Neural Network Modeling Method Research Articles

Robust modelling and prediction of thermally induced positional error based on grey rough set theory and neural networks

Thermal errors adversely affect the precision of a machine tool operation. Therefore, reduction of thermal errors is essential for improving of machining accuracy. In this paper, in order to realize real-time compensation, taking a multi-axis machine tool as an example and a method for robust modeling and predicting thermally induced positional error is proposed. First of all, the number of temperature measuring points required in thermal error’s model was reduced based on the rough set theory, which greatly reduced variable searching and modeling time. Then through grey relation theory, systematic analysis of the similarity degree between thermal error and temperature data was carried out to select sensitive temperature measuring points, and the temperature variables in the thermal error’s model were reduced from 24 to 7 after optimization, which eliminated the coupling problems. For reducing the influence of unpredictable noises, radial basis function (RBF) and back propagation (BP) neural network modeling methods were adopted to predict thermally induced positional errors, and in comparison, the prediction accuracy of RBF neural network was found superior to that of traditional BP neural network. Finally, some measured data were selected to verify the validity of the proposed method, and the results showed that prediction accuracy of the proposed thermally induced positional error model was reliable.

FORECASTING DEMAND FOR LOW COST CARRIERS IN AUSTRALIA USING AN ARTIFICIAL NEURAL NETWORK APPROACH

This study focuses on predicting Australia‘s low cost carrier passenger demand and revenue passenger kilometres performed (RPKs) using traditional econometric and artificial neural network (ANN) modelling methods. For model development, Australia‘s real GDP, real GDP per capita, air fares, Australia‘s population and unemployment, tourism (bed spaces) and 4 dummy variables, utilizing quarterly data obtained between 2002 and 2012, were selected as model parameters. The neural network used multi-layer perceptron (MLP) architecture that compromised a multi-layer feed-forward network and the sigmoid and linear functions were used as activation functions with the feed forward‐back propagation algorithm. The ANN was applied during training, testing and validation and had 11 inputs, 9 neurons in the hidden layers and 1 neuron in the output layer. When comparing the predictive accuracy of the two techniques, the ANNs provided the best prediction and showed that the performance of the ANN model was better than that of the traditional multiple linear regression (MLR) approach. The highest R-value for the enplaned passengers ANN was around 0.996 and for the RPKs ANN was round 0.998, respectively.

Novel neural network modeling method and applications

Neural networks play an important role for designing the parametric model of electromagnetic structures. The current neural network methods are unfit for a circuit model with many input variables because it is costly to extract a large number of the training data and test data to complete the highly nonlinear mapping approximation. This article proposes a new neural network modeling method-the multidimensional neural network model, which can be used to solve the issue of multivariable radiofrequency and microwave passive device modeling. The entire multidimensional neural network modeling problem is simplified into a set of neural network submodels through decomposition method. Then the submodels are combined into an equivalent model, and the final entire model is produced through the neural-network mapping model developed with the submodels and equivalent model. A microstrip hairpin filter model is developed using the proposed method. The simulation results show the correctness and the effectivity of the proposed method. © 2015 Wiley Periodicals, Inc. Int J RF and Microwave CAE 25:769-779, 2015.

Estimation of Weibull function parameters for modelling tree diameter distribution using least squares and artificial neural networks methods

For reliable forest management planning, knowledge of stand diameter distributions is valuable, since it allows, for example, calculation of merchantable volume when combined with a taper equation. Due to its flexibility, and its ability to describe a wide range of uni-modal distributions, the two-parameter Weibull function has been reported as being one of the most simple and accurate functions for modelling tree diameter distributions. However, the complex nonlinear nature of the tree-diameter distributions leads to laborious simulation of the probability density function in the Weibull distribution. Because of this weakness, an investigation was conducted using standard least squares and Levenberg–Marquardt artificial neural network method. These methods were used as inner procedures for the accurate estimation of the scale and shape parameters required in Weibull distribution modelling, using a) a method of moments and b) a maximum likelihood estimation. Data from Crimean Juniper stands grown in the Mediterranean region of Turkey was used. From the computational results it was concluded that the method which gives the most reliable estimates is the maximum likelihood estimation procedure with recovery of the Weibull distribution parmeters using the Levenberg–Marquardt artificial neural network modelling method.

