The research project aims to integrate Mean Field Games (MFG) theory, stochastic partial differential equations (SPDEs), and neural networks (NNs) to model complex industrial systems via the dynamics of interacting agent populations. The goal is to develop adaptive decision-making schemes for applications such as fault detection, predictive maintenance, logistics, and energy markets, where agents, representing system components, evolve according to stochastic dynamics influenced by mean field interactions, while neural networks update decision parameters in real time. The project follows a structured iterative cycle that combines data-driven learning, stochastic simulation, and optimization. Each iteration yields both technical and scientific outputs, facilitating technology transfer. The impact is quantitatively assessed through reference physical or industrial metrics, such as improved operational efficiency, reduced forecasting errors, and enhanced profitability in the relevant markets.
