Search

The selection of Architect/Engineer (A/E) is one of the critical decisions made by owners early in a project. Given that architects and engineers are not commodities; therefore, the assessment of A/E firms should be based on their qualifications. This paper attempts to quantitatively assess and understand the effects of A/E capabilities on project performance. To achieve this, a model was developed to identify the correlations between A/E consultant qualification, project characteristics, and project outcomes. Afterwards, it was validated using a prediction model developed using artificial neural networks implemented on a case study of 18 projects. This model connects the gap between the consultant procurement decision and its impact on the management performance and outcomes of a project. Furthermore, this paper offers an understanding of current procurement practices, as well as presenting the common evaluation criteria and their respective weights as adopted by several public owners in Alberta. [ABSTRACT FROM AUTHOR]

This paper introduces a newly developed model for automated monitoring and control of productivity in earthmoving operations. The model makes use of advancements in wireless sensing networks, Internet of things (IoT), and artificial intelligence. It utilizes data analytics and a dashboard to provide project managers with actionable data on the status of these operations in near-real time. The model consists of two modules; the first is a low-cost open-source remote sensing data acquisition module for collecting data throughout the execution of earthmoving operations. The collected data are sent to a cloud-based MySQL database, in which the second module is designed to (1) measure actual productivity in near-real-time, (2) detecting the location and condition of hauling roads, and (3) monitoring and reporting driving conditions over these roads. Artificial neural network (ANN) is used in cloud computing for analyzing the productivity to determine and prioritize causes behind experienced loss of productivity from that planned. This paper presents cloud computing over a web-based platform (Knowi®). Productivity measurement and analysis outputs are retrieved through any web browser. The work encompassed field and scaled laboratory experiments in the development and validation processes of the developed model. The laboratory experiments 1:24 scaled loader and dumping truck to simulate loading, hauling, and dumping operations. The data collected from the lab experiments and field work was used as input for the developed model. The results obtained highlight the accuracy of the developed model in recognition of the status of the hauling truck, traveled road condition, and in the estimated duration of the simulated earthmoving cycles. [ABSTRACT FROM AUTHOR]

In this paper, two artificial intelligence (AI) models (i.e., artificial neural networks (ANN) and multivariate adaptive regression splines (MARS)) were developed using cumulative percentiles from the grain-size distribution (GSD) curve as input information to predict the soil-water characteristic curve (SWCC). The importance of each input variable was testified using two different sensitivity analyses. The results show a strong correlation between the SWCC and GSD curves based on large volume of datasets. The ANN provides higher accuracy due to its unique structure; however, the MARS model facilitates in developing a regression equation that contributes to stable performance. The SWCC can be reliably predicted with the MARS regression equation using one data point from the GSD curve and bulk density information. Sensitivity analysis suggests that the prediction of the SWCC is also possible with a reasonable degree of accuracy by using single data point information from the GSD curve as an input variable. Finally, a novel AI aid design method is proposed by combining the MARS regression equation along with physico-empirical model and fitting equation that provides a rapid and reliable technique for predicting the SWCC of sands. [ABSTRACT FROM AUTHOR]

Information on lateral placement and lane indiscipline is useful in simulation of a mixed traffic stream and in identifying the distressed portion of a pavement. Despite this utility, inadequate investigation has been made to estimate the lateral placement of vehicles under prevailing traffic conditions. In a typical mixed traffic situation, vehicles having different static and dynamic characteristics take any lateral gap across the carriageway left empty by other surrounding vehicles and move in an untidy manner. This leads to variation in lateral placement of vehicles, governed by the subject vehicle type. This paper explores the potential factors that influence lateral placement of vehicles and presents an artificial neural network (ANN)-based approach to quantify lateral placement and lane indiscipline in the context of undivided urban roads. Further, sensitivity analysis revealed how different traffic parameters like traffic volume, traffic composition, and directional split influence lateral placement and lane indiscipline within each vehicle category. [ABSTRACT FROM AUTHOR]

Short-term travel time prediction is an essential input to intelligent transportation systems. Timely and accurate traffic forecasting is necessary for advanced traffic management systems and advanced traveler information systems. Despite several short-term travel time prediction approaches have been proposed in the past decade, especially for hybrid models that consist of machine learning models and statistical models, few studies focus on the over-fitting problem brought by hybrid models. The over-fitting problem deteriorates the prediction accuracy especially during peak hours. This paper proposes a hybrid model that embraces wavelet neural network (WNN), Markov chain (MAR), and the volatility (VOA) model for short-term travel time prediction in a freeway system. The purpose of this paper is to provide deeper insights into underlining dynamic traffic patterns and to improve the prediction accuracy and robustness. The method takes periodical analysis, error correction, and noise extraction into consideration and improve the forecasting performance in peak hours. The proposed methodology predicts travel time by decomposing travel time data into three components: a periodic trend presented by a modified WNN, a residual part modeled by Markov chain, and the volatility part estimated by the modified generalized autoregressive conditional heteroscedasticity model. Forecasting performance is investigated with freeway travel time data from Houston, Texas and examined by three measures: mean absolute error, mean absolute percentage error, and root mean square error. The results show that the travel times predicted by the WNN-MAR-VOA method are robust and accurate. Meanwhile, the proposed method is able to capture the underlying periodic characteristics and volatility nature of travel time data. [ABSTRACT FROM AUTHOR]