Friction Stir Processing of Mild Steel/Al2O3 Nanocomposite: Modeling and Experimental Studies

Parameters of the friction stir processing of mild steel plates were correlated with microhardness of the stir zone using artificial neural network (ANN) modeling and experimental methods. For this purpose, the number of passes, rotational speed, traverse speed, and addition of nano-sized Al2O3 powder were considered as input parameters for the ANN model, while microhardness of the center of the stir zone was obtained as the output. To examine the accuracy and capability of the model in predicting the microhardness, the effects of all ANN input parameters on the microhardness were also examined experimentally with and without the addition of Al2O3 nanopowder. For the surface nanocomposites produced, increase in the number of passes and rotation speed led to increased microhardness values, whereas higher traverse speed resulted firstly in a rise in microhardness followed by a decreased microhardness. Using optical and scanning electron microscopy, the variations in microhardness were closely discussed based on the microstructural changes. Experimental results proved to show excellent conformity with the ANN model.

Research on the Stiffness Modeling of the High Precision Positioning Platform under the Thermal Model

The heat of the positioning platform is mainly caused by conduction of each functional unit, when it affected by the influence of internal and external heat source at work, the temperature of units would change, and the condition would make the platform component stiffness change, then lead to positioning platform dynamic characteristics change. In this paper, the finite method analysis platform dynamic connection stiffness with the regularity of temperature change, and by using the grey correlation method to optimize the platform stiffness influence of the four largest temperature sampling point, through the neural network modeling method to establish high precision positioning platform stiffness model, provide theory for the research on platform of dynamic characteristics of the model.

Early prediction in collective intelligence on video users’ activity

The huge volume of available video content calls for methods that offer insight to the content without necessitating burdensome users’ extra effort or being applicable to specific types or conditions. Based on experimentation on collective users’ interactions on a controlled user-experiment, this work analyses the results collected following the argument that bell-shaped reference patterns are shown to significantly correlate with scenes of interest for each video, as designated by the viewers. Though, in order to ensure the correlation of results to bell-shaped reference patterns, aggregation of a number of users’ interactions is required. In order to overcome such an impediment and adhere to a real-case cold start scenario, we propose a stochastic transformation of the aggregated users’ interaction signal into a space defined by its correlation to the bell-shaped reference patterns that is shown to offer significant amelioration as to the percentage of users’ interaction required in order to achieve comparable results to the original users’ interaction space. Moreover, to ensure further the realistic character of the proposed scenario, given an amount of already collected users’ interaction, the interaction of new users’ is shown to be predictable using neural network time series prediction and modeling methods. The results received indicate increased accuracy on how one can predict the most important scenes from low quantity early data of users’ interactions as well as future interaction of unique users. In practice, the proposed techniques might improve both navigation within videos on the web as well as video search results with personalised video thumbnails.

Deformation Monitoring Model of Jinping-I Hydropower Station Based on NACA-BP Neural Network

Back Propagation (BP) neural network can learn and store a large number of input-output model nonlinear relationships with simple structure. Niche ant colony algorithm (NACA) combines the ant colony algorithm (ACA) with the niche technology in order to add its local search ability to ACA with preserving the intelligent search ability and robustness of ACA. To optimize predicting model establishment of the dam monitoring data, NACA and BP neural network modeling method are combined to establish a prediction model of horizontal displacement monitoring data. The traditional BP neural network prediction model is established to make a comparison with the NACA. The results show that NACA-BP neural network method can speed up the convergence rate of BP neural network and enhance local search ability and prediction accuracy.

신경망을 이용한 판온예측모델내 공정상수 설정 방법

This paper presents an algorithmic type computing technique of process coefficient in predicting model of temperature for reheating furnace and also suggests a design method of neural network model to find an adequate value of process coefficient for arbitrary operating conditions including test conditons. The proposed neural network use furnace temperature, line speed and slab information as input variables, and process coefficient is output variable. Reasonable process coefficients can be obtained by an algorithmic procedure proposed in this paper using process data gathered at test conditons. Also, neural network model output equal process coefficient under same input conditions. This means that adquate process coefficients can be found by only computing neural network model without additive test even if operating conditions vary.

