Search

In this paper we present a novel methodology, referenced as ORFEO (Ordinal classifier and Regressor Fusion for Estimating an Ordinal categorical target), to enhance the performance in ordinal classification problems for which the latent variable is observable. ORFEO is an artificial neural network model incorporating two outputs, one for ordinal classification, using the cumulative link model, and one for regression, using a linear model. Both outputs are simultaneously optimised considering a loss function that linearly combines both classification and regression losses. The main motivation behind developing the proposed approach is to enhance the performance of a standard ordinal classifier. This improvement is facilitated by considering the regression output, which allows the model to differentiate between patterns within the same category. The ORFEO model is applied to two problems in the field of marine and ocean engineering: short-term prediction of both significant wave height and flux of energy. Both problems are addressed considering four different coastal zones of the United States of America, using 13 datasets formed by buoys measurements and reanalysis data. A comprehensive comparison against 20 methodologies, including regression and nominal/ordinal classification approaches is performed, by using diverse nominal and ordinal performance metrics. Ranks achieved indicate that ORFEO outperforms all the compared methodologies in terms of all the performance measures, demonstrating the efficacy and robustness of the proposal. Finally, a statistical analysis is conducted, concluding that there are statistically significant differences across ordinal and nominal performance metrics in favour of the proposed ORFEO model. [ABSTRACT FROM AUTHOR]

Activation functions play an important role in deep learning models by introducing non-linearity to the output of a neuron, enabling the network to learn complex patterns and non-linear relationships in data and make predictions on more complex tasks. Deep learning models' most commonly used activation functions are Purelin, Sigmoid, Tansig, Rectified Linear Unit (ReLU), and Exponential Linear Unit (ELU), which exhibit limitations such as non-differentiability, vanishing gradients, and neuron inactivity with negative values. These functions are typically defined over a finite range, and their outputs are integers or real numbers. Using fractional calculus in designing activation functions for neural networks has shown promise in improving the performance of deep learning models in specific applications. These activation functions can capture more complex non-linearities than traditional integer-order activation functions, improving performance on tasks such as image classification and time series prediction. This paper focuses on deriving and testing linear and non-linear fractional-order forms of activation functions and their variants. The linear activation function includes Purelin. In contrast, the non-linear activation functions are Binary Step, Sigmoid, Tansig, ReLU, ELU, Gaussian Error Linear Unit (GELU), Hexpo, and their variants. Besides, the standard formula has been implemented and used in developing the fractional-order linear activation function. Furthermore, various expansion series, such as Euler and Maclaurin, have been used to design non-linear fractional-order activation functions and their variants. The single- and multi-layer fractional-order neural network models have been developed using the designed fractional-order activation functions. The simulation study uses developed fractional-order neural network models for predicting the Texas wind turbine systems' generated power. The performance of single and multi-layer fractional-order neural network models has been evaluated by changing the activation functions in the hidden layer while keeping the Purelin function constant at the output layer. Experiments on neural network models demonstrate that the designed fractional-order activation functions outperform traditional functions like Sigmoid, Tansig, ReLU, ELU, and their variants, effectively addressing limitations. • Activation functions are derived into fractional functions using fractional calculus. • The fractional neural network models have been made by adopting derived functions. • A Case study has been done on the Texas wind turbine with the models to predict the power. • The models' performance has been evaluated using the fractional functions. [ABSTRACT FROM AUTHOR]

Abstract: Training neural networks (NNs) is a complex task of great importance in the supervised learning area. However, performance of the NNs is mostly dependent on the success of training process, and therefore the training algorithm. This paper addresses the application of harmony search algorithms for the supervised training of feed-forward (FF) type NNs, which are frequently used for classification problems. In this paper, five different variants of harmony search algorithm are studied by giving special attention to Self-adaptive Global Best Harmony Search (SGHS) algorithm. A structure suitable to data representation of NNs is adapted to SGHS algorithm. The technique is empirically tested and verified by training NNs on six benchmark classification problems and a real-world problem. Among these benchmark problems two of them have binary classes and remaining four are n-ary classification problems. Real-world problem is related to the classification of most frequently encountered quality defect in a major textile company in Turkey. Overall training time, sum of squared errors, training and testing accuracies of SGHS algorithm, is compared with the other harmony search algorithms and the most widely used standard back-propagation (BP) algorithm. The experiments presented that the SGHS algorithm lends itself very well to the training of NNs and also highly competitive with the compared methods in terms of classification accuracy. [Copyright &y& Elsevier]

Abstract: In this paper, next-day hourly forecasts are calculated for the energy price in the electricity production market of Spain. The methodology used to achieve these forecasts is based on artificial neural networks, which have been used successfully in recent years in many forecasting applications. The days to be forecast include working days as well as weekends and holidays, due to the fact that energy price has different behaviours depending on the kind of day to be forecast. Besides, energy price time series are usually composed of too many data, which could be a problem if we are looking for a short period of time to reach an adequate forecast. In this paper, a training method for artificial neural nets is proposed, which is based on making a previous selection for the multilayer perceptron (MLP) training samples, using an ART-type neural network. The MLP is then trained and finally used to calculate forecasts. These forecasts are compared to those obtained from the well-known Box–Jenkins ARIMA forecasting method. Results show that neural nets perform better than ARIMA models, especially for weekends and holidays. Both methodologies calculate more accurate forecasts—in terms of mean absolute percentage error—for working days that for weekends and holidays. Agents involved in the electricity production market, who may need fast forecasts for the price of electricity, would benefit from the results of this study. [Copyright &y& Elsevier]

