Search

Artificial neural networks (ANNs) are generally regarded to behave as black-box systems. Recent research explores various methods that can provide an insight into the internal connections and relationships existing within the network. Various methodologies that understand the input variable contribution are analyzed in detail, and rule extraction approaches for a trained artificial neural network are addressed. To understand the contribution of input variables to rainfall-runoff model combination systems, this paper for the first time investigates knowledge extraction from artificial neural network, which is used to combine the results obtained from different competing rainfall-runoff models, using three different approaches: (1) Garson's algorithm; (2) neural interpretation diagram (NID); and (3) sensitivity analysis (SA). For the purpose of investigating knowledge extraction techniques, the trained multilayer perceptron neural network to combine the results from four different rainfall-runoff models for the Brosna Catchment located in Ireland has been chosen. The results of the three approaches obtained in this study indicate that they can be used to reduce the complexity of rainfall-runoff model combination systems by eliminating the least significant contributing input variables. Based on these approaches, the paper helps to provide guidance in the optimal number of rainfall-runoff models that best perform in a combination system. [ABSTRACT FROM AUTHOR]

This paper examines a linked simulation-optimization procedure based on the combined application of an artificial neural network (ANN) and genetic algorithm (GA) with the aim of developing an efficient model for the multiobjective management of groundwater lenses in small islands. The simulation-optimization methodology is applied to a real aquifer in Kish Island of the Persian Gulf to determine the optimal groundwater-extraction while protecting the freshwater lens from seawater intrusion. The initial simulations are based on the application of SUTRA, a variable-density groundwater numerical model. The numerical model parameters are calibrated through automated parameter estimation. To make the optimization process computationally feasible, the numerical model is subsequently replaced by a trained ANN model as an approximate simulator. Even with a moderate number of input data sets based on the numerical simulations, the ANN metamodel can be efficiently trained. The ANN model is subsequently linked with GA to identify the nondominated or Pareto-optimal solutions. To provide flexibility in the implementation of the management plan, the model is built upon optimizing extraction from a number of zones instead of point-well locations. Two issues are of particular interest to the research reported in this paper are: (1) how the general idea of minimizing seawater intrusion can be effectively represented by objective functions within the framework of the simulation-optimization paradigm, and (2) the implications of applying the methodology to a real-world small-island groundwater lens. Four different models have been compared within the framework of multiobjective optimization, including (1) minimization of maximum salinity at observation wells, (2) minimization of the root mean square (RMS) change in concentrations over the planning period, (3) minimization of the arithmetic mean, and (4) minimization of the trimmed arithmetic mean of concentration in the observation wells. The latter model can provide a more effective framework to incorporate the general objective of minimizing seawater intrusion. This paper shows that integration of the latest innovative tools can provide the ability to solve complex real-world optimization problems in an effective way. [ABSTRACT FROM AUTHOR]

In this paper, the use of a hybrid evolutionary optimization algorithm is proposed for global optimization of pavement structural parameters through inverse modeling. Shuffled complex evolution (SCE) is a population-based stochastic optimization technique combining the competitive complex evolution with the controlled random search, the implicit clustering, and the complex shuffling. Back-calculation of pavement layer moduli is an ill-posed inverse engineering problem, which involves searching for the optimal combination of pavement layer stiffness solutions in an unsmooth, multimodal, complex search space. SCE is especially considered a robust and efficient approach for global optimization of multimodal functions. A desirable characteristic of the SCE algorithm is that it uses information about the nature of the response surface, extracted using the deterministic Simplex geometric shape, to direct the search into regions with higher posterior probability. The hybrid back-calculation system described in this paper combines the robustness of the SCE in global optimization with the computational efficiency of neural networks and advanced pavement system characterization offered by employing finite-element models. This is the first time the SCE approach is applied to real-time nondestructive evaluation of pavement systems required in the routine maintenance and rehabilitation activities for sustainable transportation infrastructure. [ABSTRACT FROM AUTHOR]

This paper presents an enhanced design methodology for optimal mixture proportion of concrete composition with respect to accuracy in the case of using prediction models based on a limited database. In proposed methodology, the search space is constrained as the domain defined by a limited database instead of constructing the database covering the region represented by the possible ranges of all variables in the input space. A model for defining the search space which is expressed by the effective region in this paper and evaluating whether a mix proportion is effective is added to the optimization process, yielding highly reliable results. To demonstrate the proposed methodology, a genetic algorithm, an artificial neural network, and a convex hull were adopted as an optimum technique, a prediction model for material properties, and an evaluation model for the effective region, respectively. And then, it was applied to an optimization problem wherein the minimum cost should be obtained under a given strength requirement. Experimental test results show that the mix proportion obtained from the proposed methodology considering the regional characteristics of the database is found to be more accurate and feasible than that obtained from a general optimum technique that does not consider this aspect. [ABSTRACT FROM AUTHOR]

