Xiwei Wang

In the realm of industrial production, the condition of motors tends to deteriorate gradually over time. Among the various types of motor faults, the selection of appropriate diagnostic techniques stands out as a pivotal step in the process of motor fault identification. In this article, we introduce a fault diagnosis framework aimed at enhancing the precision and efficiency of diagnostics by perceptually matching the time–frequency features of data with the sensory fields of convolutional neural networks (CNNs). First, the time–frequency swing door algorithm (TF-SDA) is proposed to determine the optimal signal compression ratio, guided by the time–frequency attributes of the original data. Subsequently, the data are transformed into a two-dimensional image using temporal image conversion methodologies, specifically the Gramian angular field and symmetrized dot pattern. Finally, the diagnostic framework is realized by matching the CNN receptive fields with TF-SDA data features. Comparative analysis with other state-of-the-art methods and compression algorithms reveals that the proposed approach effectively isolates the optimal time–frequency components of fault characteristics while eliminating extraneous elements, thus substantially enhancing the precision and efficiency of multifault diagnosis.

The fault prediction of motors can effectively reduce the occurrence of accidents and change the post diagnose to prevention. However, the systematic errors caused by the complex signal components, such as different kinds of randomly distributed noise, make the false report or misreport inevitable, no matter what prediction model is selected. To tackle this issue, a motor fault prediction method based on fault feature extraction and signal distribution optimization is proposed in this paper. A time-frequency parameter and resolution adaptive algorithm is proposed to optimize the raw signal while reserving the fault characteristics. An extended model is developed to decomposed and reconstruct the processed signals. Then, the optimized time series is transformed to the signal distribution. Fault prediction is carried out by combining the signal distributions as the inputs of the Gated Recurrent Unit. Two data sets collected from experiments and NASA are used to validate the effectiveness of the proposed methods. The test results indicate that the proposed methods provide better performance than other state-of-art models since the unrelated components of the signal are accurately reduced and the concentration of the signal is improved. From the predictive theory point of view, achieving accurate prediction of such signals is much easier. The method can accurately fulfil both long-term and short-term fault prediction tasks.

In industrial production, the speed of the motor is constantly changing due to production demands. However, traditional diagnosis methods cannot guarantee their performance under all circumstance, especially the sharp changing speed condition. To tackle this issue, this article proposes an adaptive method that can be used for multifaults diagnosis of motor under variety working condition, especially for the dramatic changing and unstable working states. First, a time–frequency (TF) parameter and resolution adaptive algorithm is proposed for signal preprocessing. Second, the well-processed signal is resampled by TF diagram and peak search method to eliminate the effect of sudden change in speed. A proposed perception matching algorithm based on the symmetrized dot pattern and convolutional neural networks is implemented for improving the accuracy. Finally, an adaptive model with unified diagnose process is developed to improve the accuracy, efficiency, and practicality of the diagnose model. Compared with other state-of-art methods, the results show the out-performance of the proposed method under both in steady and transient state, especially in the case that the speed is changing fast.