mmWave radar-based sound sensing have attracted extensive attention due to its advantages such as simultaneous measurement of multiple sound sources, and strong noise resistance. However, limited by the mmWave reflection properties of the target sound source, existing methods can only measure targets with strong mmWave reflection characteristics at relatively short distance. This paper proposes a long-distance, source-independent sound sensing (LISS) method with mmWave radar. By designing an mmWave-acoustic passive relay, the incoming sound wave excitation is converted into modulation of the mmWave signal, enabling long-distance sound sensing without being affected by the shape or material of the sound source. Additionally, the unique structure of the relay provides directionality for sound source while maintaining omnidirectionality for incident mmWave. The LISS sensing method, the design principles, and the working mechanism of the relay are detailed in this paper. Experimental validations demonstrate that the proposed method extends sound sensing range by more than 7.5 times compared to exising methods and is not affected by the intrinsic properties of the sound source, providing a promising approach for long-range, source-independent sound sensing.

Speed fluctuations are a key issue that directly affects the motor fault diagnosis performance. Deep convolutional neural networks (DCNNet) can automatically perform fault feature extraction and classification, but the diagnostic capability of DCNNet may be reduced under speed fluctuation conditions. To address this shortcoming, this article proposes a novel method called deep nonlinear order-cyclic convolutional network (DNOCNet). First, the nonlinear order-cyclic spectrum analysis layer (NOSAL) is constructed to turn the collected signal into angular domain signals to resolve the speed fluctuations. Subsequently, a frequency domain signal-to-noise ratio-based feature fusion scheme is presented to enhance the modulation components for fault feature learning. Finally, the convolutional neural network is employed to extract fault features from NOSAL by layer-by-layer superposition, and the acquired features are fed into softmax classifier for motor fault identification. Experiments from two induction motor cases under speed fluctuation conditions are utilized to verify the effectiveness of DNOCNet.

Spectrum correlation (SC), as a typical demodulation algorithm, has been investigated for fault extraction by restraining interference components. However, SC ignores the uneven distribution of fault features in the entire frequency range, which makes the results vulnerable to interference components. To overcome these shortcomings, a novel robust modulation spectrum correlation (RMSC) gram is proposed. First, the signal is demodulated into a bispectral map display containing fundamental and modulation frequency through RMSC, and a finite impulse response filter on fundamental frequency is utilized to obtain RMSC subbands. Subsequently, the fault feature index of subbands under healthy and fault conditions is calculated, and its failure signature ratio (FSR) to generate RMSC gram is utilized. Finally, the RMSC with the maximum FSR is selected as an optimal subband, and envelope analysis is executed on the subband to obtain fault features. Simulations and experiments are performed to validate the effectiveness of RMSC in comparison with the state-of-the-art methods.

Convolutional neural network (CNN) has shown great potential in real-time gearbox monitoring. In practical engineering, due to the complex multitooth meshing motions and variable working conditions resulting in gearboxes with multiple excitation sources, and the response signals exhibit amplitude modulation and frequency modulation characteristics, which makes it difficult for CNN to obtain the fault features from complex modulated signals. To tackle these challenges, this article presents a new deep multiscale weighted modulation network (DMWMN) for gearbox fault detection under variable working conditions. First, a new frequency-domain modulation spectrum is proposed as signal processing layer in DMWMN to demodulate the modulation features from complex vibration signals. Thereafter, multibranched structure with different DMWMN slice scales is utilized to obtain fault features. The frequency domain signal-to-noise ratio-based weighted fusion method is employed to optimize the weighted coefficients in DMWMN to enhance the fault feature components. Finally, a CNN is further employed to learn features from the demodulated signals in the DMWMN layer to identify for fault classification. Experimental results prove that the DMWMN has advantages over state-of-the-art algorithms for gearbox fault feature identification under variable working conditions.

With the excellent capacity of feature representation and nonlinear mapping, deep learning with stacking deeply has aroused goer research interest in the field of intelligent fault diagnosis. However, under the case that mechanical failure signals, including gears, bearings, etc., essentially follow the excitation mechanism and modulation principle, an interpretable expression of deep learning architecture for intelligent diagnosis has been rarely discussed. Motivated by this issue, this study presents a novel interpretable multiplication convolution network (MCN), where three designed layers, including a feature separator, a feature extractor, and a classifier, are operated on spectrum samples input. Different from the conventional models, a series of multiplication filtering kernels (MFKs) are analytically designed to extract the differential modes from spectrum samples in an ex-ante interpretable way. The separated modes are stacked into a filtered mode map. A convolution layer is later used as the feature extractor to further abstract high-level feature representations. Finally, a dense decision layer is taken as the classifier for fault identification. Specially, to strengthen the sensing ability of MFKs, an anti-aliasing constraint is introduced to improve the information diversity of the separator. In essence, MCN operates in a novel framework collaborating signal processing with deep learning. Experimental results validate the effectiveness of the proposed MCN. Besides, feature map visualizations are further implemented to verify that the desired fault-sensitive modes in spectrum samples can be precisely mined, which provides the MCN with higher recognition accuracy and good ex-post interpretability. Benefiting from analytic kernel design, MCN has fewer model parameters as a lightweight efficient architecture, which shows enormous potential in the application of edge intelligent fault diagnosis. Related source codes can be available at: https://github.com/CQU-BITS/MCN-main.