Accurate prediction of recoverable train delay can support the train dispatchers' decision-making with timetable rescheduling and improving service reliability. In this paper, we present the results of an effort aimed to develop primary delay recovery (PDR) predictor model using train operation records from Wuhan-Guangzhou (W-G) high-speed railway. To this end, we first identified the main variables that contribute to delay, including dwell buffer time, running buffer time, magnitude of primary delay time, and individual sections' influence. Different models are applied and calibrated to predict the PDR. The validation results on test datasets indicate that the random forest regression (RFR) model outperforms the other three alternative models, namely, multiple linear regression (MLR), support vector machine (SVM), and artificial neural networks (ANN) regarding prediction accuracy measure. Specifically, the evaluation results show that when the prediction tolerance is less than 1 min, the RFR model can achieve up to 80.4% of prediction accuracy, while the accuracy level is 44.4%, 78.5%, and 78.5% for MLR, SVM, and ANN models, respectively. [ABSTRACT FROM AUTHOR]

The Pavement-ME can design different flexible or rigid pavement rehabilitation scenarios. A set of quantitative and qualitative input variables are considered to characterize the new and existing pavement layers and predict the damage accumulation in these layers. To evaluate the impact of the inputs on the predicted performance for the rehabilitation options, sequential analyses were performed to assess the main and interactive effects of each input. Continuous responses surface models for interactive effects were obtained using artificial neural network. Furthermore, a normalized sensitivity index was used to quantify the significance of main and interactive effects. The following rehabilitation options were included in this paper: hot mix asphalt (HMA) over HMA, HMA over jointed plain concrete pavement (JPCP), HMA over fractured JPCP, and JPCP over JPCP. The results of the study can assist the highway agencies' efforts to implementing the Pavement-ME, especially when the majority of pavement designs are related to rehabilitation of the existing highways. [ABSTRACT FROM AUTHOR]

The need for a new model capable of accurately predicting the deflection of shape memory alloy (SMA) reinforced concrete (RC) beams is clear from the results obtained in the companion paper. In the present paper, artificial neural networks (ANNs) are utilized to develop such a model. The objective is to create a design tool for computing a reduction factor β to be used in the calculation of the effective moment of inertia for SMA RC beams. First, a database was developed using the results obtained from the parametric study reported in the companion paper. The main factors affecting the moment of inertia have been considered. The network architecture that results in the optimum performance was selected and trained. After demonstrating the network’s ability to predict output data for unfamiliar input data, the network was used to develop a design chart that provides the reduction factor β as a function of the reinforcement ratio and the reinforcement modulus of elasticity. A design example is discussed to illustrate the advantages of using the developed design chart over existing models. Le besoin d’un nouveau modèle pouvant précisément prédire la flexion de poutres en béton armé avec alliage à mémoire de forme (AMF) est clairement établi à partir des résultats obtenus dans l’article complémentaire. Dans le présent article, des réseaux de neurones artificiels sont utilisés pour développer un tel modèle. L’objectif est de créer un outil de conception pour calculer un facteur de réduction β qui sera utilisé dans le calcul du moment efficace d’inertie pour les poutres en béton armé avec alliage à mémoire de forme. Une base de données a tout d’abord été développée en utilisant les résultats obtenus lors de l’étude paramétrique et rapportée dans l’article complémentaire. Les principaux facteurs affectant le moment d’inertie ont été considérés. L’architecture de réseau qui génère le rendement optimal a été choisie et formée. Après avoir démontré la capacité du réseau à prédire les données produites pour des données d’entrée non familières, le réseau a été utilisé pour développer une courbe de conception fournissant le facteur de réduction β comme fonction du rapport d’armature et du module d’élasticité de l’armature. Un exemple de conception est abordé afin d’illustrer les avantages d’utiliser la courbe de conception développée par rapport aux modèles existants. [ABSTRACT FROM AUTHOR]

Synchronization of linearly coupled neural networks with impulsive time window is one of the most challenging problems in the field of complex networks. The intrinsic property of impulse time windows is that impulse instants stochastically occur in a determined time window. In this paper, some sufficient condition in the frameworks of impulse time window and fixed impulse moments ensuring the exponential synchronization of linearly coupled neural networks are proposed by using the discretized Lyapunov function method. Moreover, some approximation algorithms are presented to compute the lower and upper bound of the impulse time window. Finally, two numerical simulations are presented to further demonstrate the validation of the proposed approach. [ABSTRACT FROM AUTHOR]

This paper is concerned with the complete periodic antisynchronization issue of memristor-based neural networks with mixed time-varying delays. Under the framework of Filippov solutions of the differential equations with discontinuous right-hand side, based on Mawhin-like coincidence theorem in set-valued analysis theory, the proof of the existence of periodic solution is presented. By applying the Lyapunov-Krasovskii functional approach, adaptive controller is designed and unknown control parameters of the slave system are determined by adaptive laws, and the complete periodic adaptive antisynchronization condition is addressed to ensure the slave system global antisynchronization with the master system. An illustrative example is given to demonstrate the effectiveness of the obtained results. [ABSTRACT FROM AUTHOR]

Predictive water main break models can assist municipalities in prioritizing the replacement and rehabilitation of water mains. The aim of the paper is to develop an artificial neural network (ANN) model to forecast water main breaks in the water distribution network of the City of Kingston, Ontario, Canada. The ANN model includes variables of diameter, age, length, and soil type to forecast breaks. Historical break data from the 1998 to 2011 period is used to develop the ANN model and forecast pipe breaks over a 5 year planning period. The mean square error, receiver operating characteristics curves, and a confusion matrix are used to evaluate the ANN model training and testing. The trained neural network correctly classified 85% of the data set at the training, validation, and testing stages. Model forecasts showed lower pipe break rates in Kingston West, Kingston Central, and Kingston East. The reduction in break rate in the Kingston system was attributed to the removal of old pipes, and the favourable performance of pipes that are in the usage phase of their life cycle. The ANN model provided Utilities Kingston with a tool to assist them in the planning and management of their water main rehabilitation program. [ABSTRACT FROM AUTHOR]

