Multilayer Perception Network Research Articles

Extracting post-nonlinear signal with reference

Constrained independent component analysis (cICA) is an important technique which can extract the desired sources from the mixtures. The post-nonlinear (PNL) mixture model is more realistic and accurate than the linear instantaneous mixture model in many practical situations. In this paper, we address the problem of extracting the desired source as the first output from the PNL mixture. The prior knowledge about the desired source, such as its rough template (reference), is assumed to be available. Two approaches of extracting PNL signal with reference are discussed. Then a novel algorithm which alternately optimizes the contrast function and the closeness measure between the estimated output and the reference signal is proposed. The inverse of the unknown nonlinear function in the PNL mixture model is approximated by the multi-layer perception (MLP) network. The correctness and validity of the proposed algorithm are demonstrated by our experiment results.

CNG-diesel engine performance and exhaust emission analysis with the aid of artificial neural network

This study investigates the use of artificial neural network (ANN) modelling to predict brake power, torque, break specific fuel consumption (BSFC), and exhaust emissions of a diesel engine modified to operate with a combination of both compressed natural gas CNG and diesel fuels. A single cylinder, four-stroke diesel engine was modified for the present work and was operated at different engine loads and speeds. The experimental results reveal that the mixtures of CNG and diesel fuel provided better engine performance and improved the emission characteristics compared with the pure diesel fuel. For the ANN modelling, the standard back-propagation algorithm was found to be the optimum choice for training the model. A multi-layer perception network was used for non-linear mapping between the input and output parameters. It was found that the ANN model is able to predict the engine performance and exhaust emissions with a correlation coefficient of 0.9884, 0.9838, 0.95707, and 0.9934 for the engine torque, BSFC, NOx and exhaust temperature, respectively.

Performance and exhaust emissions of a gasoline engine with ethanol blended gasoline fuels using artificial neural network

The purpose of this study is to experimentally analyse the performance and the pollutant emissions of a four-stroke SI engine operating on ethanol–gasoline blends of 0%, 5%, 10%, 15% and 20% with the aid of artificial neural network (ANN). The properties of bioethanol were measured based on American Society for Testing and Materials (ASTM) standards. The experimental results revealed that using ethanol–gasoline blended fuels increased the power and torque output of the engine marginally. For ethanol blends it was found that the brake specific fuel consumption (bsfc) was decreased while the brake thermal efficiency ( η b.th.) and the volumetric efficiency ( η v ) were increased. The concentration of CO and HC emissions in the exhaust pipe were measured and found to be decreased when ethanol blends were introduced. This was due to the high oxygen percentage in the ethanol. In contrast, the concentration of CO 2 and NO x was found to be increased when ethanol is introduced. An ANN model was developed to predict a correlation between brake power, torque, brake specific fuel consumption, brake thermal efficiency, volumetric efficiency and emission components using different gasoline–ethanol blends and speeds as inputs data. About 70% of the total experimental data were used for training purposes, while the 30% were used for testing. A standard Back-Propagation algorithm for the engine was used in this model. A multi layer perception network (MLP) was used for nonlinear mapping between the input and the output parameters. It was observed that the ANN model can predict engine performance and exhaust emissions with correlation coefficient ( R) in the range of 0.97–1. Mean relative errors (MRE) values were in the range of 0.46–5.57%, while root mean square errors (RMSE) were found to be very low. This study demonstrates that ANN approach can be used to accurately predict the SI engine performance and emissions.

