A scalable process aware routing architecture for resilient wireless communication in industrial and control networks

Abstract

Industrial carbon-capture plants depend on timely and reliable communication among distributed sensors, controllers, and process models. Existing studies commonly examine fault recovery, energy-aware routing, reinforcement learning, or digital-twin monitoring as separate problems, leaving limited evidence on their coordinated use in one reproducible industrial communication framework. This study presents a process-aware MeshStar wireless sensor network (WSN) that combines low-overhead star clusters with redundant mesh links among cluster routers. A tabular Q-learning controller selects recovery and traffic-management actions from network reliability, latency, congestion, residual energy, link availability, and thermal-process criticality. The framework is evaluated through OMNeT++/INET network simulation coupled with a MATLAB/Simulink carbon-capture process model. Thirty independent seeds are used for each method and operating scenario, with comparisons against static Dijkstra, RPL-MRHOF, energy-aware RPL, conventional Q-routing, and deep Q-network routing. Under the revised evaluation protocol, the proposed method achieved a 95th-percentile latency of 84.6 ms, a packet delivery ratio (PDR) of 98.32%, and a route-recovery time of 3.74 s, corresponding to improvements of 12.6%, 0.78 percentage points, and 16.1%, respectively, over the strongest baseline. Scalability, security stress, process stability, and scenario-based economic results are reported separately. Because the validation is software-dominant, the findings demonstrate comparative feasibility rather than field-deployment readiness.