A deep-learning-based fault diagnosis model with the information of time domain or frequency domain as a low-level input can effectively weaken the interference of human factors and improve the application of artificial intelligence in the field of mechanical fault diagnosis.However,the length of time domain signal is difficult to determine,and the frequency domain signal is too long with low computation efficiency.To solve the problem,frequency domain signal envelope can be extracted to get the trend of frequency information.Sparse autoencoder can be combined to construct the fault diagnosis mode.Gearbox fault diagnosis experiments were performed.The results indicate that the proposed method can effectively speed up the computation process and decrease the memory consumption while keeping the effectiveness of fault diagnosis.%直接将时域或者频域作为低层输入信息构建深度学习故障诊断模型,可以有效的削弱人为因素的干扰,进一步提高人工智能在故障诊断领域的发展.然而,低层输入的时域信号长度难以划定,而频域信号的数据长度较大,导致模型的计算效率降低.针对该问题,提出预先对低层频域信号提取包络线,得到表征频域变化态势的信息成分,接着再与稀疏自编码结合构建稀疏自编码的故障诊断模型.齿轮箱故障诊断实验证明,与原始频域输入相比,所提方法能够在保证诊断效果的同时,降低计算复杂度和所需要的存储空间.
展开▼