---

An efficient methodology is proposed to find optimal shape of arch dams on the basis of constrained natural frequencies. The optimization is carried out by virtual sub population (VSP) evolutionary algorithm employing real values of design variables. In order to reduce the computational cost of the optimization process, the arch dam natural frequencies are predicted by properly trained back propagation (BP) and wavelet back propagation (WBP) neural networks. The WBP network provides better generalization compared with the standard BP network. The numerical results demonstrate the computational merits of the proposed methodology for optimum design of arch dams.

---

Neural networks have recently been widely used to model some of the human activities in many areas of civil engineering applications. In the present paper, artificial neural networks (ANN) for predicting compressive strength of cubes and durability of concrete containing metakaolin with fly ash and silica fume with fly ash are developed at the age of 3, 7, 28, 56 and 90 days. For building these models, training and testing using the available experimental results for 140 specimens produced with 7 different mixture proportions are used. The data used in the multi-layer feed forward neural networks models are designed in a format of eight input parameters covering the age of specimen, cement, metakaolin (MK), fly ash (FA), water, sand, aggregate and superplasticizer and in another set of specimen which contain SF instead of MK. According to these input parameters, in the multi-layer feed forward neural networks models are used to predict the compressive strength and durability values of concrete. It shown that neural networks have high potential for predicting the compressive strength and durability values of the concretes containing metakaolin, silica fume and fly ash.

---

This study deals with predicting nonlinear time history deflection of scallop domes subject to earthquake loading employing neural network technique. Scallop domes have alternate ridged and grooves that radiate from the centre. There are two main types of scallop domes, lattice and continuous, which the latticed type of scallop domes is considered in the present paper. Due to the large number of the structural nodes and elements of scallop domes, nonlinear time history analysis of such structures is time consuming. In this study to reduce the computational burden radial basis function (RBF) neural network is utilized. The type of inputs of neural network models seriously affects the computational performance and accuracy of the network. Two types of input vectors: cross-sectional properties and natural periods of the structures can be employed for neural network training. In this paper the most influential natural periods of the structure are determined by adaptive neuro-fuzzy inference system (ANFIS) and then are used as the input vector of the RBF network. Results of illustrative example demonstrate high performance and computational accuracy of RBF network.

---

Heavy economic losses and human casualties caused by destructive earthquakes around the world clearly show the need for a systematic approach for large scale damage detection of various types of existing structures. That could provide the proper means for the decision makers for any rehabilitation plans. The aim of this study is to present an innovative method for investigating the seismic vulnerability of the existing concrete structures with moment resisting frames (MRF). For this purpose, a number of 2-D structural models with varying number of bays and stories are designed based on the previous Iranian seismic design code, Standard 2800 (First Edition). The seismically–induced damages to these structural models are determined by performing extensive nonlinear dynamic analyses under a number of earthquake records. Using the IDARC program for dynamic analyses, the Park and Ang damage index is considered for damage evaluation of the structural models. A database is generated using the level of induced damages versus different parameters such as PGA, the ratio of number of stories to number of bays, the dynamic properties of the structures models such as natural frequencies and earthquakes. Finally, in order to estimate the vulnerability of any typical reinforced MRF concrete structures, a number of artificial neural networks are trained for estimation of the probable seismic damage index.

---

The main contribution of the present paper is to train efficient neural networks for seismic design of double layer grids subject to multiple-earthquake loading. As the seismic analysis and design of such large scale structures require high computational efforts, employing neural network techniques substantially decreases the computational burden. Square-on-square double layer grids with the variable length of span and height are considered. Back-propagation (BP), radial basis function (RBF) and generalized regression (GR) neural networks are trained for efficiently prediction of the seismic design of the structures. The numerical results demonstrate the superiority of the GR over the BP and RBF neural networks.