Over the years, many methods have been developed to predict the concrete strength. In recent years, artificial neural networks (ANNs) have been applied to many civil engineering problems with some degree of success. In the present paper, ANN is used as an attempt to obtain more accurate concrete strength prediction based on parameters like concrete mix design, size and shape of specimen, curing technique and period, environmental conditions, etc. A total of 864 concrete specimens were cast for compressive strength measurement and verification through the ANN model. The back propagation-learning algorithm is employed to train the network for extracting knowledge from training examples. The predicted strengths found by employing ANN are compared with the actual values. The results indicate that ANN is a useful technique for predicting the concrete strength. Further, an effort to build an expert system for the problem is described in this paper. To overcome the bottleneck of intricate knowledge acquisition, an expert system is used as a mechanism to transfer engineering experience into usable knowledge through rule-based knowledge representation techniques. [ABSTRACT FROM AUTHOR]

In this paper, models for short-term prediction of influent flow rate in a wastewater-treatment plant are discussed. The prediction horizon of the model is up to 180 min. The influent flow rate, rainfall rate, and radar reflectivity data are used to build the prediction model by different data-mining algorithms. The multilayer perceptron neural network algorithm has been selected to build the prediction models for different time horizons. The computational results show that the prediction model performs well for horizons up to 150 min. Both the peak values and the trends are accurately predicted by the model. There is a small lag between the predicted and observed influent flow rate for horizons exceeding 30 min. The lag becomes larger with the increase of the prediction horizon. [ABSTRACT FROM AUTHOR]

Over the years of the study (1999–2004) presented in this paper, the City of Edmonton, Canada’s Drainage and Maintenance Department has experienced an annual increase of about 12% in the installation of water and sewer services for residential facilities. According to the current estimating procedure, a discrepancy of up to 60% exists between the estimated and actual costs of these projects. A detailed analysis of all activities involved in the installation of the water and sewer services has been carried out and is presented in this paper. The proposed methodology, which is based upon the analysis of past data obtained from the City of Edmonton’s drainage division for the period of 1999–2004, is also presented. The methodology has been incorporated into a computer module, which integrates the concept of artificial neural network (ANN) with the current estimating system used by the City of Edmonton. The following research includes a description of the algorithm used in ANN, as well as an assessment of past data obtained from the city record for over 800 jobs (cases) performed over the period of the study. [ABSTRACT FROM AUTHOR]

This paper presents the application of an improved particle swarm optimization (PSO) technique for training an artificial neural network (ANN) to predict water levels for the Heshui Watershed, China. Daily values of rainfall and water levels from 1988 to 2000 were first analyzed using ANNs trained with the conjugate gradient, gradient descent, and Levenberg-Marquardt neural network (LM-NN) algorithms. The best results were obtained from the LM-NN, and these results were then compared with those from PSO-based ANNs, including the conventional PSO neural network (CPSONN) and the improved PSO neural network (IPSONN) with passive congregation. The IPSONN algorithm improves PSO convergence by using the selfish herd concept in swarm behavior. The results show that the PSO-based ANNs performed better than the LM-NN. For models run using a single parameter (rainfall) as input, the root mean square error (RMSE) of the testing data set for IPSONN was the lowest (0.152 m) compared to those for CPSONN (0.161 m) and LM-NN (0.205 m). For multiparameter (rainfall and water level) inputs, the RMSE of the testing data set for IPSONN was also the lowest (0.089 m) compared to those for CPSONN (0.105 m) and LM-NN (0.145 m). The results also indicate that the LM-NN model performed poorly in predicting the low and peak water levels in comparison to the PSO-based ANNs. Moreover, the IPSONN model was superior to CPSONN in predicting extreme water levels. Lastly, IPSONN had a quicker convergence rate compared to CPSONN. [ABSTRACT FROM AUTHOR]

