Search

This paper presents a novel approach on motor current signature analysis (MCSA) for broken bar fault detection of induction motors (IMs), using as input the current signal measured from one of the three motor phases. Independent component analysis (ICA) is used over the Fourier-domain spectral signals obtained from the input and its autocorrelation function. The standard deviation of spectral components within a region of interest (ROI) of an ICA signal output was found to exhibit substantial differences between damaged and healthy motors. Separation of the ROI in one, two, and three sectors leads to an improved extraction of feature vectors, which are further fed into a neural network for classification purposes. The assessment of the proposed method is carried out through several experiments using two damage levels (broken bar and half broken bar) and two load motor conditions (50% and 75%), with a classification accuracy ranging from 90% to 99%. The contribution of this paper lies in a new technique of signal processing for ICA-based feature extraction in a 3-D feature space for IM fault diagnosis. [ABSTRACT FROM AUTHOR]

Current transformers (CTs) are extensively used for the measurement of electrical energy, and hence they play a crucial role in the revenue generation for power distribution utilities. The novelty of this paper is that it presents artificial neural network (ANN) application for the ratio and phase angle error corrections of CTs in the presence of harmonic distortions. In this paper, the behaviors of silicon iron alloy and nanocrystalline alloy CTs are examined by the means of a number of experimental tests under the sinusoidal and distorted waveform conditions. The relation between the nature of CTs secondary current waveforms (CTSwf) with their metrological properties, connected burdens, and nature of load currents (linear or nonlinear) is established. The training data are obtained by the nature of CTSwf in terms of amplitude errors, time shifts, and deviations caused by different harmonic contents between input and output quantities. The ANN is trained for various operating conditions. The implementation of the error correction system and data acquisition is done using dSPACE 1104 software. The experimental results illustrate that the proposed technique can be used for error compensation in case of sinusoidal and the distorted waveform composed of fundamental and many harmonics. [ABSTRACT FROM AUTHOR]

In this paper, the modeling of a thermodynamic cooling system is considered, with the emphasis on dealing with the time-varying delay. A detailed characterization of the system output is performed first in the frequency domain, which successfully yields information relating to the location in the physical structure, which causes the time-varying delay. Three approaches are compared for modeling this delay: 1) using an adaptive fractional delay; 2) using an adaptive tapped delay line; and 3) using a polynomial delay. In addition, it is shown that the use of a carefully designed perturbation signal with harmonic suppression is advantageous in this case, as the percentage output power contained in the nonexcited harmonics can serve as a measure of the relative variance of the delay since the system is found to be largely linear. [ABSTRACT FROM PUBLISHER]

With the development of the ubiquitous pedestrian position in environments where global navigation satellite system is unavailable or there is no preinstalled wireless device for positioning, the use of self-contained sensors becomes more important. Thus, this paper proposed a 3-D pedestrian position system combined with commercial microelectromechanical systems microinertial measurement unit (MIMU) and surface electromyography (SEMG) sensor. The signal of the SEMG is used mainly in two aspects: 1) to improve the recognition rate of motion pattern and 2) to obtain the more accurate 3-D velocity contrasted to MIMU. Therefore, satisfied 3-D position accuracy is obtained in two different terrain environments. [ABSTRACT FROM PUBLISHER]

The first and foremost process in tea manufacturing, withering, is the foundation for producing good quality. Moisture plays an important role in the manufacturing process of tea to get the desired quality. In this paper, a novel in situ instrumentation technique is proposed and validated experimentally for prediction of moisture loss (ML) in the withering process. In the proposed technique, ML is predicted based on the inlet and the outlet relative humidity (RH) and temperature during the process of withering. Network capable smart sensor nodes are developed for the measurement of RH and temperature at the inlet and outlet of the withering trough. Architecture of the nodes and network is described. A scaled-down prototype of an enclosed trough is developed to perform withering of tea leaves. Based on the data measured by the system, ML is predicted by using artificial neural network. Nonlinear autoregressive model with exogenous inputs is used for predicting the ML. The predicted ML is compared with the actual amount of ML measured by weight loss. A total of nine experiments are conducted for nine batches of tea leaves. The data collection, their analysis and results are reported in this paper. The observed result shows a good agreement between the predicted and actual ML. The maximum mean error in prediction is −3.6%. [ABSTRACT FROM PUBLISHER]

This paper presents a new algorithm for predicting the deviations for local timescale UTC(k) in relation to the coordinated universal time (UTC) scale by means of group method of data handling (GMDH) neural network (NN). A very important element of this algorithm significantly affecting the quality of the obtained prediction is the block of time series preparation for NN. On the basis of carried out research for NN training and predicting the deviations, three time series (TS1, TS3, and TS4) are used. The values of elements of first time series (TS1) are calculated based on data from a cesium atomic clock realizing the UTC(k) timescale and on the deviations for this timescale determined by the French for Bureau International des Poids et Mesures (BIPM) in relation to the UTC timescale. The new time series (TS3) is a complement of time series TS1 by the deviations of the UTC(k) timescale in relation to the UTC Rapid timescale. The values of elements of next new time series (TS4) are equal to the deviations for UTC(k) in relation to the UTC and UTC Rapid timescales published by the BIPM. Simulation research of the algorithm has been carried out on the example of the UTC(PL) scale. The best results of [UTC–UTC(PL)] deviation predictions, compared with the previously used method of linear regression, designated with the uncertainty of 8 ns are obtained for GMDH NN and time series TS3. [ABSTRACT FROM AUTHOR]