---

In this paper the performance of Artificial Neural Networks (ANNs) and Adaptive Neuro- Fuzzy Inference Systems (ANFIS) in simulating the inverse dynamic behavior of Magneto- Rheological (MR) dampers is investigated. MR dampers are one of the most applicable methods in semi active control of seismic response of structures. Various mathematical models are introduced to simulate the dynamic behavior of MR dampers. The Modified Bouc-Wen model is an appropriate model that has an acceptable accuracy in calculating the generated force of dampers compared to others. In this model displacement and voltage of a MR damper are known while the force generated by MR damper is considered as the unknown. Because of highly nonlinear characteristics of modified bouc-wen model determination of inverse dynamic behavior of MR dampers are generally done using ANNs and ANFIS. Since the ANNs and ANFIS have different mechanisms for emulating desired functions, their responses may be different. In this research the performance of a Back Propagation Neural Network (BPNN), Radial Basis Functions Neural Network (RBFNN) and ANFIS in estimating the inverse dynamic behavior of MR dampers are compared. The results emphasize on the advancement of ANFIS to the other methods studied in estimation of inverse dynamic behavior of MR dampers.

---

In this paper, a methodology is presented for optimum shape design of double-layer grids subject to gravity and earthquake loadings. The design variables are the number of divisions in two directions, the height between two layers and the cross-sectional areas of the structural elements. The objective function is the weight of the structure and the design constraints are some limitations on stress and slenderness of the elements besides the vertical displacements of the joints. To achieve the optimization task a variant of particle swarm optimization (PSO) entitled as quantum-behaved particle swarm optimization (QPSO) algorithm is employed. The computational burden of the optimization process due to performing time history analysis is very high. In order to decrease the optimization time, the radial basis function (RBF) neural networks are employed to predict the desired responses of the structures during the optimization process. The numerical results demonstrate the effectiveness of the presented methodology

---

The main objective of this paper is to use ant optimized neural networks to generate artificial earthquake records. In this regard, training accelerograms selected according to the site geology of recorder station and Wavelet Packet Transform (WPT) used to decompose these records. Then Artificial Neural Networks (ANN) optimized with Ant Colony Optimization and resilient Backpropagation algorithm and learn to relate the dimension reduced response spectrum of records to their wavelet packet coefficients. Trained ANNs are capable to produce wavelet packet coefficients for a specified spectrum, so by using inverse WPT artificial accelerograms obtained. By using these tools, the learning time of ANNs reduced salient and generated accelerograms had more spectrum-compatibility and save their essence as earthquake accelerograms.

---

The excavation damaged zone (EDZ) can be defined as a rock zone where the rock properties and conditions have been changed due to the processes related to an excavation. This zone affects the behavior of rock mass surrounding the construction that reduces the stability and safety factor and increase probability of failure of the structure. In this paper, a methodology was examined for computing the creation probability of damaged zone by Latin hypercube sampling based on a feed-forward artificial neural network (ANN) optimized by hybrid particle swarm optimization and genetic algorithm (HPSOGA). The HPSOGA was carried out to decide the initial weights of the neural network. A case study in a test gallery of the Gotvand dam, Iran was carried out and creation probabilities of 0.191 for highly damaged zone (HDZ) and 0.502 for EDZ were obtained.

---

In the present study, the reliability assessment of performance-based optimally seismic designed reinforced concrete (RC) and steel moment frames is investigated. In order to achieve this task, an efficient methodology is proposed by integrating Monte Carlo simulation (MCS) and neural networks (NN). Two NN models including radial basis function (RBF) and back propagation (BP) models are examined in this study. In the proposed methodology, MCS is used to estimate the total exceedence probability associated with immediate occupancy (IO), life safety (LS) and collapse prevention (CP) performance levels. To reduce the computational burden of MCS process, the required nonlinear responses of the generated structures are predicted by RBF and BP models. The numerical results imply the superiority of BP to RBF in prediction of structural responses associated with performance levels. Finally, the obtained results demonstrate the high efficiency of the proposed methodology for reliability assessment of RC and steel frame structures.

---

The Hartley transform, a real-valued alternative to the complex Fourier transform, is presented as an efficient tool for the analysis and simulation of earthquake accelerograms. This paper is introduced a novel method based on discrete Hartley transform (DHT) and radial basis function (RBF) neural network for generation of artificial earthquake accelerograms from specific target spectrums. Acceleration time histories of horizontal earthquake ground motion are obtained by the capability of learning of RBF neural network to expand the knowledge of the inverse mapping from the response spectrum to earthquake accelerogram. In the first step, Hartley transform is used to decompose earthquake accelerograms, then a RBF neural network is trained to learn to relate the response spectrum to Hartley spectrum. Finally, the generated accelerogram using inverse discrete Hartley transform is obtained from target spectrum. Approximately 200 uniformly scaled horizontal ground motion records from recent Iran’s earthquakes are used to decompose with real Hartley transform and train networks.

