Search

Deep neural networks have evolved significantly in the past decades and are now able to achieve better progression of sensor data. Nonetheless, most of the deep models verify the ruling maxim in deep learning—bigger is better—so they have very complex structures. As the models become more complex, the computational complexity and resource consumption of these deep models are increasing significantly, making them difficult to perform on resource-limited platforms, such as sensor platforms. In this paper, we observe that different layers often have different pruning requirements, and propose a differential evolutionary layer-wise weight pruning method. Firstly, the pruning sensitivity of each layer is analyzed, and then the network is compressed by iterating the weight pruning process. Unlike some other methods that deal with pruning ratio by greedy ways or statistical analysis, we establish an optimization model to find the optimal pruning sensitivity set for each layer. Differential evolution is an effective method based on population optimization which can be used to address this task. Furthermore, we adopt a strategy to recovery some of the removed connections to increase the capacity of the pruned model during the fine-tuning phase. The effectiveness of our method has been demonstrated in experimental studies. Our method compresses the number of weight parameters in LeNet-300-100, LeNet-5, AlexNet and VGG16 by 24×, 14×, 29× and 12×, respectively.

Domain adaptation aims to handle the distribution mismatch of training and testing data, which achieves dramatic progress in multi-sensor systems. Previous methods align the cross-domain distributions by some statistics, such as the means and variances. Despite their appeal, such methods often fail to model the discriminative structures existing within testing samples. In this paper, we present a sample-guided adaptive class prototype method, which consists of the no distribution matching strategy. Specifically, two adaptive measures are proposed. Firstly, the modified nearest class prototype is raised, which allows more diversity within same class, while keeping most of the class wise discrimination information. Secondly, we put forward an easy-to-hard testing scheme by taking into account the different difficulties in recognizing target samples. Easy samples are classified and selected to assist the prediction of hard samples. Extensive experiments verify the effectiveness of the proposed method.

Error factors in the encoded sun sensor (ESS) are analyzed and simulated. Based on the analysis results, an ESS error compensation model containing structural errors and fine-code algorithm errors is established, and the corresponding calibration method for model parameters is proposed. As external parameters, installation deviation between ESS and calibration equipment are introduced to the ESS calibration model, so that the model parameters can be calibrated accurately. The experimental results show that within plus/minus 60 degree of incident angle, the ESS measurement accuracy after compensation is three times higher on average than that before compensation.

Accelerated degradation testing (ADT) is an efficient technique for evaluating the lifetime of a highly reliable product whose underlying failure process may be traced by the degradation of the product’s performance parameters with time. However, most research on ADT mainly focuses on a single performance parameter. In reality, the performance of a modern product is usually characterized by multiple parameters, and the degradation paths are usually nonlinear. To address such problems, this paper develops a new s-dependent nonlinear ADT model for products with multiple performance parameters using a general Wiener process and copulas. The general Wiener process models the nonlinear ADT data, and the dependency among different degradation measures is analyzed using the copula method. An engineering case study on a tuner’s ADT data is conducted to demonstrate the effectiveness of the proposed method. The results illustrate that the proposed method is quite effective in estimating the lifetime of a product with s-dependent performance parameters.

Exchanging gradient is a widely used method in modern multinode machine learning system (e.g., distributed training, Federated Learning). Gradients and weights of model has been presumed to be safe to delivery. However, some studies have shown that gradient inversion technique can reconstruct the input images on the pixel level. In this study, we review the research work of data leakage by gradient inversion technique and categorize existing works into three groups: (i) Bias Attacks, (ii) Optimization-Based Attacks, and (iii) Linear Equation Solver Attacks. According to the characteristics of these algorithms, we propose one privacy attack system, i.e., Single-Sample Reconstruction Attack System (SSRAS). This system can carry out image reconstruction regardless of whether the label can be determined. It can extends gradient inversion attack from a fully connected layer with bias terms to attack a fully connected layer and convolutional neural network with or without bias terms. We also propose Improved R-GAP Alogrithm, which can utlize DLG algorithm to derive ground truth. Furthermore, we introduce Rank Analysis Index (RA-I) to measure the possible of whether the user’s raw image data can be reconstructed. This rank analysis derive virtual constraints Vi from weights. Compared with the most representative attack algorithms, this reconstruction attack system can recover a user’s private training image with high fidelity and attack success rate. Experimental results also show the superiority of the attack system over some other state-of-the-art attack algorithms.

Classification of faults in mechanical components using machine learning is a hot topic in the field of science and engineering. Generally, every real-world running mechanical system exhibits personalized vibration behaviors that can be measured with acceleration sensors. However, faulty samples of such systems are difficult to obtain. Therefore, machine learning methods, such as support vector machine (SVM), neural network (NNs), etc., fail to obtain agreeable fault detection results through smart sensors. A personalized diagnosis fault method is proposed to activate the smart sensor networks using finite element method (FEM) simulations. The method includes three steps. Firstly, the cosine similarity updated FEM models with faults are constructed to obtain simulation signals (fault samples). Secondly, every simulation signal is separated into sub-signals to solve the time-domain indexes to generate the faulty training samples. Finally, the measured signals of unknown samples (testing samples) are inserted into the trained SVM to classify faults. The personalized diagnosis method is applied to detect bearing faults of a public bearing dataset. The classification accuracy ratios of six types of faults are 90% and 92.5%, 87.5% and 87.5%, 85%, and 82.5%, respectively. It confirms that the present personalized diagnosis method is effectiveness to detect faults in the absence of fault samples.