Diesel engine performance and exhaust emission analysis using waste cooking biodiesel fuel with an artificial neural network

This study deals with artificial neural network (ANN) modeling of a diesel engine using waste cooking biodiesel fuel to predict the brake power, torque, specific fuel consumption and exhaust emissions of the engine. To acquire data for training and testing the proposed ANN, a two cylinders, four-stroke diesel engine was fuelled with waste vegetable cooking biodiesel and diesel fuel blends and operated at different engine speeds. The properties of biodiesel produced from waste vegetable oil was measured based on ASTM standards. The experimental results revealed that blends of waste vegetable oil methyl ester with diesel fuel provide better engine performance and improved emission characteristics. Using some of the experimental data for training, an ANN model was developed based on standard Back-Propagation algorithm for the engine. Multi layer perception network (MLP) was used for non-linear mapping between the input and output parameters. Different activation functions and several rules were used to assess the percentage error between the desired and the predicted values. It was observed that the ANN model can predict the engine performance and exhaust emissions quite well with correlation coefficient ( R) 0.9487, 0.999, 0.929 and 0.999 for the engine torque, SFC, CO and HC emissions, respectively. The prediction MSE (Mean Square Error) error was between the desired outputs as measured values and the simulated values were obtained as 0.0004 by the model.

Investigation of various artificial neural networks techniques for the prediction of inland water units' resistance

Artificial neural networks (ANNs) have been successfully used in ship hydrodynamics for pattern recognition and the prediction of resistance coefficients. Accumulated experimental data of inland water units' resistance from the Versuchsanstalt fur Binnenschiffbau Duisburg in Germany have been used before for the training of multi-layer perception (MLP) networks to reach a general expression for the prediction of inland water units and to compare the relative merits of ANNs and cubic spline interpolation in performing such a task. In this article, two models of ANNs have been created to better predict the resistance for new designs, namely the MLP and the radial basis functions. The results show that MLP records better performance compared with radial basis functions and that ANNs have large capabilities in capturing the pattern from scattered data, but the precision of the results obtained, particularly those for maximum error, should be treated with care.

Ability to forecast unsteady aerodynamic forces of flapping airfoils by artificial neural network

The ability of artificial neural networks (ANN) to model the unsteady aerodynamic force coefficients of flapping motion kinematics has been studied. A neural networks model was developed based on multi-layer perception (MLP) networks and the Levenberg–Marquardt optimization algorithm. The flapping kinematics data were divided into two groups for the training and the prediction test of the ANN model. The training phase led to a very satisfactory calibration of the ANN model. The attempt to predict aerodynamic forces both the lift coefficient and drag coefficient showed that the ANN model is able to simulate the unsteady flapping motion kinematics and its corresponding aerodynamic forces. The shape of the simulated force coefficients was found to be similar to that of the numerical results. These encouraging results make it possible to consider interesting and new prospects for the modelling of flapping motion systems, which are highly non-linear systems.

Modelling weld bead geometry using neural networks for GTAW of austenitic stainless steel

The authors analyse the importance of different weld control parameters on the weld pool geometry of gas tungsten arc welding using an online feature selection technique that suggests weld voltage and vertex–angle pair as more important than the weld voltage and torch speed pair. Using the selected features multi layer perception and radial basis function networks are developed for prediction of bead width, penetration depth, and bead area. With cross-validation the authors have extensively studied the performance of composite models (one model for all outputs) and individual models (one model for each output). The individual models are found to work better than composite models. Usually, radial basis function networks are found to work better than the multi layer perception networks. To assess the influence of weld control parameters the authors have studied the performance of both networks using different combination of inputs. Overall, the performance of the proposed models is found to be quite satisfactory.

A Comparison of BBN, ADTree and MLP in separating Quasars from Large Survey Catalogues

We compare the performance of Bayesian Belief Networks (BBN), Multilayer Perception (MLP) networks and Alternating Decision Trees (ADtree) on separating quasars from stars with the database from the 2MASS and FIRST survey catalogs. Having a training sample of sources of known object types, the classifiers are trained to separate quasars from stars. By the statistical properties of the sample, the features important for classification are selected. We compare the classification results with and without feature selection. Experiments show that the results with feature selection are better than those without feature selection. From the high accuracy found, it is concluded that these automated methods are robust and effective for classifying point sources. They may all be applied to large survey projects (e.g. selecting input catalogs) and for other astronomical issues, such as the parameter measurement of stars and the redshift estimation of galaxies and quasars.