This paper proposes a fault detection and classification method for medium voltage DC (MVDC) shipboard power systems (SPSs) by integrating wavelet transform (WT) multiresolution analysis (MRA) technique with artificial neural networks (ANNs). The MVDC system under consideration for future all-electric ships presents a range of new challenges, in particular the fault detection and classification issues addressed in this paper. The WT-MRA and Parseval's theorem are employed in this paper to extract the features of different faults. The energy variation of the fault signals at different resolution levels are chosen as the feature vectors. As a result of analysis and comparisons, the Daubechies 10 (db10) wavelet and scale 9 are the chosen wavelet function and decomposition level. Then, ANN is adopted to automatically classify the fault types according to the extracted features. Different fault types, such as short circuit faults on both dc bus and ac side, as well as ground fault, are analyzed and tested to verify the effectiveness of the proposed method. These faults are simulated in real time with a digital simulator and the data are then initially analyzed with MATLAB. The case study is a notional MVDC SPS model, and promising classification accuracy can be obtained according to simulation results. Finally, the proposed fault detection algorithm is implemented and tested on a real-time platform, which enables it for future practical use. [ABSTRACT FROM PUBLISHER]

Electromyographic (EMG) signals are utilized to analyze the neuromuscular disorders. Machine learning algorithms have been employed as a decision support system to detect neuromuscular disorders. EMG signals contain noise from different sources, such as electrical and electronic instruments and movement artifacts. In this paper, the multiscale principal component analysis (MSPCA) has been used to remove the impulsive noise from the EMG signals. Then, the dual-tree complex wavelet transform (DT-CWT) is utilized for feature extraction, and the rotation forest ensemble classifier is employed for the recognition of EMG signals. In addition, the performance of several classifiers with rotation forest has been studied. An efficient combination of DT-CWT and rotation forest achieved good performance, using tenfold cross validation regarding the total classification accuracy. Results are promising and showed that the rotation forest achieved an accuracy of 99.7% with clinical EMG signals using support vector machine and 96.6% with simulated EMG signals using the artificial neural network (ANN). [ABSTRACT FROM AUTHOR]

Fault diagnosis is vital for modern industry, and an increasing number of intelligent methods have been proposed for the fault diagnosis. However, most of the studies focus on distinguishing different fault patterns while ignoring fault deterioration. In this paper, a new hierarchical convolutional neural network (HCNN) is proposed as the two-level hierarchical diagnosis network, and it has two characteristics: 1) the fault pattern and fault severity are modeled as one hierarchical structure and 2) the fault pattern and fault severity can be estimated at the same time. Based on these, a new structure of HCNN is designed, which has two classifiers. Then, a two-stage training method is developed for HCNN to train these two classifiers at once training. The proposed HCNN is conducted on three case studies and has achieved state-of-the-art results. The results show that HCNN outperforms traditional two-layer hierarchical fault diagnosis network, and other machine learning and deep learning methods. [ABSTRACT FROM AUTHOR]

Nonselective gas sensor array has different sensitivities to different chemicals in which each gas sensor will also produce different voltage signals when exposed to an analyte with different concentrations. Therefore, the characteristics of cross sensitivities and broad spectrum of nonselective chemical sensors promote the fast development of portable and low-cost electronic nose (E-nose). Simultaneous concentration estimation of multiple kinds of chemicals is always a challengeable task in E-nose. Multilayer perceptron (MLP) neural network, as one of the most popular pattern recognition algorithms in E-nose, has been studied further in this paper. Two structures of single multiple inputs multiple outputs (SMIMO) and multiple multiple inputs single output (MMISO)-based MLP with parameters optimization in neural network learning processing using eight computational intelligence optimization algorithms are presented in this paper for detection of six kinds of indoor air contaminants. Experiments prove that the performance in accuracy and convergence of MMISO structure-based MLP are much better than SMIMO structure in concentration estimation for more general use of E-nose. [ABSTRACT FROM PUBLISHER]

With respect to automotive safety, the driver plays a crucial role. Stress level, tiredness, and distraction of the driver are therefore of high interest. In this paper, a driver state detection system based on cellular neural networks (CNNs) to monitor the driver’s stress level is presented. We propose to include a capacitive-based wireless hand detection (position and touch) sensor for a steering wheel utilizing ink-jet printed sensor mats as an input sensor in order to improve the performance. A driving simulator platform providing a realistic virtual traffic environment is utilized to conduct a study with 22 participants for the evaluation of the proposed system. Each participant is driving in two different scenarios, each representing one of the two no-stress/stress driver states. A “threefold” cross validation is applied to evaluate our concept. The subject dependence is considered carefully by separating the training and testing data. Furthermore, the CNN approach is benchmarked against other state-of-the-art machine learning techniques. The results show a significant improvement combining sensor inputs from different driver inherent domains, giving a total related detection accuracy of 92%. Besides that, this paper shows that in case of including the capacitive hand detection sensor, the accuracy increases by 10%. These findings indicate that adding a subject-independent sensor, such as the proposed capacitive hand detection sensor, can significantly improve the detection performance. [ABSTRACT FROM AUTHOR]

This paper presents two algorithms, both based on the time of flight (ToF) of scattered waves, to locate defect in an anisotropic woven-fabric carbon fiber reinforced polymer (CFRP) plate. The first algorithm uses a probabilistic approach by constructing a probability matrix. Each element of this matrix is associated with a location and represents the probability of existing a defect at the corresponding spatial coordinates of the plate. For the probability matrix method, localization results are influenced by manually chosen parameters and by the anisotropy of the CFRP plate. The second algorithm, based on artificial neural networks (ANNs), enables us to improve the accuracy of locating defects. With this ANN method, the anisotropic feature can be “learned” by the neural network from training data. The ToF of scattered waves obtained from three sensor pairs were used directly as inputs of the neural network. The spatial coordinates of the defect are the ANN outputs. The difficulty of obtaining sufficient experimental data for ANN training was surpassed by using added blocks on the surface of the plate under test to simulate defects. This scheme was validated and proved to be effective by comparing the scattered waves from a delamination and from an added block. Conclusions have been drawn by comparing the two methods. The localization results obtained by the ANN algorithm are proved to be better. [ABSTRACT FROM AUTHOR]

