Abstract Currently, the detection and identification of surface defects in polyimide foam products mainly rely on on‐site work experience, which has issues such as low detection accuracy, strong subjectivity, and low efficiency. Existing research on foam product defect detection primarily targets internal defects, lacking studies on the detection, identification, and classification of surface defects. Therefore, this article proposes a method for identifying and classifying surface defects in polyimide foam based on an improved GoogLeNet, aiming to quickly and accurately detect and identify surface defects in foam products. By optimizing the Inception blocks, introducing the ECA attention mechanism, and adding an LSTM network module, the model's recognition accuracy and generalization ability are effectively improved. In experiments, the model proposed in this article performed excellently on the foam surface defect dataset, showing a significant advantage in detection accuracy compared to other convolutional neural network models. The detection accuracy for pits and cracks reached 98.24% and 98.25%, respectively, providing a reliable reference for the detection of surface defects in industrial foam production.

Accurate first break picking is essential for tunnel microseismic monitoring. Here, we propose a self-supervised convolutional clustering picking (SCCP) method for automatically picking the first break of microseismic recordings. The time–frequency features are decomposed and reconstructed using accurate convolutional encoding and decoding under self-supervision. Then, the autoencoder output is unsupervisedly clustered into useful and invalid waveform sections employing the fuzzy$c$-means (FCMs) algorithm under long short-term memories, global attention, and self-attention constraints. Furthermore, the first point of the useful waveform is determined as the first break. Our results demonstrate that the proposed SCCP method outperforms the short-term average/long-term average (STA/LTA) and Akaike information criterion (AIC). Compared with PhaseNet, a supervised deep-learning method, the SCCP, produces similar performance without using human-labeled data. Practically, when the signal-to-noise ratio (SNR) is reduced to −6 dB, the average mean absolute error and standard deviation of the picking results remain at 1.12 and 9.19 ms, respectively.

Improving the quality of microseismic recordings is a critical step in the microseismic data processing. We introduce a wavelet-weighted online dictionary learning (WWODL) denoising strategy for noisy microseismic recordings. We develop an adaptive parameters estimation approach for tunable$Q$-factor wavelet transform (TQWT), which provides accurate periodic and nonstationary subband information from microseismic data for online dictionary learning (ODL). A sliding time window is employed to divide the obtained subbands into a series of patches of equal length, which are then assembled into a matrix and fed into the ODL. The subband kurtosis information is weighted to the constraint function of the ODL, further enhancing the sparse coding ability for each subband. Shortening the oscillation duration, an improved ODL is developed with a faster convergence speed in calculating sparse coefficients. Our results confirm that the WWODL can suppress high-frequency, low-frequency, and shared-bandwidth noises and has a minimal impact on the first arrival. The time consumption and the signal-to-noise ratio (SNR) of the WWODL are on average 1/8.675 and 56.82% higher than ODL, respectively.

Distinguishing useful microseismic signals is a critical step in microseismic monitoring. Here, we present the time series contrastive clustering (TSCC) method, an end-to-end unsupervised model for clustering microseismic signals that uses a contrastive learning network and a centroidal-based clustering model. The TSCC framework consists of two successive phases: pretraining and fine-tuning. In the pretraining phase, two random cropping augmentations are used to transform the time series microseismic data into two distinct but correlated views. Then, the multiscale temporal and instance contrasting learning are used to discriminate between negative and positive views, thus motivating the encoder to capture microseismic signal contextual information from multiple perspectives and generate distinct representations from unlabeled data. During the fine-tuning phase, the encoder weights are iteratively fine-tuned by simultaneously performing contrast learning and clustering. The corresponding loss is a weighted combination of the contrastive and clustering loss functions, which induces the encoder to learn representations that improve the clustering performance. The test results demonstrate that the proposed method can achieve better clustering accuracy (ACC) than popular clustering methods, including$k$-means, deep embedding clustering (DEC), unsupervised clustering with deep convolutional autoencoders (DCAs), and deep clustering with self-supervision (DCSS). Moreover, the TSCC model can produce results comparable to supervised deep learning approaches while requiring no labeled data, manual feature extraction, or large training datasets. In practice, the TSCC model has a clustering ACC of 98.07% and a normalized mutual information (NMI) of 86.26%.