A quantitative structure−retention relationship study has been carried out on the retention times of 63 furan and phenol derivatives using artificial neural networks (ANNs). First, a large number of descriptors were calculated using HyperChem, Mopac, and Dragon softwares. Then, a suitable number of these descriptors were selected using a multiple linear regression technique. This paper focuses on investigating the role of weight update functions in developing ANNs. Therefore, selected descriptors were used as inputs for ANNs with six different weight update functions including the Levenberg−Marquardt back-propagation network, scaled conjugate gradient back-propagation network, conjugate gradient back-propagation with Powell−Beale restarts network, one-step secant back-propagation network, resilient back-propagation network, and gradient descent with momentum back-propagation network. Comparison of the results indicates that the Levenberg−Marquardt back-propagation network has better predictive power than the other methods. [ABSTRACT FROM AUTHOR]

In this paper, robust stability analysis for neutral-type neural networks with time-varying delays and Markovian jumping parameters is conducted. By using the delay-decomposition approach, a new Lyapunov-Krasovskii functional is constructed. Based on this Lyapunov-Krasovskii functional and some stochastic stability theory, delay-dependent stability criteria are obtained in terms of linear matrix inequalities. Finally, three numerical examples are given to illustrate the effectiveness and reduced conservatism of our theoretical results. [ABSTRACT FROM AUTHOR]

This paper investigates the use of a pre-existing network of resistive strain gauges located on the girders of a single bridge span to determine the classification and estimate the weight of vehicles traveling over that span. Vehicle events on the bridge are identified automatically by a measurement filtering algorithm. Manual classification labels are then applied to a subset of these events to investigate the strain signal features that distinguish various vehicle classes. Trends in these features over time are investigated, and an estimate of vehicle weight is obtained from these features without the need for detailed knowledge of the structure's composition. Additionally, a number of neural network configurations are tested on the problem of determining vehicle class from these features. Results are tested on data from both the summer and winter seasons. Finally, estimates of vehicle weight are improved by using the classification network to filter input events. Cet article examine l'utilisation d'un réseau existant de jauges de déformation résistives situées sur les poutres maîtresses d'une travée de pont afin de déterminer le classement et estimer le poids des véhicules voyageant sur cette travée. La circulation des véhicules sur le pont est identifiée automatiquement par un algorithme de filtrage des mesures. Des étiquettes manuelles de classement sont ensuite appliquées à un sous-ensemble de la circulation afin d'examiner les caractéristiques des signaux de contrainte qui distinguent les diverses classes de véhicules. Les tendances de ces caractéristiques dans le temps sont étudiées; une estimation du poids des véhicules est obtenue à partir de ces caractéristiques sans demander une connaissance détaillée de la composition structurale. De plus, un certain nombre de configuration de réseaux neuronaux ont été mis à l'épreuve afin d'essayer de déterminer la classe de véhicule à partir de ces caractéristiques. Les résultats sont examinés par rapport aux données estivales et hivernales. Pour compléter, les estimations des poids de véhicules sont améliorées en utilisant le réseau de classification afin de filtrer les événements saisis. [ABSTRACT FROM AUTHOR]

This paper presents a methodology to correlate ground surface movements and tunnel boring machine (TBM) operation parameters. Two approaches are proposed and evaluated based on a case study of a shallow tunnel in a dense urban area. The first approach is based on a least square approximation and the second one uses an artificial neural network model. Data analysed were selected from the excavation of the subway line B tunnel in Toulouse, France, which was performed mainly by a shield TBM. Ground movements measured on the 4.7 km long contract 2 are reproduced with reasonable agreement by each of the two approaches. The amount of data (in particular for TBM operation parameters), the rather small amplitude of measured movements (a few millimetres), and the accuracy of these measurements (designed for routine construction management) make it necessary to create a pre-processing technique for the data, and a step-by-step improvement of approaches used. An elimination procedure is proposed to identify the most influential operation parameters and a sensitivity analysis shows their respective effect on ground movements. [ABSTRACT FROM AUTHOR]

In the last few decades, numerous methods have been developed for predicting the axial capacity of pile foundations. Among the available methods, the cone penetration test (CPT)-based models have been shown to give better predictions in many situations. This can be attributed to the fact that CPT-based methods have been developed in accordance with the CPT results, which have been found to yield more reliable soil properties; hence, more accurate axial pile capacity predictions. In this paper, one of the most commonly used artificial intelligence techniques, i.e., artificial neural networks (ANNs), is utilized in an attempt to develop artificial neural network (ANN) models that provide more accurate axial capacity predictions for driven piles and drilled shafts. The ANN models are developed using data collected from the literature and comprise 80 driven pile and 94 drilled-shaft load tests, as well as CPT results. The predictions from the ANN models are compared with those obtained from the most commonly used available CPT-based methods, and statistical analyses are carried out to rank and evaluate the performance of the ANN models and CPT methods. To facilitate the use of the developed ANN models, they are translated into simple design equations suitable for hand calculations. [ABSTRACT FROM AUTHOR]

