Adaptive Hybrid Controller for Real-Time Train Derailment Detection and Control

Abstract

Railway track misalignment and derailment pose a serious threat to passenger safety and property damage. This paper proposes a system that uses an ANN-driven prediction model and fuzzy-logic speed control to achieve optimal efficiency for the stated problem. A clustering model was used to identify the features with the highest information gain for a more reliable system. Data-driven learning through ANN models, including Levenberg-Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG) algorithms, applies a fuzzy control strategy for real-time speed reduction based on three critical parameters: train speed, track angle, and ANN-based misalignment prediction. The fuzzy controller optimizes speed reduction in potentially unsafe scenarios, thereby enhancing rail safety. The results show that the Levenberg-Marquardt model showcased the best accuracy for the derailment prediction. The fuzzy controller allows efficient control of the speed.