In this paper, a fault detection scheme for application to large-scale data acquisition systems is presented. The detection scheme has to process up to several hundreds of different measurements at a time and check them for consistency. The fault detection scheme works in three steps: First, principal component analysis of training data is used to determine non-sparse areas of the measurement space. Fault detection is accomplished by checking whether a new data record lies in a cluster of training data or not. Therefore, in a second step the distribution function of the available data is estimated using kernel regression techniques. In order to reduce the degrees of freedom and to determine clusters of data efficiently, in a third step the distribution function is approximated by a neural network. In order to use as few basis functions as possible a new training algorithm for ellipsoidal basis function networks is presented: New neurons are placed such that they approximate the distribution function in the vicinity of their centers up to the second order. This is accomplished by adapting the spread parameters using Taylor''s theorem. Thus, the amount of necessary parameters and the computational effort for online supervision can be reduced dramatically. An important requirement for the fault detection scheme is that it is able to automatically adapt itself to new data. The present paper also addresses this feature. It is demonstrated how a gradient optimization with algebraic constraints can be applied to adapt a pre-existing network to new data points. Numerical examples with real data show that the proposed method produces excellent results. [Copyright &y& Elsevier]

In this paper, a recurrent neural network coupled with Kalman filter is proposed to identify dynamic terms of robotic manipulator. By cooperating some inherent characteristics of robot, this network has the capability to individually identify nonlinear terms using Weighted Augmentation Error (WAE). To present the infrastructure of architecture, an adaptive scheme based on the conventional Back Propagation (BP) is firstly driven using the Gradient Descent (GD) method. Additionally, a stable adaptive updating rule is extracted from the discrete time Lyapunov candidate as an approach for the general nonlinear system identification. Then, this approach is applied to the predefined network. To experimentally validate the computational efficiency and control applicability of the proposed method, Adaptive Neural Network Based Inverse Dynamic Control (ANN-Based-IDC) is employed on a laboratory-scaled twin-rotor CE-150 helicopter. This experiment illustrates enhancement of steady-state performance from 2-to-3 times more in compared with simple PID. Moreover, disturbance rejection and robustness tests admit capability of the method for online dynamic identification in the presence of output and dynamic perturbation. [ABSTRACT FROM AUTHOR]

This paper suggests a novel three-level model-free hierarchical learning approach that solves the reference trajectory tracking problem for control systems (CSs). The new approach consists of the low level, the intermediate level and the high level, it relies on past memorized optimal input output execution patterns and adaptively merges them using a similarity measure. The low level feedback control is carried out in a novel model-free framework using a neural network (NN) controwller tuned by Virtual Reference Feedback Tuning (VRFT) in order to linearize the closed-loop CS and to match a linear reference model. The NN controller is tuned in two phases, an offline one and an online one. Nonlinear Model Predictive Control (NMPC) is first employed in the novel offline tuning phase. The online tuning phase makes next use only of the process sign in the dynamic back-propagation mechanism that updates the NN parameters. After the NN controller is trained and the feedback CS is fixed, the optimal execution patterns (input/output patterns) are defined in terms of optimal control problems, which balance control accuracy and control effort. The input/output patterns are formulated over a feedback CS and are solved in a model-free Iterative Learning Control (ILC) framework for linear time-invariant systems at the intermediate level. Once the optimal executions are learned at the intermediate level over the feedback CS, they are stored in a database (DB). Then each time a new trajectory is to be tracked as required by the high level planner, similar patterns of executions are looked-up in the DB and are merged by the weighted average sum of most similar patterns resulted from a sort algorithm using a distance metric. The proposed approach is tested on the position control of a Single Input-Single Output nonlinear aerodynamic control system and shows trajectory tracking performance improvement with respect to the case when no learnt experience is used. [ABSTRACT FROM AUTHOR]

Imbalanced problem occurs when the size of one class, i.e. the minority class, is much lower than that of the other classes, i.e. the majority classes. Conventional data level methods are employed as the preprocessing approaches to balance the datasets before the classifier learning. Since the balanced data remains unchanged during the learning process, one pre-deleted sample would never be used to train the classifier, which may result in information loss. To solve this problem, this work presents an One-sided Dynamic Undersampling (ODU) technique which adopts all samples in the training process, and dynamically determines whether a majority sample should be used for the classifier learning. Thus, ODU can dynamically undersample the majority class to balance the dataset. To validate the effectiveness of ODU, we integrate it into No-Propagation neural networks to propose an ODU No-Propagation Neural Networks (ODUNPNN). ODUNPNN takes all training samples into consideration, and dynamically undersamples majority class after each iteration, i.e. ODUNPNN integrates undersampling approach into the classifier learning process. Experimental results on both synthetic and real-world imbalance datasets demonstrate that ODUNPNN outperforms the NPNN-based algorithms, and results in comparative performance compared with LASVM-AL, EasyEnsemble, and DyS on real-world imbalance datasets. The contributions of this paper are: (1) ODUNPNN integrates undersampling approach into the classifier learning process. (2) ODUNPNN dynamically balances training data in each iteration. (3) ODU technique can be integrated into other classification learning machines. [ABSTRACT FROM AUTHOR]

