A digital twin-based multi-objective optimization method for technical schemes in process industry

To solve problems of technical objective conflicts, untimely feedback, and inaccurate decision making, a multi-objective optimization method was developed for technical schemes in process industry based on digital twin. According to actual requirements, a multi-objective optimization problem was established with key technical parameters and constraints determined. A digital twin model was built to achieve a real-time interaction between physical and virtual production systems and assist in optimization. For optimization efficiency and adaptability to different working conditions, a dynamic acquisition function-based Bayesian method was adopted to fit original complicated functions and obtain Pareto optimal solutions. A rapid decision-making was carried out by comprehensively utilizing dynamic parameter set correlation evaluation, entropy weight method, and technique for order preference by similarity to ideal solution. A typical case study of optical fiber preform drawing process was presented to prove the method effectiveness. Results demonstrated that the proposed method was valuable for actual application with better product quality (an improvement by 0.71% for preform rod diameter precision control), higher utilization rate of raw materials (an increase by 3.64%), and lower total energy consumption (a reduction by 4.37%).