Adaptive neuro fuzzy system for modelling and prediction of distance pantograph catenary in railway transportation

Abstract

Abstract. This research presents an adaptive neuro-fuzzy system which is used in the prediction of the distance between the pantograph and contact line of the electrical locomotives used in railway transportation. In railway transportation any incident that occurs in the electrical system can have major negative effects: traffic interrupts, equipment destroying. Therefore, a prediction as good as possible of such situations is very useful. In the paper was analyzing the possibility of modeling and prediction the variation of the distance between the pantograph and the contact line using intelligent techniques IntroductionThe problem of modeling is an important step in any activity that requires the identification of an engineering solution to a real problem. Solving a new problem involves: identifying and developing a model; using techniques to find an optimal solution based on the developed model (this can be done by simulation); using the optimal solution to the needs of the real problem Modeling and simulation are the use of statistical or dynamic (timely) models to obtain data and decision making information. Modeling and simulation help reduce costs and increase product and system quality. The modeling and simulation technique is cheaper and more secure than real-world experiments in many practical situations. It is also sometimes closer to reality than traditional experiments and allowing the configuration of environmental parameters to the values encountered in the application domain. Simulation is (often) much faster than a real process, allowing for an efficient analysis of different situations (the most unfavorable case.) Once is identified and developed a model, this can be used many times.The present paper uses modeling and simulation techniques based on fuzzy logic and neural calculus having the prediction of a time series, namely the variation in time of the catenary pantograph distance to electric locomotives. For the validation of the identified model, results of measurements of electrical quantities performed on a laboratory model were used. The catenary pantograph system has a dynamic oscillatory behavior and can be disturbed by certain turbulences during operation. It has been observed that at speeds of over 180 km / h the contact between the pantograph and the catenary can be interrupted and electric arcs can occur. There is no adequate mathematical model describing this behavior. Figure 1 shows the pantograph and figure 2 shows the aliasing of this movement.

Evaluation of Different Fault Diagnosis Methods and Their Applications in Vehicle Systems

“…The accurate rate is 95.44%. (20), OSF in the inverter of the right motor (19), complete demagnetization of the rotor in the left motor (9), phase-to-phase short circuit in the right motor (21), phase-to-phase short circuit in the left motor (8), SSF in the inverter of the left motor (7), OSF in the inverter of the left motor (6).…”

“…Combining some methods has also been considered a way to improve existing methods. For example, neural networks and fuzzy logic could be used together [9,10].…”

“…Figure 19. An example of the result of fault diagnosis for PMSM, which has eight fault states (Blue line: the predicted label with CNN model, red line: the original fault label): non-fault (0), SSF in the inverter of the right motor(20), OSF in the inverter of the right motor(19), complete demagnetization of the rotor in the left motor(9), phase-to-phase short circuit in the right motor(21), phase-to-phase short circuit in the left motor(8), SSF in the inverter of the left motor(7), OSF in the inverter of the left motor(6).…”

ANFIS-Net for Automatic Detection of COVID-19

“…For each membership function, a rule will be generated, so that if the total numbers of variables increase, the generated rules will be increased exponentially causing a problem of dimensionality and this technique is sufficient for small data set. 21 , 39 , 40 . On the other hand, subtractive clustering is another technique where each data point is labeled as a cluster center and the number of generated rules will be independent on the number of variables and will be linear to the data points 40 , 41 .…”