In this paper, we present a decentralized load estimation approach to support real-time volt/var optimization in distribution networks with high penetration of distributed generation. The decentralized load estimation scheme relies on local information and on a limited amount of information from neighbor areas. We defined a procedure for measurement selection and feeder partitioning so as to limit communication needs and allow prioritizing the estimation accuracy for selected loads. The measurement selection and feeder partitioning methods are based on the mutual information and minimum-redundancy-maximum-relevance concepts. We tested the proposed approach with a modified radial IEEE 34-node test feeder and a weakly meshed IEEE 123-node test feeder. The effectiveness of the method is validated using a hardware-in-the-loop simulation platform. [ABSTRACT FROM AUTHOR]

This paper introduces a soft sensor (SS) for the estimation of the deflection of a polymeric mechanical actuator. The actuator is based on ionic polymer-metal composites (IPMCs). Applications of IPMCs have been proposed in fields such as robotics, surgery, and aerospace, to mention the most interesting ones. In such application fields, both the complexity and the size of the actuating system are of chief importance. An SS can be, therefore, preferred to hardware measuring the actuator output, for estimating the actuator motion. Also, low-order models are of interest to limit the computational load, which can be a constraint in real-time applications. To this aim, several data-driven nonlinear finite-impulse response (NFIR) models have been investigated. Data, used for the model identification, have been acquired, in controlled environmental conditions, by using swept signals as the input to the IPMC actuator. Linear and nonlinear models, based on principal component analysis, shallow, and deep neural networks (NNs), have been investigated, for different model orders. The best results have been obtained by an SS based on a fifth-order NFIR model, implemented by a deep belief NN. [ABSTRACT FROM AUTHOR]

A novel measurement approach for power-flow analysis in medium-voltage (MV) networks, based on load power measurements at low-voltage level in each secondary substation (SS) and only one voltage measurement at the MV level at primary substation busbars, was proposed by the authors in previous works. In this paper, the method is improved to cover the case of temporary unavailability of load power measurements in some SSs. In particular, a new load power estimation method based on artificial neural networks (ANNs) is proposed. The method uses historical data to train the ANNs and the real-time available measurements to obtain the load estimations. The load-flow algorithm is applied with the estimated load powers, and the MV network state variables are obtained. The proposed method is validated for the real MV distribution network of the island of Ustica. The loads of selected SSs are estimated for two full days of different seasons. In comparison with previous works, satisfactory results are obtained in terms of uncertainty in the calculated power flows, thus suggesting the applicability of the proposed method for real-time monitoring of MV distribution networks. [ABSTRACT FROM AUTHOR]

As an effective way to integrate the information contained in multiple medical images with different modalities, medical image fusion has emerged as a powerful technique in various clinical applications such as disease diagnosis and treatment planning. In this paper, a new multimodal medical image fusion method in nonsubsampled shearlet transform (NSST) domain is proposed. In the proposed method, the NSST decomposition is first performed on the source images to obtain their multiscale and multidirection representations. The high-frequency bands are fused by a parameter-adaptive pulse-coupled neural network (PA-PCNN) model, in which all the PCNN parameters can be adaptively estimated by the input band. The low-frequency bands are merged by a novel strategy that simultaneously addresses two crucial issues in medical image fusion, namely, energy preservation and detail extraction. Finally, the fused image is reconstructed by performing inverse NSST on the fused high-frequency and low-frequency bands. The effectiveness of the proposed method is verified by four different categories of medical image fusion problems [computed tomography (CT) and magnetic resonance (MR), MR-T1 and MR-T2, MR and positron emission tomography, and MR and single-photon emission CT] with more than 80 pairs of source images in total. Experimental results demonstrate that the proposed method can obtain more competitive performance in comparison to nine representative medical image fusion methods, leading to state-of-the-art results on both visual quality and objective assessment. [ABSTRACT FROM AUTHOR]

This paper proposes a sparse representation and fully connected neural network (SRFCNN) methodology for residential ac series arc fault identification. The SRFCNN method captures the features of idiographic signal using sparse coding and conducts intelligent feature learning and classification through neural network (NN). The methodology is elaborately structured by a pretreatment layer, a sparse representation layer, and a decision layer. The processing procedure includes the following main steps. A set of customized bases are first designed via dictionary learning to behave as precise feature descriptors rather than choosing a fixed dictionary. The pretreated input samples are accordingly transformed into a series of sparse coefficients and then fed into the NN for recognition. Experimental results indicate that the method works well in distinguishing the signature of the arcing state from that of normal state within a certain single load type, as well as recognizing the two states of various load types (resistive, inductive, capacitive, switching, etc.) all together. The general classification accuracy of every half-cycle signal in the 10 tested classes could reach 94.3% and be probably higher. The methodology’s good generality and reliability for arc fault identification show its potential in other unsupervised signal feature recognition and fault detection. [ABSTRACT FROM AUTHOR]

Iris-based biometric systems identify individuals by comparing the characteristics of the iris captured by suited sensors. When reflections are present in the iris image, the portion of the iris covered by the reflections should not be considered in the comparison since it may produce erroneous matches. This paper presents an adaptive design methodology for reflection detection and location in iris biometric images based on inductive classifiers, such as neural networks. In particular, this paper proposes a set of features that can be extracted and measured from the iris image and that can effectively be used to achieve an accurate identification of the reflection position using a trained classifier. In addition, the use of radial symmetry transform (RST) is presented to identify the reflections in iris images. The proposed design methodology is general and can be used in any biometric system based on iris images. [ABSTRACT FROM AUTHOR]

This paper proposes a simple and novel method of designing and developing high-linearity linear variable differential transformer (LVDT)-based displacement sensing systems. Conventionally, precise adjustment of windings is made to enhance the linearity range of LVDTs. The tedious job of pitch adjustment of windings of LVDTs can be overcome by using the proposed method. A functional link artificial neural network has been successfully used in this paper for nonlinear compensation of the LVDT. The effectiveness of the proposed method is demonstrated through computer simulation with the experimental data of two different LVDT. The complete algorithm with the practical setup for the development of a linear LVDT is presented in this paper. [ABSTRACT FROM AUTHOR]

