Search

Using advanced machine learning techniques as an alternative to conventional double-entry volume equations, a regression model of the inside-bark volume (dependent variable) for standing Eucalyptus globulus trunks (or main stems) has been built as a function of the following three independent variables: age, height and outside-bark diameter at breast height (DBH). The experimental observed data (age, height, outside-bark DBH and inside-bark volume) for 142 trees ( E. globulus ) were measured and a nonlinear model was built using a data-mining methodology based on support vector machines (SVM) and multilayer perceptron networks (MLP) for regression problems. Coefficients of determination and Furnival’s indices indicate the superiority of the SVM with a radial kernel over the allometric regression models and the MLP.

Accurate wind speed forecasting is important in power grid security, power system management, operation and market economics. However, most research has focused only on improving either accuracy or stability, with few studies addressing the two issues, simultaneously. Therefore, we proposed a novel combined model based on multi-objective particle swarm optimization, which is applied to optimize the key parameters of the echo state network. Most combined wind speed forecasting methods just use the combination theory to combine individual methods, this paper uses echo state network to combine the intermediate wind speed forecasting results of three artificial neural networks. Moreover, a new dataset division mechanism based on the moving window is applied in this paper. Firstly, the length of the input data is changed from 5 to 15 for 1-step, 2-step and 3-step wind speed forecasting, after that, the optimal length of the input vector can be got. And then we apply this optimal length of the input vector to another dataset for further verifying the proposed method. In order to verify the forecasting effectiveness of the proposed forecasting model, the 80/min wind speed data of M2 tower of the National Wind Power Technology Center of the United States were taken as an example. The experimental results indicate that the proposed algorithm is superior to the other ten comparative models in prediction accuracy and stability, and it also performs better than the combined model that we have proposed before.

The paper addresses the problem of lumpy demand forecasting which is typical for spare parts. Several prediction methods are presented in the paper - traditional techniques based on time series and advanced methods which use artificial neural networks. The paper presents a new hybrid spares demand forecasting method dedicated to mining companies. The method combines information criteria, regression modeling and artificial neural networks. The paper also discusses simulation research related to efficiency assessment of the chosen variable selection methods and its application in the newly developed forecasting method. The assessment of this method is conducted by a comparison with traditional methods and is based on selected forecast errors.

This paper deals with the finite-time synchronization issue of time-varying delayed neural networks (DNNs) with discontinuous activations. Based on master-slave concept, several sufficient conditions are given to guarantee the finite-time synchronization of discontinuous DNNs. In order to control the synchronization error to converge zero in a finite time, we design three classes of novel switching state-feedback controllers which involve time-delays and discontinuous factors. The analysis in this paper employs the extended differential inclusion theory, the famous finite-time stability theorem, inequality techniques and generalized Lyapunov approach. Moreover, the upper bounds of the settling time of synchronization are estimated. Finally, the validity of proposed design method and theoretical results are illustrated by numerical examples.

Sequential processes appear naturally in all types of industries. Biotechnology is a good example of such schemes. Wastewater treatment using microbiological activity is a particular case having all the characteristics of sequential methods. Sulfate reduction as pre-treatment followed by the decomposition of sulfated compounds using adapted microorganisms is the sequential nonlinear process with state constrains analyzed in this paper. Modeling this procedure is still a difficult task because the number of elements involved in the reaction. This paper presents an adaptive algorithm to obtain a suitable model of this process using continuous neural networks. The adaptive model preserves the sequential nature of the process as well as the bounded nature of all states. The neural network is proposed as a system identifier in terms of the hybrid systems theory. The Lyapunov stability method is used to demonstrate the convergence of the identifier states to the real concentrations of the microbiological system. Experimental results and their corresponding simulation using the adaptive model based on neural networks confirm the theoretical results described in this paper.

Simulation is generally used to study non-deterministic problems in industry. When a simulation process finds the solution to an NP-hard problem, its efficiency is lowered, and computational costs increase. This paper proposes a stochastic dynamic lot-sizing problem with asymmetric deteriorating commodity, in which the optimal unit cost of material and unit holding cost would be determined. This problem covers a sub-problem of replenishment planning, which is NP-hard in the computational complexity theory. Therefore, this paper applies a decision system, based on an artificial neural network (ANN) and modified ant colony optimization (ACO) to solve this stochastic dynamic lot-sizing problem. In the methodology, ANN is used to learn the simulation results, followed by the application of a real-valued modified ACO algorithm to find the optimal decision variables. The test results show that the intelligent system is applicable to the proposed problem, and its performance is better than response surface methodology.

