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OA

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.