual test for more than two years, the actual test shows that the accuracy rate of occupied track and counting the number of wheels reaches 100%.The theoretical research, simulation and actual test in this paper demonstrate that the proposed method is feasible, accurate and effective, and its reliability, security and efficience can meet the design requirements. Moreover, The Multi-fault caused by track occupied is a problem to be solved. Further researches and experiments show that this proposed method applied to this paper has certain promotion value closely combined with production practice. At the same time, this method can provide a certain experiences and reference for the research of theory and application in other field.