摘要详情

ID / 提交时间

75 / 2025-05-14 19:02:28

标题

A CNN-BiLSTM-Based Evaluation Model for Steady-State Bearing Fault Diagnosis Method

关键字

CNN-BiLSTM; Rolling Bearings; Envelope Analysis; Fault Feature Coefficient; Evaluation Model.

主题及专题

6、Artificial Intelligence for EOM

状态

全文录用

作者

建文曾 / 北京航空航天大学

波平肖 / 北京航空航天大学

麟马 / 北京航空航天大学

乃超王 / 北京航空航天大学

燕燕靳 / 北京航空航天大学

摘要

The examination of rolling bearing fault characteristics is crucial for guaranteeing the state of rotating machinery. This paper presents an evaluation model based on CNN-BiLSTM to assess the effectiveness of fault frequency diagnostic methods under steady-state conditions. The model focuses on envelope spectrum analysis, It employs fault feature coefficients (FFC) to label data, The CNN extracts time-frequency features, while the BiLSTM learns sequence features, enabling a binary classification of the diagnosis success. The experimental outcomes means that the model have a high accuracy in effectiveness evaluation.Comparisons show CNN-BiLSTM outperforms CNN-LSTM in precision, recall and F1-score. despite longer training times. This research offers a reliable approach for assessing fault diagnosis methods, facilitating the choice of suitable diagnostic techniques for analyzing bearing vibration signals.