Search

A framework for the detection and identification of machine faults through vibration measurements and higher order statistics (HOS) analysis is presented. As traditional signal processing techniques are based on the nonparametric magnitude analysis of vibration signals, in this paper, two different state-of-the-art HOS-based methods, namely, a nonparametric phase-analysis approach and a parametric linear or nonlinear modeling approach are used for machine fault diagnostic analysis. The focus of this paper is on the application of the techniques, not on the underlying theories. Each technique is described briefly and is accompanied by an experimental discussion on how it can be applied to classify the synthetic mechanical and electrical faults of induction machines compared with their normality. Promising results were obtained which show that the presented methodologies are possible approaches to perform effective preventive maintenance in rotating machinery. Index Terms--ARMA models, blind identification, higher order statistics, machine fault detection, quadratic models, third-order cumulants, third-order spectra.

Chaotic modulation is an important spread spectrum (SS) technique amongst chaotic communications. The logistic chaotic signal acts as the modulation signal in this paper. An adaptive demodulator based on the radial basis function (RBF) neural network is proposed. The demodulator makes use of the good approximate capacity of RBF network for a nonlinear dynamical system. Using the proposed adaptive learning algorithm, the source message can be recovered from the received SS signal. The recovering procedure is on line and adaptive. The simulated examples are included to demonstrate the new method. For the purpose of comparison, the extended-Kalman-filter-based (EKF) demodulator was also performed. The results indicate that the mean square error (MSE) of the recovered source signal by the proposed demodulator is significantly reduced, especially for the SS signal with a higher signal-to-noise ratio (SNR). Index Terms--Adaptive demodulator, chaos, RBF neural network, spread spectrum.

A novel approach to blindly estimate kernels of any discrete- and finite-extent quadratic models in higher order cumulants domain based on artificial neural networks is proposed in this paper. The input signal is assumed an unobservable independently identically, distributed random sequence which is viable for engineering practice. Because of the properties of the third-order cumulant functions, identifiability of the nonlinear model holds, even when the model output measurement is corrupted by a Gaussian random disturbance. The proposed approach enables a nonlinear relationship between model kernels and model output cumulants to be established by means of neural networks. The approximation ability of the neural network with the weights-decoupled extended Kalman filter training algorithm is then used to estimate the model parameters. Theoretical statements and simulation examples together with practical application to the train vibration signals modeling corroborate that the developed methodology is capable of providing a very promising way to identify truncated Volterra models blindly. Index Terms--Blind identification, neural networks, quadratic kernels, quadratic nonlinear models, third-order cumulants, weights decoupled extended Kalman filter training algorithm.

This paper proves that any finite time trajectory of a given n-dimensional dynamical continuous system with input can be approximated by the internal state of the output units of a continuous-time recurrent neural network (RNN). The proof is based on the idea of embedding the n-dimensional dynamical system into a higher dimensional one. As a result, we are able to confirm that any continuous dynamical system can be modeled by an RNN. Index Terms--Approximation, continuous-time recurrent neural networks, dynamical system.

In this paper, an orthogonal wavelet-based neural network (OWNN) is proposed. In the proposed OWNN both the orthogonal scaling functions and the corresponding mother wavelets are combined as the nonlinear activation function. The OWNN is applied to identify a Wiener-type cascade dynamical model. A linear autoregressive moving average (ARMA) model is used as the dynamic subsystems and the OWNN is employed as the nonlinear static subsystem. A Wiener model identification algorithm is formed by combining the proposed OWNN with the conventional least squares method. Index Terms--Neural networks, orthogonal wavelets, Wiener model.

In this paper, a real-time iterative learning control (ILC) approach for a nonlinear continuous-time system using recurrent neural networks (RNN's) with time-varying weights is presented. Two RNN's are utilized in the ILC system. One is used to approximate the nonlinear system and another is used to mimic the desired system response. The ILC rule is obtained by combining the two RNN's to form a neural network control system. Also, a kind of iterative RNN's training algorithm is developed based on the two-dimensional (2-D) system theory. An RNN using the proposed 2-D training algorithm is able to approximate any trajectory to a very high degree of accuracy. Simulation results show that the proposed ILC approach is very efficient. The newly developed 2-D RNN's training algorithms provides a new dimension to the application of RNN's in a nonlinear continuous-time system. Index Terms--Approximation, continuous-time iterative learning control, real-time training algorithm, recurrent neural networks, two-dimensional system.

