| Abstract:The intelligent transformation of the paper industry involves solving dynamic and real-time problems in numerous high-dimensional mathematical models. Due to the nonlinearity, multidimensionality, and uncertainty inherent in paper production systems, the mathematical models describing these processes often consist of extensive sets of equations. Furthermore, frequent production fluctuations and transitions necessitate efficient and frequent resolution of these complex model sets to meet the demands of dynamic production optimization. Developing solvers tailored to the challenges of paper production models is key to addressing this issue. This study focuses on the characteristics of nonlinear, non-convex problems in paper production models and designs a global optimization solver for nonlinear, multidimensional paper production systems based on the trust-region interior-point method and TikTak multi-start optimization algorithm. The proposed solver achieves efficient resolution of complex production constraints and uncertain initial conditions. Results demonstrate that the solver achieved a 100% success rate in finding the global optimum in a paper drying section optimization case, with an average solving time of 0.81 seconds per instance, exhibiting high robustness. Additionally, in a paper energy system optimization case, the solver successfully reduced computational resource usage by 59.67% and computation time by 9.67%. |