A fully automated data-driven methodology for the detection of pipe bursts and other events that induce similar abnormal pressure/flow variations (e.g., unauthorized consumptions) at the district metered area (DMA) level has been recently developed by the authors. This methodology works by simultaneously analyzing the data coming on-line from all the pressure and/or flow sensors deployed in a DMA. It makes synergistic use of several self-learning artificial intelligence (AI) and statistical techniques. These include (1) wavelets for the de-noising of the recorded pressure/flow signals; (2) artificial neural networks (ANNs) for the short-term forecasting of pressure/flow signal values; (3) statistical process control (SPC) techniques for the short-term and long-term analysis of the burst/other event-induced pressure/flow variations; and (4) Bayesian inference systems (BISs) for inferring the probability that a pipe burst/other event has occurred in the DMA being studied, raising the corresponding detection alarms, and provide information useful for performing event diagnosis. This paper focuses on the (re)calibration of the above detection methodology with the aim of improving the forecasting performances of the ANN models and the classification performances of the BIS used to raise the detection alarms (i.e., DMA-level BIS). This is achieved by using (1) an Evolutionary Algorithm optimization strategy for selecting the best ANN input structures and related parameter values to be used for training the ANN models, and (2) an Expectation Maximization strategy for (re)calibrating the values in the conditional probability tables (CPTs) of the DMA-level BIS. The (re)calibration procedure is tested on a case study involving several DMAs in the U.K. with real-life pipe bursts/other events, engineered pipe burst events (i.e., simulated by opening fire hydrants), and synthetic pipe burst events (i.e., simulated by arbitrarily adding 'burst flows' to an actual flow signal). The results obtained illustrate that the new (re)calibration procedure improves the performance of the event detection methodology in terms of increased detection speed and reliability. [ABSTRACT FROM AUTHOR]

The parametric approach to eliminating the temperature-caused modal variability in vibration-based structural damage detection requires a correlation model between the modal properties and environmental temperatures. This paper examines the generalization capability of neural network models, established using long-term monitoring data, for correlation between the modal frequencies and environmental temperatures. A total of 770 h modal frequency and temperature data obtained from an instrumented bridge are available for this study, which are further divided into three sets: training data, validation data, and testing data. A two-layer back-propagation neural network (BPNN) is first trained using the training data by the conventional training algorithm, in which the number of hidden nodes is optimally determined using the validation data. Then two new BPNNs are configured with the same data by applying the early stopping technique and the Bayesian regularization technique, respectively. The reproduction and prediction capabilities of the two new BPNNs are examined in respect of the training data and the unseen testing data, and compared with the performance of the baseline BPNN model. This study indicates that both the early stopping and Bayesian regularization techniques can significantly ameliorate the generalization capability of BPNN-based correlation models, and the BPNN model formulated using the early stopping technique outperforms that using the Bayesian regularization technique in both reproduction and prediction capabilities. [ABSTRACT FROM AUTHOR]

Forecasts of future events are required in many activities associated with planning and operation of the components of a water resources system. For the hydrologic component, there is a need for both short term and long term forecasts of streamflow events in order to optimize the system or to plan for future expansion or reduction. This paper presents a comparison of different artificial neural networks (ANNs) algorithms for short term daily streamflow forecasting. Four different ANN algorithms, namely, backpropagation, conjugate gradient, cascade correlation, and Levenberg–Marquardt are applied to continuous streamflow data of the North Platte River in the United States. The models are verified with untrained data. The results from the different algorithms are compared with each other. The correlation analysis was used in the study and found to be useful for determining appropriate input vectors to the ANNs. [ABSTRACT FROM AUTHOR]

An efficient training and pruning methodology based on the H∞ filtering algorithm is proposed for artificial neural networks (ANNs). ANNs are first trained by the H∞ filtering algorithm and then some unimportant weights are removed based on the training. The results presented in the paper show that the proposed method provides better pruning results of the network without losing its generalization capacity. It also provides a robust training algorithm for given arbitrary network structures. The usefulness and effectiveness of the proposed methodology are demonstrated in developing an ANN model of a hysteretic structural system. [ABSTRACT FROM AUTHOR]

The paper presents a methodology for arriving at optimum bid markups in static competitive bidding environments by use of neurofuzzy systems and integrated multidimensional risk analysis algorithms. The neurofuzzy framework enables integration of the objective and subjective factors found in the underlying decision-making process, and serves as the stepping stone for the generation of the multidimensional probability distribution function that governs competitive bidding. Subsequent bid optimization is achieved by employing a multidimensional risk analysis algorithm. [ABSTRACT FROM AUTHOR]

In this paper, the microfauna distribution data of a contact stabilization process were used in a neural network system to model and predict the biological activity of the effluent. Five uncorrelated components of the microfauna were used as the artificial neural network model input to predict the dehydrogenase activity of the effluent (DAE) using back-propagation and general regression algorithms. The models’ optimum architectures were determined for the back-propagation neural network (BPNN) model by varying the number of hidden layers, hidden transfer functions, test set size percentages, and initial weights. Comparison of the two model prediction results showed that the genetic general regression neural network model demonstrated the ability to calibrate the multicomponent microfauna, and yielded reliable DAE close to that resulting from direct experimentation, and thus was judged superior to BPNN models. [ABSTRACT FROM AUTHOR]

