Generalized bilinear Koopman realization from input-output data for multi-step prediction with metaheuristic optimization of lifting function and its application to real-world industrial system
A new AI modeling technique uses global optimization to automate the design of complex 'lifting functions' for industrial systems.
Researchers Shuichi Yahagi and team developed a new Generalized Bilinear Koopman Realization model. It uses a metaheuristic algorithm to automatically optimize Radial Basis Function (RBF) lifting functions, moving beyond manual design. The model, validated on a complex diesel engine airpath system, achieved significantly higher predictive accuracy than traditional linear Koopman models. This enables more reliable long-term, multi-step predictions for nonlinear industrial processes where full sensor data is unavailable.
Why It Matters
Automates complex AI model design for better predictions in manufacturing, energy, and automotive systems without needing full sensor coverage.