Search published articles


Showing 17 results for Neural Network

Misaghi F., Mohammadi K., Mousavizadeh M.h.,
Volume 1, Issue 1 (9-2003)
Abstract

In the present paper, ANN is used to predict the tidal level fluctuations, which is an important parameter in maritime areas. A time lagged recurrent network (TLRN) was used to train the ANN model. In this kind of networks, the problem is representation of the information in time instead of the information among the input patterns, as in the regular ANN models. Two sets of data were used to test the proposed model. San Francisco Bay tidal levels were used to test the performance of the model as a predictive tool. The second set of data was collected in Gouatr Bay in southeast of Iran. This data set was used to show the ability of the ANN model in predicting and completing of data in a station, which has a short period of records. Different model structures were used and compared with each other. In addition, an ARMA model was used to simulate time series data to compare the results with the ANN forecasts. Results proved that ANN can be used effectively in this field and satisfactory accuracy was found for the two examples. Based on this study, an operational real time environment could be achieved when using a trained forecasting neural network.
Mohammad T. Dastorani, Nigel G. Wright,
Volume 2, Issue 3 (9-2004)
Abstract

In this study, an artificial neural networks (ANN) was used to optimise the results obtained from a hydrodynamic model of river flow prediction. The study area is Reynolds Creek Experimental Watershed in southwest Idaho, USA. First a hydrodynamic model was constructed to predict flow at the outlet using time series data from upstream gauging sites as boundary conditions. The model, then was replaced with an ANN model using the same inputs. Finally a hybrid model was employed in which the error of the hydrodynamic model is predicted using an ANN model to optimise the outputs. Simulations were carried out for two different conditions (with and without data from a recently suspended gauging site) to evaluate the effect of this suspension in hydrodynamic, ANN and the hybrid model. Using ANN in this way, the error produced by the hydrodynamic model was predicted and thereby, the results of the model were improved.
S.n. Moghaddas Tafreshi, Gh. Tavakoli Mehrjardi, S.m. Moghaddas Tafreshi,
Volume 5, Issue 2 (6-2007)
Abstract

The safety of buried pipes under repeated load has been a challenging task in geotechnical engineering. In this paper artificial neural network and regression model for predicting the vertical deformation of high-density polyethylene (HDPE), small diameter flexible pipes buried in reinforced trenches, which were subjected to repeated loadings to simulate the heavy vehicle loads, are proposed. The experimental data from tests show that the vertical diametric strain (VDS) of pipe embedded in reinforced sand depends on relative density of sand, number of reinforced layers and height of embedment depth of pipe significantly. Therefore in this investigation, the value of VDS is related to above pointed parameters. A database of 72 experiments from laboratory tests were utilized to train, validate and test the developed neural network and regression model. The results show that the predicted of the vertical diametric strain (VDS) using the trained neural network and regression model are in good agreement with the experimental results but the predictions obtained from the neural network are better than regression model as the maximum percentage of error for training data is less than 1.56% and 27.4%, for neural network and regression model, respectively. Also the additional set of 24 data was used for validation of the model as 90% of predicted results have less than 7% and 21.5% error for neural network and regression model, respectively. A parametric study has been conducted using the trained neural network to study the important parameters on the vertical diametric strain.
Kourosh Behzadian, Abdollah Ardeshir, Zoran Kapelan, Dragan Savic,
Volume 6, Issue 1 (3-2008)
Abstract

A novel approach to determine optimal sampling locations under parameter uncertainty in a water distribution system (WDS) for the purpose of its hydraulic model calibration is presented. The problem is formulated as a multi-objective optimisation problem under calibration parameter uncertainty. The objectives are to maximise the calibrated model accuracy and to minimise the number of sampling devices as a surrogate of sampling design cost. Model accuracy is defined as the average of normalised traces of model prediction covariance matrices, each of which is constructed from a randomly generated sample of calibration parameter values. To resolve the computational time issue, the optimisation problem is solved using a multi-objective genetic algorithm and adaptive neural networks (MOGA-ANN). The verification of results is done by comparison of the optimal sampling locations obtained using the MOGA-ANN model to the ones obtained using the Monte Carlo Simulation (MCS) method. In the MCS method, an equivalent deterministic sampling design optimisation problem is solved for a number of randomly generated calibration model parameter samples.The results show that significant computational savings can be achieved by using MOGA-ANN compared to the MCS model or the GA model based on all full fitness evaluations without significant decrease in the final solution accuracy.
Shahriar Afandizadeh, Jalil Kianfar,
Volume 7, Issue 1 (3-2009)
Abstract

This paper presents a hybrid approach to developing a short-term traffic flow prediction model. In this

approach a primary model is synthesized based on Neural Networks and then the model structure is optimized through

Genetic Algorithm. The proposed approach is applied to a rural highway, Ghazvin-Rasht Road in Iran. The obtained

results are acceptable and indicate that the proposed approach can improve model accuracy while reducing model

structure complexity. Minimum achieved prediction r2 is 0.73 and number of connection links at least reduced 20%

as a result of optimization.


