Intelligence-based safety decision models for train traction control systems

In this thesis, two intelligence-based safety decision models for train traction control systems are proposed. These models are to prove the effectiveness of a modern method for speed sensor vehicles in a communication-based train control system (CBTC). Fuzzy theory and Bayesian decision theory have been modeled to learn and to classify the vehicle traction conditions using a pattern recognition concept. The proposed models are original and formulated for such integrated and complex systems like automatic train protection (ATP) and automatic train operation (ATO). In the intelligent format, the train traction’s patterns are extracted and applied on speed sensors’ input to classify the train traction. The error and risk of traction misclassification is also calculated to reduce the impact and exposure of safety and hazards. The proposed safety models are suitable for such a decision system due to processing the manageable number of state of nature (i.e., slip/spin, normal and slide), features (speed and acceleration) and having the prior knowledge of the vehicle’s behaviour which can be collected either from field tests or lab simulations. Both models involve a mathematical problem which can be solved in any programming language and to be used in the on-board or embedded computers. The conceptual models are applied to a hypothetical case study with promising results.