Glyben is an artificial soil comprised of bentonite and glycerin that is used in scaled physical model testing. The dynamic properties of glyben are strongly influenced by the percent glycerin by mass in the artificial soil, temperature, time after mixing, shear strain amplitude, excitation frequency, and confining stress. This paper describes the development and testing of a modular neural network (MNN) that is suitable for predicting the dynamic properties of glyben. The MNN architecture comprises an input layer, two expert modules (neural networks) linked by a gating network, and an output layer. The MNN is trained using 124 datasets obtained from the literature and tested as part of the current study to evaluate its accuracy. It is shown that the developed MNN can adequately predict the dynamic properties of glyben. [ABSTRACT FROM AUTHOR]

The dynamic modulus (|E*|) is one of the primary hot-mix asphalt (HMA) material property inputs at all three hierarchical levels in the new Mechanistic-empirical pavement design guide (MEPDG). The existing |E*| prediction models were developed mainly from regression analysis of an |E*| database obtained from laboratory testing over many years and, in general, lack the necessary accuracy for making reliable predictions. This paper describes the development of a simplified HMA |E*| prediction model employing artificial neural network (ANN) methodology. The intelligent |E*| prediction models were developed using the latest comprehensive |E*| database that is available to researchers (from National Cooperative Highway Research Program Report 547) containing 7400 data points from 346 HMA mixtures. The ANN model predictions were compared with the Hirsch |E*| prediction model, which has a logical structure and a relatively simple prediction model in terms of the number of input parameters needed with respect to the existing |E*| models. The ANN-based |E*| predictions showed significantly higher accuracy compared with the Hirsch model predictions. The sensitivity of input variables to the ANN model predictions were also examined and discussed. Le module dynamique (|E*|) est l’une des principales propriétés du mélange d’asphalte à chaud utilisée en entrée aux trois niveaux hiérarchiques dans le nouveau guide de conception mécanisto-empirique des chaussées (« Mechanistic-empirical pavement design guide »). Les modèles actuels de prédiction de |E*| sont principalement développés à l’aide d’analyses de régression de bases de données de |E*| obtenues à partir de tests en laboratoire sur de nombreuses années. Ils manquent en général de la précision nécessaire pour faire des prédictions fiables. Cet article décrit la mise au point d’un modèle de prédiction simplifié du |E*| du mélange d’asphalte à chaud qui emploie la méthodologie des réseaux de neurones artificiels (RNA). Les modèles de prédiction intelligents de |E*| ont été développés en utilisant la base de données complète de |E*| la plus récente (rapport no 547 du National Cooperative Highway Research Program) à la disposition des chercheurs et contenant 7400 exemples provenant de 346 mélanges d’asphalte à chaud. Les prédictions de |E*| à partir du modèle des RNA ont été comparées à celles du modèle de Hirsch, lequel a une structure logique et un modèle de prédiction relativement simple en termes du nombre de paramètres nécessaires en entrée comparativement aux autres modèles de prédiction de |E*| existants. Les prédictions de |E*| fondées sur le modèle des RNA ont montré une précision significativement plus élevée comparativement à celles du modèle de Hirsch. La sensibilité des variables d’entrée aux prédictions du modèle des RNA a aussi été examinée et discutée. [ABSTRACT FROM AUTHOR]

Adaptive traffic control systems (ATCS) are designed to calculate traffic signal timings in real time to accommodate current traffic demand changes. A conventional off-line computer-based design procedure that uses iterative evaluations to select alternatives may not be appropriate for ATCS due to its unstable searching time. Search-free analytical procedures that directly find solutions have been noted for ATCS for this reason. This paper demonstrates (i) the shortcomings of an analytical cycle-length design model, specifically COSMOS, in its ability to generate satisfactory solutions at various saturation levels and (ii) an artificial neural network (ANN) based model that can overcome these shortcomings. The ANN-based model consistently yielded cycle lengths that ensure a proper operational target volume to capacity (v/c) ratio, whereas the use of the analytical model resulted in unstable target v/c ratios that might promote congestion. [ABSTRACT FROM AUTHOR]

Efforts have been made in this paper to backcalculate the in situ elastic moduli of asphalt pavement from synthetically derived falling weight deflectometer (FWD) deflections at seven equidistant points. An artificial neural network (ANN) is used as a tool for backcalculation in this work. The ANN is observed to backcalculate layer moduli, both from normal as well as noisy deflection basins, with better accuracy compared with other software, namely, EVERCALC and ExPaS. EVERCALC is a backcalculation software downloaded from the Internet and ExPaS is a backcalculation algorithm developed in-house, based on a “search and expand” approach. Work have been extended further to develop ANN models that can predict a possible rigid layer at the bottom of the pavement and can directly predict the remaining life of the pavement without backcalculating the layer moduli. Finally, a reliability analysis is performed to quantify the performance of backcalculation using an ANN. [ABSTRACT FROM AUTHOR]

In this paper, a new genetic programming (GP) approach for predicting settlement of shallow foundations is presented. The GP model is developed and verified using a large database of standard penetration test (SPT) based case histories that involve measured settlements of shallow foundations. The results of the developed GP model are compared with those of a number of commonly used traditional methods and artificial neural network (ANN) based models. It is shown that the GP model is able to learn, with a very high accuracy, the complex relationship between foundation settlement and its contributing factors, and render this knowledge in the form of a function. The attained function can be used to generalize the learning and apply it to predict settlement of foundations for new cases not used in the development of the model. The advantages of the proposed GP model over the conventional and ANN based models are highlighted. [ABSTRACT FROM AUTHOR]