Froth Flotation is one of the fundamental processes used in iron ore attainment. It involves various physical–chemical reactions, and its multivariate control is not trivial. Traditionally, plant operators get quality measurements from laboratory analysis of sampled material, which can take hours to be published. In this paper, we present alternatives that eliminate both the necessity to periodically sample material as well as the delay associated with laboratory analysis: a predictive machine learning model working as a soft sensor. Two different classes of models were developed, exploring both deep and shallow approaches of machine learning, with a deep neural network architecture constructed for the particular task at hand. All models yielded accurate predictions, with results favoring the deep neural network when a larger volume of data is available. [ABSTRACT FROM AUTHOR]

Data based modeling applications require the complete automation of model creation of, in general, nonlinear processes. Numerous model architectures are available to approximate complex data structures, however, creating and selecting the best model provides a present challenge for application. Model ensembles provide excellent strategies if the prediction performance of the models can be assessed in some way. Information criteria approaches, trading off goodness of fit criteria against the average variance error of the model structures, are theoretically applicable, but demonstratively perform poorly in settings with sparse data sets and complex model structures, settings which are present for many industrial purposes. This paper proposes and discusses a methodology how to regularise the weight calculation building model ensembles. Thereby, the weights trade off goodness of fit criteria against empirical measures of risk, increasing the model ensemble prediction performance. The methodology is independent of the model class, applicable to all black box models, and improves prediction performance for a variety of data in several industrial application areas. [ABSTRACT FROM AUTHOR]

Extreme learning machine (ELM), a simple single-hidden-layer feed-forward neural network with fast implementation, has been successfully applied in many fields. This paper proposes an ELM with a constructional structure (CS-ELM) for improving the performance of ELM in dealing with regression problems. In the CS-ELM, there are some partial input subnets (PISs). The first step in designing the PISs is to divide the data-attribute-space into several sub-spaces through using an improved extension clustering algorithm (IECA). The input data attributes in the same sub-space can build a PIS and the similar information of the data attributes is stored in the corresponding PIS. Additionally, a double parallel structure is applied in the CS-ELM, in which there is a special channel that directly connects the input layer neurons to the output layer neurons. In this regard, the proposed procedure can be called ELM with a data-attribute-space-oriented double parallel (DASODP) structure (DASODP–ELM). To test the validity of the proposed method, it is applied to 4 regression applications. The experimental results indicate that, compared with ELM, DASODP–ELM with less number of parameters can achieve higher regression precision in the generalization phase. [ABSTRACT FROM AUTHOR]

The most important problems with exploiting artificial neural networks (ANNs) are to design the network topology, which usually requires an excessive amount of expert’s effort, and to train it. In this paper, a new evolutionary-based algorithm is developed to simultaneously evolve the topology and the connection weights of ANNs by means of a new combination of grammatical evolution (GE) and genetic algorithm (GA). GE is adopted to design the network topology while GA is incorporated for better weight adaptation. The proposed algorithm needs to invest a minimal expert’s effort for customization and is capable of generating any feedforward ANN with one hidden layer. Moreover, due to the fact that the generalization ability of an ANN may decrease because of overfitting problems, the algorithm utilizes a novel adaptive penalty approach to simplify ANNs generated through the evolution process. As a result, it produces much simpler ANNs that have better generalization ability and are easy to implement. The proposed method is tested on some real world classification datasets, and the results are statistically compared against existing methods in the literature. The results indicate that our algorithm outperforms the other methods and provides the best overall performance in terms of the classification accuracy and the number of hidden neurons. The results also present the contribution of the proposed penalty approach in the simplicity and generalization ability of the generated networks. [ABSTRACT FROM AUTHOR]

The grape phenolic maturity is one of the most important parameters to determine the optimal time for harvest. In this paper we propose an innovative methodology for the problem of how this task is performed today. In particular, the method consists in analyzing seed images using pattern recognition methodology, and classifying them in immature, mature and over mature states through a supervised learning neural network. The methodology presented gives objective information about maturity, which is useful for deciding the moment when the harvest should be performed. [ABSTRACT FROM AUTHOR]

Abstract: In recent decades, and in order to develop applications covering several areas of knowledge, different researchers have been performing hardware implementations around paradigms such as fuzzy systems, neural networks or systems resulting from the hybridization of the previous two systems, known as neuro–fuzzy systems. Applications have been performed on different types of devices and/or platforms. The point of view of this paper is focused on a hardware taxonomy (devices where the applications have been implemented) and highlights the characteristics of the different applications covering the aforementioned paradigms done over the last two decades, and the beginning of the current decade. Special mention is made up of reconfigurable devices. [Copyright &y& Elsevier]

Abstract: In this paper, we introduce a novel authentication and discrimination system using artificial neural networks with RGB histograms. Concisely, we collect key codes and inter key times for passwords in register and login steps, colorize the keys to generate RGB histograms and train the histograms with neural networks to determine a password keystroke interval. After training determined number of histograms, our proposed system has ability to test future logins and reject the histograms which are outside of the interval. Our proposed system have successfully granted 90% of real user attempts and rejected 90% of frauds. [Copyright &y& Elsevier]

