Search

Concrete confined using fiber-reinforced polymer (FRP) composites experience significant enhancements in strength and strain. For the seismic retrofitting of existing reinforced concrete (RC) structures, a large rupture strain (LRS) FRP (i.e., polyethylene terephthalate and naphthalate, denoted as PET and PEN respectively), with a larger rupture strain of more than 5%, is a promising alternative to conventional FRPs with a rupture strain of less than 3%. The majority of analytical models on the stress–strain behavior of FRP-confined concrete under axial compression have focused largely on concrete confined with the traditional FRP material. Analytical research on LRS FRP-confined concrete is, however, limited. Moreover, all existed stress–strain models were determined based on theoretical analysis and test data fitting. In this paper, the artificial neural networks (ANN) method is employed to build a confinement model directly from experimental data to predict the different components of the stress–strain response. A test database consisting of 226 axial compression tests on LRS FRP-confined concrete specimens is used. The test results, in terms of full confined stress–strain response, strength, strain, FRP rupture strain, and dilation response were investigated. Predictive expressions and practical ANN models for the strength, strain, and shape of an axial stress–strain response are provided. Existing models for LRS FRP-confined concrete were also evaluated. The results of the existing and proposed models report that the proposed methods achieve significantly better results. Concrete confined using fiber-reinforced polymer (FRP) composites experience significant enhancements in strength and strain. For the seismic retrofitting of existing reinforced concrete (RC) structures, a large rupture strain (LRS) FRP (i.e., polyethylene terephthalate and naphthalate, denoted as PET and PEN respectively), with a larger rupture strain of more than 5%, is a promising alternative to conventional FRPs with a rupture strain of less than 3%. The majority of analytical models on the stress–strain behavior of FRP-confined concrete under axial compression have focused largely on concrete confined with the traditional FRP material. Analytical research on LRS FRP-confined concrete is, however, limited. Moreover, all existed stress–strain models were determined based on theoretical analysis and test data fitting. In this paper, the artificial neural networks (ANN) method is employed to build a confinement model directly from experimental data to predict the different components of the stress–strain response. A test database consisting of 226 axial compression tests on LRS FRP-confined concrete specimens is used. The test results, in terms of full confined stress–strain response, strength, strain, FRP rupture strain, and dilation response were investigated. Predictive expressions and practical ANN models for the strength, strain, and shape of an axial stress–strain response are provided. Existing models for LRS FRP-confined concrete were also evaluated. The results of the existing and proposed models report that the proposed methods achieve significantly better results.

In this paper, a machine learning-based approach has been proposed to integrate artificial intelligence during the designing of fiber-reinforced polymeric composites. With the help of the proposed approach, an artificial neural network (ANN) model has been developed to achieve the targeted filler content for cotton fiber/polypropylene composite while satisfying the required targeted properties. Previously obtained experimental data sets were trained on the TensorFlow backend using Keras library in Python, followed by hyperparameter tuning and k-fold cross-validation method for acquiring a better performing model to predict the amount of targeted filler content. The developed approach proved to be very efficient and reduced the time and effort of the material characterization for numerous samples, and it will help materials designers to design their future experiments effectively. The developed approach in this paper can be extended for other composite materials if the necessary experimental data are available to train the ANN model.

Fiber Reinforced Polymer has been widely used in the retrofit of existing structures and the construction of new structures. The ultimate conditions and stress–strain model of FRP-confined composites are critical to structural design and prediction of structural response, especially under extreme loads such as earthquakes. In this paper, a data-driven neural network prediction model for ultimate conditions and stress–strain constitutive relation of FRP-confined concrete is proposed, and the validity and accuracy of the model are verified. A uniaxial compression database containing 169 FRP-confined normal concrete cylinders is collected from the open literature, and the quality of the database is examined and evaluated in detail. Based on the feed forward neural network technology, a prediction model for the ultimate conditions of FRP-confined normal concrete cylinders is established. Configurations and hyper-parameters of the network are carefully analyzed, and the optimal model is used for prediction and comparison. Besides, a uniaxial stress–strain model for FRP-confined concrete is established using a neural network with a recursive structure. The prediction accuracy of the proposed model is proven to be superior to the existing design-oriented models. The data-driven neural network prediction models developed in this paper can provide a rapid prediction and design for FRP-confined composites.