M.h. Vahidnia, A.a. Alesheikh, A. Alimohammadi, F. Hosseinali,
Volume 7, Issue 3 (9-2009)
Abstract

Landslides are major natural hazards which not only result in the loss of human life but also cause economic burden on the society. Therefore, it is essential to develop suitable models to evaluate the susceptibility of slope failures and their zonations. This paper scientifically assesses various methods of landslide susceptibility zonation in GIS environment. A comparative study of Weights of Evidence (WOE), Analytical Hierarchy Process (AHP), Artificial Neural Network (ANN), and Generalized Linear Regression (GLR) procedures for landslide susceptibility zonation is presented. Controlling factors such as lithology, landuse, slope angle, slope aspect, curvature, distance to fault, and distance to drainage were considered as explanatory variables. Data of 151 sample points of observed landslides in Mazandaran Province, Iran, were used to train and test the approaches. Small scale maps (1:1,000,000) were used in this study. The estimated accuracy ranges from 80 to 88 percent. It is then inferred that the application of WOE in rating maps’ categories and ANN to weight effective factors result in the maximum accuracy.
F. Rezaie Moghaddam, Sh. Afandizadeh, M. Ziyadi,
Volume 9, Issue 1 (3-2011)
Abstract

In spite of significant advances in highways safety, a lot of crashes in high severities still occur in highways. Investigation of influential factors on crashes enables engineers to carry out calculations in order to reduce crash severity. Therefore, this paper deals with the models to illustrate the simultaneous influence of human factors, road, vehicle, weather conditions and traffic features including traffic volume and flow speed on the crash severity in urban highways. This study uses a series of artificial neural networks to model and estimate crash severity and to identify significant crash-related factors in urban highways. Applying artificial neural networks in engineering science has been proved in recent years. It is capable to predict and present desired results in spite of limited data sets, which is the remarkable feature of the artificial neural networks models. Obtained results illustrate that the variables such as highway width, head-on collision, type of vehicle at fault, ignoring lateral clearance, following distance, inability to control the vehicle, violating the permissible velocity and deviation to left by drivers are most significant factors that increase crash severity in urban highways.


M. Karamouz, M. Fallahi, S. Nazif, M. Rahimi Farahani,
Volume 10, Issue 4 (12-2012)
Abstract

Runoff simulation is a vital issue in water resource planning and management. Various models with different levels of accuracy

and precision are developed for this purpose considering various prediction time scales. In this paper, two models of IHACRES

(Identification of unit Hydrographs And Component flows from Rainfall, Evaporation and Streamflow data) and ANN (Artificial

Neural Network) models are developed and compared for long term runoff simulation in the south eastern part of Iran. These

models have been utilized to simulate5-month runoff in the wet period of December-April. In IHACRES application, first the

rainfall is predicted using climatic signals and then transformed to runoff. For this purpose, the daily precipitation is downscaled

by two models of SDSM (Statistical Downscaling Model) and LARS-WG (Long Ashton Research Station-Weather Generator). The

best results of these models are selected as IHACRES model input for simulating of runoff. In application of the ANN model,

effective large scale signals of SLP(Sea Level Pressure), SST(Sea Surface Temperature), DSLP and runoff are considered as model

inputs for the study region. The performances of the considered models in real time planning of water resources is evaluated by

comparing simulated runoff with observed data and through SWSI(Surface Water Scarcity Index) drought index calculation.

According to the results, the IHACRES model outperformed ANN in simulating runoff in the study area, and its results are more

likely to be comparable with the observed values and therefore, could be employed with more certainty.


H. Ziari, H. Divandari,
Volume 11, Issue 2 (6-2013)
Abstract

Pavement permanent deformations due to lack of shear strength in mixture are a major reason of rutting. Any simple test to determine mixtures resistance to permanent deformation isn’t distinguished in the 1st level of SUPERPAVE mix design method and prevalent methods for evaluating mixture rut resistance are expensive and time-consuming. Two aggregate types, gradations, asphalt cements and filler types were used in this research to present a prediction model for rutting based on flow number. A mathematical model to estimate flow number of dynamic creep test was developed using model parameters and gyratory compaction slope. The model is validated using Neural Network and Genetic Algorithm and makes it possible to evaluate mixtures shear strength while optimum asphalt content is being determined in laboratory. So not only there is no need to expensive test instruments of rutting or dynamic creep but a remarkable time saving in mix design procedure is achievable.
M. H. Baziar, A. Saeedi Azizkandi,
Volume 11, Issue 2 (11-2013)
Abstract

