A Development Method for Industrial Automation Based on Digital Twins with Edge Preprocessing and an Event Bus: Experimental Evaluation of Latency, Delivery, and Recovery

The paper proposes a method for developing industrial automation based on digital twins for IIoT, combining edge telemetry preprocessing, an event message bus, and a digital twin layer for monitoring and control loops. The architecture includes edge modules (filtering, normalization, window aggregation, and buffering), a device connectivity platform (Mainflux), an event bus (NATS) for asynchronous delivery with backpressure, DT orchestration and rules (OpenRemote), time-series storage (TimescaleDB), and Web UI/2D–3D visualization. For quantitative evaluation, a comparison was performed against three alternatives: centralized processing without edge, point-to-point integrations without a message bus, and a monitoring-oriented variant without a full-fledged DT platform. Experiments demonstrate that the proposed solution provides higher throughput (at a load of about 2000 msg/s, approximately 1400 msg/s of received messages are achieved versus ~1150 msg/s for P2P, ~1000 msg/s for the centralized scheme, and ~800 msg/s for monitoring-only) and significantly lower tail latencies (p95 about 0.50 s versus ~1.5 s, ~3.0 s, and ~6.8–7.0 s, respectively). Fault injection tests confirm the rapid recovery of key components (NATS restart ~6 s, edge ~8 s, DT-manager ~12 s, Mainflux ~15 s), while DB failure constitutes the dominant contribution to recovery time (~28 s). The obtained results demonstrate that the combination of edge analytics and event-based decoupling enhances scalability, resilience to network degradation, and the architecture's suitability for near real-time DT control.