This study builds a large database from an extensive survey of existing single-lap shear tests on fiber-reinforced polymer (FRP)-concrete interfacial bonds, comprising 969 test results. Twenty shear-bond strength models published over the past 20 years have been collected and analyzed. These models take into account the effects of the concrete compressive strength, concrete width, FRP elastic modulus, FRP thickness, FRP width and FRP bond length on the ultimate bond strength of the FRP-concrete interface. This paper evaluates the predictive accuracy of the 20 collected models and finds that these models have limited accuracy. To accurately predict the bond strength of the FRP-concrete interface, this paper employs the back propagation neural networks (BPNN) method to train and test the database and builds an artificial neural networks (ANN) model that consists of weighted values, biases and transfer functions. The ANN model test conducts 84 training iterations and selects the optimal combination of input nodes. The accuracy of the developed ANN model is higher (i.e., lower predictive error) than that of the existing bond strength models in the literature. Furthermore, this paper develops an explicit user-friendly formula based on the trained ANN model. The proposed formula estimates and validates the 969 bond strength results, and the predictions using the explicit equation fit the test data very well with small error. As such, the formula can be easily applied during practical designs instead of the implicit processes in the ANN model.

In this paper, the use of Artificial Neural Networks to predict the absorbed energy in the composite plates impacted with low velocity is described. The impact response of a composite laminate depends upon various factors such as thickness, stacking sequence and number of layers. These factors are identified in an earlier study using the sensitivity analysis. These factors have the most prominent effect on the impact resistance of the composite plates. These are studied here with the help of design of experiments so that a suitable data set is obtained. The ability to solve a large number of simulations using FEA gives an advantage in the design optimization with the help of DOE (Design of Experiments). During the study different variations of these factors were tried and the response in terms of the absorbed energy was estimated. The simulation results were then used along with the ANN (Artificial Neural Networks) to fit a function to estimate the amount of absorbed energy. The results from the DOE follow the intuition that the increase of thickness and number of layers increase the performance of the composite plates. The ANN model is trained such that it is able to predict with an acceptable accuracy range the amount of absorbed energy for different configurations of input variables. The paper discusses the codification of input variables so that they can be used to train ANN model. Also, the use of differential evolution algorithm is discussed which is used to select the best possible ANN model based on the maximum error and root mean square error of the ANN models.

This paper presents an integrated method for damage detection of composite structures using their vibration responses, wavelet transform and artificial neural networks (ANN) identification. Structural damage feature proxy vectors are constructed and calculated based on energy variation of the structural vibration responses decomposed using wavelet package before and after the occurrence of structural damage. The ANN are applied to establish the mapping relationship between structural damage feature proxy and damage status (location and severity). The results of crack damage detection for PVC sandwich plates show that the method developed in this paper can be applied to online structural damage detection and health monitoring for various industrial structures.

In this paper, an efficient method for performing minimum weight optimization of composite laminates using artificial neural network (ANN) based surrogate models is proposed. By predicting the buckling loads of the laminates using ANN the use of time-consuming buckling evaluations during the iterative optimization process are avoided. Using for the first time lamination parameters, laminate thickness and other dimensional parameters as inputs for these ANN models significantly reduces the number of models required and therefore computational cost of considering laminates with many different numbers of layers and total thickness. Besides, as the stacking sequences are represented by lamination parameters, the number of inputs of the ANN models is also significantly reduced, avoiding the curse of dimensionality. Finite element analysis (FEA) is employed together with the Latin hypercube sampling (LHS) method to generate the database for the training and testing of the ANN models. The trained ANN models are then employed within a genetic algorithm (GA) to optimize the stacking sequences and structural dimensions to minimize the weight of the composite laminates. The advantages of using ANN in predicting buckling load is proved by comparison with other machine learning methods, and the effectiveness and efficiency of the proposed optimization method is demonstrated through the optimization of flat, blade-stiffened and hat-stiffened laminates.

