Chemical production data anomaly discrimination method and system based on multilevel feature decomposition algorithm

Through the combination of multi-level feature decomposition and LSTM deep learning model, the accuracy and early warning reliability of chemical production data abnormal detection are solved, and the safety, reliable early warning and early risk detection of chemical production processes are achieved.

CN120372508APending Publication Date: 2025-07-25ANHUI UNIVERSITY OF TECHNOLOGY
0 Cites 0 Cited by

Patent Information

Application Number
CN202510451314.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-04-11
Publication Date
2025-07-25

AI Technical Summary

Technical Problem

The existing chemical production data abnormal detection and safety warning systems cannot fully consider the deviation of multiple different parameters, and it is difficult to accurately determine whether the equipment has potential risks, and it is impossible to issue effective warnings for different levels of risks, resulting in insufficient early warning reliability.

Method used

The multi-level feature decomposition algorithm is used to decompose chemical production data, combine with the LSTM deep learning model to predict, and judge the abnormal data by calculating the total anomaly evaluation value, and send corresponding early warning signals based on different degrees of deviation evaluation value.

Benefits of technology

It improves the accuracy and real-timeness of abnormal judgment of chemical production data, can promptly detect potential risks, achieve safe and reliable early warning, and improves the stability and efficiency of the production process.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a chemical production data anomaly discrimination method and system based on a multilevel feature decomposition algorithm, and belongs to the technical field of chemical production safety management. The method comprises the steps of data acquisition and preprocessing; performing feature decomposition on the preprocessed time sequence data Yt by adopting a multi-level feature decomposition algorithm SFDL, and decomposing the time sequence data Yt into a seasonal component St, a trend component Tt and a residual component Rt; further performing multi-level decomposition on the residual component Rt through a multi-level regression model; predicting the current production data based on an LSTM deep learning model to obtain a predicted value Xs of the production data; and according to the qualified value X, the measured value Xi and the predicted value Xs of the production data, calculating a total abnormal evaluation value D of the parameters so as to judge the abnormal condition of the chemical production data. According to the method, the multi-level characteristic decomposition and the deep learning model are combined, so that the accuracy and the real-time performance of abnormal judgment of the chemical production data can be effectively improved, and safe and reliable early warning of the chemical production process is realized.
Need to check novelty before this filing date? Find Prior Art