Abstract: The transmission of real-time multimedia streams requires service guarantees, such as limited packet loss, minimum bandwidth and low delay and jitter, to ensure a good quality of experience (QoE) for viewers. The spatial and temporal redundancy of videos is addressed by coding algorithms that reduce the amount of information necessary to represent the images. As a consequence, multimedia traffic commonly presents variable bit rate behavior and self-similar characteristics. Although the reduction in bandwidth requirements is highly desirable, the burstiness of traffic leads to problems in network design and performance prediction. Even a low level of packet loss could severely affect the viewer QoE. In this paper, we propose a real-time packet payload classifier, implemented with artificial neural network (ANN) to be used at network routers. A priority packet discard strategy can be implemented to avoid discarding packets that carry the most relevant information for image reconstruction, thus improving the perceived quality. This approach does not require changes at the video source to classify outgoing packets. The ANN was employed because of its good capacity in temporal series recognition and the possibility of its implementation in real-time systems due to its low computational complexity. The video traces used for training and validation were encoded with H.264/MPEG-4 Advanced Video Coding and are publicly available. The priority packet discard strategy was tested through computational simulations. The QoE was estimated comparing the peak signal-to-noise ratio (PSNR) of original and the received frames of video, and the results indicate that the proposed method improves the QoE. The implementation does not require packet payload processing and can be performed with network layer information only. [Copyright &y& Elsevier]

Abstract: This paper addresses the problem of adaptive neural sliding mode control for a class of multi-input multi-output nonlinear system. The control strategy is an inverse nonlinear controller combined with an adaptive neural network with sliding mode control using an on-line learning algorithm. The adaptive neural network with sliding mode control acts as a compensator for a conventional inverse controller in order to improve the control performance when the system is affected by variations in its entire structure (kinematics and dynamics). The controllers are obtained by using Lyapunov's stability theory. Experimental results of a case study show that the proposed method is effective in controlling dynamic systems with unexpected large uncertainties. [Copyright &y& Elsevier]

Multi-step time-series prediction is an essential supportive step for decision-makers in several industrial areas. Artificial intelligence techniques, which use a neural network component in various forms, have recently frequently been used to accomplish this step. However, the complexity of the neural network structure still stands up as a critical problem against prediction accuracy. In this paper, a method inspired by the proportional–integral–derivative (PID) control approach is investigated to enhance the performance of neural network models used for multi-step ahead prediction of periodic time-series information while maintaining a negligible impact on the complexity of the system. The PID-based method is applied to the predicted value at each time step to bring that value closer to the real value. The water demand forecasting problem is considered as a case study, where two deep neural network models from the literature are used to prove the effectiveness of the proposed boosting method. Furthermore, to prove the applicability of this PID-based booster to other types of periodic time-series prediction problems, it is applied to enhance the accuracy of a neural network model used for multi-step forecasting of hourly energy consumption. The comparison between the results of the original prediction models and the results after using the proposed technique demonstrates the superiority of the proposed method in terms of prediction accuracy and system complexity. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents a constrained Model Predictive Control (MPC) strategy enriched with soft-control techniques as neural networks and fuzzy logic, to incorporate self-tuning capabilities and reliability aspects for the management of drinking water networks (DWNs). The control system architecture consists in a multilayer controller with three hierarchical layers: learning and planning layer, supervision and adaptation layer, and feedback control layer. Results of applying the proposed approach to the Barcelona DWN show that the quasi-explicit nature of the proposed adaptive predictive controller leads to improve the computational time, especially when the complexity of the problem structure can vary while tuning the receding horizons. [Copyright &y& Elsevier]

Abstract: Modelling complex dynamic mechanical systems, such as PEMFC, without any physical models is a difficult challenge but it could allow the monitoring of endurance tests of fuel cell systems. Neural networks are recognised as powerful numerical tools for predicting complex and nonlinear dynamic behaviours. They require only data limited to experimental inputs and outputs but the choice of an adapted architecture is critical. This paper presents a method for defining a neural network architecture optimised for the fuel cell systems. The associated experimental conditions specifying the vibration tests to train and validate were defined. They consist of swept sinus as well as random excitation forces. The resulting simulations are presented and analysed. [Copyright &y& Elsevier]

Abstract: This paper investigates the development and experimental implementation of an adaptive dynamic nonlinear model inversion control law for a Twin Rotor MIMO System (TRMS) using artificial neural networks. The TRMS is a highly nonlinear aerodynamic test rig with complex cross-coupled dynamics and therefore represents the control challenges of modern air vehicles. A highly nonlinear 1DOF mathematical model of the TRMS is considered in this study and a nonlinear inverse model is developed for the pitch channel of the system. An adaptive neural network element is integrated thereafter with the feedback control system to compensate for model inversion errors. The proposed on-line learning algorithm updates the weights and biases of the neural network using the error between the set-point and the real output. The real-time response of the method shows a satisfactory tracking performance in the presence of inversion errors caused by model uncertainty. The approach is therefore deemed to be suitable to apply real-time to other nonlinear systems with necessary modifications. [Copyright &y& Elsevier]

