The Artificial Intelligence and Predictive Analytics research line develops advanced AI-driven methodologies for context-aware system monitoring, failure prognostics, and real-time decision-making in complex dynamic systems. Building on two decades of advances in Prognostics and Health Management (PHM), this line integrates machine learning, Bayesian inference, multimodal data fusion, and reinforcement learning to enable predictive and proactive operation strategies across cyber-physical systems.
Our focus moves beyond traditional data analytics toward the development of computationally efficient, uncertainty-aware algorithms capable of fusing heterogeneous information sources (sensor signals, maintenance records, textual reports, images, and environmental data) into coherent latent representations. These representations support real-time estimation of system degradation, probabilistic characterization of event timing, and informed decision-making under operational and computational constraints.
Research:
- Context-Aware Multimodal AI: Development of advanced machine learning architectures capable of fusing heterogeneous data sources into unified latent representations that enhance degradation modeling and predictive accuracy.
- Real-Time Probabilistic Prognostics: Design of computationally efficient Bayesian algorithms for failure prediction under uncertainty, enabling real-time estimation of Time-of-Event (ToE) distributions while operating within industrial-scale computational constraints.
- Uncertainty-Aware Decision Intelligence: Development of rigorous Prognostic Decision-Making (PDM) frameworks using machine learning and probabilistic search strategies to optimize maintenance planning, mission reconfiguration, and operational control under uncertainty.
- Scalable Deployment in Complex Systems: Integration of predictive and decision-making modules into large-scale cyber-physical environments, ensuring interpretability, robustness, and practical deployability across mission-critical infrastructure.
Industrial and Technological Impact: The technology developed in this line enables predictive maintenance, mission-aware decision-making, and resilience enhancement in critical infrastructure and industrial systems. Applications span mining (e.g., mills, crushers, conveyor systems, heavy mobile equipment), energy systems, robotics, electromobility, and autonomous platforms. These solutions reduce operational costs, extend asset life cycles, mitigate catastrophic failures, and enhance safety in high-value industrial processes.
By embedding probabilistic reasoning and contextual awareness into AI-driven monitoring systems, this research line supports the transition toward Industry 5.0, where intelligent systems collaborate with human operators under transparent, interpretable, and uncertainty-aware frameworks. The resulting methodologies position AC3E as a national reference in safe, resilient, and computationally efficient AI for mission-critical infrastructure.