Due to its critical impact and significant destructive nature during and after seismic events, soil liquefaction and liquefactioninduced

lateral ground spreading have been increasingly important topics in the geotechnical earthquake engineering field

during the past four decades. The aim of this research is to develop an empirical model for the assessment of liquefaction-induced

lateral ground spreading. This study includes three main stages: compilation of liquefaction-induced lateral ground spreading

data from available earthquake case histories (the total number of 525 data points), detecting importance level of seismological,

topographical and geotechnical parameters for the resulted deformations, and proposing an empirical relation to predict

horizontal ground displacement in both ground slope and free face conditions. The statistical parameters and parametric study

presented for this model indicate the superiority of the current relation over the already introduced relations and its applicability

for engineers.


K. J. Tu, Y. W. Huang,
Volume 11, Issue 4 (12-2013)
Abstract

The decisions made in the planning phase of a building project greatly affect its future operation and maintenance (O&M) cost. Recognizing the O&M cost of condominiums’ common facilities as a critical issue for home owners, this research aims to develop an artificial neural network (ANN) O&M cost prediction model to assist developers and architects in effectively assessing the impacts of their decisions made in the planning phase of condominium projects on future O&M costs. A regression cost prediction model was also developed as a benchmark model for testing the predictive accuracy of the ANN model. Six critical building design attributes (building age, number of apartment units, number of floors, average sale price, total floor area, and common facility floor area) which are usually available in the project planning phase, were identified as the input factors to both models and average monthly O&M cost as the output factor. 55 of the 65 existing condominium properties randomly selected were treated as the training samples whose data were used to develop the ANN and regression models the other ten as the test samples to compare and verify the predictive performance of both models. The study results revealed that the ANN model delivers more accurate and reliable cost prediction results, with lower average absolute error around 7.2% and maximum absolute error around 16.7%, as compared with the regression model. This study shows that ANN is an effective method in predicting building O&M costs in the project planning phase. Keywords: Project management, Facility management, Common facilities, Cost modeling.
H. Shahnazari, M. A. Shahin, M. A. Tutunchian,
Volume 12, Issue 1 (1-2014)
Abstract

Due to the heterogeneous nature of granular soils and the involvement of many effective parameters in the geotechnical behavior of soil-foundation systems, the accurate prediction of shallow foundation settlements on cohesionless soils is a complex engineering problem. In this study, three new evolutionary-based techniques, including evolutionary polynomial regression (EPR), classical genetic programming (GP), and gene expression programming (GEP), are utilized to obtain more accurate predictive settlement models. The models are developed using a large databank of standard penetration test (SPT)-based case histories. The values obtained from the new models are compared with those of the most precise models that have been previously proposed by researchers. The results show that the new EPR and GP-based models are able to predict the foundation settlement on cohesionless soils under the described conditions with R2 values higher than 87%. The artificial neural networks (ANNs) and genetic programming (GP)-based models obtained from the literature, have R2 values of about 85% and 83%, respectively which are higher than 80% for the GEP-based model. A subsequent comprehensive parametric study is further carried out to evaluate the sensitivity of the foundation settlement to the effective input parameters. The comparison results prove that the new EPR and GP-based models are the most accurate models. In this study, the feasibility of the EPR, GP and GEP approaches in finding solutions for highly nonlinear problems such as settlement of shallow foundations on granular soils is also clearly illustrated. The developed models are quite simple and straightforward and can be used reliably for routine design practice.
O. Farzaneh, F. Askari, J. Fatemi,
Volume 12, Issue 4 (12-2014)
Abstract

AWT IMAGEPresented is a method of two-dimensional analysis of the active earth pressure due to simultaneous effect of both soil weight and surcharge of strip foundation. The study’s aim is to provide a rigorous solution to the problem in the framework of upper-bound theorem of limit analysis method in order to produce some design charts for calculating the lateral active earth pressure of backfill when loaded by a strip foundation. A kinematically admissible collapse mechanism consisting of several rigid blocks with translational movement is considered in which energy dissipation takes place along planar velocity discontinuities. Comparing the lateral earth forces given by the present analysis with those of other researchers, it is shown that the results of present analysis are higher (better) than other researchers’ results. It was found that with the increase in AWT IMAGE, the proportion of the strip load (q) which is transmitted to the wall decreases. Moreover, Increasing the friction between soil and wall ( AWT IMAGE) will result in the increase of effective distance ( AWT IMAGE). Finally, these results are presented in the form of dimensionless design charts relating the mechanical characteristics of the soil, strip load conditions and active earth pressure.