Drought is a natural phenomenon that occurs in many places on the planet and may cause considerable damage. Selection of an integrated index for quantifying drought severity is a challenge for decision makers in developing water resources and operation management policies. In this study, the standardized precipitation index, water surface supply index, and Palmer drought severity index have been combined to develop an integrated index, called the hybrid drought index (HDI), using associated damage of drought events. Application of the HDI in drought severity prediction has been examined using two different types of artificial neural networks, namely, a probabilistic neural network and a multilayer perceptron network. These models have been selected due to their special characteristics that are suitable for prediction schemes. The proposed algorithm for developing HDI and drought prediction has been applied to the “Gavkhooni/Zayandeh-rud” basin in the central part of Iran. The results show the merits of each model in prediction of drought severity and model adaptation. The results also show the significant value of the proposed algorithm in formulation of a combined index for drought prediction. [ABSTRACT FROM AUTHOR]

Problems in geotechnical engineering are full of uncertain, vague, and incomplete information. In most instances, successfully solving such problems depends on experts’ knowledge and experience. The primary object of this research was to develop an evolutionary fuzzy neural inference system (EFNIS) to imitate the decision-making processes in the human brain in order to facilitate geotechnical expert decision making. First, an evolutionary fuzzy neural inference model (EFNIM) was constructed by combining the genetic algorithm (GA), fuzzy logic (FL), and neural network (NN). In the proposed model, GA is primarily concerned with optimizing parameters required in the fuzzy neural network; FL with imprecision and approximate reasoning; and NN with learning and curve fitting. This research then integrates the EFNIM with an object-oriented computer technique to develop an EFNIS. Finally, the potential to apply the proposed system to practical geotechnical decision making is validated using two real problems, namely estimating slurry wall duration and selecting retaining wall construction methods. [ABSTRACT FROM AUTHOR]

In China, normal height, which is the height above the geoid calculated using the mean normal gravity along the plumb line, is used in engineering applications. However, the adjusted global positioning system (GPS) height is the height above the surface of the WGS-84 ellipsoid. Thus, it is necessary to convert a GPS height into a normal height. Normally, the conicoid fitting method (CFM) and the neural network method (NNM) are used for this purpose in China, but each has its own advantages and disadvantages. After studying these two methods, a new method (CF&NNM) is conceived. The structure of the back-propagation neural network and detailed algorithm of the CFM, NNM, and CF&NNM are discussed. The procedure of the new method is introduced. A practical engineering example is used to study these three different methods. The results by the three methods are listed. It is demonstrated that the CF&NNM could produce better results than either the CFM or the NNM in deriving the normal height from the GPS height. The theory of the CF&NNM is analyzed. [ABSTRACT FROM AUTHOR]

Surge vessels provide efficient protection against low and high transient pressures in water distribution systems. However, they can be quite expensive, and any reduction in surge vessel size can significantly reduce surge protection costs. Graphical and other heuristic methods reported in literature are limited to sizing surge vessels for simple rising mains. Attempts to use more structured optimization techniques have been largely unsuccessful because of their impractical computational requirements. This article proposes a robust framework for developing surge protection design tools and demonstrates the usefulness of the framework through an example surge vessel sizing tool. The essence of the proposed framework is in the identification of key transient response parameters that influence surge vessel characteristics from seemingly unmanageable transient response data. This parameterization helps the sizing tool to exploit the similarity between transient responses of small pipe networks and subsections of large pipe networks. The framework employs a knowledge-base of transient pressures and flows derived from several small network models and corresponding optimal surge vessel sizes obtained from genetic algorithm (GA) optimizers. Key transient response parameters were identified from this knowledge-base and used as input variables for a neural network model along with the associated surge vessel characteristics as output variables. The trained neural network model was successfully applied for complex pipe networks to obtain optimal surge vessel sizes for transient protection. Neural network model predictions were compared with optimal surge vessel characteristics, and their performance was evaluated by transient simulation models to assess the efficiency of the proposed framework and sizing tool. Application of this framework has the advantage of developing surge protection sizing tools independent of network system schematics and boundary conditions. [ABSTRACT FROM AUTHOR]