---

Structural vibration control is one of the most important features in structural engineering. Real-time information about seismic resultant forces is required for deciding module of intelligent control systems. Evaluation of lateral forces during an earthquake is a complicated problem considering uncertainties of gravity loads amount and distribution and earthquake characteristics. An artificial neural network (ANN) has been trained in this article to estimate these forces. This ANN was trained on the results of time history analysis of a three-story building under 702 different loadings. Results of numerical examples verify that the trained ANN can predict the expected forces with negligible deviations.

---

Vibration based techniques of structural damage detection using model updating method, are computationally expensive for large-scale structures. In this study, after locating precisely the eventual damage of a structure using modal strain energy based index (MSEBI), To efficiently reduce the computational cost of model updating during the optimization process of damage severity detection, the MSEBI of structural elements is evaluated using properly trained cascade feed-forward neural network (CFNN). In order to achieve an appropriate artificial neural network (ANN) model for MSEBI evaluation, a set of feed-forward artificial neural networks which are more suitable for non-linear approximation, are trained. All of these neural networks are tested and the results demonstrate that the CFNN model with log-sigmoid hidden layer transfer function is the most suitable ANN model among these selected ANNs. Moreover, to increase damage severity detection accuracy, the optimization process of damage severity detection is carried out by particle swarm optimization (PSO) whose cost function is constructed based on MSEBI. To validate the proposed solution method, two structural examples with different number of members are presented. The results indicate that after determining the damage location, the proposed solution method for damage severity detection leads to significant reduction of computational time compared to finite element method. Furthermore, engaging PSO algorithm by efficient approximation mechanism of finite element (FE) model, maintains the acceptable accuracy of damage severity detection.

---

Optimal locations of the actuators for frame active control are investigated in this article. The aim is to minimize the structural drifts by employing several actuators. By utilizing genetic algorithm, the appropriate locations of the actuators are determined. They should be placed in locations where they can minimize the maximum structural drift. To explore the capability of the proposed techniques, the response of a 20-story building is controlled using three suggested methods. Furthermore, two different concepts are considered for comparing the performance of the authors' approaches. One is based on the maximum responses of the structure, and the other is according to the magnitudes of the actuators' forces. All findings prove the efficiency of the recommended strategies.

---

In the present study, two different data-driven models, artificial neural network (ANN) and multiple linear regression (MLR) models, have been developed to predict the 28 days compressive strength of concrete. Seven different parameters namely 3/4 mm sand, 3/8 mm sand, cement content, gravel, maximums size of aggregate, fineness modulus, and water-cement ratio were considered as input variables. For each set of these input variables, the 28 days compressive strength of concrete were determined. A total number of 140 input-target pairs were gathered, divided into 70%, 15%, and 15% for training, validation, and testing steps in artificial neural network model, respectively, and divided into 85% and 15% for training and testing steps in multiple linear regression model, respectively. Comparing the testing steps of both of the models, it can be concluded that the artificial neural network model is more capable in predicting the compressive strength of concrete in compare to multiple linear regression model. In other words, multiple linear regression model is better to be used for preliminary mix design of concrete, and artificial neural network model is recommended in the mix design optimization and in the case of higher accuracy requirements.

---

Shear wave velocity (V_s) is known as one of the fundamental material parameters which is useful in dynamic analysis. It is especially used to determine the dynamic shear modulus of the soil layers. Nowadays, several empirical equations have been presented to estimate the shear wave velocity based on the results from Standard Penetration Test (SPT) and soil type. Most of these equations result in different estimation of V_s for the same soils. In some cases a divergence of up to 100% has been reported. In the following study, having used the field study results of Urmia City and Artificial Neural Networks, a new correlation between V_s and several simple geotechnical parameters (i.e. Modified SPT value number (N₆₀), Effective overburden stress, percentage of passing from Sieve #200 (Fc), plastic modulus (PI) and mean grain size (d₅₀)) is presented. Using sensitivity analysis it is been shown that the effect of PI in V_s prediction is more than that of N₆₀ in over consolidated clays. It is also observed that Fc has a high influence on evaluation of shear wave velocity of silty soils.