A. Kaveh, R. Ghaffarian,
Volume 13, Issue 1 (3-2015)
Abstract

The main aim of this paper is to find the optimum shape of arch dams subjected to multiple natural frequency constraints by using an efficient methodology. The optimization is carried out by charged system search algorithm and its enhanced version. Computing the natural frequencies by Finite Element Analysis (FEA) during the optimization process is time consuming. In order to reduce the computational burden, Back Propagation (BP) neural network is trained and utilized to predict the arch dam natural frequencies. It is demonstrated that the optimum design obtained by the Enhanced Charged System Search using the BP network is the best compared with the results of other algorithms. The numerical results show the computational advantageous of the proposed methodology.
Masoud Ahmadi , Hosein Naderpour , Ali Kheyroddin ,
Volume 15, Issue 2 (3-2017)
Abstract

Concrete filled steel tube is constructed using various tube shapes to obtain most efficient properties of concrete core and steel tube. The compressive strength of concrete is considerably increased by the lateral confined steel tube in circular concrete filled steel tube (CCFT). The aim of this study was to present an integrated approach for predicting the steel-confined compressive strength of concrete in CCFT columns under axial loading based on large number of experimental data using artificial neural networks. Neural networks process information in a similar way the human brain does. Neural networks learn by example. The main parameters investigated in this study include the compressive strength of unconfined concrete (f'c), outer diameter (D) and length (L) of column, wall thickness (t) and tensile yield stress (fy) of steel tube. Subsequently, using the reliable network, empirical equations are developed for the confinement effect. The results of proposed model are compared with recently existing model on the basis of the experimental results. The findings demonstrate the precision and applicability of the empirical approach to determine capacity of CCFT columns.


Sohrab Karimi, Hossein Bonakdari, Azadeh Gholami, Amir Hossein Zaji,
Volume 15, Issue 2 (3-2017)
Abstract

Dividing open channels are varied types of open channel structures used to provide water for irrigation channels, agriculture and wastewater networks. In the present study the mean velocity is calculated in different dividing angles within the branches channel through the use of artificial Neural Network (ANN) and coputational fluid dynamices (CFD) models. First the ANSYS-CFX model is used to simulate the flow pattern within the branch with a 90° angle. The results of the CFX model correspond fairly well to the results of the experimental model with Mean Absolute Percentage Error (MAPE) of 5%. After verifying, two CFX model are generated in 30° and 60° angle in different width ratios of 0.6, 0.8, 1, 1.2, and 1.4, and the mean velocities are obtained by flowmeter. Following that ANN model trained and tested through the use of a set of experimental and CFX datas. The comparison showed that the ANN model has an acceptable level of accuracy in predicting the dividing open channel mean flow velocity with mean value R2 of 0.93. Comparing the results indicated that ANN model with the MAPE of 1.8% performs better in 0.8 m width ratio. Also, in this width ratio the MAPE are equal to 1.58, 1.87, and 2.04 % in 30°, 60°, and 90° deviation angles respectively and therefore the model performs better in 30° angle.


Mohsen Poor Arab Moghadam, Parham Pahlavani,
Volume 15, Issue 7 (10-2017)
Abstract

Traffic simulation is a powerful tool for analyzing and solving several transportation issues and traffic problems. However, all traffic micro-simulation models require a suitable car-following model to show the real situation in the best way possible. Several car-following models have been proposed. An obvious disadvantage of the former models is the great number of parameters which are difficult to calibrate. Moreover, any change in these parameters creates considerable disturbances. In this paper, a car-following model was proposed using the Epsilon -Support Vector Regression method whose output is the acceleration of the following car. Radial Basis Function was used as the kernel of the ε-SVR method, and the model parameters were tuned using the Grid Search method. The best values for the parameters were obtained. Furthermore, linear scaling in the interval [-1, 1] was used for both the training and testing input data, and the method was proven to more accurate than the case where no scaling was used. Accordingly, a car-following model with the mean squared error equal to 0.005 and the squared correlation coefficient equal to 0.98 was proposed using the function estimation method through the ε-SVR method. Finally, the ε-SVR output was compared with the results of the well-known car-following models, including Helly linear model, the GHR model, and the Gipps model. It was shown that, when using the scaling and parameters tuning techniques, the proposed method was more accurate compared to all three of those models. Moreover, a function fitting Artificial Neural Network was ran for this purpose and the outputs showed that the result of the ε-SVR method is better than that of the function fitting method by the proposed ANN. Implementing a microscopic validation of the proposed model showed that it can be used in the drivers’ assistance devices and other purposes of Intelligent Transportation Systems.



Page 1 from 1     

© 2019 All Rights Reserved | International Journal of Civil Engineering

Designed & Developed by : Yektaweb