【摘要】 齿轮箱是工业动力传输的核心组件，其故障诊断技术对传动系统运行可靠性、稳定性至关重要。多传感器数据融合技术在故障诊断中备受关注，但数据可靠性低和特征信息不全面问题严重影响齿轮箱健康状态诊断的准确性。为此，提出一种基于多传感器自适应融合的齿轮箱故障诊断方法。首先，通过一致性检验剔除异常数据，实现单传感器数据预处理；基于此，采用最小误差准则的自适应加权融合策略，有效融合多传感器数据；其次，通过自注意力机制自适应加权融合数据与单传感器预处理后数据，并利用卷积神经网络进行健康状态分类。通过齿轮箱故障数据案例验证，该方法在确保数据可靠性的同时，能充分利用多传感器数据中的故障信息，有效识别不同健康状态，使诊断精度显著提高，为齿轮箱健康管理和故障预测提供了有效的解决方案。更多还原

【Abstract】 Gearboxes are the core components for industrial power transmission, and their fault diagnosis technology is crucial for the reliability and stability of the transmission system operation. Multi-sensor data fusion technology has attracted much attention in fault diagnosis, but issues of low data reliability and incomplete feature information severely affect the accuracy of the gearbox health status diagnosis. Therefore, this paper proposed a fault diagnosis method for gearboxes based on adaptive fusion of multi-sensor data. First of all, abnormal data was eliminated through consistency testing to realize preprocessing of single-sensor data. On this basis, an adaptive weighted fusion strategy using the minimum error criterion was employed to effectively integrate the multi-sensor data. Secondly, the integrated data and pre-processed data from singlesensor was adaptively weighted based on self-attention mechanism, and the convolutional neural network was utilized to classify the health status of gearbox. This method was verified through a gearbox fault experimental dataset. This method ensures data reliability, and fully utilizes fault information from multi-sensor data to effectively and accurately identify and diagnose different health states. This work provides an effective solution for gearbox health management and fault prediction.更多还原