This paper investigates the dynamics of a class of recurrent neural networks where the neural activations are modeled by discontinuous functions. Without presuming the boundedness of activation functions, the sufficient conditions to ensure the existence, uniqueness, global exponential stability and global convergence of state equilibrium point and output equilibrium point are derived, respectively. Furthermore, under certain conditions we prove that the system is convergent globally in finite time. The analysis in the paper is based on the properties of M-matrix, Lyapunov-like approach, and the theories of differential equations with discontinuous right-hand side as introduced by Filippov. The obtained results extend previous works on global stability of recurrent neural networks with not only Lipschitz continuous but also discontinuous neural activation functions.

The paper presents a comparative study on representative methods for model-based and model-free control of flexible-link robots. Model-based techniques for the control of flexible-link robots can come up against limitations when an accurate model is unavailable, due to parameters uncertainty or truncation of high order vibration modes. On the other hand, several research papers argue that suitable model-free control methods result in satisfactory performance of flexible-link robots. In this paper two model-free approaches of flexible-link robot control are examined: (i) energy-based control, and (ii) neural adaptive control. The performance of the aforementioned methods is compared to the inverse dynamics model-based control, in a simulation case study for planar 2-DOF manipulators.

When modeling neural networks in a real world, not only diffusion effect and fuzziness cannot be avoided, but also self-inhibitions, interconnection weights, and inputs should vary as time varies. In this paper, we discuss the dynamical behaviors of delayed reaction–diffusion fuzzy cellular neural networks with varying periodic self-inhibitions, interconnection weights as well as inputs. By using Halanay’s delay differential inequality, M -matrix theory and analytic methods, some new sufficient conditions are obtained to ensure the existence, uniqueness, and global exponential stability of the periodic solution, and the exponentially convergent rate index is also estimated. In particular, the traditional assumption on the differentiability of the time-varying delays is no longer needed. The methodology developed in this paper is shown to be simple and effective for the exponential periodicity and stability analysis of neural networks with time-varying delays. Two examples are given to show the usefulness of the obtained results that are less restrictive than recently known criteria.

In this paper, we present a general class of BAM neural networks with discontinuous neuron activations and impulses. By using the fixed point theorem in differential inclusions theory, we investigate the existence of periodic solution for this neural network. By constructing the suitable Lyapunov function, we give a sufficient condition which ensures the uniqueness and global exponential stability of the periodic solution. The results of this paper show that the Forti’s conjecture is true for BAM neural networks with discontinuous neuron activations and impulses. Further, a numerical example is given to demonstrate the effectiveness of the results obtained in this paper.

In this paper, a new concept called α-inverse Lipschitz function is introduced. Based on the topological degree theory and Lyapunov functional method, we investigate global convergence for a novel class of neural networks with impulses where the neuron activations belong to the class of α-inverse Lipschitz functions. Some sufficient conditions are derived which ensure the existence, and global exponential stability of the equilibrium point of neural networks. Furthermore, we give two results which are used to check the stability of uncertain neural networks. Finally, two numerical examples are given to demonstrate the effectiveness of results obtained in this paper.

This paper describes the mathematical modeling and control of the tension and the speed of moving web in a web-fed machine. In this paper the web control system is structured as a two-input and two-output system. To lessen the effects of system uncertainties, such as system parameters, unmodeled errors, and external disturbance uncertainties in the web control system, a sliding mode control (SMC) is proposed, with an estimator based on a recurrent neural network, which is applied to estimate system uncertainties. Some simulation results are presented to show that the proposed control scheme is feasible for a web-fed machine.