Multi-Aspect Target Discrimination Using Hidden Markov Models and Neural Networks

This paper presents a new multi-aspect pattern classification method using hidden Markov models (HMMs). Models are defined for each class, with the probability found by each model determining class membership. Each HMM model is enhanced by the use of a multilayer perception (MLP) network to generate emission probabilities. This hybrid system uses the MLP to find the probability of a state for an unknown pattern and the HMM to model the process underlying the state transitions. A new batch gradient descent-based method is introduced for optimal estimation of the transition and emission probabilities. A prediction method in conjunction with HMM model is also presented that attempts to improve the computation of transition probabilities by using the previous states to predict the next state. This method exploits the correlation information between consecutive aspects. These algorithms are then implemented and benchmarked on a multi-aspect underwater target classification problem using a realistic sonar data set collected in different bottom conditions.

Quick Seismic Response Estimation of Prestressed Concrete Bridges Using Artificial Neural Networks

Seismic early warning has been very important and has become feasible in Taiwan. Perhaps because of the lack of quick and reliable estimations of the induced structural response, however, the triggering criteria of almost all of the existing earthquake protection or early warning systems in the world are merely based on the collected or estimated data of the ground motion, without any information regarding the structural response. This paper presents a methodology of generating quick seismic response estimations of a prestressed concrete (PC) bridge using artificial neural networks (ANNs), which may be incorporated in a seismic early warning system for the bridge. In the methodology ANNs were applied to model the critical structural response of a PC bridge subjected to earthquake excitation of various magnitudes along various directions. The objective was to implement a well-trained network that is capable of providing a quick prediction for the critical response of the target bridge. The well-known multilayer perception (MLP) networks with back propagation algorithm were employed. A simple augmented form of MLP that can be quantitatively determined was proposed. These networks were trained and tested based on the analytical data obtained from the nonlinear dynamic finite fiber element analyses of the target PC bridge. The augmented MLPs were found to be much more efficient than the MLPs in modeling the critical bending moments of the piers and girder of the PC bridge.

Some applications of neural networks in microwave modeling

This paper presents some applications of neural networks in the microwave modeling. The applications are related to modeling of either passive or active structures and devices. Modeling is performed using not only simple multilayer perception network (MLP) but also advanced knowledge based neural network (KBNN) structures.

A Neural Network Implementation For Data Assimilation Using MPI

The meteorological data assimilation process can be described as a procedure that uses observational data to improve the weather forecast produced by means of a mathematical model. Traditional metlods include deterministic and probabilistic approaches. Included in the latter class, Kahnan filtering has the advantage of minimizing tie error propagated during the data assimilation. However, this method demands a heavy computational power as it usually deals with meteorological parameters for a huge number of grid points. Recently, neural networks have been proposed as a new method for meteorological data assimilation by employing a Multilayer Perception network with backpropagation learning to emulate Kalrnan filtering at a lower computational cost. In order to further reduce that cost and to overcome performance bottlenecks of the training phase, code opdmizations were performe& making use of classical, hardware-independent techniques and Fortran 95 intrinsic functions, thus eliminating inefficient routines. A former attempt to parallelize the code and run it in a 4-processor shared memory machine, made use of HPF (High Performance Fortran) dwectives imbedded in the optimized code. Performance was poor and therefore a new version of the Fortran 95 code was developed using Message Passing Interface library calls. This code was executed in a 16-node multicomputer showing good speed-up up to 4 processors.