This paper presents a new fault diagnosis method for analog circuits. The proposed method extracts the original signals from the output terminals of the circuits under test (CUTs) by a data acquisition board and finds the kurtoses and entropies of the signals, which are used to measure the high-order statistics of the signals. The entropies and kurtoses are then fed to a neural network as inputs for further fault classification. The proposed method can detect and identify faulty components in an analog circuit by analyzing its output signal with high accuracy and is suitable for nonlinear circuits. Preprocessing based on the kurtosis and entropy of signals for the neural network classifier simplifies the network architecture, reduces the training time, and improves the performance of the network. The results from our examples showed that the trochoid of the entropies and kurtoses is unique when the faulty component's value varies from zero to infinity; thus, we can correctly identify the faulty components when the responses do not overlap. Applying this method for three linear and nonlinear circuits, the average accuracy of the achieved fault recognition is more than 99%, although there are some overlapping data when tolerance is considered. Moreover, all the trochoids converge to one point when the faulty component is open-circuited, and thus, the method can classify not only soft faults but also hard faults. [ABSTRACT FROM AUTHOR]

In this paper, a virtual instrument for the estimation of octane number in the gasoline produced by refineries is introduced. The instrument was designed with the aim of replacing measuring hardware during maintenance operations. The virtual instrument is based on a nonlinear moving average model, implemented by using multilayer perceptron neural networks. Stacking approaches are adopted to improve the estimation performance of the instrument. Classical linear algorithms of model aggregation are compared in the paper with a nonlinear strategy, based on the neural combination of a set of first-level neural estimators. The validity of the proposed approach is verified by comparison with the performance of both linear and nonlinear modeling techniques. The designed virtual instrument has been implemented by a large refinery in Sicily, which supplied the data used during the design phase. [ABSTRACT FROM AUTHOR]

Inertial-navigation system (INS) and global position system (GPS) technologies have been widely applied in many positioning and navigation applications. INS determines the position and the attitude of a moving vehicle in real time by processing the measurements of three-axis gyroscopes and three-axis accelerometers mounted along three mutually orthogonal directions. GPS, on the other hand, provides the position and the velocity through the processing of the code and the carrier signals of at least four satellites. Each system has its own unique characteristics and limitations. Therefore, the integration of the two systems offers several advantages and overcomes each of their drawbacks. The integration of INS and GPS is usually implemented utilizing the Kalman filter, which represents one of the best solutions for INS/GPS integration. However, the Kalman filter performs adequately only under certain predefined dynamic models. Alternatively, this paper suggests an INS/GPS integration method based on artificial neural networks (ANN) to fuse uncompensated INS measurements and differential GPS (DGPS) measurements. The proposed method suggests two different architectures: the position update architecture (PUA) and the position and velocity PUA (PVUA). Both architectures were developed utilizing multilayer feed-forward neural networks with a conjugate gradient training algorithm. [ABSTRACT FROM AUTHOR]

This paper presents a new approach for compensating nonlinear errors of a load cell, based on the optimal neural network with an augmented Lagrange multiplier (ALMNN). The load cell has serious nonlinear errors, which lower the accuracy of the weighing results. In this proposed approach, first, we construct the constraints for training the neural network (NN) using the load cell’s prior knowledge, i.e., the monotonic increasing property of a load cell’s input-output function. Second, we create the augmented Lagrange function with a multiplier and a penalty factor to serve as an NN’s performance index, and then deduce the ALMNN’s detailed training algorithm. The penalty factor plays an important role in ALMNN, thus the reasonable range of numerical value is discussed in this paper. Meanwhile, the weighing principle of a strain gauge load cell is introduced and its nonlinear error model is established. Finally, the experimental results demonstrate the effectiveness of ALMNN. By comparing with the conventional data induction method (DINN, i.e., the training of NN not with the prior knowledge but only the data samples), ALMNN significantly improves the NN’s generalization ability, especially in the case of lacking training samples. In addition, the applicability of ALMNN is verified by three additional experiments. [ABSTRACT FROM PUBLISHER]

This paper introduces a new data-driven bottom-up monitoring approach for distribution systems. In this approach, local estimations of the subsections into which the system is split are performed independently, thus leading to a scalable architecture. The monitoring approach is focused only on the estimation of voltage magnitude rather than the complete state of the system. This reduces the measurement requirements significantly, thus addressing economical and technical concerns for existing systems, while staying open to accommodating further incremental improvements in the available data and data quality. The estimation of each section is realized via an artificial neural network (ANN), for which a set of parameterizations is available to cope with different operating conditions. The estimation convergence is achieved even with relatively few measurements, although accuracy varies depending on the available measurements. At the Medium Voltage (MV) level, where reconfiguration is common, a configuration identification unit chooses the right ANN, the one trained for the actual network configuration. The estimation process is computationally simple and can be executed on low-cost hardware, as demonstrated in this paper by the implementation on a BeagleBone Black board. To demonstrate the concept, a prototype and a laboratory setup have been developed. The experimental test results are presented both for an Low Voltage distribution system and an MV distribution system. [ABSTRACT FROM PUBLISHER]

Early fault detection technique is crucial to reduce the machine downtime and has high impact on a wide variety of industrial applications. However, early fault detection is still subject to the following challenges: 1) extracting features from incipient fault signals; 2) detecting anomalies with considering sequential data correlation; and 3) enhancing the reliability of fault alarm. In this paper, we introduce a novel deep-structured framework to solve the early fault detection problem. First, the system variation is measured with the deviation value generated by a current feature extraction model using deep neural network (DNN) and a distribution estimator based on the long short-term memory (LSTM) network. DNN has the ability of representing a complicated and intrinsic distribution for data, which is suitable for handling the early fault data masked by heavy noise, and LSTM is able to discover temporal dependencies in high-dimensional sequential data, which allows distribution estimator making use of previous context information as well as makes the distribution estimator more robust to warp along the time axis. Second, a circular indirect alarm assessment strategy is designed for collecting deviation values and confirming the fault appearance only when a specified confidence level is reached. Experimental results on the typical real-world bearing data sets demonstrate the effectiveness and the reliability of our model. [ABSTRACT FROM AUTHOR]