This paper is to explore the dynamical analysis and active control synthesis for improving seaport operations through system optimization. First, the multi-dimensional system of the fractional Lotka-Volterra model is utilized to investigate its underlying dynamics of port competition and cooperation for container throughput. The nonlinear dynamical behaviors are intensely examined by using eigenvalue evaluation, bifurcation analysis and time series analysis to illustrate the stability of the competition and collaboration model. With the system theory based on fractional order calculus, various dynamic properties of ports system have been discovered. Based on the dynamical analysis, adaptive fractional-order sliding mode control with artificial neural network algorithm has been realized for port growth rate improvement against disruptions. Furthermore, the proposed approach is successfully validated by the case study of major Korean seaports for improving port operations through neural network prediction. The decision making policies implemented by control algorithms are guaranteed to improve the port growth rate against disruptions. Specifically, the novel management strategy is to provide managerial insights and innovative solutions for seaport authorities who can make more efficient strategic planning for handling risk management in maritime logistics.

Accurate carbon price forecast plays a vital role in energy conservation, emission reduction and environmental protection. In previous studies, more attention was focused on the prediction accuracy and stability, while the problem of disharmony between the prediction model and the data pattern is usually ignored. Considering the matching utility with deeper understanding of data and model, this paper proposes a novel approach to forecast carbon price, which combines the data preprocessing mechanism, decomposition technology, forecast module with selection and matching strategy and ensemble model based on an original hybrid optimization algorithm. According to a comprehensive evaluation index in consideration of several evaluation perspectives, the optimal parameter structures of the three forecast models are selected in this framework. Then, the data components decomposed by variational mode decomposition are reconstructed into three novel range entropy series with different levels of complexity by range entropy. As a result, the matching relation between the three forecasting models and the three range entropy series is correspondingly established. Additionally, a feedback neural network optimized by hybrid optimization algorithm, which persists more superiorities of reasonable weight assignment than the usual ensemble method, is initially used to synthesize three forecasting results of range entropy series. The carbon price data from four different trading markets in China is used to test the novel approach and the experimental results indicate that it does enhance the performance of carbon price forecasting, and provide a convincing tool for the operation and investment of the carbon markets.

This paper studies the globally repetitive learning leader-following consensus for a class of unknown nonlinear multi-agent systems with uncertain periodic time-varying parameters. Neural networks are used as the feed-forward compensators to model some unknown nonlinear system functions. According to the different distributions of agents in the network topology, and by learning periodic uncertainties, a new hybrid repetitive learning control protocol is presented based on Lyapunov theory. Meanwhile, the proposed learning control protocol can guarantee that all the followers can track the leader asymptotically. Moreover, the boundedness of all the signals is ensured. Finally, three simulation examples are provided to verify the effectiveness of the proposed scheme.

This paper presents complete steps for the construction and solution algorithm for a projection-based reduced order model coupled with an artificial neural net model. The reduced model is constructed based on the Galerkin projection of the equations governing the physical processes on reduced subspaces obtained by the Proper Orthogonal Decomposition method. The projection is done using numerical discretisation schemes implemented in the commonly used OpenFOAM® platform. The neural net model is trained using the Nonlinear AutoRegressive eXogenous model network (NARX) architecture. The reduced order models are demonstrated for true predictions of incompressible flows at low Reynolds numbers driven by various boundary conditions. Compared with the full CFD simulations, this model shows excellent agreement while only requiring a fraction of the computational resources.

The paper is aimed at enhancing computational performance for optimizing the material distribution of tri-directional functionally graded (FG) plates. We exploit advantages of using a non-uniform rational B-spline (NURBS) basis function for describing material distribution varying through all three directions of functionally graded (FG) plates. Two-dimensional free vibration and buckling behaviors of multi-directional (1D, 2D and 3D) FG plates analyzed by using a combination of generalized shear deformation theory (GSDT) and isogeometric analysis (IGA) is first proposed. This approach can help to save a significant amount of computational cost while still ensure the accuracy of the solutions. The effectiveness and reliability of the present method are demonstrated by comparing it to other methods in the literature. The obtained results are in excellent agreement with the reference ones. More importantly, data sets consisting of input-output pairs are randomly generated from the analysis process through iterations for the training process in deep neural networks (DNN). DNN is utilized as an analysis tool to supplant finite element analysis to reduce computational cost. By using DNN, behaviors of the multi-directional FG plates are directly predicted from those material distributions. Optimal material distributions of tri-directional FG plates under free vibration or compression in various volume fraction constraints are found by using modified symbiotic organisms search (mSOS) algorithm for the first time. Moreover, an isogeometric multimesh design technique is also used to diminish a large number of design variables in optimization. Optimal results obtained by DNN are compared with those of IGA to verify the effectiveness of the proposed method.