Abstract: The paper presents the application of wavelet transformation and neural network ensemble to the accurate forecasting of the daily average concentration of particulate matter of diameter up to 10μm (PM10). Few neural predictors are applied: the multilayer perceptron, radial basis function, Elman network and support vector machine as well as one linear ARX model. They are used for prediction in combination with wavelet decomposition, forming many individual prediction results that will be combined in an ensemble. The important role in presented approach fulfills the wavelet transformation and the integration of this ensemble. We have proposed solution applying the additional neural network responsible for the final forecast (integration of all particular prediction results). The numerical experiments for prediction of the daily concentration of the PM10 pollution in Warsaw are presented. They have shown good overall accuracy of prediction in terms of all investigated measures of quality. [Copyright &y& Elsevier]

Abstract: The search capabilities of the Differential Evolution (DE) algorithm – a global optimization technique – make it suitable for finding both the architecture and the best internal parameters of a neural network, usually determined by the training phase. In this paper, two variants of the DE algorithm (classical DE and self-adaptive mechanism) were used to obtain the best neural networks in two distinct cases: for prediction and classification problems. Oxygen mass transfer in stirred bioreactors is modeled with neural networks developed with the DE algorithm, based on the consideration that the oxygen constitutes one of the decisive factors of cultivated microorganism growth and can play an important role in the scale-up and economy of aerobic biosynthesis systems. The coefficient of mass transfer oxygen is related to the viscosity, superficial speed of air, specific power, and oxygen-vector volumetric fraction (being predicted as function of these parameters) using stacked neural networks. On the other hand, simple neural networks are designed with DE in order to classify the values of the mass transfer coefficient oxygen into different classes. Satisfactory results are obtained in both cases, proving that the neural network based modeling is an appropriate technique and the DE algorithm is able to lead to the near-optimal neural network topology. [Copyright &y& Elsevier]

Abstract: This paper discusses the application of the virtual reference tuning (VRT) techniques to tune neural controllers from batch input–output data, by particularising nonlinear VRT and suitably computing gradients backpropagating in time. The flexibility of gradient computation with neural networks also allows alternative block diagrams with extra inputs to be considered. The neural approach to VRT in a closed-loop setup is compared to the linear VRFT one in a simulated crane example. [Copyright &y& Elsevier]

Abstract: Decision-making frequently involves identifying how to change input parameters in a given process in order to effect a directed change in the process output. Artificial neural networks have been used extensively to model business and manufacturing processes and there are several existing neural network-based influence measures that allow a decision-maker to assess the relative impact of each variable on process performance. The purpose of this paper is to review those neural network-based measures of variable influence, and to identify the combination of those measures that results in a comprehensive approach to characterizing variable influence within a trained neural network model. We then demonstrate how this comprehensive approach can be used as a tool to guide decision makers in dynamic process control. [Copyright &y& Elsevier]

Abstract: In this paper, a fusion approach to determine inverse kinematics solutions of a six degree of freedom serial robot is proposed. The proposed approach makes use of radial basis function neural network for prediction of incremental joint angles which in turn are transformed into absolute joint angles with the assistance of forward kinematics relations. In this approach, forward kinematics relations of robot are used to obtain the data for training of neural network as well to estimate the deviation of predicted inverse kinematics solution from the desired solution. The effectiveness of the fusion process is shown by comparing the inverse kinematics solutions obtained for an end-effector of industrial robot moving along a specified path with the solutions obtained from conventional neural network approaches as well as iterative technique. The prominent features of the fusion process include the accurate prediction of inverse kinematics solutions with less computational time apart from the generation of training data for neural network with forward kinematics relations of the robot. [ABSTRACT FROM AUTHOR]

Abstract: In this paper a new color space, called the RGB color ratio space, is proposed and defined according to a reference color such that an image can be transformed from a conventional color space to the RGB color ratio space. Because a color in the RGB color ratio space is represented as three color ratios and intensity, the chrominance can be completely reserved (three color ratios) and the luminance can be de-correlated with the chrominance. Different from traditional distance measurement, a road color model is determined by an ellipse area in the RGB ratio space enclosed by the estimated boundaries. A proposed adaptive fuzzy logic in which fuzzy membership functions are defined according to estimated boundaries is introduced to implement clustering rules. Therefore, each pixel will have its own fuzzy membership function corresponding to its intensity. A basic neural network is trained and used to achieve parameters optimization. The low computation cost of the proposed segmentation method shows the feasibility for real time application. Experimental results for road detection demonstrate the robustness to intensity variation of the proposed approach. [Copyright &y& Elsevier]

Abstract: This paper addresses the problem of fault detection and isolation in railway track circuits. A track circuit can be considered as a large-scale system composed of a series of trimming capacitors located between a transmitter and a receiver. A defective capacitor affects not only its own inspection data (short circuit current) but also the measurements related to all capacitors located downstream (between the defective capacitor and the receiver). Here, the global fault detection and isolation problem is broken down into several local pattern recognition problems, each dedicated to one capacitor. The outputs from local neural network or decision tree classifiers are expressed using the Dempster–Shafer theory and combined to make a final decision on the detection and localization of a fault in the system. Experiments with simulated data show that correct detection rates over 99% and correct localization rates over 92% can be achieved using this approach, which represents a major improvement over the state of the art reference method. [Copyright &y& Elsevier]

Abstract: In this paper, a novel approach to control a single generator, connected to an infinite bus, is presented. Modifying published results for nonlinear identification using recurrent neural networks, a block controllable neural identifier is proposed, based on this neural model a control law is derived, which combines sliding modes and block control. The neural identifier and the proposed control law allows to reject external disturbances caused by generator terminal short circuits and mechanical power variations. Applicability of the approach is tested via simulations. [Copyright &y& Elsevier]