Hopfield Neural Network Modeling Method Based on Subset KPCA with FCMC Clustering

In order to improve the performance of nonlinear modeling, a Hopfield neural network modeling method based on Subset Kernel Principal Components Analysis (SubKPCA) with Fuzzy C-Means Clustering (FCMC) is proposed. The simulation result shows that, the performance of the proposed method is better than that of Hopfield neural network based on KPCA. It also is effective and feasible to establish the model for the estimation of missing flight data.

Hysteresis Neural Network Modeling and Compensation of Piezoelectric Actuator

By studying on the nonlinear hysteresis characteristics of piezoelectric actuators,this paper proposes a neural network modeling method based on polynomial fitting algorithm and a compound control method for compensation of the hysteresis.Simulation shows that the fitting error of neural network model is 1.42%. According to the developed hysteretic model,PID and feed-forward control methods are applied to the system.The result is that the tracking relative error of control system is 1.59%,so the tracking precision of system is improved significantly.This indicates that the neural network model reflects the hysteresis characteristics of piezoelectric actuators accurately,and this control method is an effective compensation control for hysteresis in piezoelectric actuators.

Application of BP Neural Network and GPRS in Design of Supply Water Management Information System in Campus

The prediction problem of water according to the characteristics of campus, the general steps combined with neural network modeling method, BP neural network was selected as a basis for modeling. In this paper, the monitoring system GPRS technology and campus water supply network are combined, and the establishment of a burst pipe fault analysis model. The paper presents application of BP neural network and GPRS in design of supply water management information system in campus. Experimental results show that the proposed method has high efficiency.

Research of EDM Titanium Alloy TC11 Based on GM(1,N) Model and BP Neural Network

In order to predict the influences of electrical discharge machining (EDM) and enhance material removal rate (MRR). A model was presented for predictions of MRR in EDM TC11 test piece based on the construction method of gray GM(1,N) model and BP neural network model (GNNM).The results show that the GNNM is credible with the maximum absolute error 3.58%, the minimum absolute error 0.01% and the mean error 0.53%were obtained .The results indicated that the model can reflect the technological law of EDM TC11 and successfully predict its processing speed .The paper provides references and basis for selecting processing parameters of EDM TC11 and has practical significance.

Bitterness intensity prediction of berberine hydrochloride using an electronic tongue and a GA-BP neural network.

The aim of this study was to predict the bitterness intensity of a drug using an electronic tongue (e-tongue). The model drug of berberine hydrochloride was used to establish a bitterness prediction model (BPM), based on the taste evaluation of bitterness intensity by a taste panel, the data provided by the e-tongue and a genetic algorithm-back-propagation neural network (GA-BP) modeling method. The modeling characteristics of the GA-BP were compared with those of multiple linear regression, partial least square regression and BP methods. The determination coefficient of the BPM was 0.99965±0.00004, the root mean square error of cross-validation was 0.1398±0.0488 and the correlation coefficient of the cross-validation between the true and predicted values was 0.9959±0.0027. The model is superior to the other three models based on these indicators. In conclusion, the model established in this study has a high fitting degree and may be used for the bitterness prediction modeling of berberine hydrochloride of different concentrations. The model also provides a reference for the generation of BPMs of other drugs. Additionally, the algorithm of the study is able to conduct a rapid and accurate quantitative analysis of the data provided by the e-tongue.

Rapid ULSI Interconnect Reliability Analysis Using Neural Networks

Neural network modeling method is introduced for analyzing ultralarge scale integration (ULSI) interconnect reliability for the first time. By training the simulation data from ANSYS (a finite-element tool), a neural network model is developed, where the prediction of ULSI interconnect reliability can be more effectively done. The proposed technique is useful for integrated circuit design since it can produce a database of interconnect layouts with reliability comparison for a given circuit. From the database, we can know the relative reliability of interconnect layout at any given temperature or current rapidly. Through this proposed technique, we can also derive the allowable temperature and current range of a circuit to ensure given reliability criteria.