As frost accumulates on the heat exchanger surface with time, system operating performance will be dramatically degraded, and limit its use in climates susceptible to frost formation. A novel self-adaptive control strategy of frost prevention and retardation for air source heat pumps (ASHP) is introduced in this paper. The control strategy relies on a new thermodynamic model, which involves a Dimensionless Artificial Neural Network (DANN) correlation model describing frost accumulation for ASHP on the air-side of the fin-and-tube heat exchanger. The dimensionless parameters of this DANN model, including the ambient conditions, 6 commonly used refrigerants, and the geometric parameters of the heat exchanger, are considered in the model. To enhance the reliability of DANN, we develop a self-adaptive algorithm, including determining the optimal transfer algorithm and selecting the number of neurons in the hidden layer, for the DANN model. Results show a limited relative error (7.55%) between calculated values and experimental data, which help researchers and manufacturers analyze the complicated frosting process and design the new ASHPs more reasonably in different regions with different ambient conditions.

In this paper, we consider a special kind of time-dependent nonlinear system that includes different-level subsystems (i.e., a subsystem with respect to time-dependent solution and a subsystem with respect to its time derivative). To solve this kind of time-dependent nonlinear system, an equivalency between different-level subsystems is proposed and termed as zeroing equivalency based on the zeroing neural network method. This kind of time-dependent nonlinear system can be solved in a discrete-time manner with future-instant solution predicted at current instant based on the zeroing equivalency and a discretization formula. Prediction ability can perfectly meet the requirements of real-time computation, which is crucial in engineering fields. In addition, different-level algorithms are further employed to solve the problem of robotic system control considering additional restrictions. The position tracking control and end-effector orientation control are simultaneously achieved with even fault tolerance by utilizing the proposed algorithms.

The use of X-Ray Computed Tomography scanners to better characterize rock properties behavior at micro-scale is becoming increasingly common in oil industry. In this paper, we propose a new approach based on modeling X-Ray Computed Tomography images in terms of 2D textures in order to predict rock properties and classify main textures along core samples. First, we implement a parametric model of textures based on a multi-scale analysis to extract main representative textural descriptors. Then, we use Kohonen unsupervised classification technique to find main representative textures and classify core samples images. In addition, we simulate several rock properties such as porosity, density, formation factor and volume of clay along cores using a neural network system. Finally, we compare our simulation results with experimental real data and discuss main advantages and limitations of our approach.

Short-term power load forecasting is receiving increasing attention, especially because of intrinsic difficulties and practical applications. In this paper, the novel integrated approaches, combining longitudinal data selection (LDS), singular spectrum analysis (SSA) technique, adaptive particle swarm optimization based on gravitational search algorithm (APSOGSA) and the nonlinear multi-layer perceptron neural network (NMLPNN), were proposed for the short-term power load forecasting. Firstly, the LDS, which guarantees that the input and output data have the same properties to ensure abundant performance. Then, the SSA technique is used for identifying and extracting the trend and seasonality of power load time series. Finally, the NMLPNN, which is optimized by the APSO, GSA, and APSOGSA, is utilized to deal with the irregularity and volatility of the power load. These integrated methods are applied to forecast half-hour power load data from New South Wales, Queensland and Singapore. By comparison of the obtained experimental results, the proposed SSA-APSOGSA-NMLPNN integrated method indicates the superiority and promising performance and has a good robustness. (C) 2015 Elsevier Inc. All rights reserved.

Two models for the forecasting of the triangular fuzzy-number (TF) series are presented in this paper. One is the triangular fuzzy-number grey model (TFGM (1, 1)). In this model, the fundamental equation of GM (1, 1) has been improved to be available for the application on the TF series. However, TFGM (1, 1) can be only applicable to the forecasting of the TF series with weak fluctuation because the essence of GM (1, 1) is to match the raw series with an exponential-type curve. In order to make it applicable to the forecasting of the fluctuating TF series, the neural network model is introduced to amend TFGM (1, 1). In the process of amendment, the TF series has been transformed into three real number series in order to avoid the disorder of the relative positions of the three boundary points of the triangular fuzzy number. Then the other model, the neural network TFGM (1, 1) (NNTFGM (1, 1)), is presented. The prediction of Consumer Price Indexes and the power load of one district of China illustrate that for the fluctuating TF series, the forecast accuracy of NNTFGM (1, 1) is higher than that of TFGM (1, 1).