Construction contracts are becoming more complicated, and the increase in complexity of construction processes, documents, and conditions of contracts has contributed to a higher possibility of disputes and conflicting interpretations. The judicial system has been the means for dispute resolution for claims that cannot be solved through other means such as negotiation and arbitration. Knowledge of previous outcomes of judicial processes will both inform participants in a dispute and increase the likelihood of a less-expensive out-of-court dispute-resolution process. This paper presents a methodology to classify, categorize, and analyze Canadian case-law construction claims. In total, 567 Canadian construction court cases have been collected from 10 different sources and are classified into 12 categories that follow the Canadian Construction Documents Committee (CCDC) standard construction contract document CCDC 2-1994. The proposed methodology is implemented in a computer-integrated system called the Canadian construction claim tracker (CCCT), which consists of one central database and three modules, namely a statistical module, a prediction module, and a classification module. The CCCT provides its users with easy and quick access to past case-law claim information. [ABSTRACT FROM AUTHOR]

The paper deals with two approaches to optimizing pavement condition surveys for the urban pavement network of the City of Winnipeg, Manitoba. First, a nonparametric statistical test was applied to assess the transverse variability of the data. The test compared the ratings for one lane with those of all lanes of each segment. The test concluded that the medians of the two groups are equal at a 92% confidence interval and that there are observed biases in the data. The bias can be eliminated if the surveyed lane is selected randomly. The second approach was to predict visual survey scores from automated (laser-based) measurement of rut depth and international roughness index (IRI). A resilient back-propagation algorithm was selected, and the Kappa coefficient was used to examine the strength of the agreement. The results showed that only moderate agreement was achieved and that additional data elements are required to improve the predictive ability of the model. [ABSTRACT FROM AUTHOR]

IIn this paper the capability of two different feed-forward back-propagation neural network algorithms, namely Levenberg-Marquardt and gradient-descent, in solving complex nonlinear problems is utilized for suspended sediment prediction. The monthly streamflow and suspended sediment data from two stations, Palu and Çayağzi, in the Firat Basin in Turkey are used as case studies. The first part of the study involves the prediction of sediment data for the two stations. The second part of the study focuses on the prediction of the downstream station sediment data using upstream data. The effect of the periodicity on model performance is also investigated in each application. [ABSTRACT FROM AUTHOR]

Marquees are temporary light structures that are connected to the ground by small anchors that act in tension and are designed to resist uplift forces. Due to the temporary nature of these structures, little, if any, attention is given to the pullout capacity of the anchors used to secure them. Failures of such structures are not rare and have resulted in deaths and tens of thousands of dollars of damage. This paper reports on a series of 119 in situ anchor pullout tests conducted on rough mild steel anchors of various lengths, cross-sectional shapes, and areas. Comparison tests are carried out to investigate the impact of the factors affecting the pullout capacity of small anchors. Six methods that determine the axial pile capacity directly from cone penetration test (CPT) data are presented and used to calculate the pullout capacity of small ground anchors. The capacities obtained from these CPT-based methods are compared with predictions from a recently developed artificial neural network (ANN) model. The actual pullout loads are compared with predictions from the CPT and ANN methods, and statistical analyses are carried out to evaluate and rank their performance. The results indicate that the ANN-based method provides superior predictions of the pullout capacity of small ground anchors, whereas the Schmertmann method provides the best performance of the CPT-based techniques examined. [ABSTRACT FROM AUTHOR]

Different relationships have been proposed by codes and researchers for predicting the shear capacity of members without transverse reinforcement. In this paper, the applicability of the artificial neural network (ANN) technique as an analytical alternative to existing methods for predicting this shear capacity is investigated using a critically reviewed and agreed upon database of experimental work that serves as a basis of comparison and (or) assessment of existing and new relationships. Both ANN and eight different codes and researcher's predictions of the shear capacity of the specimens of the database were compared. The ANN predictions are much superior to those of any of the current available relationships. [ABSTRACT FROM AUTHOR]

Uncertainty inevitably affects the performance of a construction project. The ability to identify uncertainty and quantitatively evaluate its impact on the project performance can significantly enhance the accuracy and reliability of a project plan. This paper describes a methodology to classify, model, and reduce uncertainty. This method was implemented within an integrated simulation framework capable of capturing physical features of the facility to be constructed, uncertainty in the working environment, and the construction processes. A "virtual" steel fabrication shop model was developed using the proposed framework to model uncertainty in a steel fabrication shop environment. The artificial neural network (ANN) modeling technique, which plays a significant role in the simulation model, was used to model activity duration based on the identified productivity influencing factors and data collected through a time study project conducted in the fabrication shop. The virtual shop model with the embedded ANN model proved to be more accurate than the traditional approach of modeling activity duration using statistical distributions. [ABSTRACT FROM AUTHOR]

Traditional methods of settlement prediction of shallow foundations on granular soils are far from accurate and consistent. This can be attributed to the fact that the problem of estimating the settlement of shallow foundations on granular soils is very complex and not yet entirely understood. Recently, artificial neural networks (ANNs) have been shown to outperform the most commonly used traditional methods for predicting the settlement of shallow foundations on granular soils. However, despite the relative advantage of the ANN based approach, it does not take into account the uncertainty that may affect the magnitude of the predicted settlement. Artificial neural networks, like more traditional methods of settlement prediction, are based on deterministic approaches that ignore this uncertainty and thus provide single values of settlement with no indication of the level of risk associated with these values. An alternative stochastic approach is essential to provide more rational estimation of settlement. In this paper, the likely distribution of predicted settlements, given the uncertainties associated with settlement prediction, is obtained by combining Monte Carlo simulation with a deterministic ANN model. A set of stochastic design charts, which incorporate the uncertainty associated with the ANN method, is developed. The charts are considered to be useful in the sense that they enable the designer to make informed decisions regarding the level of risk associated with predicted settlements and consequently provide a more realistic indication of what the actual settlement might be. [ABSTRACT FROM AUTHOR]