Edge computing is an emerging paradigm that offloads the computations and analytics workloads onto the Internet of Things (IoT) edge devices to accelerate the computation efficiency, reduce the channel occupation of signal transmission, and reduce the storage and computation workloads on the cloud servers. These distinct merits make it a promising tool for IoT-based machine signal processing and fault diagnosis. This article reviews the edge computing methods in signal processing-based machine fault diagnosis from the aspects of concepts, state-of-the-art methods, case studies, and research prospects. In particular, the lightweight designed algorithms and application-specific hardware platforms of edge computing in the typical fault diagnosis procedures, including signal acquisition, signal preprocessing, feature extraction, and pattern recognition, are reviewed and discussed in detail. The review provides an insight into the edge computing framework, methods, and applications, so as to meet the requirements of IoT-based machine real-time signal processing, low-latency fault diagnosis, and high-efficient predictive maintenance.

Intelligent fault detection of rotating machines is essentially a pattern classification issue. At the same time, effectively obtaining fault features from the measured signals is a key step to timely diagnose the health status of rotating machinery and evaluate the results of fault classification. To accurately obtain effective fault information to enhance fault accuracy, this article proposes a novel fault detection scheme based on cyclic morphological modulation spectrum (CMMS) and hierarchical Teager permutation entropy (HTPE). In this scheme, first, CMMS was developed to analyze the measured signal to obtain a series of CMMS slices with different frequency bands, which solved the deficiencies of manual empirical selection of frequency band bandwidth in the traditional cyclic modulation spectrum. Subsequently, by integrating Teager energy operator into hierarchical permutation entropy (HPE), an improved feature selection method named hierarchical Teager permutation entropy (HTPE) is presented to obtain fault information of different frequency band slices, which can improve the fault feature extraction capability of HPE. Finally, the acquired HTPE-based vectors are integrated into the extreme learning machine classifier to achieve fault classification of rotating machinery under different working conditions. The proposed scheme is validated by experimental cases including cylindrical roller bearings and planetary gearboxes. The analysis results indicate that the proposed scheme not only can effectively obtain the fault features, but also accurately realize the classification and recognition of the fault mode. In addition, the proposed scheme can achieve higher detection accuracy than some existing algorithms.

Intelligent machine fault diagnosis technique has recently exploded interest in digital health, energy power, and industrial maintenance. Collecting machine fault data in engineering practice is usually costly, causing big challenges in the intelligent diagnosis of most fresh-from-the-factory machines that are missing fault data. Inspired by the main idea of the digital twin, in this article, we propose a dynamic model-embedded intelligent machine fault diagnosis framework. Specifically, a machine dynamic modeling and parameter identification approach is introduced for building the machine digital model. The digital model is used to predict machine fault data from the measured healthy vibration signals. Furthermore, a parameterized convolutional neural network structure is designed for learning optimization features and recognizing the machine's healthy state. Experimental investigation demonstrates the effectiveness of the framework. The results show that the proposed framework enables intelligent machine fault diagnosis without fault data. The diagnosis effect outperforms supervised, unsupervised, and small-sample learning approaches and can be generalized to another load and speed with acceptable accuracy.

The transient features caused by a local fault are of vital importance for bearing fault diagnosis in an intelligent industry. Due to the uncertainty of fault forms and nonstationarity of operating conditions, the fault feature distribution influenced by the physical dynamic response of actual defect is always complex and irregular with morphological differences. This will bring embarrassments for an accurate fault diagnosis. Motivated by this, a convolution sparse self-learning (CSSL) is proposed in this article to accomplish an adaptive feature enhancement. In the view of image sparse processing, the representation for desired morphological structures is promoted by a two-dimensional optimizing approach with manifold sensing. From a randomly selected fragment, the time-frequency manifold learning is first applied to mine the latent structures. The image entropy is then introduced to adaptively output the optimal one as a sensing kernel. Therewith, this kernel is used to operate a shift-invariant sparse analysis on raw time-frequency image. Combining this rebuilt image with the raw phase, an enhanced signal is finally synthesized. In this manner, the desired transient morphology can be automatically mined, which is consistent with the physical dynamic response. Practical defective bearing data are analyzed to illustrate the effectiveness of the proposed method. Specifically, a comparison further illustrates that the proposed CSSL is superior in the morphological transient features enhancement.

Digital signal processing algorithms are widely adopted in motor bearing fault diagnosis. However, most algorithms are developed on desktop platforms, and their focus is on the analysis of offline captured signals. In this paper, a simple and easily implemented algorithm running on an embedded system is proposed for the online fault diagnosis of motor bearing. The core part of the algorithm is a stochastic-resonance-based adaptive filter that realizes signal denoising and adaptation of the filter coefficient. Processed by the filter, the period of the purified signal is obtained, and then the fault type of the motor bearing is identified. The proposed method has distinct merits, such as low computational cost, online implementation, contactless measurement, and availability for various speed motors. This paper provides a simple, flexible, and effective solution for conducting motor bearing diagnosis on an embedded/portable device. The algorithm proposed is validated by a brushless dc motor and a brushed dc motor fabricating with defective/healthy support bearings.