This paper investigates incorporation of the neural network (NN) into the vector tracking loop (VTL) for improving the Global Positioning System (GPS) positioning quality during satellite signal blockages. The VTL architectures of the GPS receivers provide several important advantages as compared to the scalar tracking loop (STL) architectures. The tracking and navigation modules in the VTL are not independent anymore as compared to those in the traditional STL since the user’s position can be determined by using the information from the satellites with stronger signals and the weak signals in turn can then be predicted on the basis of the states of the user. Signal blockage and reflections from buildings and other large, solid objects can lead to accuracy degradation. One of the merits of the VTL is that the tracking loop can be assisted from the navigation solution. The NN is used to bridge the GPS signal and prevent the error growth due to signal outage from contaminating the entire tracking loop. In the training stage, the output information from the discriminator and navigation filter is adopted as the inputs of the neural network. The NN is employed for predicting adequate numerical control oscillator (NCO) inputs, i.e., providing better prediction of residuals for the Doppler frequency and code phase in order to maintain regular operation of the navigation system. Two types of networks involved are the radial basis function neural network (RBFNN) and the adaptive network-based fuzzy inference system (ANFIS). Performance evaluation for the NN aided architecture as compared to the unaided VTL is carried out. The results show that the VTL with the support of neural networks can effectively provide improved performance during GPS signal blockages.

In this paper, the problem of pth moment exponential stability analysis is considered for stochastic delayed hybrid systems with Levy noise. By the technique of stochastic analysis as well as M-matrix theories, we propose some sufficient conditions to guarantee the exponential stability of the system. The criterion of mean square exponential stability for delayed neural networks with Levy noise is derived as well. A numerical example is provided to show the usefulness of the proposed exponential stability criterion.

Thermal errors are often quoted as being the largest contributor to CNC machine tool errors, but they can be effectively reduced using error compensation. The performance of a thermal error compensation system depends on the accuracy and robustness of the thermal error model and the quality of the inputs to the model. The location of temperature measurement must provide a representative measurement of the change in temperature that will affect the machine structure. The number of sensors and their locations are not always intuitive and the time required to identify the optimal locations is often prohibitive, resulting in compromise and poor results. In this paper, a new intelligent compensation system for reducing thermal errors of machine tools using data obtained from a thermal imaging camera is introduced. Different groups of key temperature points were identified from thermal images using a novel schema based on a Grey model GM (0, N) and fuzzy c-means (FCM) clustering method. An Adaptive Neuro-Fuzzy Inference System with fuzzy c-means clustering (FCM-ANFIS) was employed to design the thermal prediction model. In order to optimise the approach, a parametric study was carried out by changing the number of inputs and number of membership functions to the FCM-ANFIS model, and comparing the relative robustness of the designs. According to the results, the FCM-ANFIS model with four inputs and six membership functions achieves the best performance in terms of the accuracy of its predictive ability. The residual value of the model is smaller than ±2 μm, which represents a 95% reduction in the thermally-induced error on the machine. Finally, the proposed method is shown to compare favourably against an Artificial Neural Network (ANN) model.

This paper represents the first study to use the grey model (GM) for predicting CO 2 , SO 2 and O 2 in the emissions from a medical incinerator. The artificial neural network (ANN) was also employed for comparison. Four control parameters were served as the input variables. The results indicated that two control parameters of temperature highly influenced air pollutant emissions. The minimum mean absolute percentage errors of 3.70%, 6.11% and 1.08% for CO 2 , SO 2 and O 2 could be achieved using GMs, meanwhile the minimum root mean squared errors for three air pollutant were 0.1660, 2.4521 and 0.2112. The control parameters could be applied to the prediction of air pollutant emissions. It also revealed that GM could predict the air pollutant emissions even though emission data were not sufficient.

In this paper, we introduce a new type neural networks by superpositions of a sigmoidal function and study its approximation capability. We investigate the multivariate quantitative constructive approximation of real continuous multivariate functions on a cube by such type neural networks. This approximation is derived by establishing multivariate Jackson-type inequalities involving the multivariate modulus of smoothness of the target function. Our networks require no training in the traditional sense.