Model-based dimensionless neural networks for fin-and-tube condenser performance evaluation

The paper presents a dimensionless neural network modeling method for the fin-and-tube refrigerant-to-air condensers which are widely used in air-cooled refrigeration and heat pump systems. The model-based dimensional analysis method is applied to develop the dimensionless Pi-groups for the condenser performance. The three-layer perceptron neural network is served as the performance model using the dimensionless Pi-groups as its inputs and outputs. Compared with a well-validated tube-by-tube first-principle model, the standard deviations of trained dimensionless neural networks are 0.66%, 4.83% and 0.11% for the heating capacity, the refrigerant pressure drop and the air pressure drop, respectively. The accuracy is also consistent with the previously developed dimensional neural networks. Furthermore, independent model validation using different refrigerants shows that the dimensionless models have good potential in predicting the condenser performance if the Pi-groups were in the range of training data.

A Neural Learning Algorithm of Blind Separation of Noisy Mixed Images Based on Independent Component Analysis

Blind source separation problem has recently received a great deal of attention in signal processing and unsupervised neural learning. In the current approaches, the additive noise is negligible so that it can be omitted from the consideration. To be applicable in realistic scenarios, blind source separation approaches should deal evenly with the presence of noise. In this contribution, we proposed approaches to independent component analysis when the measured signals are contaminated by additive noise. A noisy multiple channels neural learning algorithm of blind separation is proposed based on independent component analysis. The data have no noise are used to whiten the noisy data, and the windage wipe off technique is used to correct the infection of noise, a neural network model having denoise capability is adopted to recover some original signals from their noisy mixtures observed by the same number of sensors. And a relaxation factor is introduced into the iteration algorithm, thus the new algorithm can implement convergence. Computer simulations and experiment results prove the feasibility and validity of the neural network modeling and control method based on independent component analysis, which can renew the original images effectively.

Comparison of Kinetic-based and Artificial Neural Network Modeling Methods for a Pilot Scale Vacuum Gas Oil Hydrocracking Reactor

An artificial neural network (ANN) and kinetic-based models for a pilot scale vacuum gas oil (VGO) hydrocracking plant are presented in this paper. Reported experimental data in the literature were used to develop, train, and check these models. The proposed models are capable of predicting the yield of all main hydrocracking products including dry gas, light naphtha, heavy naphtha, kerosene, diesel, and unconverted VGO (residue). Results showed that kinetic-based and artificial neural models have specific capabilities to predict yield of hydrocracking products. The former is able to accurately predict the yield of lighter products, i.e. light naphtha, heavy naphtha and kerosene. However, ANN model is capable of predicting yields of diesel and residue with higher precision. The comparison shows that the ANN model is superior to the kinetic-base models. © 2013 by Authors, Published by BCREC Group. This is an open access article under the CC BY-SA License (https://creativecommons.org/licenses/by-sa/4.0)

Prediction Modeling of Hot Metal Silicon Content in Blast Furnace Based on PSO-LSSVM

The hot metal silicon content is important for the quality of the iron, but also as an indicator of the thermal state of the furnace. In order to stable operation of the blast furnace, a model is needed to predict hot metal silicon content more accurately. Towards this goal, a model based on least square support vector machine (LSSVM) is developed to model and predict hot metal silicon content, particle swarm optimization (PSO) is adopted to search the optimal set of the LSSVM model parameters. Prediction experiment is conducted based on the data obtained from a blast furnace past operation records in a steel tube plant and being preprocessed in several different ways. The experimental results show that the proposed methods yielded more accurate predictions than the neural network modeling methods.

Research and Realization of Hardware Back-Propagation Neural Network Based on FPGA

In order to meet the requirements of real-time and embedded of industrial field, a reconfigurable Back-Propagation neural network based on FPGA has been implemented on Xilinx's Spartan-3E (XC3S250E) chip which has 250000 gate. First the optimal network structure and weights were gotten by a variable structure of BP neural network algorithm. Then an improved hardware approaching method of excitation function was put forward, and the maximum error was 1.58% by simulation and comparative analysis on the error. Finally hardware co-imitation and timing simulation was token based on a reasonable choice of data accuracy, and then the hardware BP neural network algorithm was been downloaded and implemented on FPGA. This method has better accuracy and speed, it is an effective method of BP neural network modeling based on hardware, and lays the foundation for the hardware realization of other neural network and embedded image processing.