机械与电子2025，Vol.43Issue(7)：3-8,15,7.
基于注意力机制的多向传感器数据融合齿轮箱故障诊断方法
Fault Diagnosis Method for Gearbox Based on Attention Mechanism and Multi-directional Sensor Data Fusion
摘要
Abstract
To address the issue of reduced fault diagnosis accuracy in gearbox systems due to informa-tion loss from using a single sensor,this paper proposes a multi-sensor data fusion fault diagnosis method for gearboxes based on an attention mechanism.First,based on the sensitivity of vibration signals to ran-dom pulses,the comprehensive cliff entropy measure is used as a weighting standard to fuse vibration sig-nals collected from multiple sensors,achieving information complementarity and fusion.Then,a light-weight CA-GAPNet fault diagnosis model based on 2DCNN is constructed,and the fused signals are con-verted into GAF images as input to the model,ultimately achieving efficient fault diagnosis of the gearbox.Experimental results show that,compared with single-sensor and other commonly used algorithms,the proposed method demonstrates superior diagnostic performance in gearbox fault diagnosis.关键词
齿轮箱/故障诊断/格拉姆角场/多向传感器/注意力机制Key words
gearbox/fault diagnosis/Gramian angular summation/multi-sensor/attention mechanism分类
机械制造引用本文复制引用
段耀斌,王城宇,张国谋,王志峰,徐伯梁,万书亭..基于注意力机制的多向传感器数据融合齿轮箱故障诊断方法[J].机械与电子,2025,43(7):3-8,15,7.基金项目
国家自然科学基金资助项目(52275109) （52275109）
河北省自然科学基金资助项目(E2022502007) （E2022502007）
