Methods and systems for detecting abnormal states in the synthesis reaction process of pharmaceutical intermediates

By dividing the drug intermediate synthesis reaction process into stages and decoupling features, and combining multi-scale dilated convolution and channel attention mechanisms, the problem of identifying abnormal states in the drug intermediate synthesis reaction process, which is difficult to identify in existing technologies, is solved, and higher-precision anomaly detection and early warning are achieved.

CN122135828AActive Publication Date: 2026-06-02SHANDONG INST FOR FOOD & DRUG CONTROL
View PDF 4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHANDONG INST FOR FOOD & DRUG CONTROL
Filing Date
2026-05-07
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing technologies struggle to distinguish between normal control and regulation and real abnormal reactions during the synthesis of drug intermediates. Furthermore, early, subtle anomalies involving multiple variables are difficult to detect in advance, leading to false alarms and low model training accuracy.

Method used

By dividing the reaction process into different stages, dynamic time warping and feature decoupling are performed using the coupling strength matrix. Combined with multi-scale dilated convolution and channel attention mechanisms, the collaborative relationship between variables is explicitly modeled. Anomaly detection is performed using stage-aware feature aggregation and global temporal aggregation methods.

Benefits of technology

It improves the ability to distinguish subtle anomalies, reduces interference from stage switching and process rhythm differences, and significantly enhances the accuracy of abnormal state identification and early warning capabilities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122135828A_ABST
    Figure CN122135828A_ABST
Patent Text Reader

Abstract

This invention relates to a method and system for detecting abnormal states in the synthetic reaction process of pharmaceutical intermediates, belonging to the field of abnormal state detection technology. It includes the following steps: collecting and labeling reaction process data to construct a dataset; dividing the process into stages based on stage coding, completing variable coupling modeling, dynamic time warping alignment, and feature decoupling within each stage to obtain a reconstructed feature matrix; extracting abnormal patterns across different time spans through temporal convolution and multi-scale dilated convolution, combining channel attention to enhance key features to obtain an encoded feature matrix, performing stage-aware feature aggregation and global temporal aggregation to obtain the final classification feature vector, inputting four independent stage classification heads to output the classification result; and jointly optimizing the training process using labeled smoothed cross-entropy loss and focus loss. This invention can eliminate batch process rhythm differences, distinguish between normal regulation and true anomalies, and accurately capture early, subtle anomalies.
Need to check novelty before this filing date? Find Prior Art