Neural Network Metamodel Research Articles

New sampling scheme for neural network-based metamodelling with application to air pollutant estimation

Purpose A new method for the design of experiments (DOE) or sampling technique is proposed, using a distance weight function and the k-means theory. The radial basis function neural network metamodelling approach is used to evaluate the performance of the proposed DOE by using an n-degree of test function, applied to the complex nonlinear problem of spatial distribution of air pollutants. A comparison study is included to analyse the performance of the proposed technique against available methods such as the n-level full fractional design method and the Latin Hypercube Design method. Method For one design objective and n number of input design variables, a set of input-output training dataset are { } n j m i x x x x x x X i j j j i ,.., 1 , ,.., 1 ) ,.., , ;....; ,..., , ) ( ) 2 ( ) 1 ( ) ( 1 ) 2 ( 1 ) 1 ( 1 = = = and { } m i y y y Y i ,..., 2 , 1 ,..., , ) ( ) 2 ( ) 1 ( = = , where m is the

A neural network meta-model of roll-drafting process

This paper presents a neural network based meta‐model of the roll‐drafting process which was elaborated on the basis of a discrete‐event simulation model presented previously. The GRNN, RBF, MLP3, and MLP4 networks were trained, tested and compared. The training set for the neural networks was obtained from the discrete event simulation model, which is characterized by a beta‐distributed velocity change point and satisfies the first and second limit schemes of the roll‐drafting process. A comparative analysis of the four different types of meta‐models led to the conclusion that the MLP4 meta‐model provides the best prediction of roll‐drafted fibrous material irregularity. Testing of the developed meta‐model has shown a high‐level coincidence between this meta‐model and the DES model of the roll‐drafting process.

A Framework for Analysis of Production Authorization Card‐Controlled Production Systems

A framework for the analysis of manufacturing systems operating under a production authorization card (PAC) system is outlined. The PAC system provides a single model, which encompasses a broad variety of control strategies, including Kanban and CONWIP. This paper describes a framework for the performance analysis and comparison of both specific and families of control strategies. The framework starts with system performance measures estimated by simulation. These simulations in turn provide training data for neural network metamodels. The metamodels allow for a variety of analysis and optimization approaches, including the construction of optimal policy curves, which can provide considerable insight into the systems under study.

Modeling impact of choice complexity on production rate in mixed-model assembly system

The increasing variety of products complicates the mixed-model assembly process and affected the mixed-model assembly system in terms of product quality and productivity. Choice complexity comes from the process of making choices for various assembly operations due to the product variety and impacts production rate which is the performance measure of the system. The choice complexity is measured with information entropy, and the relational expression between choice complexity and error rate is analyzed by means of those research finds on average reaction time and speed-accuracy trade-off. The main achievement of our study is establishing an artificial neural network meta-model for the impact of choice complexity on production rate. The meta-model performs better than a multiple linear regression meta-model in terms of experiment results and appears to be the optimal model of the impact of choice complexity on production rate in the mixed-model assembly system.

Optimization of Helicopter Rotor Using Polynomial and Neural Network Metamodels

This study aims to determine optimal locations of dual trailing-edge flaps and blade stiffness to achieve minimum hub vibration levels in a helicopter, with low penalty in terms of required trailing-edge flap control power. An aeroelastic analysis based on finite elements in space and time is used in conjunction with an optimal control algorithmtodeterminethe flaptimehistoryforvibrationminimization.Usingtheaeroelasticanalysis,itisfoundthat the objective functions are highly nonlinear and polynomial response surface approximations cannot describe the objectives adequately. A neural network is then used for approximating the objective functions for optimization. Pareto-optimal points minimizing both helicopter vibration and flap power are obtained using the response surface and neural network metamodels. The two metamodels give useful improved designs resulting in about 27% reduction in hub vibration and about 45% reduction in flap power. However, the design obtained using response surface is less sensitive to small perturbations in the design variables.

A hybrid artificial neural network: computer simulation approach for scheduling a flow shop with multiple processors

Depending on the characteristics of the manufacturing system and production objectives, dispatching rules have different efficiencies. In this regard, a multiattribute combinatorial dispatching (MACD) decision problem for scheduling a flow shop with multiple processors environment is presented in this paper. We propose a hybrid artificial neural network (ANN) simulation approach as a valid and superior alternative for solving the MACD decision problem. ANNs are one of the commonly used meta-heuristics and are a proven tool for solving complex optimisation problems. The hybrid approach is capable of modelling a non-linear and stochastic problem. Feed forward, multilayered neural network meta-models were trained through the back propagation learning algorithm to provide a complex MACD problem. The solution quality is illustrated by a case study from a multilayer ceramic capacitor manufacturing plant. The manufacturing lead times produced by the hybrid ANN simulation model turned out to be as valid and super...

