RCS plate shock resistance prediction method based on improved CNN-BiLSTM neural network

Through the improved CNN-BiLSTM neural network combined with the lightweight attention network, the data shortage in the RCS board impact resistance study was solved, and an efficient and reliable prediction method was realized, which was suitable for safety assessment of important structures such as nuclear power plants.

CN120337725APending Publication Date: 2025-07-18CHINA NAT CHEM CONSTR INVESTMENT GRP CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510357519.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-25
Publication Date
2025-07-18

AI Technical Summary

Technical Problem

In the prior art, there is insufficient research on the impact resistance performance of RCS plates, there is a lack of comprehensive and effective test data and fast and reliable prediction methods, and the existing finite element model fails to accurately consider the interaction force between concrete and rear steel plate.

Method used

The data set was established through the impact-resistant numerical simulation calculation of the RCS board, and the improved CNN-BiLSTM neural network model was used to train and predict the failure mode of the RCS board under the impact. Combined with the lightweight attention network optimization feature extraction, the accuracy and reliability of the model were improved.

Benefits of technology

It realizes efficient and reliable impact resistance prediction of RCS plates, adapts to a variety of working conditions, has high efficiency and low cost, and is not restricted by site and personnel. The numerical analysis results are highly consistent with the test results, providing higher reliability and applicability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120337725A_ABST
    Figure CN120337725A_ABST
Patent Text Reader

Abstract

The invention discloses an RCS plate impact resistance prediction method based on an improved CNN-BiLSTM neural network, and the method comprises the steps: carrying out the simulation calculation of different concrete compressive strength, steel bar yield strength, steel bar reinforcement ratio, rear steel plate thickness, rear steel plate yield strength and RCS plate thickness corresponding failure modes of an RCS plate at different impact speeds through the impact resistance numerical simulation of the RCS plate; establishing a data set based on the RCS plate impact resistance numerical simulation result and historical data; and S2, training an improved CNN-BiLSTM neural network model through the data set established in S2, and predicting the failure mode of the RCS plate under the impact effect through the finally trained model. According to the method, the failure modes corresponding to the RCS plate under different working conditions are calculated through RCS plate impact resistance numerical simulation, the failure modes serve as a training database of the improved CNN-BiLSTM neural network, the impact resistance of the RCS plate is predicted, efficiency is high, and cost is low.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Cited By

  • Deep learning-based worst working condition identification method and system

    CN121658870A