Although studied for years, due to their dynamic nature, research in the field of mobile ad hoc networks (MANETs) has remained a vast area of interest. Since once distributed, there will be less to no plausibility of recharge, energy conservation has become one of the pressing concerns regarding this particular type of network. In fact, one of the main obligations of designers is to make efficient use of these scarce resources. There has been tremendous work done in different layers of protocol stack in order to intensify energy conservation. To date, numerous topology control algorithms have been proposed, however, only a few have used meta-heuristics such as genetic algorithms, neural networks and/or learning automata to overcome this issue. On the other hand, since nodes are mobile and thus in a different spatial position, as time varies, we can expect that by regulating time intervals between topology controls, one may prolong the network’s lifetime. The main initiative of this paper is to intensify energy conservation in a mobile ad hoc network by using weighted and learning automata based algorithms. The learning automata, regulates time intervals between which the topology controls are done. The represented learning automata based algorithm uses its learning ability to find appropriate time-intervals so that the nodes would regulate the energy needed in order to exchange the information to their neighbors, accordingly. Moreover, at first we have represented two weighted based algorithms which extend two prominent protocols, namely K-Neigh and LMST. Then these algorithms are combined with a learning based algorithm which regulates time intervals between which the topology controls are done. In comparison with approaches that are based on periodic topology controls, proposed approach shows enhanced results. On the other hand, considering the learning ability of the learning automata based algorithms, composition of the aforementioned algorithms has been proven to be enhanced, in the respect of energy consumed per data transmitted, over those compared with.

In this paper an alternative approach for identification problems is discussed. Unlike existing methods, this new approach combines in a general way finite differences and function approximation and is herein used for the identification of a particular system in structural dynamics, that is the damped Duffing oscillator subject to a swept-sine excitation. The solution obtained by means of the proposed method has been compared with the one obtained by a neural network. The present method gives better results at a low computational cost, with the advantage of solutions in explicit form. Besides, it is possible to prove that the solutions are stable and that from this new approach one can deduce, as a particular case, the approximation previously proposed by other authors.

In this paper, vibrational resonance in excitable neuron populations with synapses is investigated by numerical simulation. In particular, the effect of the hybrid synapses on the signal detection and transmission in neural system is studied. Different topologies from regular and random networks to small-world networks are considered to analyze the dependence of vibrational resonance on the network structure and parameters. It is shown that there exists an optimal amplitude of high-frequency driving, enhancing the response of coupled neuron populations to a subthreshold signal. We find that chemical synaptic coupling is more efficient than the electrical coupling in signal detection and electrical synaptic coupling is better in signal transmission. Neuron populations with hybrid synapses compromise the merits of the two types of coupling and have an advantage in information communication.

Relative Radiometric Normalization is often required in remote sensing image analyses particularly in the land cover change detection process. Normalization process minimizes the radiometric differences between two images caused by inequalities in the acquisition conditions rather than changes in surface reflectance. A wide range of RRN methods have been developed to adjust linear models. This paper proposes an automated Relative Radiometric Normalization (RRN) method to adjust a non-linear model based on an Artificial Neural Network (ANN) and unchanged pixels. The proposed method includes the following stages: (1) automatic detection of unchanged pixels based on a new idea that uses CVA (Change Vector Analysis) method, PCA (Principal Component Analysis) transformation and K-means clustering technique, (2) evaluation of different architectures of perceptron neural networks to find the best architecture for this specific task and (3) use of the aforementioned network for normalizing the subject image. The method has been implemented on two images taken by the TM sensor. Experimental results confirm the effectiveness of the presented technique in the automatic detection of unchanged pixels and minimizing imaging condition effects (i.e., atmosphere and other effective parameters).

This paper intended to offer an architecture of artificial neural networks (NNs) for finding approximate solution of a second kind linear Fredholm integral equations system. For this purpose, first we substitute the N -th truncation of the Taylor expansion for unknown functions in the origin system. By applying the suggested neural network for adjusting the real coefficients of given expansions in resulting system. The proposed NN is a two-layer feed-back neural network such that it can get a initial vector and then calculates it’s corresponding output vector. In continuance, a cost function is defined by using output vector and the target outputs. Consequently, the reported NN using a learning algorithm that based on the gradient descent method, will adjust the coefficients in given Taylor series. Eventually, we have showed this method in comparison with existing numerical methods such as trapezoidal quadrature rule provides solutions with good generalization and high accuracy. The proposed method is illustrated by several examples with computer simulations.

