Dynamic routing with deterministic scheduling and self-healing recovery for industrial thermal storage networks in a hybrid physical emulation testbed

Abstract

Industrial thermal-storage systems require communication mechanisms that can prioritize control traffic according to both network conditions and the physical urgency of the associated thermal process. This study proposes a Thermal-State-Aware Dynamic Routing (TSDR) method that integrates thermal criticality with link delay, jitter, packet-delivery probability, congestion, energy cost, and fault risk in a constrained multi-criteria path-selection framework. The proposed method is deterministic and rule-based; it combines multi-criteria routing, deterministic queue scheduling, threshold-based fault detection, precomputed backup paths, route hysteresis, and explicit recovery-confirmation rules, and it does not employ a trained artificial-intelligence or machine-learning model. Evaluation was conducted in a hybrid physical–emulation Hardware-in-the-Loop environment comprising three physical 500-L thermocline tanks, 45 measurement devices, PLC-based control and industrial networking hardware, 14 physical communication endpoints, 146 NS-3-emulated nodes, and synchronized MATLAB/Simulink and reduced-order CFD thermal models. A total of 480 independently initialized runs were completed across four routing architectures and ten operating or fault conditions. Compared with the strongest baseline, namely TSN-enabled network-state adaptive routing, the proposed method reduced P95 end-to-end latency from 98 to 64 ms (34.7%), jitter from 32 to 24 ms ( 25.0 % ), and fault-recovery time from 6.1 to 2.9 s ( 52.5 % ). Packet Delivery Ratio increased from 97.3 % to 99.1 % , while deadline misses decreased from 5.8 % to 1.9 % . Physical temperature-tracking RMSE decreased from 0.81 to 0.62 ∘ C , Energy Utilization Factor increased from 0.84 to 0.95 , and measured auxiliary power decreased by 9.1 % . Mixed-effects analyses identified statistically significant differences under the evaluated HIL conditions, with moderate-to-large effect sizes for the principal outcomes. Emulation-based scalability tests maintained the 100 ms routing deadline at 640 logical nodes; this result is EMU evidence and does not represent a physically deployed 640-node system. Under the base techno-economic scenario, the projected incremental investment produced an NPV of approximately USD 104,000, an IRR of 41.8 % , a simple payback of 2.32 years, and an estimated annual reduction of 68.3 tCO 2 e . These financial and environmental outcomes are scenario-based projections, while the technical findings apply only to the evaluated hybrid HIL configuration.