A novel approach for analog fault diagnosis based on neural networks and improved kernel PCA

We have developed a neural-network-based fault diagnosis approach of analog circuits using maximal class separability based kernel principal components analysis (MCSKPCA) as preprocessor. The proposed approach can detect and diagnose faulty components efficiently in the analog circuits by analyzing their time responses. First, using wavelet transform to preprocess the time responses obtains the essential and reduced candidate features of the corresponding response signals. Then, the second preprocessing by MCSKPCA further reduces the dimensionality of candidate features so as to obtain the optimal features with maximal class separability as inputs to the neural networks. This simplifies the architectures reasonably and reduces the computational burden of neural networks drastically. The simulation results show that our resulting diagnostic system can classify the faulty components of analog circuits under test effectively and achieves a competitive classification performance.     

Fault tolerant machine learning for nanoscale cognitive radio

We introduce a machine learning-based classifier that identifies free radio channels for cognitive radio. The architecture is designed for nanoscale implementation, under nanoscale implementation constraints; we do not describe all physical details but believe future physical implementation to be feasible. The system uses analog computation and consists of cyclostationary feature extraction and a radial basis function network for classification. We describe a model for nanoscale faults in the system, and simulate experimental performance and fault tolerance in recognizing WLAN signals, under different levels of noise and computational errors. The system performs well under expected non-ideal manufacturing and operating conditions.     

Application of extreme learning machine for series compensated transmission line protection

This paper proposes a new approach based on combined Wavelet Transform-Extreme Learning Machine (WT-ELM) technique for fault section identification (whether the fault is before or after the series capacitor as observed from the relay point), classification and location in a series compensated transmission line. This method uses the samples of fault currents for half cycle duration from the inception of fault. The features of fault currents are extracted by first level decomposition of the current samples using discrete wavelet transform (DWT) and the extracted features are applied as inputs to ELMs for fault section identification, classification and location. The feasibility of the proposed method has been tested on a 400 kV, 300 km series compensated transmission line for all the ten types of faults using MATLAB simulink. On testing 28,800 fault cases with varying fault resistance, fault inception angle, fault distance, load angle, percentage compensation level and source impedance, the performance of the proposed method has been found to be quite promising. The results also indicate that the proposed method is robust to wide variation in system and operating conditions.     

Analyzing the techniques that improve fault tolerance of aggregation trees in sensor networks

Sensor networks are finding significant applications in large scale distributed systems. One of the basic operations in sensor networks is in-network aggregation. Among the various approaches to in-network aggregation, such as gossip and tree, including the hash-based techniques, the tree-based approaches have better performance and energy-saving characteristics. However, sensor networks are highly prone to failures. Numerous techniques suggested in the literature to counteract the effect of failures have not been carefully analyzed. In this paper, we focus on the performance of these tree-based aggregation techniques in the presence of failures. First, we identify a fault model that captures the important failure traits of the system. Then, we analyze the correctness of simple tree aggregation with our fault model. We then use the same fault model to analyze the techniques that utilize redundant trees to improve the variance. The impact of techniques for maintaining the correctness under faults, such as rebuilding or locally fixing the tree, is then studied under the same fault model. We also do the cost-benefit analysis of using the hash-based schemes which are based on FM sketches. We conclude that these fault tolerance techniques for tree aggregation do not necessarily result in substantial improvement in fault tolerance.     

Active fault detection and control: Unified formulation and optimal design

A new unified formulation of the active fault detection and control problem for discrete-time stochastic systems and its optimal solution are proposed. The problem formulation stems from the optimal stochastic control problem and includes important special cases: an active detector and controller, an active detector and input signal generator, and an active detector with a given input signal generator. The optimal solution is derived using the so-called closed loop information processing strategy. This strategy respects the influence of the current decision and/or input on the future behavior of the observed system, allows penalizing future wrong decisions, and improves the quality of fault detection. The proposed formulation and obtained solution also provide better understanding of the active fault detection and its relation to the optimal stochastic control. The results are illustrated in numerical examples.     

Implementing uniform reliable broadcast with binary consensus in systems with fair-lossy links

When implementing multivalued consensus using binary consensus, previous algorithms assume the availability of uniform reliable broadcast, which is not implementable in systems with fair-lossy links. In this paper, we show that with binary consensus we can implement uniform reliable broadcast directly in systems with fair-lossy links, and thus the separate assumption of the availability of uniform reliable broadcast is not necessary. We further prove that, when binary consensus instances are available only as black boxes, any implementation of uniform reliable broadcast in the fair-lossy link model requires the invocation of an infinite number of binary consensus instances even if no process ever broadcasts any messages, and this is true even when multivalued consensus is used.     

Energy efficient redundant configurations for real-time parallel reliable servers

Modular redundancy and temporal redundancy are traditional techniques to increase system reliability. In addition to being used as temporal redundancy, with technology advancements, slack time in a system can also be used by energy management schemes to save energy. In this paper, we consider the combination of modular and temporal redundancy to achieve energy efficient reliable real-time service provided by multiple servers. We first propose an efficient adaptive parallel recovery scheme that appropriately processes service requests in parallel to increase the number of faults that can be tolerated and thus system reliability. Then we explore schemes to determine the optimal redundant configurations of the parallel servers to minimize system energy consumption for a given reliability goal or to maximize system reliability for a given energy budget. Our analysis results show that small requests, optimistic approaches, and parallel recovery favor lower levels of modular redundancy, while large requests, pessimistic approaches and restricted serial recovery favor higher levels of modular redundancy.

Effects of the number of hidden nodes used in a structured-based neural network on the reliability of image classification

A structured-based neural network (NN) with backpropagation through structure (BPTS) algorithm is conducted for image classification in organizing a large image database, which is a challenging problem under investigation. Many factors can affect the results of image classification. One of the most important factors is the architecture of a NN, which consists of input layer, hidden layer and output layer. In this study, only the numbers of nodes in hidden layer (hidden nodes) of a NN are considered. Other factors are kept unchanged. Two groups of experiments including 2,940 images in each group are used for the analysis. The assessment of the effects for the first group is carried out with features described by image intensities, and, the second group uses features described by wavelet coefficients. Experimental results demonstrate that the effects of the numbers of hidden nodes on the reliability of classification are significant and non-linear. When the number of hidden nodes is 17, the classification rate on training set is up to 95%, and arrives at 90% on the testing set. The results indicate that 17 is an appropriate choice for the number of hidden nodes for the image classification when a structured-based NN with BPTS algorithm is applied.     

Short-term fault prediction based on support vector machines with parameter optimization by evolution strategy

Support vector machines (SVMs) are the effective machine-learning methods based on the structural risk minimization (SRM) principle, which is an approach to minimize the upper bound risk functional related to the generalization performance. The parameter selection is an important factor that impacts the performance of SVMs. Evolution Strategy with Covariance Matrix Adaptation (CMA-ES) is an evolutionary optimization strategy, which is used to optimize the parameters of SVMs in this paper. Compared with the traditional SVMs, the optimal SVMs using CMA-ES have more accuracy in predicting the Lorenz signal. The industry case illustrates that the proposed method is very successfully in forecasting the short-term fault of large machinery.