Abstract: This paper is a study of the application of artificial neural networks to the problem of calculating error compensation values for axis positioning on a machine tool. The primary focus is on the development of a neural network-based system that could be implemented and integrated into the open architecture control system of an actual machine. A number of neural network architectures were examined for applicability to the problem and one was selected and implemented on the actual machine. Positioning error compensation capabilities were evaluated using industry standard equipment and procedures, and the results obtained were compared with the capabilities of standard error compensation routines in machine tool controls. [Copyright &y& Elsevier]

Abstract: This paper presents an identification method based on artificial neural networks, which can be used for the robust fault detection. In particular, a problem of the multi-layer perceptron neural network uncertainty estimation with application of the outer bounding ellipsoid algorithm is considered. The mathematical description of the model uncertainty enables designing robust fault detection system, which effectiveness was verified with the DAMADICS benchmark. [Copyright &y& Elsevier]

Abstract: In this paper, a combustion temperature retrieval approximation for high-resolution infrared ground-based measurements has been developed based on a multilayer perceptron (MLP) technique. The introduction of a selection subset of features is mandatory due to the problems related to the high dimensionality data and the worse performance of MLPs with this high input dimensionality. Principal component analysis is used to reduce the input data dimensionality, selecting the physically important features in order to improve MLP performance. The use of a priori physical information over other methods in the chosen feature''s phase has been tested and has appeared jointly with the MLP technique as a good alternative for this problem. [Copyright &y& Elsevier]

This paper presents a special rule base extraction analysis for optimal design of an integrated neural-fuzzy process controller using an “impact assessment approach.” It sheds light on how to avoid some unreasonable fuzzy control rules by screening inappropriate fuzzy operators and reducing over fitting issues simultaneously when tuning parameter values for these prescribed fuzzy control rules. To mitigate the design efforts, the self-learning ability embedded in the neural networks model was emphasized for improving the rule extraction performance. An aeration unit in an Aerated Submerged Biofilm Wastewater Treatment Process (ASBWTP) was picked up to support the derivation of a solid fuzzy control rule base. Four different fuzzy operators were compared against one other in terms of their actual performance of automated knowledge acquisition in the system based on a partial or full rule base prescribed. Research findings suggest that using bounded difference fuzzy operator (O b) in connection with back propagation neural networks (BPN) algorithm would be the best choice to build up this feedforward fuzzy controller design. [Copyright &y& Elsevier]

In this paper, a new approach for steering a binocular head is presented. This approach is based on extracting the expert''s knowledge in order to improve the behaviour of the classical control strategies. This is carried out without inserting new elements in the system. Neuro-fuzzy techniques have been chosen in order to reach this target. As a result, a more friendly robotic system is achieved. [Copyright &y& Elsevier]

Abstract: Automatic discrimination of speech and music is an important tool in many multimedia applications. The paper presents an effective approach based on an adaptive network-based fuzzy inference system (ANFIS) for the classification stage required in a speech/music discrimination system. A new simple feature, called warped LPC-based spectral centroid (WLPC-SC), is also proposed. Comparison between WLPC-SC and the classical features proposed in the literature for audio classification is performed, aiming to assess the good discriminatory power of the proposed feature. The vector length used to describe the proposed psychoacoustic-based feature is reduced to a few statistical values (mean, variance and skewness). With the aim of increasing the classification accuracy percentage, the feature space is then transformed to a new feature space by LDA. The classification task is performed applying ANFIS to the features in the transformed space. To evaluate the performance of the ANFIS system for speech/music discrimination, comparison to other commonly used classifiers is reported. The classification results for different types of music and speech signals show the good discriminating power of the proposed approach. [Copyright &y& Elsevier]

Abstract: This paper involves experimentation on coating process of metal substrates in an electrostatic fluidized bed (EFB). Several operational parameters were covered like coating time, applied voltage and gas flow rate fed to the fluidized bed. First, a design of experiment (DOE) approach was used to define the experimental campaign and a general linear model based on analysis of variance (ANOVA) was used to elaborate and interpret the influence of all the operational parameters on coating thickness trends. Second, the experimental data were modelled using artificial neural networks. Different neural networks and training algorithms were employed to find the best technique to predict the coating thickness trends. The reliability of the best neural network solutions was checked by comparing them with a built ad hoc regression model. The multi-layer perceptron (MLP) neural network trained with back-propagation (BP) algorithm was found to be the fittest model. Besides, a genetic algorithm (GA) was also employed to improve the capability of MLP model to provide the best fit of experimental results all over the investigated ranges. Finally, a verification experimental plan was performed and a related analytical model was developed to check the reliability of the neural network model with GA to predict the whole coating thickness trends according to the operational parameters. A comparison between the neural network model and an analytical model was also carried out. [Copyright &y& Elsevier]

Abstract: This paper recapitulates the results of a long research on a family of artificial intelligence (AI) methods—relying on, e.g., artificial neural networks and search techniques—for handling systems with high complexity, high number of parameters whose input or output nature is partly unknown, high number of dependencies, as well as uncertainty and incomplete measurement data. Aside from classical modelling, basic problem solving and optimization techniques are presented. Finally, a novel submodel decomposition method is shown with an extended feature selection algorithm highlighted, along with possibilities of further development. Examples of practical application are shown to illustrate the viability of the methods. [Copyright &y& Elsevier]