Metamodeling of aircraft infrared signature dispersion

Existing computer simulations of aircraft InfraRed Signature (IRS) do not account for the dispersion induced by uncertainty on input data such as aircraft aspect angles and meteorological conditions. As a result, they are of little use to estimate the detection performance of optronic systems: in that case, the scenario encompasses a lot of possible situations that must indeed be addressed, but cannot be singly simulated. In this paper, a three-step methodological approach for predicting simulated IRS dispersion of imperfectly known aircraft is proposed. The first step is a sensitivity analysis. The second step consists in a Quasi-Monte Carlo survey of the code output dispersion. In the last step, a neural network metamodel of the IRS simulation code is constructed. It will allow carrying out thorough computationally demanding tasks, such as those required for optimization of an optronic sensor. This method is illustrated in a typical scenario, namely an air-to-ground full-frontal attack by a generic combat aircraft, and gives satisfactory estimation of the infrared signature dispersion.

NEURAL NETWORK META-MODELING OF STEAM ASSISTED GRAVITY DRAINAGE OIL RECOVERY PROCESSES

Production of highly viscous tar sand bitumen using Steam Assisted Gravity Drainage (SAGD) with a pair of horizontal wells has advantages over conventional steam flooding. This paper explores the use of Artificial Neural Networks (ANNs) as an alternative to the traditional SAGD simulation approach. Feed forward, multi-layered neural network meta-models are trained through the Back-Error-Propagation (BEP) learning algorithm to provide a versatile SAGD forecasting and analysis framework. The constructed neural network architectures are capable of estimating the recovery factors of the SAGD production as an enhanced oil recovery method satisfactorily. Rigorous studies regarding the hybrid static-dynamic structure of the proposed network are conducted to avoid the over-fitting phenomena. The feed forward artificial neural network-based simulations are able to capture the underlying relationship between several parameters/operational conditions and rate of bitumen production fairly well, which proves that ANNs are suitable tools for SAGD simulation.

Neural network metamodeling for cycle time-throughput profiles in manufacturing

This paper proposed a neural network (NN) metamodeling method to generate the cycle time (CT)–throughput (TH) profiles for single/multi-product manufacturing environments. Such CT–TH profiles illustrate the trade-off relationship between CT and TH, the two critical performance measures, and hence provide a comprehensive performance evaluation of a manufacturing system. The proposed methods distinct from the existing NN metamodeling work in three major aspects: First, instead of treating an NN as a black box, the geometry of NN is examined and utilized; second, a progressive model-fitting strategy is developed to obtain the simplest-structured NN that is adequate to capture the CT–TH relationship; third, an experiment design method, particularly suitable to NN modeling, is developed to sequentially collect simulation data for the efficient estimation of the NN models.

A flexible artificial neural network–fuzzy simulation algorithm for scheduling a flow shop with multiple processors

One of the most popular approaches for scheduling manufacturing systems is dispatching rules. Different types of dispatching rules exist, but none of them is known to be globally the best. A flexible artificial neural network–fuzzy simulation (FANN–FS) algorithm is presented in this study for solving the multiattribute combinatorial dispatching (MACD) decision problem. Artificial neural networks (ANNs) are one of the commonly used metaheuristics and are a proven tool for solving complex optimization problems. Hence, multilayered neural network metamodels and a fuzzy simulation using the α-cuts method were trained to provide a complex MACD problem. Fuzzy simulation is used to solve complex optimization problems to deal with imprecision and uncertainty. The proposed flexible algorithm is capable of modeling nonlinear, stochastic, and uncertain problems. It uses ANN simulation for crisp input data and fuzzy simulation for imprecise and uncertain input data. The solution quality is illustrated by two case studies from a multilayer ceramic capacitor manufacturing plant. The manufacturing lead times produced by the FANN–FS model turned out to be superior to conventional simulation models. This is the first study that introduces an intelligent and flexible approach for handling imprecision and nonlinearity of scheduling problems in flow shops with multiple processors.

