Fault Detection, Diagnosis & Condition Monitoring

ISACLAB designs data-driven and physics-informed diagnostic systems that catch equipment faults before they escalate into failures. Our methods span nonlinear NARX and neuro-fuzzy models, convolutional neural networks, and Bayesian prognostic algorithms, applied to heavy-duty gas turbines, steam turbines, boilers, compressors, and rotating machinery such as bearings and rotors. We combine physics-based process models with machine learning to build robust residual-generation and sensor-fusion frameworks that isolate faults under noise, uncertainty, and sensor anomalies — work published in ISA Transactions, IEEE Sensors Journal, and IEEE/ASME Transactions on Mechatronics, and applied directly with power-plant and refinery partners.