Abstract: In this paper, a hybrid neural network model, based on the integration of fuzzy ARTMAP (FAM) and the rectangular basis function network (RecBFN), which is capable of learning and revealing fuzzy rules is proposed. The hybrid network is able to classify data samples incrementally and, at the same time, to extract rules directly from the network weights for justifying its predictions. With regards to process systems engineering, the proposed network is applied to a fault detection and diagnosis task in a power generation station. Specifically, the efficiency of the network in monitoring the operating conditions of a circulating water (CW) system is evaluated by using a set of real sensor measurements collected from the power station. The rules extracted are analyzed, discussed, and compared with those from a rule extraction method of FAM. From the comparison results, it is observed that the proposed network is able to extract more meaningful rules with a lower degree of rule redundancy and higher interpretability within the neural network framework. The extracted rules are also in agreement with experts’ opinions for maintaining the CW system in the power generation plant. [Copyright &y& Elsevier]

Abstract: In this paper, classifiers based on Multi-Layer Perceptrons and Support Vector Machines are used in order to classify seismic events that occurred in metropolitan France. The results are exploited in the software RAMSES to help the seismic analysts to conduct efficiently the revision of the weekly French seismic bulletin. With 96.5% of good classification, and less than 7% of the events emphasized for verification, RAMSES strikingly improves the speed of the revision. [Copyright &y& Elsevier]

Abstract: The design of Non-Destructive-Testing systems for fault detection in long and not accessible pipelines is an actual task in the industrial and civil environment. At this purpose, the diagnosis based on the propagation of guided ultrasonic waves along the pipes offers an attractive solution for the fault identification and classification. The authors studied this problem by means of suitable Artificial Neural Network models. Numerical techniques have been used to simulate the guided wave propagation in the pipes. In particular, the finite element method has been used to model different kinds of pipes and faults, and to obtain several returning echoes containing the faults information. Torsional wave modes have been used as excitation waves. The obtained signals have been processed in order to reduce the data dimensionality, and to extract suitable features. The features selected from the signals can be further processed in order to limit the size of the Neural Network models without loss of information. At this purpose, the principal component analysis has been investigated. Finally, the selected features have been used as input for the neural network models. In this paper, traditional feed-forward, multi-layer perceptron networks have been used to obtain the information on size and location of localized notches. [Copyright &y& Elsevier]

Abstract: This paper deals with the problem of finding the optimum load allocation on machines and apparatuses in complex Cogeneration Heat and Power (CHP) plants. A methodology based on Neural Networks (NN) has been developed. A database has been populated by using a real plant simulator. Two kinds of plant neural models have been trained, the first consists in an Identification Neural Model (INM) that provides a “picture” of the actual plant status by using monitoring data as input; the second consists in an Optimum Load Allocation Neural Model (OLANM) whose inputs are boundary conditions and outputs the Degrees of Freedom corresponding to the optimum operation set points. To reduce the relevant computational effort required to populate the training databases a sequential chain of neural models has been arranged. The methodology has been applied to a typical industrial cogeneration plant installed in Turin (Italy). Results are presented and discussed. [Copyright &y& Elsevier]

Abstract: The paper deals with the identification of recurrent neural networks (RNNs) for simulating the air–fuel ratio (AFR) dynamics into the intake manifold of a spark ignition (SI) engine. RNN are derived from the well-established static multi layer perceptron feedforward neural networks (MLPFF), that have been largely adopted for steady-state mapping of SI engines. The main contribution of this work is the development of a procedure that allows identifying a RNN-based AFR simulator with high generalization and limited training data set. The procedure has been tested by comparing RNN simulations with AFR transients generated using a nonlinear-dynamic engine model. The results show how training the network making use of inputs that are uncorrelated and distributed over the entire engine operating domain allows improving model generalization and reducing the experimental burden. Potential areas of application of the procedure developed can be either the use of RNN as virtual AFR sensors (e.g. engine or individual AFR prediction) or the implementation of RNN in the framework of model-based control architectures. [Copyright &y& Elsevier]

Abstract: Voltage stability has become a major concern among the utilities over the past decade. With the development of FACTS devices, there is a growing interest in using these devices to improve the stability. In this paper a method using parallel self-organizing hierarchical neural network (PSHNN) is proposed to estimate the loadability margin of the power system with static var compensator (SVC). Limits on reactive generations are considered. Real and reactive power injections along with firing angle of SVC and bus voltage at which SVC is connected, are taken as input features. To improve the performance of network, K-means clustering is employed to form the clusters of patterns having similar loadability margin. To reduce the number of input features in each cluster, system entropy information gain method is used and only those real and reactive power injections, which affect the loadability margin most, are selected. Separate PSHNN is trained for each cluster. The proposed method is implemented on IEEE-30 bus and IEEE-118 bus system. Once trained, the network produces the output, with accuracy and speed. The computation time is also independent of the system size and the load pattern. [Copyright &y& Elsevier]