Optimization of manufacturing systems using a neural network metamodel with a new training approach

In this study, two manufacturing systems, a kanban-controlled system and a multi-stage, multi-server production line in a diamond tool production system, are optimized utilizing neural network metamodels (tst_NNM) trained via tabu search (TS) which was developed previously by the authors. The most widely used training algorithm for neural networks has been back propagation which is based on a gradient technique that requires significant computational effort. To deal with the major shortcomings of back propagation (BP) such as the tendency to converge to a local optimal and a slow convergence rate, the TS metaheuristic method is used for the training of artificial neural networks to improve the performance of the metamodelling approach. The metamodels are analysed based on their ability to predict simulation results versus traditional neural network metamodels that have been trained by BP algorithm (bp_NNM). Computational results show that tst_NNM is superior to bp_NNM for both of the manufacturing systems.

A prediction interval-based approach to determine optimal structures of neural network metamodels

Neural networks have been widely used in literature for metamodeling of complex systems and often outperform their traditional counterparts such as regression-based techniques. Despite proliferation of their applications, determination of their optimal structure is still a challenge, especially if they are developed for prediction and forecasting purposes. Researchers often seek a tradeoff between estimation accuracy and structure complexity of neural networks in a trial and error basis. On the other hand, the neural network point prediction performance considerably drops as the level of complexity and amount of uncertainty increase in systems that data originates from. Motivated by these trends and drawbacks, this research aims at adopting a technique for constructing prediction intervals for point predictions of neural network metamodels. Space search for neural network structures will be defined and confined based on particular features of prediction intervals. Instead of using traditional selection criteria such as mean square error or mean absolute percentage error, prediction interval coverage probability and normalized mean prediction interval will be used for selecting the optimal network structure. The proposed method will be then applied for metamodeling of a highly detailed discrete event simulation model which is essentially a validated virtual representation of a large real world baggage handling system. Through a more than 77% reduction in number of potential candidates, optimal structure for neural networks is found in a manageable time. According to the demonstrated results, constructed prediction intervals using optimal neural network metamodel have a satisfactory coverage probability of targets with a low mean of length.

The scope of artificial neural network metamodels for precision casting process planning

Precision investment casting process planning has been tackled in the past according to experience. Recently, casting simulation software is being increasingly used to predict product quality by implementing ‘what-if’ scenarios. Input parameters include relatively simple factors such as mould temperature, melting temperature, casting material. They also include factors whose influence is more complex to quantify, such number and location of feeding points, diameter and length of inflow channels, angle of channel with respect to the main sprue axis. Simulation results cannot help the engineer for workpieces other than the one simulated. In this paper a series of feedforward artificial neural network (ANN) models is presented aiming at such generalisation. To achieve this, a large number of software simulation runs were conducted for a number of different small parts, with varying runner geometry and casting conditions. The parameters characterising part geometry have been chosen to be surface area and volume-to-area ratio. The different ANN models predictive capabilities are reflected to the respective training and generalisation errors. A user-friendly interface has been conducted for model execution in a complete application, whose main virtue is expandability.

Controllers optimization for a fluid mixing system using metamodelling approach

Offline optimization of controller parameters for complex non-linear processes can be time consuming, even with high performance computers. This paper demonstrates how Metamodelling techniques can be utilized to quickly tune the controller parameters for a nonlinear process. The process used in this study is the mixing process which is a multivariable and intrinsically non-linear plant. The Radial Basis Function Neural Network Metamodel founded a good approximation to the optimum controller parameters in this case. This paper proposes an intuitive methodology to use only a small fraction of the design space to create a Radial Basis Function Neural Network Metamodel that is good enough to optimize the system. Comparisons were made between the controllers optimized using the Metamodelling technique and the original large space design. (Received in May 2008, accepted in October 2008. This paper was with the authors 1 month for 1 revision.)

Coupling product and manufacturing system design

This work examines the pathways leading from preliminary product and process development first (Stage 1) to detailed product design, with emphasis on analysis-based configuration issues (Stage 2) and then, to manufacturing system design focusing on performance prediction (Stage 3). Stage 1 starts with an enhanced QFD exercise that setting both functional and manufacturing requirements, resulting in a product configurations table. Stage 2 enables generation of systematic variations of the best main configuration using parametric CAD software and DfM/DfA assessment tools. In addition, FEA is conducted on a limited number of product variations and results are used to train a neural network meta-model which acts as an evaluator of functional performance. Stage 3 simulates manufacturing system performance based on individual manufacturing processes decided in the previous two stages considering equipment, material handling options, production rate drivers, etc. Feedback is generated mainly within stages and in exceptional circumstance between stages, too. The methodology is demonstrated using as case-study a meteorological mast.

