Design of a Computer-Based Condition Monitoring System for Gas Turbine

Nigeria's power sector faces persistent electricity shortages despite substantial installed gas turbine capacity, primarily due to reactive maintenance strategies that fail to detect incipient faults early. This study presents the design, development, and techno-economic validation of a Computer-Based Condition Monitoring System (CBCMS) for a GE Frame-5 (MS5001) gas turbine operating at a Nigerian power generation facility. The system integrates industrial data acquisition via Modbus TCP and MQTT protocols, workflow orchestration with Node-RED, lightweight machine learning analytics using ML5.js, and real-time visualization with P5.js. A feedforward artificial neural network regression (ANN) model was trained on 50,698 preprocessed operational records to establish a dynamic "healthy baseline" for performance prediction. A hybrid outlier detection method combining 3-sigma and quartile-based limits was applied during preprocessing. The final multivariate ANN model, incorporating compressor discharge pressure, compressor inlet temperature, inlet guide vane angle, fuel valve position, fuel flow rate, and fuel lower heating value, achieved a coefficient of determination (R² = 0.99). Techno-economic analysis across three effectiveness scenarios (5%, 10%, and 20% reduction in downtime and breakdown costs) yielded net present values ranging from ₦797.5 million to ₦3.35 billion, with payback periods as short as 3–10 weeks. The work demonstrates a scalable, cost-effective pathway for transitioning from reactive to predictive maintenance, with applicability to gas turbines, compressors, pumps, and other rotating equipment, offering substantial economic returns suitable for resource-constrained energy and industrial sectors in emerging economies.