---

This paper deals with the effect of fiber aspect ratio of steel fibers on shear strength of steel fiber reinforced concrete deep beams loaded with shear span to depth ratio less than two using the artificial neural network technique. The network model predicts reasonably good results when compared with the equation proposed by previous researchers. The parametric study involves deep beams of M55 grade concrete with fiber volume fraction 0.5% to 2% of fiber aspect ratio ranging from 50 to 100 and longitudinal steel percentage varying from 0% to 2.5%. The analysis reveals that the fiber aspect ratio also affects the shear strength and needs to be combined with fiber volume fraction.
 

---

This study investigates the prediction model of compressive strength of self–compacting concrete (SCC) by utilizing soft computing techniques. The techniques consist of adaptive neuro–based fuzzy inference system (ANFIS), artificial neural network (ANN) and the hybrid of particle swarm optimization with passive congregation (PSOPC) and ANFIS called PSOPC–ANFIS. Their performances are comparatively evaluated in order to find the best prediction model. In this study, SCC mixtures containing different percentage of nano SiO₂ (NS), nano–TiO₂ (NT), nano–Al₂O₃ (NA), also binary and ternary combining of these nanoparticles are selected. The results indicate that the PSOPC–ANFIS approach in comparison with the ANFIS and ANN techniques obtains an improvement in term of generalization and predictive accuracy. Although, the ANFIS and ANN techniques are a suitable model for this purpose, PSO integrated with the ANFIS is a flexible and accurate method due tothe stronger global search ability of the PSOPC algorithm.

---

Optimization techniques can be efficiently utilized to achieve an optimal shape for arch dams. This optimal design can consider the conditions of the economy and safety simultaneously. The main aim is to present an applicable and practical model and suggest an algorithm for optimization of concrete arch dams to enhance their seismic performance. To achieve this purpose, a preliminary optimization is accomplished using PSO procedure in the first stage. Capabilities of Ansys Parametric Design Language (APDL) are applied for modeling the Dam-Foundation-Reservoir system. In the second stage with training the neural network, Group Method of Data Handling (GMDH) and replacement of Ansys analyst, optimal results have been achieved with the lowest error and less number of iteration respectively. Then a real world double-arch dam is presented to demonstrate the effectiveness and practicality of the PSO-GMDH. The numerical results reveal that the proposed method called PSO-GMDH provides faster rate and high searching accuracy to achieve the optimal shape of arch concrete dams and the modification and optimization of shape have a quite important role in increasing the safety against dynamic design loads.

---

Reinforced concrete (RC) slabs exhibit complexities in their structural behavior under load due to the composite nature of the material and the multitude and variety of factors that affect such behavior. Current methods for determining the load-deflection behavior of reinforced concrete slabs are limited in scope and are mostly dependable on the results of experimental tests. In this study, an alternative approach using Artificial Neural Network (ANN) model is produced to predict the load-deflection behavior of a two-way RC slab. In the study, 30 sets of RC slab specimens of sizes 700mm x 600mm x 75mm were cast, cured for 28days using the sprinkling method of curing and tested for deflection experimentally by applying loads ranging from 10kN to 155kN at intervals of 5kN. ANN model was then developed using the neural network toolbox of ANN in MATLAB version R2015a using back propagation algorithm. About 54% of the RC specimens were used for the training of the network while 23% of the sets were used for validation leaving the remaining 23 % for testing the network. The experimental test results show that the higher the applied load on the slab, the higher the deflection. The result of the ANN model shows a good correlation between the experimental test and the predicted results with training, validation and test correlation coefficients of 0.99692, 0.98921 and 0.99611 respectively. It was also found that ANN model is quite efficient in determining the deflection of 2-way RC slab. The predicted accuracy of performance value for the load-deflection set falls at 96.67% of the experimental load-deflection with a 0.31% minimum error using the Microsoft spreadsheet model. As such the comprehensive spreadsheet tool created to incorporate the optimum neural network. The spreadsheet model uses the Microsoft version 2013 excel tool software and can be used by structural engineers for instantaneous access to the prediction if any aspect of a concrete slab behavior given minimal data to describe the slab and the loading condition.