In this paper, a new hybrid YUKI-ANN is implemented for Structural Health Monitoring (SHM) of laminated composite plates. A finite element model is constructed and used to identify damage at five randomly selected elements in a laminated composite plate. The process of damage identification is divided into two main steps, namely damage localization and quantification. During the localization step, damaged elements are identified using a damage indicator based on Modal Strain Energy change ratio (MSEcr). After excluding the healthy el-ements, the level of damage is predicted using ANN modified with four optimization algorithms: Arithmetic Optimizing Algorithm (AOA), Balancing Composite Motion Optimization (BCMO), Particle Swarm Optimization (PSO), and YUKI algorithm. The performance of these optimization algorithms is evaluated, and it is found that the YUKI algorithm (YA) outperforms AOA, BCMO, and PSO algorithms without exception. YA gives better predictions with lower errors when compared to PSO and BCMO algorithms with equivalent computational time. In some cases, AOA provides slightly better predictions than YA, but the computational time for these predictions is eight times more that of YA. If both performance and efficiency are considered, YA seems to be the best choice. Find the MATLAB code for YUKI-ANN at https://github.com/Brahim-Benaissa/YUKI_ANN.

This short communication comments on a series of papers using artificial neural networks published by Guessasma and co-workers in structures journals. The issues discussed include the size of the database for training a neural network, database enlargement for training a neural network, and extrapolation based on modelling results. The paper includes criticisms of other mistakes in those previously published papers.

The triangulation method has been an effective tool for estimating the source locations of various acoustic emissions (AE). However, it is hard to guarantee high accuracy in composite structures due to their anisotropy. Also, the conventional methods basically require the precise arrival time data of AE signals. Because most of commercial fiber optic sensing systems applicable to real structures have limited measurement performance, the arrival times cannot be clearly identified. In this paper, the magnitudes of fiber optic sensor signals were used for estimating the distances between each sensor and impact location. In order to obtain higher correlation between the magnitude and distance, the signal near the roughly estimated arrival time was used for calculating the magnitude. Then, through the neural network training , the accuracy of estimating the distances from the signal magnitudes could be enhanced. Finally, the triangulation method was applied for localizing the impact sources. As a result, our suggested triangulation method showed the acceptable localization results about the non-trained impact points. Because the input data for this method could be reliably obtained from the commercial fiber optic sensing system, it can be useful for constructing a simple impact monitoring system for the real composite structures.

Laminated composites constitute a crucial family of advanced materials for the development of high-performance components. A classic challenge with this composite family is delamination . Lately, vibration-based methods, augmented by dynamic response signal processing and soft computing techniques, have become central for the non-destructive detection of delamination. To date, several studies have been put forth in relation to tackling the problem. Despite this, the concurrent determination of the presence and severity of delamination remains a challenge for structures more complicated than simple beam-like structures. Grounded on the combination of random forests (RF) and natural frequency-shift damage assessment methods, this paper presents an alternative strategy for the simultaneous predictions of severity and location parameters of delamination in composite plates . The study commences with the establishment of a robust finite element (FE) procedure for the inclusion of delamination in composite plates. Subsequently, the FE procedure was validated with good agreement against published results. With high-dimensional datasets generated from the FE procedure, four RF models with different architectures are developed and compared (including a comparison with a deep neural network model). Relying only on four natural frequencies, the predictive strength of the RF model with optimally tuned hyperparameters is assessed via the FE-generated and experimentally obtained datasets. Results indicate that the proposed RF-based method is capable of accurately predicting five parameters quantifying the location and severity of delamination in composite plates. The best RF model produces a coefficient of correlation as high as 0.996 for the location and size parameters and up to 90% accuracy for classifying the delaminated interface.

In this paper, we propose a hybrid machine learning method to predict the thermal conductivity of polymeric nanocomposites (PNCs). Therefore, a combination of artificial neural network (ANN) and particle swarm optimization (PSO) is applied to estimate the relationship between variable input and output parameters. The ANN is used for modeling the composite while PSO improves the prediction performance through an optimized global minimum search. We select the thermal conductivity of the fibers and the matrix, the kapitza resistance, volume fraction and aspect ratio as input parameters. The output is the macroscopic (homogenized) thermal conductivity of the composite. The results show that the PSO significantly improves the predictive ability of this hybrid intelligent algorithm, which outperforms traditional neural networks.