Using neural networks to model and predict amplitude dependent damping in buildings

In this paper, artificial neural networks, a new kind of intelligent method, are employed to model and predict amplitude dependent damping in buildings based on our full-scale measurements of buildings. The modelling method and procedure using neural networks to model the damping are studied. Comparative analysis of different neural network models of damping, which includes multi-layer perception network (MLP), recurrent neural network, and general regression neural network (GRNN), is performed and discussed in detail. The performances of the models are evaluated and discussed by tests and predictions including self-test, one-lag prediction and multi-lag prediction of the damping values at high amplitude levels. The established models of damping are used to predict the damping in the following three ways : (1) the model is established by part of the data measured from one building and is used to predict the another part of damping values which are always difficult to obtain from field measurements : the values at the high amplitude level. (2) The model is established by the damping data measured from one building and is used to predict the variation curve of damping for another building. And (3) the model is established by the data measured from more than one buildings and is used to predict the variation curve of damping for another building. The prediction results are discussed.

Feature analysis: Neural network and fuzzy set theoretic approaches

In this paper a new scheme of feature ranking and hence feature selection using a Multilayer Perception (MLP) Network has been proposed. The novelty of the proposed MLP-based scheme and its difference from another MLP-based feature ranking scheme have been analyzed. In addition we have modified an existing feature ranking/selection scheme based on fuzzy entropy. Empirical investigations show that the proposed MLP-based scheme is superior to the other schemes implemented.

A novel signal validation method applied to a stochastic process

A novel methodology is presented for the purposes of modelling the process dynamics of neutron flux during induced boiling of the coolant in a nuclear reactor environment. The proposed approach is based on the utilisation of radial basis functions from the point of view of function approximation, regularisation, and noisy interpolation. Data-driven multiple-input/multiple-output radial basis function networks are subjected to faulty input signals and their signal validation capabilities are studied. Networks based on radial basis functions employ training procedures that are substantially faster than the methods used in training multi-layer perception networks. Auto-associative schemes of the proposed technique are shown capable of identifying faulty process signal measurements and provide the operator with an estimate of the actual process value. The data utilised for the purposes of this project are obtained from an experimental facility constructed to simulate boiling phenomena and positioned next to the core of the IRI research reactor.

DISCRIMINATIVE NONLINEAR DIMENSIONALITY REDUCTION FOR IMPROVED CLASSIFICATION

Multi-Layer Perception (MLP) neural networks have been used extensively for classification tasks. Typically, the MLP network is trained explicitly to produce the correct classification as its output. For speech recognition, however, several investigators have recently experimented with an indirect approach: a unique MLP predictive network is trained for each class of data, and classification is accomplished by determining which predictive network serves as the best model for samples of unknown speech. Results from this approach have been mixed. In this report, we compare the direct and indirect approaches to classification from a more fundamental perspective. We show how recent advances in nonlinear dimensionality reduction can be incorporated into the indirect approach, and we show how the two approaches can be integrated in a novel MLP framework. We further show how these new MLP networks can be usefully viewed as generalizations of Learning Vector Quantization (LVQ) and of subspace methods of pattern recognition. Lastly, we show that applying these ideas to the classification of temporal trajectories can substantially improve performance on simple tasks.

Artificial Neural Networks for Wear Estimation

Results of our investigations in feature selection and wear estimation for cutting tools by turning machines using artificial neural networks are presented. The features describing the cutting process are calculated from solid borne sound (vibration) signals. Furthermore, the performance of a Multilayer Perception Network with back propagation learning rule, a Learning Vector Quantization (LVQ) and a Condensed Nearest Neighbour Network (CNNN) used for wear estimation are analysed and the results are compared.

Iiistological image understanding by error backpropagation

The paper describes a non trivial application from histopathology, and its implementation using a multi-layer perception network. We show that with careful network design, the backpropagation learning procedure is an effective way of training neural networks for histological image understanding. The choice of squashing function is an important design issue in achieving fast convergence and good generalisation performance. We evaluate the use of symmetric and asymmetric squashing functions in the learning procedure, and show that symmetric functions yield faster convergence and 100% generalisation performance. We derive analytic results to show the conditions under which symmetric squashing functions yield faster convergence, and to quantify the upper bounds on the convergence improvement.