The current study was undertaken to analyze the performance of three back-propagation training algorithms of artificial neural network (ANN) along with a multiple linear regression model (MLR) for transient simulation of short/midterm soil temperature (TS). Each ANN used a different type of learning algorithm (gradient descent, conjugate gradient, and Levenberg-Marquardt) as methods for daily and weekly TS prediction. The analysis was performed as a case study using three meteorological parameters [air temperature (TA), net radiation (NR), and relative humidity (RH)] and two hydrological variables [precipitation (P) and runoff (Q)] as input data for a region in Wyoming. Pearson and cross correlation analyses were applied to investigate the relationships between input and target values to determine several input combinations. The correlation analysis indicated that the meteorological parameters of TA and NR were reasonably correlated with TS. Based on the input combinations, 14 models were constructed for each of the daily/weekly MLR and ANNs. The accuracy of the predictions was evaluated by the RMS error (RMSE), mean absolute error (MAE), and Nash-Sutcliffe coefficients between the measured and predicted TS values. The Levenberg-Marquardt ANN method was found to provide a more accurate prediction than the other two types of ANNs and the MLR model. Furthermore, it was found that weekly structure of ANNs performance surpasses that of any other daily structures. Although results illustrated that the soil temperature is a function of meteorological variables, the measured TS is also shown to be positively influenced by hydrological parameters-a fact that has not been pointed out by correlation analyses. [ABSTRACT FROM AUTHOR]

In the analysis or design process of reinforced concrete structures, the peak stress and strain in plain concrete under triaxial stress are critical. However, the nonlinear behavior of concrete under triaxial stresses is very complicated; modeling its behavior is therefore a complicated task. In the present study, several radial basis function neural network (RBFN) models have been developed for predicting peak stress and strain in plain concrete under triaxial stress. For the purpose of constructing the RBFN models, 56 records including normal- and high-strength concretes under triaxial loads were retrieved from literature for analysis. The K-means clustering algorithm and the pseudoinverse technique were employed to train the network for extracting knowledge from training examples. Besides, the performance of the developed RBFN models was estimated by the method of three-way data splits and K-fold cross-validation. On the other hand, a comparative study between the RBFN models and existing regression models was made. The results demonstrate the versatility of RBFN in constructing relationships among multiple variables of nonlinear behavior of concrete under triaxial stresses. Moreover, the results also show that the RBFN models provided better accuracy than the existing parametric models, both in terms of root-mean-square error and correlation coefficient. [ABSTRACT FROM AUTHOR]

The interbasin water transfer project is an alternative to balance the nonuniform temporal and spatial distribution of water resources and water demands, especially in arid and semi arid regions. A water transfer project can be executed if it is environmentally and economically justified. In this study, the feasibility of two interbasin water transfer projects from Karoon River in the western part of Iran to the central part of the country is investigated. An optimization model with an economic objective function to maximize the net benefit of the interbasin water transfer projects is developed. The planning horizon of the model is 23 years (the length of historical data); and it is solved using genetic algorithm. In order to consider environmental impacts of water transfer projects, a water quality simulation model has been used. Then, an Artificial Neural Network model is trained based on the simulation results of a river water quality model in order to be coupled with the optimization model. The outputs of the optimization model are the value of economic gain of the sending (Karoon) basin to offset the loss of agricultural income and environmental costs. The optimal polices for water transfer during the planning horizon has been generated using the coupled simulation-optimization model. Then, operating rules are developed using a K Nearest Neighborhood model for the real time water transfer operation. The results show the significant value of using the proposed algorithm and economic evaluation for water transfer projects. [ABSTRACT FROM AUTHOR]

Generalized models for predicting the storage–yield–reliability functions of surface water reservoirs are developed using multiple linear regression and multilayer perceptron, artificial neural networks (ANNs). Linear regression was used to model the total capacity using the over-year capacity as one of the inputs. However, the ANNs were used to simultaneously model directly the intrinsically nonlinear over-year and total (i.e., within-year+over-year) capacity-yield-reliability functions. The inputs used for the ANNs were basic runoff and systems variables such as the coefficient of variation of annual and monthly runoff, minimum monthly runoff, the demand ratio, and reliability. The results showed that all the models performed well during development and when tested with independent data sets. Both models offer avenues for predicting reservoir capacity at gauged sites without the expense of time-series based simulation alternatives. [ABSTRACT FROM AUTHOR]

A new approach for predicting local scour downstream of grade-control structures based on neural networks is presented. An explicit neural networks formulation (ENNF) is developed using a transfer function (sigmoid) and optimal weights obtained from a training process. A genetic algorithm was used to optimize the neural network architecture and the optimal weights for input and output parameters were obtained using the Levenberg–Marquardt back-propagation algorithm. Experimental data available in the literature, including large-scale results were used for training and validation of the proposed model. The predictive performance of the ENNF was found superior to other regression-based equations and the robustness of ENNF was evaluated using field data. [ABSTRACT FROM AUTHOR]

