Study On The Gap Sensor Of High-Speed Maglev Train And Its Modeling Method

As an idea vehicle on land, high-speed maglev trains have particular characteristics, including high speed, lower energy consumption, low noise, safety and comfort. The gap sensor plays an important role for electromagnetic suspension system which is a critical component of high-speed maglev train. So it is very important to study and impove the performance of the gap sensor for the development and implement of the maglev vehicle technology.With the rapid development of the artificial intelligence, the method of modeling based on artificial intelligence is used in sensor modeling widely. No study was found about the high-speed maglev train gap sensor with artificial intelligence modeling technology. Inoder to impove the measure precision and reduce the manufacturing cost, the intelligence modeling technology about gap sensor should be studied.Aiming to solve the problems of nonlinear characteristic, temperature drift and the slot effect with intelligence modeling technology, the following aspects are discussed and studied:Firstly, studies are carried about the gap sensor environment, work principle and the characteristics. Based on investigation and analysis about the equivalent circuit and the mesuer circuit of the inductace sensor, the main temperature drift reasons are analyzed. The simulation is carried about magnetic field of the detection coil, and the simulation results show that the fields are different clearly when the sensor is put in different position in a tooth-groove period of stator railway. The fluctuation of equivalent inductace leads to the slot effect. A method is proposed to determine the gap as well as the slot position. The gap coil is consisted of a figure of eight coil. The slot coil comprises two figure of eight coils. No influence is introduced by the position sensor to the gap sensor and no influence is introduced by the gap sensor to the position sensor too. The gap and slot position can be achieved simultaneously.Secondly, the data are abtained from experiments on temperature drift and the slot effect. All the preprocessed data are separated into training pattern set and testing set for the simulation modeling. A method is proposed that the gap sensor is corrected with two models in series based on artificial modeling. The signal sampling system is designed. The ambient temperature is obtained by the thermal resistor putted on the gap sensor coil. The nonlinear and temperature drift error are calibrated by the temperature compensation model. The slot effect is compensated by the tooth-groove checking model. The tooth-groove position signal is measured by the position sensor and the position sensor coils are layout at the same room with the gap coils. The final output of the compensator is undependent on the temperature and tooth-groove positon and the ouput value is equal to the fact gap. The space is effectively used by position sensor and gap sensor.Finally, the temperature and position compensation are implemented by RBF neural network, T-S fuzzy network and the least square support vector machine respectively. RBF neural networks are trained by the the gradient descent algorithm. The process is reduced by improving the learning algorithms with the momentum. The T-S fuzzy networks are trained by the BP algorithms. The corresponding process can be reduced by inproving the self-adjusted learning rate. The ls-svm has a signifieant advantage of the lower computational complexity than the other support vector machine formulations using linear or nonlinear mathematical programming. Therefore, the ls-svm has shown an excellent classification or regression performance in many applications. The simulation temperature and tooth-goove models are achieved by ls-svm. The comparative analysis is carried out on the different modeling methods. The combination forecasting models are composed of a RBF netework and a ls-svm model. The combination models may compensate the sensor with lesser error than both RBF and ls-svm. The compensation scheme may be used in the other error check of the gap sensor. The hardware resource may be saved by including this technology.