In this paper, a practical fault diagnosis algorithm is presented to realize multiple open-circuit fault diagnosis for photovoltaic (PV) inverters. By the feature analysis for the output currents of normal and fault states, a fault diagnosis algorithm is developed, which is composed of multistate data processing (MSDP) block, subsection fluctuation analysis (SSFA) block, and the artificial neural network (ANN) block. First, the MSDP block is used to distinguish the different feature data and to adopt the data processing scheme, which improves the smoothness of the main data, retains the main fault features, and removes the influence of load change. Second, the SSFA block is used to extract the data feature, which can be used to accurately distinguish the different states of any switch. Finally, ANN is used by combining with the proposed MSDP and SSFA to implement intelligent classification, in which the dependency and the number of thresholds can be reduced. Comparing with the existing fault classification algorithms, the proposed algorithm is simple and stable to realize the multiple switches fault diagnosis for PV inverters. It does not require additional hardware equipment, and it can reduce the complexity of design and realization. Finally, the effectiveness of the fault diagnosis algorithm is verified by the experimental results. [ABSTRACT FROM AUTHOR]

Considering various health conditions under varying operational conditions, the mining sensitive feature from the measured signals is still a great challenge for intelligent fault diagnosis of spindle bearings. This paper proposed a novel energy-fluctuated multiscale feature mining approach based on wavelet packet energy (WPE) image and deep convolutional network (ConvNet) for spindle bearing fault diagnosis. Different from the vector characteristics applied in intelligent diagnosis of spindle bearings, wavelet packet transform is first combined with phase space reconstruction to rebuild a 2-D WPE image of the frequency subspaces. This special image can reconstruct the local relationship of the WP nodes and hold the energy fluctuation of the measured signal. Then, the identifiable characteristics can be further learned by a special architecture of the deep ConvNet. Other than the traditional neural network architecture, to maintain the global and local information simultaneously, deep ConvNet combines the skipping layer with the last convolutional layer as the input of the multiscale layer. The comparisons of clustering distribution and classification accuracy with six other features show that the proposed feature mining approach is quite suitable for spindle bearing fault diagnosis with multiclass classification regardless of the load fluctuation. [ABSTRACT FROM AUTHOR]

Measuring solar irradiance allows for direct maximization of the efficiency in photovoltaic power plants. However, devices for solar irradiance sensing, such as pyranometers and pyrheliometers, are expensive and difficult to calibrate and thus seldom utilized in photovoltaic power plants. Indirect methods are instead implemented in order to maximize efficiency. This paper proposes a novel approach for solar irradiance measurement based on neural networks, which may, in turn, be used to maximize efficiency directly. An initial estimate suggests the cost of the sensor proposed herein may be price competitive with other inexpensive solutions available in the market, making the device a good candidate for large deployment in photovoltaic power plants. The proposed sensor is implemented through a photovoltaic cell, a temperature sensor, and a low-cost microcontroller. The use of a microcontroller allows for easy calibration, updates, and enhancement by simply adding code libraries. Furthermore, it can be interfaced via standard communication means with other control devices, integrated into control schemes, and remote-controlled through its embedded web server. The proposed approach is validated through experimental prototyping and compared against a commercial device. [ABSTRACT FROM PUBLISHER]

This paper presents a novel artificial neural network (ANN)-based near-field microwave nondestructive testing technique for defect detection and classification in nonceramic insulators (NCI). In this paper, distribution class 33-kV NCI samples with no defects, air voids in silicone rubber and fiber glass core, cracks in the fiberglass core, and small metallic inclusion between the fiber core and shank were inspected. The microwave inspection system uses an open-ended rectangular waveguide sensor operating in the near-field at a frequency of 24 GHz. A data acquisition system was used to record the measured data. ANN was trained to classify the different types of defects. The results showed that all defects were detected and classified correctly with high recognition rates. [ABSTRACT FROM AUTHOR]

This paper presents a new approach of compensating truck scale's weighing errors based on prior knowledge and neural network ensembles (PKNNEs). Truck scale is a typical nonlinear system and it is fussy and labor intensive to compensate the weighing errors with the conventional method, leading to low accuracy of the weighing results. The general idea of this proposed approach consists of building individual neural networks (NNs) and designing the constraint conditions for optimizing neural network ensembles (NNEs) with the prior knowledge of the truck scale. First, three uncorrelated individual NNs are created by using the step-distribution characteristics of the truck scale's permissible maximum weighing error. Second, the constraint conditions for training the individual NNs are constructed by using the ideal weighing model and its derivatives, which can significantly improve the generalization ability of NNs, especially when the training samples are few or lacking. The detailed design procedure of this proposed method is given, the weighing principle of truck scale is discussed, and its weighing error models are found in this paper. Experimental results demonstrate the effectiveness of this method, and the testing results of a truck scale with PKNNEs in the field show that it meets the requirement for the weighing accuracy of medium-class scale defined by OIML R76 “nonautomatic weighing instruments.” [ABSTRACT FROM AUTHOR]

Liquor characteristics of black cut, tear, and curl tea mostly depend on two biochemical components like theaflavin (TF) and thearubigin (TR). Evaluation of tea quality can be done efficiently by estimating the concentration of TF and TR without using biochemical tests as it takes much time, which requires laborious effort for sample preparation, storage, and measurement. Moreover, the required instruments for this test are very costly. In this paper, we have proposed an efficient method of TF and TR prediction in a given tea sample using electronic tongue (ET) signal. Combinations of transformed features, like discrete cosine transform, Stockwell transform (ST), and singular value decomposition, of ET signals are fused to develop regression models to predict the contents of TF, TR, and TR/TF. Three different regression models such as artificial neural network, vector-valued regularized kernel function approximation, and support vector regression are used to evaluate the performance of the proposed method. High prediction accuracy using fusion of features ensures the effectiveness of the proposed method for prediction of TF and TR using ET signals. [ABSTRACT FROM PUBLISHER]