An optimal groundwater remediation design problem generally requires determination of the location of extraction and injection wells and their pumping and injection rates once the well locations are selected. According to many researchers, the determination of an optimal well location is even more important than the optimal pumping rate in groundwater remediation problems. The objective of the study is to apply a neural network embedded Monte Carlo approach to determine potential well locations for in situ bioremediation of contaminated groundwater. The methodology developed in this study has three important components: a bioremediation simulation model, an artificial neural network, and an application of Monte Carlo simulations. This method has been further applied to an in situ bioremediation problem, for which data are adopted from the available literature. The results show that the proposed approach can successfully identify potential well locations from a set of preselected well locations, which can be further used in optimization algorithms to identify optimal well locations and the corresponding pumping rates. The advantage of this approach is that it reduces the size of the problem considerably by eliminating redundant well locations and hence the computational burden involved. Furthermore, the computational burden of Monte Carlo simulations is managed within a practical time frame by replacing the bioremediation simulation model with a trained neural network. [ABSTRACT FROM AUTHOR]

Ice condition forecasts are very important for preventing ice disasters. Because of the complexity of ice conditions, traditional methods could hardly give accurate prediction in the ice condition forecast, especially for the meandering rivers such as the Yellow River, while the artificial neural networks (ANNs) have an obvious advantage over other traditional methods for forecasting ice conditions. An ANN model based on feed-forward back-propagation and improved by the Levenberg-Marquardt algorithm is applied to forecast the ice conditions of the Yellow River in the Inner Mongolia region. The forecast results in the winter of 2004–2005 are in good agreement with the measured ones. Simulation also shows that the ANN model is superior to the multiple linear regression model and the GM (0,1) model. [ABSTRACT FROM AUTHOR]

We propose an intelligent tool for terrestrial transportation design based on the underlying concept of adaptive meshing. Vehicle routes are deformable visual patterns which cover the geographic area adapting shapes to distributed demands or specific places. In this way the system allows dynamic interaction and continuous visual feedback for the designer constructing and evaluating a transportation bus network. To assist the designer, an optimization framework merging neural networks and evolutionary algorithms permits one to position services and create routes among them automatically. The transport system optimizer built on a geographic information system is illustrated by application to a real life case of clustering and routing for the transportation of the 780 employees of an enterprise. [ABSTRACT FROM AUTHOR]

Applications of artificial neural networks in the field of aeration phenomena in surface aerators, which are not geometrically similar, are explored to predict reaeration rates under varying dynamic as well as geometric conditions. The primary network for prediction is a feed forward network with nonlinear elements. The network consists of an input layer, an output layer, a hidden layer, and the nonlinear transfer function in each processing element. The network requires supervised learning and the learning algorithm is the back-propagation. As back-propagation learning is affected by local minima, and to get over this aspect various other modifications have been suggested like Levenberg-Marquardt, quasi-Newton, conjugate-gradient, etc. The present study suggests that the Levenberg-Marquardt modification is a very efficient algorithm in comparison with others like quasi-Newton and conjugate-gradient. In the situations when the dimension of the input vector is large, and highly correlated, it is useful to reduce the dimension of the input vectors. An effective procedure for performing this operation is principal component analysis. The best prediction performance is achieved when the data are preprocessed using principal components analysis before they are fed to a back-propagated neural network, but at the cost of losing the physical significance of experimental data. The model thus developed can be used to predict the reaeration rate for different sizes of geometric elements (like rotor diameter, sizes of rotor, aerators’ geometry, water depth, etc.) under various dynamic conditions, i.e., the speed of the rotor. [ABSTRACT FROM AUTHOR]

Runoff simulation and prediction in watersheds is an important and essential step in water management systems, safety yield computations, environmental disposal, design of flood control structures, and so on. In this study, the runoff records of Linshan Watershed, Sichuan Province, PRC, during 1984–1993 are presented and used as samples for predictions. The time-delay neural network (TDNN) model combined with a genetic algorithm is proposed and used to predict the nonlinear relationship and to analyze the characteristics of runoff time series in the Linshan Watershed area. Based on analyzing the whole runoff process—for example, the average, maximum, and standard deviation—during said period, the equal length for training and testing is defined. The optimum TDNN structure of August 20, 2001 has been obtained by gradually increasing the time delay to avoid the limitations of the TDNN model. Comparisons between training and testing show that the forecasting model of the runoff level using TDNN combined with genetic algorithms is generally satisfactory and effective, with slight underpredictions at some points. [ABSTRACT FROM AUTHOR]