One-way-broadcast-based flooding time synchronization algorithms are commonly used in wireless-sensor networks (WSNs). However, the packet delay and clock drift pose a challenge to accuracy, as they entail serious by-hop error accumulation problems in the WSNs. To overcome this, a rapid-flooding multibroadcast time synchronization with real-time delay compensation (RDC-RMTS) is proposed in this article. By using a rapid-flooding protocol, flooding latency of the referenced time information is significantly reduced in the RDC-RMTS. In addition, a new joint clock skew-offset maximum-likelihood estimation (MLE) is developed to obtain the accurate clock parameter estimations and the real-time packet delay estimation. Moreover, an innovative implementation of the RDC-RMTS is designed with an adaptive clock offset estimation. The experimental results indicate that the RDC-RMTS can easily reduce the variable delay and significantly slow the growth of by-hop error accumulation. Thus, the proposed RDC-RMTS can achieve accurate time synchronization in large-scale complex WSNs.

Identifying useful microseismic events is one of the key steps in monitoring tunnel rockbursts. Here, we propose a strong noise-tolerance deep learning (SNTDL) network for the automatic classification of noisy microseismic events. The training set, validation set, and test set of the SNTDL network consist of 27,989 unfiltered microseismic recordings. First, to comprehensively characterize the microseismic events, we extract 10 weakly correlated features of the microseismic recordings as the input of the SNTDL network. Then, the skip connection and concatenation structure are added to this network, which can further enhances its generalization ability. Additionally, the SNTDL, AlexNet, Inception, Visual Geometry Group, and ResNet are compared using the synthetic microseismic recordings with different signal-noise ratios. The results demonstrate that the SNTDL network has a higher accuracy and stronger noise-tolerance capability than the other approaches. Application to a dataset collected from a different construction environment confirms that the SNTDL network can still achieve an accurate classification result, which further verifies that the proposed network has a reliable generalization performance.

We proposed a novel wavelet threshold denoising method based on the discrete wavelet transform for noisy microseismic recordings. This algorithm can simultaneously suppress both the high- and low-frequency noises of the microseismic recordings and further highlight the first break of the noisy microseismic recordings. First, we design an adaptive threshold calculation method based on all the wavelet coefficients in different decomposition levels. Then, a continuous and derivable thresholding function is constructed. By adjusting the preset shape adjustment parameters, a more reasonable thresholding function is determined to achieve the denoising processing for highlighting the first break of the microseismic recordings. Finally, different comparison tests are performed to evaluate the performance of the developed method, and the results indicate that the new method can achieve a more effective denoising result. Application on the real microseismic recordings further confirms that the proposed method can make the corresponding first break more clear.

Average consensus theory is intensely popular for building time synchronization in wireless sensor network (WSN). However, the average consensus-based time synchronization algorithm is based on the iteration that poses challenges for efficiency, as they entail high communication cost and long convergence time in large-scale WSN. Based on the suggestion that the greater the algebraic connectivity the faster the convergence, a novel multihop average consensus time synchronization (MACTS) is developed with innovative implementation in this article. By employing multihop communication model, it shows that virtual communication links among multihop nodes are generated and algebraic connectivity of the network increases. Meanwhile, a multihop controller is developed to balance the convergence time, accuracy, and communication complexity. Moreover, the accurate relative clock offset estimation is yielded by delay compensation. Implementing the MACTS based on the popular one-way broadcast model and taking multihop over short distances, we achieve hundreds of times the MACTS convergence rate compared to average TimeSync (ATS).

Accurate and fast-convergent time synchronization is very important for wireless sensor networks. The flooding time synchronization converges fast, but its transmission delay and by-hop error accumulation seriously reduce the synchronization accuracy. In this article, a rapid-flooding multiple one-way broadcast time-synchronization (RMTS) protocol for large-scale wireless sensor networks is proposed. To minimize the by-hop error accumulation, the RMTS uses maximum likelihood estimations for clock skew estimation and clock offset estimation, and quickly shares the estimations among the networks. As a result, the synchronization error resulting from delays is greatly reduced, while faster convergence and higher-accuracy synchronization is achieved. Extensive experimental results demonstrate that, even over 24-hops networks, the RMTS is able to build accurate synchronization at the third synchronization period, and moreover, the by-hop error accumulation is slower when the network diameter increases.