Complex electromechanical systems have a high probability of failure, which may lead to relatively severe loss in the event of actual malfunction. Moreover, current state monitoring equipment is not universally applicable. Therefore, this paper proposes a method for integrating visual and auditory information to monitor the operating state of an electromechanical device without the aid of built-in sensors. First, the Hilbert–Huang transform is used to decompose the audiovisual waveform in order to obtain the time–frequency information of the signal. Then, the signal characteristics are selected and correlation analysis is adopted to filter the features. Finally, the residual features are input into a generalized regression neural network to obtain the degradation model of the device. In the process of establishing the model, the optimal prediction results are selected by using stochastic resonance to improve the accuracy of the model. Moreover, to reduce the single prediction error, this paper proposes a method for correcting the prediction results by combining historical data. The effectiveness of the proposed method is experimentally verified using a turbojet engine, where the flame fluctuation and sound are specified as the audiovisual information.

复杂的机电系统发生故障的可能性很高，如果发生实际故障，可能会导致相对严重的损失。而且，当前状态监视设备不是普遍适用的。因此，本文提出了一种在不借助内置传感器的情况下集成视觉和听觉信息以监视机电设备的运行状态的方法。首先，使用希尔伯特-黄（Hilbert-Huang）变换分解视听波形，以获得信号的时频信息。然后，选择信号特征并进行相关分析以过滤特征。最后，将残差特征输入到广义回归神经网络中以获得设备的退化模型。在建立模型的过程中，通过使用随机共振选择最佳预测结果，以提高模型的准确性。此外，为减少单个预测误差，本文提出了一种结合历史数据对预测结果进行校正的方法。使用涡轮喷气发动机通过实验验证了该方法的有效性，其中将火焰波动和声音指定为视听信息。