Design and Implementation of an Intelligent SCADA System for Marine Engine Predictive Maintenance Using OT-IT Convergence

Abstract

In the maritime industry, the transition from Time-Based Maintenance (TBM) to Predictive Maintenance (PdM) is increasingly critical to meet stringent environmental regulations and minimize operational downtime. However, implementing data-driven diagnostics on marine vessels faces significant challenges due to the "OT-IT gap" and the hardware constraints of legacy shipboard systems. This study proposes a cost-effective, Edge-centric Intelligent SCADA architecture that integrates a Siemens S7-1200 PLC with a Hybrid Artificial Neural Network (ANN) via an SQL-based middleware. To ensure compatibility with standard low-configuration Industrial PCs (IPCs), the system employs a deliberate asynchronous sampling strategy (T_scan = 3 s), decoupling heavy predictive analysis from real-time monitoring. This design allows the diagnostic module to operate seamlessly as a background service without overburdening the control loop. Additionally, an optimized Bit-wise State Encoding (Bit-packing) protocol is implemented to address bandwidth constraints. This method compresses diagnostic vectors into atomic integer payloads, ensuring data integrity during transmission. The system was validated through Hardware-in-the-Loop (HIL) simulations and field trials on a Daihatsu 6PSHTb-26D training ship. Experimental results demonstrate that the architecture maintains a standard supervision latency of < 450 ms. Meanwhile, the AI-based fault detection operates with a stable latency of 3.1 s to 4.9 s—a designed safety margin that is sufficient for marine propulsion dynamics while optimizing computational resources. Furthermore, the Bit-packing protocol reduces communication overhead by approximately 90% compared to traditional tag-based methods, confirming the system’s feasibility as a robust retrofit solution.