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.

Vibrations carry a wealth of useful physical information in various fields. Identifying the multi-source vibration information generally requires a large number of sensors and complex hardware. Compressive sensing has been shown to be able to bypass the traditional sensing requirements by encoding spatial physical fields, but how to encode vibration information remains unexplored. Here we propose a randomized resonant metamaterial with randomly coupled local resonators for single-sensor compressed identification of elastic vibrations. The disordered effective masses of local resonators lead to highly uncorrelated vibration transmissions, and the spatial vibration information can thus be physically encoded. We demonstrate that the spatial vibration information can be reconstructed via a compressive sensing framework, and this metamaterial can be reconfigured while maintaining desirable performance. This randomized resonant metamaterial presents a new perspective for single-sensor vibration sensing via vibration transmission encoding, and potentially offers an approach to simpler sensing devices for many other physical information.

The fault-induced impulses with uneven amplitudes and durations are always accompanied with amplitude modulation and (or) frequency modulation, which leads to that the acquired vibration/acoustic signals for rotating machine fault diagnosis always present nonlinear and nonstationary properties. Such an effect affects precise fault detection, especially when the impulses are submerged in heavy background noise. To address this issue, a nonstationary weak signal detection strategy is proposed based on a time-delayed feedback stochastic resonance (TFSR) model. The TFSR is a long-memory system that can utilize historical information to enhance the signal periodicity in the feedback process, and such an effect is beneficial to periodic signal detection. By selecting the proper parameters including time delay, feedback intensity, and calculation step in the regime of TFSR, the weak signal, the noise, and the potential can be matched with each other to an extreme, and consequently a regular output waveform with low-noise interference can be obtained with the assistant of the distinct band-pass filtering effect. Simulation study and experimental verification are performed to evaluate the effectiveness and superiority of the proposed TFSR method in comparison with a traditional stochastic resonance (SR) method. The proposed method is suitable for detecting signals with strong nonlinear and nonstationary properties and (or) being subjected to heavy multiscale noise interference.