This paper presents a case study of a digital discontinuity mapping system used as a rock mass characterization tool in an underground hard rock mine. This mapping system allows for a fast acquisition of information that can best characterize the geological structural regime without exposing workers to potentially unsafe conditions. This method can be used to overcome some of the shortcomings of traditional mapping methods, such as limited access to rock exposures. Photographic images of the exposed rock mass are introduced into a software package that has been developed to extract potential discontinuity traces using detection algorithms. Detected features that do not describe discontinuity traces are removed from the images using artificial neural networks. Operator intervention can improve the reliability of the system by linking incomplete discontinuity segments. This developed process results in the construction of a discontinuity trace map that can be used for rock mass characterization purposes. The system was employed to construct discontinuity trace maps of twenty 1.8 m by 1.8 m mapping windows from two locations in an underground hard rock mine. The ability of the system to quantify the geomechanical characteristics of the rock mass was evaluated by comparing the results with those of manually drawn discontinuity trace maps. The results of this study have helped to evaluate the digital face mapping system and identify its limitations. [ABSTRACT FROM AUTHOR]

This paper presents an artificial neural network (ANN) model that predicts the efficiency of pile groups installed in cohesionless soil and subjected to axial loading. The model accounts for the planar geometry of the group (pile diameter, pile spacing, and pile arrangement) and incorporates the effect of pile installation, pile length, cap condition, soil condition, and type of loading on the group efficiency. The results produced by the proposed ANN model compared well with the available results of laboratory and field tests. The ANN model is a viable design tool that assists foundation engineers in predicting the pile group efficiency in an accurate and realistic manner. In addition, this model can be easily updated to incorporate new data and accommodate new design parameters. [ABSTRACT FROM AUTHOR]

This paper suggests the use of wavelet multiresolution analysis (WMRA) as a reliable tool for digital signal processing in structural health monitoring (SHM) systems. A damage occurrence detection algorithm using WMRA augmented with artificial neural networks (ANN) is described. The suggested algorithm allows intelligent monitoring of structures in real time. The probability of damage occurrence is determined by evaluating the wavelet norm index (WNI) representing the energy of a signal describing the change in the system dynamics due to damage. An example application of the proposed algorithm is presented using a finite element simulated structural dynamics of a prestressed concrete bridge. The new algorithm showed very promising results. [ABSTRACT FROM AUTHOR]

Conceptual cost estimates play a crucial role in initial project decisions, although scope is not finalized and very limited design information is available during early project stages. In this paper, the advantages and disadvantages of the current conceptual cost estimation methods are discussed and the use of regression, neural network, and range estimation techniques for conceptual cost estimation of building projects are presented. Historical cost data of continuing care retirement community projects were compiled to develop regression and neural network models. Three linear regression models were considered to identify the significant variables affecting project cost. Two neural network models were developed to examine the possible need for nonlinear or interaction terms in the regression model. Prediction intervals were constructed for the regression model to quantify the level of uncertainty for the estimates. Advantages of simultaneous use of regression analysis, neural networks, and range estimation for conceptual cost estimating are discussed. [ABSTRACT FROM AUTHOR]

This paper presents an application of neural networks to the extraction of bare soil surface parameters such as roughness and soil moisture content using synthetic aperture radar (SAR) satellite data. It uses a fast learning algorithm for training a multilayer feedforward neural network using the Kalman filter technique. Two different databases (theoretical and empirical) were used for the learning stage. Each database was configured as single and multiangular sets of input data (data acquired at two different incidence angles) that are compatible with data from one and two satellite images, respectively. All the configurations are trained and then evaluated using RADARSAT-1 and simulated data. The empirical (measured) database with the multiangular set of input data configuration had the best accuracy with a mean error of 1.54 cm for root mean square (rms) height of the surface roughness and 2.45 for soil dielectric constant in the study area. Based on these results the proposed approach was applied on RADARSAT-1 images from the Chateauguay watershed area (Quebec, Canada) and the final results are presented in the form of roughness and humidity maps. [ABSTRACT FROM AUTHOR]

Structural reliability methods are often used to evaluate the failure performance of geotechnical structures. A common approach is to use the first-order reliability method. Its popularity results from the mathematical simplicity of the method, since only second moment information (mean and coefficient of variation) on the random variables is required. The probability of failure is then assessed by an index known commonly as the reliability index. One critical aspect in determining the reliability index is the explicit definition of the limit state surface of the system. In a problem involving multi-dimensional random variables, the limit state surface is the boundary separating the safe domain from the "failure" (or lack of serviceability) domain. In many complicated and nonlinear problems where the analyses involve the use of numerical procedures such as the finite element method, this surface may be difficult to determine explicitly in terms of the random variables, and therefore the limit state can only be expressed implicitly rather than in a closed-form solution. It is proposed in this paper to use an artificial intelligence technique known as the back-propagation neural network algorithm to model the limit state surface. First, the failure domain is found through repeated point-by-point numerical analyses with different input values. The neural network is then trained on this set of data. Using the optimal weights of the neural network connections, it is possible to develop a mathematical expression relating the input and output variables that approximates the limit state surface. Some examples are given to illustrate the application and accuracy of the proposed approach. [ABSTRACT FROM AUTHOR]

