Intermittent process hierarchical optimization method based on twin automigration model

Abstract

A batch process hierarchical optimization method based on a twin auto-migration model comprises the following steps: in upper layer optimization, generating a modeling data set by using a DODE method, and establishing a global RSM; introducing MA to compensate the model, solving an optimization problem based on RSM to obtain a suboptimal solution satisfying convergence, and taking the suboptimal solution as an initial optimization point of lower-layer optimization; in the lower-layer optimization, establishing a local model near a suboptimal solution through PLS and SVR; in consideration of the similar but different characteristics of the two models, dynamically combining functions of the two models into a twin self-migration model through weights in an online learning process; the operation track is adjusted through inter-batch self-adjustment optimization according to the gradient information; meanwhile, judging whether the lower-layer optimization process has the capability of realizing an expected target or not, if not, returning to the upper layer and searching for a suboptimal solution for the lower layer again; and if yes, continuing lower-layer optimization until the target is met. According to the method, stable and efficient optimization of the intermittent process can be realized under the conditions of no similar process data and less data.

Description

technical field [0001] The invention belongs to the technical field of industrial production process operation optimization, and in particular relates to a layered optimization method for batch processes based on a twin self-migration model. Background technique [0002] At this stage, industrial technology is constantly developing. In order to meet the needs of industrial production such as small batches, high added value, and multiple varieties, batch processes have gradually attracted attention and are often used in many industries such as chemical industry, metal processing, and semiconductor integrated circuit production. In order to maximize the production efficiency of the batch process, it is necessary to have a deep understanding of the process mechanism and characteristics, so as to change the conditions that are unfavorable to the production efficiency in time, then it is necessary to establish an accurate mathematical model for analysis and optimization. At the s...

AI Technical Summary

Problems solved by technology

Faced with this situation, there may not be similar processes for some processes, so traditional transfer learning will not be applicable

[0005] Model quality is the prerequisite for accurate optimization of the batch process, but due to interference factors such as data quality and various modeling assumptions, the established mathematical model is different from the actual process, that is, the plant model mismatch phenomenon

Due to the existence of this phenomenon, the final optimal solution based on the mathematical model is not the optimal solution

Images