Intelligent optimization method and system for preparation process of traditional chinese medicine extract based on reinforcement learning

What is AI technical title?
AI technical title is built by Patsnap AI team. It summarizes the technical point description of the patent document.
By generating virtual samples within the safety boundaries of the traditional Chinese medicine extract preparation process and fusing them with historical data for training, and combining the constraints and differences of the real production environment for debugging, the problem of cross-task knowledge reuse between virtual pre-training and real production was solved, and efficient optimization and safe learning of the traditional Chinese medicine extract preparation process were achieved.

CN122290741APending Publication Date: 2026-06-26CHONGQING THREE GORGES VOCATIONAL COLLEGE

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: CHONGQING THREE GORGES VOCATIONAL COLLEGE
Filing Date: 2026-02-06
Publication Date: 2026-06-26

AI Technical Summary

Technical Problem

In existing technologies, the cross-task knowledge reuse of traditional Chinese medicine extract preparation processes based on reinforcement learning suffers from low efficiency and security risks, and it is difficult to achieve efficient optimization.

Method used

Virtual state-action samples are generated within the safety boundary of the preparation process of traditional Chinese medicine extracts. These samples are then fused with historical production data to form an initial training set. The reinforcement learning agent is pre-trained offline and then connected to a real production environment. The action output is constrained within the limits of the equipment and the quality standards. The data distribution difference is determined by the maximum mean difference. This allows for strategy debugging and knowledge base storage, enabling rapid reuse of knowledge across tasks.

Benefits of technology

It significantly improves the security and efficiency of online learning, shortens the start-up cycle for process optimization of new medicinal materials or new optimization targets, and achieves efficient optimization of the preparation process of traditional Chinese medicine extracts.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122290741A_ABST

Patent Text Reader

Abstract

This application provides a method and system for intelligent optimization of the preparation process of traditional Chinese medicine extracts based on reinforcement learning. An initial virtual training set is constructed, and a reinforcement learning agent collects data across multiple consecutive production batches to obtain a real interaction dataset. The data distribution difference between the virtual training set and the real interaction dataset is determined. Based on the data distribution difference and a distribution difference threshold, the policy network of the reinforcement learning agent is adjusted. The parameters of the policy network and the corresponding optimization objectives of the preparation process of the traditional Chinese medicine extract are quantified into description vectors. The description deviation between the new description vector of the traditional Chinese medicine extract and all description vectors in the policy knowledge base is determined. Then, based on the initial batch data of the new traditional Chinese medicine extract, the historical policy with the smallest description deviation is fine-tuned to initiate a new round of optimization of the preparation process of the traditional Chinese medicine extract. Using the scheme of this application, intelligent optimization of the preparation process of traditional Chinese medicine extracts can achieve rapid reuse of knowledge across tasks.

Need to check novelty before this filing date? Find Prior Art