This paper evaluates the potential of using artificial neural networks to model ice parameters related to ice jams in the St. Lawrence River navigation channel through Lake St. Pierre. The artificial neural networks mapped environmental conditions onto ice parameters through multilayer feed-forward networks. The ice parameters include velocity, thickness, concentration, and unit discharge. The input to the network is based on two meteorological parameters: wind velocity and air temperature. The Levenberg–Marquardt algorithm with Bayesian regularization is used to train the feed-forward network. The artificial neural networks adequately modelled the ice parameters. The predicted ice velocity, thickness, and unit discharge were very satisfactory, but ice concentration was not. Methods to improve forecasting (particularly of ice concentration) are suggested. [ABSTRACT FROM AUTHOR]

In this paper, a neural network approach is used to describe the mechanical behavior of unsaturated soils. A sequential architecture was chosen for the network, that is, a multilayer perception network with feedback capability. The input layer consisted of nine neurons, where six of them represented the initial soil conditions and the remaining three neurons were continuously updated for each increment of axial strain based on outputs from the previous increment. The output layer consisted of three neurons representing values of deviatoric stress, volumetric strain, and change in suction at the end of each increment. Next, a database was developed from triaxial test results available in the literature. The database was used to train and test the network. Neural network simulations were compared with experimental results. The comparison indicates the good performance of the proposed network for predicting the mechanical behavior of unsaturated soils. Moreover, the trained network was employed to simulate other stress paths not present in the database to model the so-called "collapse phenomena." The results were promising. [ABSTRACT FROM AUTHOR]

Compaction curves (or density–moisture relationships) of cohesive soils are essential components for establishing practical and reliable criteria for effective control of field compaction. In this paper, modules built from empirical models for simulating the compaction curves of cohesive soils based on easily measured basic soil properties and compaction energy were developed using both statistical regression and artificial neural networks (ANNs) techniques. A large number of compaction curves pertaining to a wide variety of fine-grained soils were collected and used in modeling. The developed modules were able to predict compaction curves of soils with good accuracy, with the ANN-based module outperforming the statistical-based analog. The compaction modules were utilized to inquire about the compactibility behavior of fine-grained soils in relation to their properties and the compaction energy used. Besides their use as independent compaction curve predictors, the compaction modules can be used as supplementary units in numerical models for solving geotechnical engineering problems and as tools useful in preliminary design phases and feasibility studies.Key words: cohesive soils, compaction curve, modeling, neural networks, regression.Les courbes de compactage (ou relations poids volumique sec – teneur en eau) des sols cohérents sont des composantes essentielles dans la définition de critères pratiques et fiables pour le contrôle du compactage sur le terrain. Dans cet article, des modules mis au point pour simuler les courbes de compactage des sols cohérents à partir de propriétés fondamentales des sols et de l'énergie de compactage ont été développés au moyen de la régression statistique de même que des techniques de réseaux de neurones artificiels (ANNs). Un grand nombre de courbes de compactage provenant d'un large éventail de sols à grains fins ont été colligés et utilisés dans la modélisation. Les modules développés ont pu prédire les courbes de compactage de sols avec une bonne précision; mais le module basé sur le ANN a offert une meilleure performance que l'analogie basée sur la statistique. Les modules de compactage ont été utilisés pour se renseigner sur le potentiel de compactage des sols fins en relation avec leurs propriétés et l'énergie de compactage utilisée. En plus de leur utilisation comme moyen indépendant de prédire la courbe de compactage, les modules de compactage peuvent être utilisés comme des éléments supplémentaires dans les modèles numériques pour résoudre des problèmes géotechniques, et comme des outils utiles dans les phases préliminaires de conception et dans les études de faisabilité.Mots clés : sols cohérents, courbe de compactage, modélisation, réseaux de neurones, régression.[Traduit par la Rédaction] [ABSTRACT FROM AUTHOR]

To improve drinking water quality while reducing operating costs, many drinking water utilities are investing in advanced process control and automation technologies. The use of artificial intelligence technologies, specifically artificial neural networks, is increasing in the drinking water treatment industry as they allow for the development of robust nonlinear models of complex unit processes. This paper highlights the utility of artificial neural networks in water quality modelling as well as drinking water treatment process modelling and control through the presentation of several case studies at two large-scale water treatment plants in Edmonton, Alberta. [ABSTRACT FROM AUTHOR]

This paper describes the authors' recent research on the theories about internal workings of feed-forward artificial neural networks (ANNs). Knowledge of the inner workings of ANNs progresses with the evolution of ANN modelling techniques. In the early stages, neural networks consisted of perceptrons and were easy to interpret. Since the emergence of backpropagation learning in the 1980s, ANN models have become much more complex. Even today, many users still consider ANN models as black-box type models. In this paper, a systematic approach is demonstrated to study the behaviours of ANN models in an attempt to open up the ANN "black box". The authors emphasise that feed-forward ANN models are functions, and use a graphical interpretation technique to open the ANN models and show the detailed activities of each inner component through several case studies. These studies include how the training process changes the internal components of an ANN model, how noise impacts the training process, and how the sensitivity of ANN models is affected by the training data. [ABSTRACT FROM AUTHOR]

In the past few years, artificial neural networks (ANNs) have been used in describing and modelling wastewater treatment processes. Artificial neural network models can be identified without a detailed knowledge of the kinetics of the system to be modelled. Also, ANN models can potentially contain a great deal of information about the system itself, including the same type of information contained in conventional deterministic models. The fact that these models can be continuously updated with minimal resource requirements makes them very attractive for application in a real-time control scenario. In the current paper, applications of ANNs in the field of wastewater treatment performance prediction are reviewed. In addition, this paper presents a case study that reports some comprehensive modelling work to develop nonlinear neural network prediction models for the Gold Bar Wastewater Treatment Plant (GBWWTP), the largest sewage treatment facility in Edmonton, Alberta. [ABSTRACT FROM AUTHOR]