A new approach for measuring the shape and surface of an object observed from a single camera is proposed. The proposed approach is based on using the neural networks as a parameteric representation of the three-dimensional object and the shape-from-shading problem is formulated as the minimization of an intensity error function with respect to the network weights. Experimental results demonstrate that our proposed methodology exhibits high efficiency and accuracy for measuring and inspecting the product's surface in the manufacturing industry. Index Terms--Heuristic global learning algorithm, neural networks, shape from shading.

A novel recursive algorithm for identifying orders and parameters of ARMA models driven by a sequence of non-Gaussian random signals is investigated. The input sequence is assumed to be unobservable and the conditions are based on properties of the model output cumulants of the third order. In every cycle of updating the model order, the proposed algorithm minimizes a quadratic cost function to determine the parameters. The novelty of the approach is that the model orders and parameters are all estimated without a priori knowledge; the system is blind. The identification process is said to be total because the model parameters together with the model order are estimated in the same process. Owing to its order-recursive nature, the proposed algorithm requires little computational complexity and exhibits fast convergence behavior. Simulation results verify that Gaussian noises present at the output do not have noticeable effects on the identifiability and the accuracy of estimation. Index Terms--Autoregressive moving average models, blind and total identification, convergence, order recursion, third-order cumulants.

In this paper, we present a neural-based reflectance model of which the physical parameters of the reflectivity under different lighting conditions are interpreted by the network weights. The idea of our method is to optimize a proper reflectance model by an effective learning algorithm and to recover the object surface by a simple shape from shading recursive algorithm with this resulting model. Experimental results, including synthetic and real images, were performed to demonstrate the performance of the proposed method for practical applications. Index Terms--Heuristic learning algorithm, neural-based reflectance model, shape from shading.

Health monitoring of machines and electronic devices as well as their manufacturing processes plays a significant role in improving their safety, reliability, and their effective lifetime. The Mahalanobis distance (MD)-based approach is a powerful health monitoring method that has recently received more interest from industrial and academic communities. It has been extensively studied and widely applied in many fields since the early 2000s. This study briefly surveys recent developments and implementation issues of the MD-based approach in system health monitoring. It provides comprehensive references for researchers, highlights relevant papers, and predicts future research trends. [ABSTRACT FROM AUTHOR]

Two textual metrics "Frequency Rank" (FR) and "Intimacy" are proposed in this paper to measure the word using and collocation characteristics which are two important aspects of text style. The FR, derived from the local index numbers of terms in a sentences ordered by the global frequency of terms, provides single-term-level information. The Intimacy models relationship between a word and others, i.e. the closeness a term is to other terms in the same sentence. Two textual features "Frequency Rank Ratio (FRR)" and "Overall Intimacy (OI)" for capturing language variation are derived by employing the two proposed textual metrics. Using the derived features, language variation among documents can be visualized in a text space. Three corpora consisting of documents of diverse topics, genres, regions, and dates of writing are designed and collected to evaluate the proposed algorithms. Extensive simulations are conducted to verify the feasibility and performance of our implementation. Both theoretical analyses based on entropy and the simulations demonstrate the feasibility of our method. We also show the proposed algorithm can be used for visualizing the closeness of several western languages. Variation of modern English over time is also recognizable when using our analysis method. Finally, our method is compared to conventional text classification implementations. The comparative results indicate our method outperforms the others. [ABSTRACT FROM AUTHOR]

Dementia is one of the most common neurological disorders among the elderly. Identifying those who are of high risk suffering dementia is important for early diagnosis in order to slow down the disease progression and help preserve some cognitive functions of the brain. To achieve accurate classification, significant amount of subject feature information are involved. Hence identification of demented subjects can be transformed into a pattern classification problem. In this letter, we introduce a graph based semi-supervised learning algorithm for Medical Diagnosis by using partly labeled samples and large amount of unlabeled samples. The new method is derived by a compact graph that can well grasp the manifold structure of medical data. Simulation results show that the proposed method can achieve better sensitivities and specificities compared with other state-of-art graph based semi-supervised learning methods. [ABSTRACT FROM AUTHOR]