The low-mass loading gas cyclone separator has two performance parameters, the pressure drop and the collection efficiency (cut-off diameter). In this paper, a multi-objective optimization study of a gas cyclone separator has been performed using the response surface methodology (RSM) and CFD data. The effects of the inlet height, the inlet width, the vortex finder diameter and the cyclone total height on the cyclone performance have been investigated. The analysis of design of experiment shows a strong interaction between the inlet dimensions and the vortex finder diameter. No interaction between the cyclone height and the other three factors was observed. The desirability function approach has been used for the multi-objective optimization. A new set of geometrical ratios (design) has been obtained to achieve the best performance. A numerical comparison between the new design and the Stairmand design confirms the superior performance of the new design. As an alternative approach for applying RSM as a meta-model, two radial basis function neural networks (RBFNNs) have been used. Furthermore, the genetic algorithms technique has been used instead of the desirability function approach. A multi-objective optimization study using NSGA-II technique has been performed to obtain the Pareto front for the best performance cyclone separator.

This paper introduces an artificial neural network (ANN) application to a hot strip mill to improve the model’s prediction ability for rolling force and rolling torque, as a function of various process parameters. To obtain a data basis for training and validation of the neural network, numerous three dimensional finite element simulations were carried out for different sets of process variables. Experimental data were compared with the finite element predictions to verify the model accuracy. The input variables are selected to be rolling speed, percentage of thickness reduction, initial temperature of the strip and friction coefficient in the contact area. A comprehensive analysis of the prediction errors of roll force and roll torque made by the ANN is presented. Model responses analysis is also conducted to enhance the understanding of the behavior of the NN model. The resulted ANN model is feasible for on-line control and rolling schedule optimization, and can be easily extended to cover different aluminum grades and strip sizes in a straight-forward way by generating the corresponding training data from a FE model.

Bone adaptation models are often solved in the forward direction, meaning that the response of bone to a given set of loads is determined by running a bone tissue adaptation model. The model is generally solved using a numerical technique such as the finite element model. Conversely, one may be interested in the loads that have resulted in a given state of bone. This is the inverse of the former problem. Even though the inverse problem has several applications, it has not received as much attention as the forward problem, partly because solving the inverse problem is more difficult. A nonlinear system identification technique is needed for solving the inverse problem. In this study, we use artificial neural networks for prediction of tissue adaptation loads from a given density distribution of trabecular bone. It is shown that the proposed method can successfully identify the loading parameters from the density distribution of the tissue. Two important challenges for all load prediction algorithms are the non-uniqueness of the solution of the inverse problem and the inaccuracies in the measurement of the morphology of the tissue. Both challenges are studied, and it is shown that the load prediction technique proposed in this paper can overcome both.

In this paper, two types of recurrent neural network (RNN) are comparatively developed and exploited for the online solution of the well-known Lyapunov matrix equation with the stationary and nonstationary coefficients. Based on a new design method, the resultant Zhang neural networks (ZNN) are generalized and presented to solve the stationary and nonstationary problems with accuracy and efficiency. For comparison, the conventional gradient-based neural networks (GNN) are also used for the same problems. Computer simulation results show that, when used to solve the whether stationary or nonstationary problems, the convergence performance of ZNN solvers are superior than that of GNN solvers.

Supplier selection and evaluation is a complicated and disputed issue in supply chain network management, by virtue of the variety of intellectual property of the suppliers, the several variables involved in supply demand relationship, the complex interactions and the inadequate information of suppliers. The recent literature confirms that neural networks achieve better performance than conventional methods in this area. Hence, in this paper, an effective artificial intelligence (AI) approach is presented to improve the decision making for a supply chain which is successfully utilized for long-term prediction of the performance data in cosmetics industry. A computationally efficient model known as locally linear neuro-fuzzy (LLNF) is introduced to predict the performance rating of suppliers. The proposed model is trained by a locally linear model tree (LOLIMOT) learning algorithm. To demonstrate the performance of the proposed model, three intelligent techniques, multi-layer perceptron (MLP) neural network, radial basis function (RBF) neural network and least square-support vector machine (LS-SVM) are considered. Their results are compared by using an available dataset in cosmetics industry. The computational results show that the presented model performs better than three foregoing techniques.