The accurate detection and classification of power quality (PQ) disturbances in power systems is a key step to determine the causes of these events before any proper countermeasure could be taken. This paper presents a new algorithm for detection and classification of PQ disturbances based on the combination of double-resolution S-transform (DRST) and directed acyclic graph support vector machines (DAG-SVMs). The proposed method first employs DRST for an effective feature extraction from power signals. Then, the DAG-SVMs are used to predict the classes of PQ disturbances. The DRST not only has better time–frequency localization and stronger robustness but also reduces the computational complexity without losing the useful information of the original signal in comparison with the traditional S-transform. Through the combined use of DRST and DAG-SVMs, the algorithm can be easily implemented in embedded real-time applications. Finally, the implementation of the proposed algorithm in a digital signal processor + advanced reduced instruction set computing machine-based hardware test platform is introduced. The effectiveness of the proposed method is demonstrated by means of computer simulations and practical experiments with single and combined PQ disturbances. [ABSTRACT FROM PUBLISHER]

In this paper, a novel fault diagnosis approach based on sparse representation and support vector machine (SVM) is proposed. By keeping excluding subdictionaries and representing the testing sample using the sparse representation model, the confidence of each class is computed through a Hough voting progress using the computed sparse coding to show how much chance the testing sample belongs to a certain class. And then, the maximal confidence is selected as the highest confidence of every class. If the maximal confidence is high enough, the corresponding class label is taken as the diagnosis outcome. Otherwise, an SVM process is applied to classes with higher confidence factors to give the final outcome. An integrated framework from feature selection to fault type classification is designed and used in the roller bearing fault diagnosis. The result shows that the accuracy of our approach can reach 91.7%, 2% higher than SVM approach, and 7% higher than the conventional sparse-representation-based classification. Finally, an experiment based on the sparse representation model alone was undertaken. The result shows that using a confidence threshold, this model can give a compromised diagnosis between diagnosis accuracy and the number of samples received to diagnosis, which is suitable in areas where high diagnosis accuracy is required such as nuclear and aerospace industries. [ABSTRACT FROM PUBLISHER]

This paper presents research that uses discrete wavelet transform (DWT) and radial basis neural network for automatic classification. The flow rate of a wet-etching fabrication facility for a single wafer can be analyzed automatically. The electrical signal of a flow meter is collected and decomposed by means of DWT. The signal power of the coefficients processed by the DWT is fed into the radial basis neural network for initial classification. A digital filter for post signal processing and a user-defined threshold value are applied; calculations for successful identification rate take place at the final step. The research results are applicable to automatic identification functions for in situ fabrication monitoring. [ABSTRACT FROM PUBLISHER]

In this paper, we present a novel feature-based neural network (NN) approach for estimation of blood pressure (BP) from wrist oscillometric measurements. Unlike previous methods that use the raw oscillometric waveform envelope (OMWE) as input to the NN, in this paper, we propose to use features extracted from the envelope. The OMWE is mathematically modeled as a sum of two Gaussian functions. The optimum parameters of this model are found by minimizing the least squares error between the model and the OMWE using the Levenberg–Marquardt algorithm and are used as features. Two separate feed-forward NNs (FFNNs) are then designed to estimate the systolic and diastolic BPs using these features. The FFNNs are trained using the resilient backpropagation learning algorithm and tested on a data set of BP measurements recorded from 85 subjects. The performance is then compared with that of the conventional maximum amplitude algorithm, adaptive neuro-fuzzy inference system, and already published NN-based methods. It is found that the proposed approach achieves lower values of mean absolute error and standard deviation of error in the estimation of BP. In addition, the proposed approach has the following advantages: lower complexity with respect to the design parameters, smaller training data set, and lower computational load. [ABSTRACT FROM PUBLISHER]

The detection of Swerling 0 targets in movement in sea-ice Weibull-distributed clutter by neural networks (NNs) is presented in this paper. Synthetic data generated for typical sea-ice Weibull parameters reported in the literature are used. Due to the capability of NNs for learning the statistical properties of the clutter and target signals during a supervised training, high clutter reduction rates are achieved, reverting on high detection performances. The proposed NN-based detector is compared with a reference detector proposed in the literature that approximates the Neyman–Pearson (NP) detector. The results presented in the paper allow empirically demonstrating how the NN-based detector outperforms the detector taken as reference in all the cases under study. It is achieved not only in performance but also in robustness with respect to changes in sea-ice Weibull-distributed clutter conditions. Moreover, the computational cost of the NN-based detector is very low, involving high signal processing speed. [ABSTRACT FROM PUBLISHER]

Today's technology allows the construction of complex experimental apparatus with reasonable budget. This paper supplies detailed guidelines for constructing a low-cost scaled automobile platform for research and education in vehicle dynamics and control. It summarizes the knowledge gathered when designing, building, and evaluating a model car, which was deployed in a real-world environment for testing an electronic stability control (ESC) algorithm. The model car was built using off-the-shelf hardware and open-source software. Data from a variety of onboard sensors are fused in real time so as to deliver accurate measurements to the ESC algorithm, whereas sensor fault diagnosis is achieved at the same time through an innovative approach based on artificial neural networks (NNs). The detailed presentation of this case study provides a road map on how a researcher can build effective experimental automotive platforms for research and educational purposes. [ABSTRACT FROM AUTHOR]