Although the artificial neural network (ANN) approach has been used in various disciplines of engineering since the 1980s, its use in the oil sands extraction industry is quite new and has great potential. This paper demonstrates two important ANN modelling techniques that will be very useful to the oil sand industry through a case study. First, the ANN pattern recognition approach is illustrated to categorize a large oil sand processing database consisting of experimental data from the last 20 years. Second, within each category, the authors demonstrate how to follow a general protocol to build ANN models that are capable of predicting the primary recovery and the primary froth quality for the oil sand treatment process with fairly good accuracy. To verify the reliability of the ANN models, besides the regular statistical and graphical analysis, the authors also conducted a sensitivity analysis to test the response logic, parameter interaction, and extrapolation capability of the ANN models. As shown in the paper, these tests are both satisfactory and interesting. With sufficient accuracy and robustness in performance, these ANN models can be used to evaluate both the qualitative and the quantitative response of the oil sand treatment process to the known key parameters, which, in turn, can then be used to optimize the oil sand treatment process. Furthermore, these ANN models can be linked with the geological survey results to estimate the production potential of an oil sand ore field and to manage the cost of stockpiling the chemicals for the treatment process. [ABSTRACT FROM AUTHOR]

This paper examines the application of the artificial neural network (ANN) modeling technique to model a lime softening process at a full-scale drinking water treatment facility. The modeling was done for the Rossdale Water Treatment Plant (WTP) operated by EPCOR Water Services Inc. in Edmonton, Alberta. It was determined that ANN can model a lime clarifier accurately and with superior performance to other modeling methods. During the development stage, a prediction of alum clarifier pH also becomes necessary, and a very accurate inferential (virtual) sensor for pH was developed using ANN. The ANN models were also integrated with the Supervisory Control and Data Acquisition (SCADA) system of the plant so that real-time predictions of lime doses and effluent total hardness could be monitored. It was shown that the performance of the ANN models that were developed using average daily values for the parameters also work well when they are executed in real time. [ABSTRACT FROM AUTHOR]

Artificial neural network (ANN) technology has evolved from the experimental stage into actual industrial applications. To achieve this significant transition, careful planning and adjustment are required. This paper illustrates such an example in the water treatment industry. The project objective is to upgrade the ANN models from a previous research project and install the system on-line in the Rossdale Water Treatment Plant in Edmonton, Alberta, Canada, to forecast raw water colour one day ahead. The article discusses the important issues and techniques to upgrade the neural network model to the actual application. Furthermore, sufficient communication is also required between the designers and the users to address the applicability and user friendly issues in model implementation. Failure in communication can render the whole process ineffective. Possible improvements are also recommended for the future on-line applications. [ABSTRACT FROM AUTHOR]

This paper reports on the application of artificial neural network (ANN) techniques for predicting the concentration of trihalomethanes (THMs) in finished water at the E.L. Smith Water Treatment Plant (WTP) in Edmonton, Alberta, Canada. The formation of THMs in finished water involves many complex chemical reactions and interactions that are difficult to model using conventional methods. The formation of THMs has been found to be correlated to raw and treated water quality characteristics such as colour, pH, and temperature and chemical addition such as chlorine, alum, and powder activated carbon (PAC). Three models were derived using raw water, post clarification water, and a combination of raw and post clarification water parameter inputs. The model that most successfully predicted the concentration of THMs in finished water is the model that uses clarifier effluent parameter inputs. This model can be used at the E.L. Smith WTP for early detection of potentially high THM concentrations in finished water and gives plant operators enough advanced warning to reduce THM precursors. With an adequate understanding of water treatment plant processes and THM formation potential it will be fairly easy for any water treatment facility, which has a few years of historical plant data, to develop its own ANN model for predicting the formation of THM in finished water. [ABSTRACT FROM AUTHOR]

Pilot-scale testing is widely used in the drinking water supply industry to test treatment theories, develop new processes, and enhance process operations. The data sets derived from pilot testing are usually small owing to financial and time considerations. The analysis of such data is extremely complex, since treatment processes are highly complex and nonlinear. Multiple regression analysis is widely considered to be the best available technology for analysing data collected from pilot-scale experiments in the drinking water supply industry. Unfortunately, this technique is limited in its ability to handle the combinations of fixed and random variables that are characteristic of water treatment processes. This paper demonstrates the applicability and advantages of artificial network modelling for pilot-scale data analysis. Data collected at two separate pilot-scale facilities are analysed using the artificial neural network (ANN) technique and multiple regression methods, and performance assessments of the two are made. [ABSTRACT FROM AUTHOR]

Virtually all water utilities are looking at improving the operation of their plants to keep control of costs and to meet stringent water quality regulations. Better process control and automation of the plants can help achieve these goals. However, traditional control techniques such as proportional–integral–derivative (PID) can be inadequate when automating certain water treatment processes such as turbidity, organics, or hardness removal in a clarification process. Advanced process control techniques are alternatives to mitigate this impediment. At the cornerstone of many advanced process control techniques is a model of the process being controlled, which can be developed using artificial neural networks (ANNs). This paper describes various advanced process control techniques, the potentially large role of ANN models in implementing these techniques, and issues and solutions when using ANN in a real-time control system. [ABSTRACT FROM AUTHOR]