Classical inverse and metamodel approach in identification of nanomaterials properties

System of nanocoatings deposited by using physical vapour deposition is planned to be applied in the artificial heart prosthesis. Material models of these coatings are necessary for numerical design of these products and are crucial for accuracy of simulations. The objective of the present work is identification of parameters of material model of nanocoatings using two methods based on the nanoindentation test data: the classical inverse analysis and the artificial neural network metamodel. The inverse analysis is preceded by the development of FEM model dedicated to the nanoindentation test for the system of nanocoatings. The parameters of individual coating of the system are evaluated. In the second approach to decrease the computation cost, the metamodel is suggested. The metamodelling method is based on the artificial neural network technique. The achieved results confirm the usefulness of the presented solution in the identification of the material properties for the system of nanocoatings.

Evolutionary Tuning Method for PID Controller Parameters of a Cruise Control System Using Metamodeling

For long time the optimization of controller parameters uses the well-known classical method such as the Ziegler-Nichols and the Cohen-Coon tuning techniques. Despite its effectiveness, these off-line tuning techniques can be time consuming especially for a case of complex nonlinear system. This paper attempts to show a great deal on how Metamodeling techniques can be utilized to tune the PID controller parameters quickly. Note that the plant use in this study is the cruise control system with 2 different models, which are the linear model and the nonlinear model. The difference between both models is that the disturbances were taken into consideration for the nonlinear model, but in the linear model the disturbances were assumed as zero. The Radial Basis Function Neural Network Metamodel is able to prove that it can minimize the time in tuning process as it is able to give a good approximation to the optimum controller parameters in both models of this system.

Stochastic sampling design using a multi-objective genetic algorithm and adaptive neural networks

This paper presents a novel multi-objective genetic algorithm (MOGA) based on the NSGA-II algorithm, which uses metamodels to determine optimal sampling locations for installing pressure loggers in a water distribution system (WDS) when parameter uncertainty is considered. The new algorithm combines the multi-objective genetic algorithm with adaptive neural networks (MOGA–ANN) to locate pressure loggers. The purpose of pressure logger installation is to collect data for hydraulic model calibration. Sampling design is formulated as a two-objective optimization problem in this study. The objectives are to maximize the calibrated model accuracy and to minimize the number of sampling devices as a surrogate of sampling design cost. Calibrated model accuracy is defined as the average of normalized traces of model prediction covariance matrices, each of which is constructed from a randomly generated sampling set of calibration parameter values. This method of calculating model accuracy is called the ‘full’ fitness model. Within the genetic algorithm search process, the full fitness model is progressively replaced with the periodically (re)trained adaptive neural network metamodel where (re)training is done using the data collected by calling the full model. The methodology was first tested on a hypothetical (benchmark) problem to configure the setting requirement. Then the model was applied to a real case study. The results show that significant computational savings can be achieved by using the MOGA–ANN when compared to the approach where MOGA is linked to the full fitness model. When applied to the real case study, optimal solutions identified by MOGA–ANN are obtained 25 times faster than those identified by the full model without significant decrease in the accuracy of the final solution.

Neural network-based simulation metamodels for predicting probability distributions

Simulation is an important tool for supporting decision-making under uncertainty, particularly when the system under consideration is too complex to evaluate analytically. The amount of time required to generate large numbers of simulation replications can be prohibitive, however, necessitating the use of a simulation metamodel in order to describe the behavior of the system under new conditions. The purpose of this study is to examine the use of neural network metamodels for representing output distributions from a stochastic simulation model. A series of tests on a well-known simulation problem demonstrate the ability of the neural networks to capture the behavior of the underlying systems and to represent the inherent uncertainty with a reasonable degree of accuracy.

The design and analysis of complex system neural network metamodels

The basic task of work is to research possibility of using the neural networks as a metamodels for three complex systems: dispatching system (planning system of transport routes), multicommodity network and bank system with several cash registers. The results of constructing the metamodels for these systems are presented. The principles of the factorial experimental design creation for the neural network learning are described. The perceptron with one hidden layer and sigmoid activation function is used as the metamodel.