Artificial Intelligence-Enhanced Modeling, Control, and Fault Diagnosis in a Digital Twin Framework for Coal Coking Processes

Driven by carbon neutrality goals and the demand for greener industrial production, this thesis develops an intelligent framework to support the digital transformation of the coal coking process. It integrates numerical simulations, data-driven prediction methods, advanced neural network architectures, reinforcement learning-based process control, and Transformer-based fault diagnosis to address key challenges in heat transfer and volatile matter generation modeling, operational regulation, and system reliability. The proposed framework improves modeling accuracy, control performance, and fault response under complex and nonlinear conditions. Overall, it provides a unified foundation for intelligent operation and digital twin development in the coking industry.