Most hydrological processes are nonlinear in nature. Although there have been many successful applications of artificial neural networks (ANNs) to capture these nonlinear relationships, there are cases when ANNs have not been able to predict flow extremes (low and high flows) accurately. In a more general sense, ANNs have not performed well when data are clustered. The objective of this study is to demonstrate the influence of clustering on neural network performance by constructing a cluster-based conjunction model based on clustering, neural networks, and genetic algorithm (GA). The performance of the GA-trained cluster-based model is compared to that of the Bayesian regularization back-propagation algorithm, the Levenberg–Marquatrdt algorithm, and a regular GA-trained ANN model. The cluster-based neural network model was tested on (1) chaotic time series data (the Henon map); (2) cross-correlated monthly streamflow data. Results from the study indicate that the cluster-based neural network model offers a promising alternative to its conventional counterparts in mapping fragmented input–output relationships. From threshold analysis it is found that the cluster-based neural network model was effective, compared to its counterparts, in capturing the dynamics of high flows. Improvement in clustering accuracy is shown to improve the performance of the cluster-based neural network model. [ABSTRACT FROM AUTHOR]

Applications of artificial neural networks in simulation and forecasting of hydrologic systems have a long record and generally promising results. Most of the earlier applications were based on the back-propagation (BP) feed-forward method, which used a trial-and-error to determine the final network parameters. The minimal resource allocation network (MRAN) is an on-line adaptive method that automatically configures the number of hidden nodes based on the input–output patterns presented to the network. Numerous MRAN applications in various fields such as system identification and signal processing demonstrated flexibility of the MRAN approach and higher or similar accuracy with more compact networks, compared to other learning algorithms. This research introduces MRAN and assesses its performance in hydrologic applications. The technique was applied to an agricultural watershed in central Illinois to predict daily runoff and nitrate–nitrogen concentration, and the predictions were more accurate compared to the BP model. [ABSTRACT FROM AUTHOR]

Nondestructive testing (NDT) methods are widely used for the performance evaluation of flexible pavements. Falling weight deflectometer (FWD), which measures time-domain deflections resulting from applied impulse loads, is the most popular technique among all NDT methods. The evaluation of the FWD data requires the inversion of mechanical pavement properties using a backcalculation tool that includes both a forward pavement response model and an optimization algorithm. Neural networks (NNs) have also emerged as alternative tools that can be employed for pavement backcalculation problems relative to their real-time processing abilities. However, there have been no comprehensive analyses in previous studies that focus on the learning algorithm and the architecture of a NN model, which considerably affect backcalculation results. In this study, 284 different NN models were developed using synthetic training and testing databases obtained by layered elastic theory. Results indicated that both the learning algorithm and network architecture play important roles in the performance of the NN based backcalculation process. [ABSTRACT FROM AUTHOR]

The majority of artificial neural network (ANN) applications to water resources data employ the feed-forward back-propagation (FFBP) method. This study used an ANN algorithm, the generalized regression neural network (GRNN), for intermittent river flow forecasting and estimation. GRNNs were superior to FFBP in terms of the selected performance criteria. The GRNN simulations do not face the frequently encountered local minima problem in FFBP applications, and GRNNs do not generate forecasts or estimates that are not physically plausible. Preliminary analysis of statistics such as auto- and cross correlation, which explained variance by multilinear regression and the Akaike criterion for the autoregressive moving average (ARMA) model of corresponding order, were found quite informative in determining the number of nodes in the input layer of neural networks. [ABSTRACT FROM AUTHOR]

This technical note applies hybrid models of neural networks (NN) and genetic algorithms (GA) to cost estimation of residential buildings to predict preliminary cost estimates. Data used in the study are for residential buildings constructed from 1997 to 2000 in Seoul, Korea. These are used in training each model and evaluating its performance. The models applied were Model I, which determines each parameter of a back-propagation network by a trial-and-error process; Model II, which determines each parameter of a back-propagation network by GAs; and Model III, which trains weights of NNs using genetic algorithms. The research revealed that optimizing each parameter of back-propagation networks using GAs is most effective in estimating the preliminary costs of residential buildings. Therefore, GAs may help estimators overcome the problem of the lack of adequate rules for determining the parameters of NNs. [ABSTRACT FROM AUTHOR]

Research for advanced traveler information systems (ATIS) has been focused on urban roads. However, research for short-term traffic prediction on all categories of highways is needed, as highway agencies expect to implement intelligent transportation systems across their jurisdictions. In this study, genetic algorithms were used to design time delay neural network (TDNN) models as well as locally weighted regression models to predict short-term traffic for six rural roads from Alberta, Canada. These roads are from various trip-pattern groups and functional classes. Refined TDNN models developed in this study can limit most average errors less than 10% for all study roads. Refined regression models show even higher accuracy. Average errors for the refined regression models are less than 2% for roads with stable patterns. Even for roads with unstable patterns, average errors are below 4%, and the 95th percentile errors are less than 7%. It is believed that such accurate predictions would be useful for highway agencies to implement statewide ATIS. [ABSTRACT FROM AUTHOR]