An analog neural network (NN) was designed, implemented in 0.35-μm complementary metal-oxide-semiconductor technology, and tested to increase the distance measurement interval of a phase-shift laser rangefinder and to classify different types of surfaces for varying distances and incidence angles. This paper focuses on the ability of the NN to remove the indecision on the distance value deduced from the phase-shift measurement. The NN architecture is a multilayer perceptron. (MLP) with two inputs, three processing neurons in the hidden layer, and one output neuron. The amplified and filtered photoelectric signal provided by the rangefinder is set at one input. The NN is trained so that its output voltage is proportional to the distance for easy evaluation. By combining both measurements coming from the rangefinder and the NN, it is possible to obtain a resolution of 50 μm on a distance interval [0.5 m; 3.2 m], whereas the rangefinder measurement range width is limited to 0.9 m. This paper presents the complete system, concentrating more on the training phase of the implemented NN and on the experimental results. [ABSTRACT FROM AUTHOR]

This paper presents a new approach to the analysis of the metrological properties of a measuring chain with neural networks used for the reconstruction of an input signal of a nonlinear sensor. The general idea of this approach consists of building such an error model, which contains both the error sources arising in all the elements of the measuring chain with neural reconstruction and a description of the error propagation from the input to the output of the chain. The error source extraction requires the decomposition of the static and dynamic properties of the sensor, which results in the realization of the reconstruction in two stages independently. This paper shows how to use a built-up error model, which is described in probabilistic categories, for the analysis of the metrological properties of the measuring chain and to calculate uncertainty at every stage of the analysis. Theoretical considerations have been illustrated by the results obtained from measurement and simulation experiments. [ABSTRACT FROM AUTHOR]

Fault detection and diagnosis (FDD) is crucial for stable, reliable, and safe operation of industrial equipment. In recent years, deep learning models have been widely used in data-driven FDD methods because of their automatic feature learning capability. In general, these models are trained on historical sensor data, and therefore, it is very difficult to meet the real-time requirement of online FDD applications. Since transfer learning can solve different but similar problems in the target domain efficiently and effectively with the knowledge learned from the source domain, this paper proposes an online fault diagnosis method based on a deep transfer convolutional neural network (TCNN) framework. The TCNN framework is made up of an online CNN based on LeNet-5 and several offline CNNs with a shallow structure. First, time-domain signal data are converted into images that contain abundant fault information and are suitable as the input of CNN. Then, the online CNN is constructed to automatically extract representative features from the converted images and classify faults. Finally, in order to improve the real-time performance of the online CNN, several offline CNNs are also constructed and pretrained on related data sets. By directly transferring the shallow layers of the trained offline CNNs to the online CNN, the online CNN can significantly improve the real-time performance and successfully address the issue of achieving the desired diagnostic accuracy within limited training time. The proposed method is validated on two bearing data sets and one pump data set, respectively. The prediction accuracy of the proposed method using three data sets are 99.88%, 99.13%, and 99.98%, respectively. The experimental results also indicate that the improvement of accuracy is 19.21% for the motor bearing case, 29.82% for the rolling mill bearing case, and 33.26% for the pump case during the early stage of learning. [ABSTRACT FROM AUTHOR]

Surface flaw inspection is of great importance for quality control in the field of manufacture. In this paper, a novel surface flaw inspection algorithm is proposed based on adaptive multiscale image collection (AMIC) using convolutional neural networks. First, the inspection networks are pretrained with ImageNet data set. Second, the AMIC is established, which consists of adaptive multiscale image extraction and with-contour local extraction from training images. Through the AMIC, the training data set is greatly augmented, and labels of images can be accomplished automatically without artificial consumption. Then, transfer learning is performed with the AMIC established from training data set. Finally, an automatic surface flaw inspection instrument for large-volume metal components embedded with the proposed inspection algorithm is designed. Experiments with small metal components are performed to analyze the influence of parameters, and comparative experiments are carried out. The inspecting precisions for indentation, scratch, and pitted surface of the proposed method are 97.3%, 99.5%, and 100%, respectively. The experimental results demonstrate the effectiveness of the proposed method in the detection of various surface flaws. [ABSTRACT FROM AUTHOR]

In recent years, using distributed fiber-optic sensors based on Brillouin scattering, for monitoring pipelines, tunnels, and other constructional structures have gained huge popularity. However, these sensors have a low signal-to-noise ratio (SNR), which usually increases their measurement error. To alleviate this issue, ensemble averaging is used which improves the SNR but in return increases the measurement time. Reducing the noise by averaging requires hundreds or thousands of scans of the optical fiber; hence averaging is usually responsible for a large percent of the entire system latency. In this paper, we propose a novel method based on artificial neural network for SNR enhancement and measurement acceleration in distributed fiber-optic sensors based on the Brillouin scattering. Our method takes the noisy Brillouin spectrums and improves their SNR by 20 dB, which reduces the measurement time significantly. It also improves the accuracy of the Brillouin frequency shift estimation process and its latency by more than 50% in comparison with the state-of-the-art software and hardware solutions. [ABSTRACT FROM AUTHOR]

This paper presents a comprehensive approach to accurate and efficient modeling of microwave active devices such as metal semiconductor field effect transistors (MESFETs) using artificial neural networks (ANNs). A radial basis function (RBF)-ANN model is developed for S-parameters and equivalent circuit parameters (ECPs) of MESFETs. The training and testing data for these models are obtained from the mea- sured two-port scattering parameters and extracted ECPs of a 0.25 × 200 μm (4 × 50 μm) gallium arsenide MESFET. A four-input eight-output ANN is used to model the S-parameters of a microwave MESFET versus bias, temperature, and frequency, and a three-input eight-output ANN is used to model the ECPs of a microwave MESFET versus bias and temperature. Comparisons of measured and modeled data are presented, and the results show very good agreement. The average relative errors using the RBF-ANN models for the S-parameters and ECPs were 0.81% and 0.77%, respectively, which both represent about 60% reduction in error when compared to backpropagation ANN models of similar parameters of the same device. [ABSTRACT FROM AUTHOR]

