abstract
- This work presents the design, development, and deployment of a human-centered cyber-physical system (HCPS) for forecasting the ON potential in cathodic protection systems for critical hydrocarbon-transport infrastructure. The proposal is built on a hybrid architecture that combines long short-term memory (LSTM) networks with Mamdani-type fuzzy logic, enabling the capture of complex spatiotemporal patterns while managing the uncertainty inherent to nonlinear, stochastic systems subject to environmental perturbations. We introduce a multiscale preprocessing pipeline for reconstructed ON-potential time series to feed the model, optimized via adaptive dynamic weighting. In tandem, the research contributes two advances: (i) the definition of a fuzzy recursive operator, and (ii) the formulation of a recursive equation modeling the dynamic behavior of the ON potential. The procedure achieved an experimentally validated forecasting accuracy of 83 %. This multidisciplinary approach not only improves prediction and operational resilience but also reframes the role of artificial intelligence in high-criticality industrial environments. The proposed system lays a robust foundation for future developments in prescriptive maintenance, extreme-scenario simulation, and technology transfer to other critical infrastructures.