An integrative computational methodology is developed for the management of nonpoint source pollution from watersheds. The associated decision support system is based on an interface between evolutionary algorithms (EAs) and a comprehensive watershed simulation model, and is capable of identifying optimal or near-optimal land use patterns to satisfy objectives. Specifically, a genetic algorithm (GA) is linked with the U.S. Department of Agriculture’s Soil and Water Assessment Tool (SWAT) for single objective evaluations, and a Strength Pareto Evolutionary Algorithm has been integrated with SWAT for multiobjective optimization. The model can be operated at a small spatial scale, such as a farm field, or on a larger watershed scale. A secondary model that also uses a GA is developed for calibration of the simulation model. Sensitivity analysis and parameterization are carried out in a preliminary step to identify model parameters that need to be calibrated. Application to a demonstration watershed located in Southern Illinois reveals the capability of the model in achieving its intended goals. However, the model is found to be computationally demanding as a direct consequence of repeated SWAT simulations during the search for favorable solutions. An artificial neural network (ANN) has been developed to mimic SWAT outputs and ultimately replace it during the search process. Replacement of SWAT by the ANN results in an 84% reduction in computational time required to identify final land use patterns. The ANN model is trained using a hybrid of evolutionary programming (EP) and the back propagation (BP) algorithms. The hybrid algorithm was found to be more effective and efficient than either EP or BP alone. Overall, this study demonstrates the powerful and multifaceted role that EAs and artificial intelligence techniques could play in solving the complex and realistic problems of environmental and water resources systems. [ABSTRACT FROM AUTHOR]

A backpropagation artificial neural network (ANN) model has been developed to predict the liquefaction cyclic resistance ratio (CRR) of sands using data from several laboratory studies involving undrained cyclic triaxial and cyclic simple shear testing. The model was verified using data that was not used for training as well as a set of independent data available from laboratory cyclic shear tests on another soil. The observed agreement between the predictions and the measured CRR values indicate that the model is capable of effectively capturing the liquefaction resistance of a number of sands under varying initial conditions. The predicted CRR values are mostly sensitive to the variations in relative density thus confirming the ability of the model to mimic the dominant dependence of liquefaction susceptibility on soil density already known from field and experimental observations. Although it is common to use mechanics-based approaches to understand fundamental soil response, the results clearly demonstrate that non-mechanistic ANN modeling also has a strong potential in the prediction of complex phenomena such as liquefaction resistance. [ABSTRACT FROM AUTHOR]

Monitoring structural integrity of pavements is a central task of pavement management systems toward needs analysis and the subsequent design, prioritization, and optimization of pavement maintenance and rehabilitation projects. Nondestructive testing (NDT) methods, including falling weight deflectometer (FWD), are the most widely used monitoring approach. The FWD device creates an impulse load on the pavement surface, and the resulting pavement surface deflections are captured using geophones at a number of distances from the load. Various backcalculation methods have been proposed to calculate pavement structural properties from FWD surface deflection measurements. However, no unique technique has proved to yield a globally optimum solution to this complex, nondifferentiable problem. This study explores development of an effective and reliable backcalculation strategy with attention to variable layer thicknesses. A synthetic database of typical three-layer flexible pavement structures is created by a three-dimensional finite-element method (FEM) program. To replace computationally intensive FEM routines, artificial neural networks (ANNs)-massively parallel computing systems-are trained and tested using the synthetic data. To minimize the error between the FWD-measured deflections and ANN-calculated deflections, a restart covariance matrix adaptation evolution strategy (CMA-ES) is implemented. This strategy is superior to most available evolutionary algorithms (EAs) in efficient, effective, and reliable optimization of complex test functions. Testing of the developed methodology (RCMA-BC) on the synthetic database demonstrates its effectiveness and reliability in backcalculating moduli and surface layer thickness. However, RCMA-BC cannot reliably backcalculate base thickness because the forward calculation routine does not have significant sensitivity to this parameter. Additionally, the RCMA-BC models and backcalculation software are applied to data from the Federal Highway Administration (FHWA) Long-Term Pavement Performance (LTPP) database; RCMA-BC exhibits consistently lower errors in deflections. The RCMA-BC backcalculation results are demonstrated to be independent of seed values. The backcalculated layer thicknesses are also compared to available ground-penetrating radar (GPR) and coring information, showing better agreement of results on thinner surface layers within the considered pavement sections. [ABSTRACT FROM AUTHOR]