In the oil industry, many applications require the measurement of more than one liquid level interface, often in challenging environments. In this paper, a new ultrasonic-based technique has been developed to determine oil, emulsion, and water levels in oil tanks. It consists of a compact and programmable ultrasound-based multilayer level measurement device for which a feed-forward neural network is implemented. The advantages of this method over current methods include contactless distance measurement, higher accuracy, lower cost, simpler setup, and not using ionizing radiation. The other advantage of this technique over light-based methods is its insensitivity to a dusty and smoky environment and independence of the object material and surface. Preliminary experiments have been conducted on the device. In this paper, the design and operating parameters of the device are discussed, and evidence of satisfactory performance is given. [ABSTRACT FROM AUTHOR]

Packet-based data networks such as British Telecommunications (BT) plc's switched multimegabit data service (SMOS) have performance that varies with customer behavior, network design, and network provision. This paper describes a system researched to monitor such networks from the customer's perspective, but on behalf of the operator. The information generated allows the operator to determine whether the network is performing correctly or whether changes or customer load will require the network to be changed in some way. Such networks are increasingly important as many parts of the Internet are built on similar technologies. The paper specifically considers an architecture that encompasses all the main functionality involved in performance management of networks. Consideration is given to the management and dissemination of the data collected. A database to enhance network data mining performance is described. A method to identify the periods of significance in the measurement data based on a statistical test and a neural network is incorporated into a realization of the architecture. In addition, a novel mechanism for the real-time presentation of the network performance data in an operator-friendly manner is given. [ABSTRACT FROM AUTHOR]

This paper investigates a human walking behavior- based mobile robot control system. The Intelligent Space (iSpace) is a distributed sensory system, which is the background infrastructure to observe human walking in a limited area. The observation of human walking behavior is applied to train fuzzy-neural networks (FNN). The trained FNNs are applied to approximate the obstacle avoidance behavior of human walking. The paper introduces the iSpace and the mobile agents, which are mobile robots, utilizing the intelligence of the iSpace. The observed and trained human walking behaviors are applied to control the mobile agent in a human-robot shared environment. Experimental results demonstrate the effectiveness of the FNN-based control system. [ABSTRACT FROM AUTHOR]

This paper is concerned with the design and implementation of active vibration control based on fuzzy logic and neural networks (NNs). The controller is used to dampen the end-point vibration in a single-link flexible manipulator mounted on a two degrees-of-freedom platform. The inputs to the fuzzy logic controller (FLC) are the angular position of the hub and the end point deflection of the flexible beam. A NN predicting the deflection was obtained using a set of three strain gauge pairs mounted on the beam and an linear-variable differential transformer placed at the tip. This paper discusses how to build the rule base for the flexible beam based on the relation between the angular displacement of the hub and the end-point deflection, as well as the approach that was used to develop the NN model. Several experimental results showing the effectiveness of the FLC in reducing the end-point vibration of the flexible beam are presented. [ABSTRACT FROM AUTHOR]

In-service inspection of complex systems such as nuclear steam generator (SG) tubes and their surrounding support structures is challenged by overlapping degradation modes. In these complex systems the simultaneous and accurate measurement of more than two interdependent parameters is difficult using standard statistical regression analysis tools. Recently, artificial neural networks (ANNs) have been investigated for dealing with the complex relation between inspection data and defect properties. In this paper, pulsed eddy current data were obtained using a single driver with an array of eight pick-up coils configured for inspection of Alloy-800 SG tube fretting, accompanied by tube offset within a simulated corroding ferromagnetic support structure. Time-voltage data were processed by a modified principal component analysis (MPCA) to reduce data dimensionality, and MPCA scores were input into an ANN that simultaneously targeted four parameters associated with support structure hole size, tube off-centering in two dimensions, and fret depth. The neural network was trained, tested, and validated on experimental data and provided estimates within 2% of hole inner diameter (ID) and 3% of fret depth targets. The estimates of hole ID and tube position were further improved when fret depth was used as an input, as might occur if fret depth inspection results are available. [ABSTRACT FROM PUBLISHER]

Gas-liquid two-phase flows are present in nature and in different industrial activities alike, such as the chemical, petroleum, and nuclear industries. In this type of flow, the liquid and gas phases assume different spatial configurations inside the pipe, called flow patterns. The mathematical modeling of slug flow comprises from simple steady-state models to more complex models for transient regimes. Those models require closure relationships, e.g., empirical correlations and statistical distributions of characteristic flow parameters. In this paper, a model based on artificial neural networks (ANNs) for predicting the two-phase slug flow behavior is proposed. With this ANN model, the parameters that characterize the flow are extracted from the time series of void fractions obtained experimentally. The variables of interest are superficial velocities of the fluids, liquid slug and gas bubble lengths, and the bubble translational velocity and their standard deviations. The knowledge and understanding of those parameters will improve the characterization of the intermittent slug flows and will also provide information on the development of physical models that describe this phenomenon, such as the unit cell models, the drift flux model and the slug tracking model. In general, the estimation models based on ANNs showed good results compared with reference values obtained experimentally. The results show that the estimation models present a mean square error below 2%. The methodology presented here, combining experimentally obtained void fraction time series and ANN, is an appropriated method to infer flow parameters and thus to support slug flow characterization. [ABSTRACT FROM PUBLISHER]

In this paper, a battery health prognostics system is developed based on Bayesian-inference probabilistic (BIP) indication and state-space model (SSM) that integrates logistic regression (LR) and particle filtering (PF). In this system, generative topographic mapping is constructed to model distribution of multisensor data from healthy battery under an assumption that predictable fault patterns are not available. BIP is developed as a quantification indication of battery state-of-health. BIP is capable of offering failure probability for the monitored batteries, which has intuitionist explanation related to health degradation state. SSM is used for modeling health propagation of battery on the time flow, where LR and PF are integrated to predict remaining useful life of the battery. The experimental results on a lithium–ion battery testbed illustrate the potential applications of the proposed system as an effective tool for battery health prognostics. [ABSTRACT FROM PUBLISHER]