With the ability to deal with high non-linearity, artificial neural networks (ANNs) and support vector machines (SVMs) have been widely studied and successfully applied to time series prediction. However, good fitting results of ANNs and SVMs to nonlinear models do not guarantee an equally good prediction performance. One main reason is that their dynamics and properties are changing with time, and another key problem is the inherent noise of the fitting data. Nonlinear filtering methods have some advantages such as handling additive noises and following the movement of a system when the underlying model is evolving through time. The present paper investigates time series prediction algorithms by using a combination of nonlinear filtering approaches and the feedforward neural network (FNN). The nonlinear filtering model is established by using the FNN’s weights to present state equation and the FNN’s output to present the observation equation, and the input vector to the FNN is composed of the predicted signal with given length, then the extended Kalman filtering (EKF) and Unscented Kalman filtering (UKF) are used to online train the FNN. Time series prediction results are presented by the predicted observation value of nonlinear filtering approaches. To evaluate the proposed methods, the developed techniques are applied to the predictions of one simulated Mackey–Glass chaotic time series and one real monthly mean water levels time series. Generally, the prediction accuracy of the UKF-based FNN is better than the EKF-based FNN when the model is highly nonlinear. However, comparing from prediction accuracy and computational effort based on the prediction model proposed in our study, we draw the conclusion that the EKF-based FNN is superior to the UKF-based FNN for the theoretical Mackey–Glass time series prediction and the real monthly mean water levels time series prediction.

Data classification is an important area of data mining. Several well known techniques such as decision tree, neural network, etc. are available for this task. In this paper we propose a Kalman particle swarm optimized (KPSO) polynomial equation for classification for several well known data sets. Our proposed method is derived from some of the findings of the valuable information like number of terms, number and combination of features in each term, degree of the polynomial equation etc. of our earlier work on data classification using polynomial neural network. The KPSO optimizes these polynomial equations with a faster convergence speed unlike PSO. The polynomial equation that gives the best performance is considered as the model for classification. Our simulation result shows that the proposed approach is able to give competitive classification accuracy compared to PNN in many datasets.

This paper presents a new technique for model order reduction (MOR) that is based on an artificial neural network (ANN) prediction. The ANN-based MOR can be applied for different scale systems with substructure preservation. In the proposed technique, the ANN is implemented for predicting the unknown elements of the reduced order model. Prediction of the ANN architecture is based on minimizing the cost function obtained by the difference between the actual and desired system behaviour. The ANN prediction process is pursued while maintaining the full order substructure in the reduced model. The proposed ANN-based model order reduction method is compared to recently published work on MOR techniques. Simulation results verify the validity of the new MOR technique.

The computational cost associated with the use of high-fidelity computational fluid dynamics (CFD) models poses a serious impediment to the successful application of formal sensitivity analysis in engineering design. Even though advances in computing hardware and parallel processing have reduced costs by orders of magnitude over the last few decades, the fidelity with which engineers desire to model engineering systems has also increased considerably. Evaluation of such high-fidelity models may take significant computational time for complex geometries. In many engineering design problems, thousands of function evaluations may be required to undertake a sensitivity analysis. As a result, CFD models are often impractical to use for design sensitivity analyses. In contrast, surrogate models are compact and cheap to evaluate (order of seconds or less) and can therefore be easily used for such tasks. This paper discusses and demonstrates the application of several common surrogate modelling techniques to a CFD model of flocculant adsorption in an industrial thickener. Results from conducting sensitivity analyses on the surrogates are also presented.

In this paper, we consider high-order recurrent neural networks with a class of general activation functions. By using some mathematical analysis techniques, we establish new results to ensure that all solutions of the networks converge exponentially to zero point.

In this paper minimization of the chromatic error in the data reduction pipeline of the Gaia mission is presented by applying polynomial cellular neural networks (PCNN). We introduce generalized PCNN model which enables us to solve the nonlinear approximation task. The advantage of the newly proposed method is in solving large-size image processing problem of diagnosis of astrometric chromaticity in real time. Rigorous stability analysis of the PCNN is presented by using the method of Lyapunov’s finite majorizing equations. The simulation results show a linear relation between the output of the proposed PCNN model and the chromaticity values as the target data.