Abstract: This paper describes two data analysis techniques adopted in a Decision Support System (DSS) that aids users in predicting oil production of an infill well. The system generates predictions in the form of a possible range of cumulative production and length of production life of an infill well. The system also shows the worst and best case scenarios based on different production curves so that the expert can examine the plots of predicted production rates for each existing well and decide which model gives the best fit. The production curve of each individual well was mathematically modeled so that production values beyond the historical data can be produced. Decline curve estimation and neural network approaches were adopted for data analysis in the system. The system was tested with data from two groups of wells from two different fields in Saskatchewan, Canada. Observations on the suitable duration that the historical data set should cover and a comparison among different curve estimation and neural network models are presented. [Copyright &y& Elsevier]

Jet engines are nonlinear dynamical systems for which an exact mathematical model cannot be used for estimator design, because it is either not available or so complex that it does not fit the necessary assumptions. Thus, classical analytical tools for studying standard system properties like observability, which is very important in estimator design, cannot be directly applied. Generally, for practical jet engine applications, the designer faces two closely related problems: first, given a non-measurable parameter, find the minimal set of estimator inputs that facilitates achieving a satisfactory estimation performance (input selection); second, given a predetermined set of inputs, derive an “observability” measure that characterizes the estimation feasibility of a specific non-measurable parameter. In this paper, techniques for solving these two problems are developed and applied to estimator design for jet engine thrust, stall margins, and an unmeasurable state. [Copyright &y& Elsevier]

This paper presents a neural decoder for trellis coded modulation (TCM) schemes. Decoding is performed with Radial Basis Function Networks and Multi-Layer Perceptrons. The neural decoder effectively implements an adaptive Viterbi algorithm for TCM which learns communication channel imperfections. The implementation and performance of the neural decoder for trellis encoded 16-QAM with amplitude imbalance are analyzed. [Copyright &y& Elsevier]

This paper discusses the application of neural networks to the white tracking adjustment of television receivers during production. High quality levels of tracking for the color temperature 8000 K were obtained with four-layer (7-10-10-6) network. The network input set consists of brightness level, high and low luminance levels, and “x” and “y” coordinates on the chromaticity diagram for both high and low luminance. The network output set consists of recommended adjustments for brightness, red, green, and blue cutoffs, and green and blue gains. The network was trained using the back-propagation algorithm. The experimental study has shown that the application of neural networks has reduced the testing time which has led to an increase in production rate. [Copyright &y& Elsevier]

In this paper multilayer neural networks are used to control the balancing of a base-excited inverted pendulum. The pendulum has 2 degrees of rotational freedom and the base-point moves freely in three-dimensional space. The goal is to apply control torques to keep the pendulum in a prescribed orientation in spite of disturbing base-point movement. A control algorithm is proposed that utilizes a set of neural networks to compensate for the effect of system''s nonlinearities. These networks are updated on-line, according to a learning algorithm, which guarantees the stability of the closed-loop system. Furthermore, since the pendulum''s base-point movement is considered unmeasurable, a novel neural inverse model is employed to estimate it from measurable variables. The proposed neural controller has been tested through simulations. Its performance has also been compared with the performance of the most recently developed control technique on the same problem. It is shown that the proposed neural controller produces fast, yet well maintained damped responses with reasonable control torques and without a knowledge of the model or model parameters. Additionally, the developed controller does not require measurement of the base-point accelerations, which are difficult to obtain. The work presented here benefits practical problems such as the study of stable locomotion of human upper body and bipedal locomotion. [Copyright &y& Elsevier]

In this paper, a hybrid intelligent tracking control approach is developed to address optimal tracking problems for a class of nonlinear discrete-time systems. The generalized value iteration algorithm is utilized to attain the admissible tracking control with off-line training, while the on-line near-optimal control method is established to enhance the control performance. It is emphasized that the value iteration performance is improved by introducing the acceleration factor. By collecting the input–output data of the unknown system plant, the model neural network is constructed to provide the partial derivative of the system state with respect to the control law as the approximate control matrix. A novel computational strategy is introduced to obtain the steady control of the reference trajectory. The critic and action neural networks are utilized to approximate the cost function and the tracking control, respectively. Considering approximation errors of neural networks, the stability analysis of the specific systems is provided via the Lyapunov approach. Finally, two numerical examples with industrial application backgrounds are involved for verifying the effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]

This article relates to the topic of the escalator demand response potential. Previous studies mapped escalators as an unrealized potential for additional demand response. The decrease of the nominal speed is the proposed method of reducing the power consumption of an escalator that comes at the cost of passenger travel time and queuing. This work proposes a solution to a problem of selecting appropriate escalators from a large pool to accommodate the target of power curtailment at a minimum cost and highlights the escalator features that constitute the best demand response candidates. The paper compares four methods which differ in calculation speed and accuracy. The primal solution is the earlier developed and enhanced simulation-based model. The random forest and the neural network models provide a solution trained on the output of the simulation-based model aiming to enhance the calculation speed. Finally, all of the developed solutions are compared to the random selection of escalators. The comparison of the proposed statistical approaches shows that the random forest outperforms the neural networks with a maximum error in the prediction of the overall costs in the range of 10.5% of the simulation-based model solution, while the neural network solution lies within 10%–58%, depending on the targeted value of the power reduction. Statistical approaches enable performing predictions for different times of the day and for new escalator populations without the need for time-demanding simulations. Comparison to the random selection of escalators demonstrates that the proposed models generally outperform the random selection at least seven-fold. [ABSTRACT FROM AUTHOR]