In this paper, two-stage approaches are proposed to study damage detection, localization and quantification in Functionally Graded Material (FGM) plate structures. Metal and Ceramic FGM plates are considered using three different composite materials: Al/Al 2O3, Al/ZrO2-1, and Al/ZrO2-2 . The FGM plates are modelled using IsoGeometric Analysis (IGA), which is more efficient than the classical Finite Element Method (FEM). Using a power-law distribution of the volume fractions of the plate constituents, the material properties of the plates are expected to vary continuously through their thickness. Improved damage indicator based on Frequency Response Function (FRF) is employed to predict the damaged elements in the first stage. A robust and efficient Improved Artificial Neural Network using Arithmetic Optimization Algorithm (IANN-AOA) is implemented for damage quantification problem in the second stage. The main idea is based on eliminating the healthy elements from the numerical model by the improved indicator. Next, collected data from damaged element based on damage index of an improved indicator is used as input and damage level as output. To prove the robustness of IANN-AOA, a Balancing Composite Motion Optimization (BCMO) is considered to improve ANN (IANN-BCMO) and is used for comparison. The results show that the improved indicator can predict the damaged elements with high precision. For damage quantification, IANN-AOA provides more accurate results than IANN-BCMO.

This paper provides a novel establishment method of the variable-fidelity surrogate model (VFSM) driven by transfer learning for shell buckling prediction problems such as buckling prediction of variable-stiffness composite shells, aiming at achieving higher prediction accuracy than the traditional VFSM constructed by the bridge function. The transfer learning based variable-fidelity surrogate model (TL-VFSM) includes two steps. In the first step, low-fidelity model and high-fidelity model points are calculated and collected into the source dataset and target dataset for transfer learning, respectively. Based on the source dataset, a Deep Neural Network (DNN) model is initialized. Then, the hyperparameter tuning of the DNN model is carried out based on Bayesian Optimization (BO), which uses the target dataset as the validation set. After the hyperparameter pre-tuning, the pre-trained DNN model is obtained. In the second step, early layers of the pre-trained DNN model are remained, while the last layer is replaced. Then, another hyperparameter tuning is carried out to fine-tune the pre-trained DNN model using BO. After that, the transfer learning is finished and the re-trained DNN model is obtained. Through above two steps, the TL-VFSM is constructed. The effectiveness and efficiency of the proposed TL-VFSM are verified by three test functions and two engineering examples of variable-stiffness composite shells and hierarchical stiffened shells. Example results indicate that, using the same computational cost, the proposed TL-VFSM achieves higher prediction accuracy than traditional VFSMs constructed by the bridge function, and it can be a potential method for time-consuming buckling prediction problems with multi-source data or variable fidelities.

Composite hat-stiffened panels are typical composite structures that embody the concept of high strength and light weight. The large number of geometric and material parameters are involved in the design of composite hat-stiffened panels, which needs a computationally efficient analysis tool. This paper deals with the development of an analytical and computationally efficient analysis tool using artificial neural networks for predicting the buckling and ultimate loads of composite hat-stiffened panels under in-plane shear. First, the training, validation, and test sets of the ANN are prepared using finite element analysis . Then, an autoencoder was employed to compress the original characteristics, and a back propagation neural network was established to predict the buckling and ultimate loads. At last, the performance and generalization ability of the ANN are examined based on the test set. The results show that the trained ANN can accurately and efficiently predict the buckling and ultimate loads of composite hat-stiffened panels under in-plane shear, and will be very useful in the practical engineering design.

In this paper, we present a machine learning (ML) method to search for geometric patterns of magnetoelectric multi-phase composites with optimal magnetoelectric coupling properties. The 2D finite element method is used to calculate the coupling coefficients and build the database for the training and testing of ML algorithms . Both the convolution neural network (CNN) and artificial neural network (ANN) algorithms are used as ML algorithms in this work. By investigating the effects of network parameters, such as training data density, iteration number and batch size, we construct the networks with proper parameters and good prediction accuracy. We present the predicted geometric patterns by two methods, CNN and ANN, and compare them with the FEM patterns. Two types of magnetoelectric composites, two -volume fractions of magnetostrictive phase and two system sizes are considered to investigate and improve the prediction efficiency. The presented results show that by using the ML methods, it can well predict the coupling effect and rank optimal patterns . The results also prove the feasibility that by using the ML method, we can accurately predict the coupling performance with very limited data. We aim to predict the optimal patterns instead of the best pattern. Therefore, this work demonstrates the feasibility and offers a new perspective way for the design and optimization of magnetoelectric multi-phase composites by using the ML algorithm.

The existing models for predicting the ultimate axial strain of FRP-confined concrete cylinders are mainly derived from the regression analyses on small datasets. Such models usually targeted more specific use cases and could give inaccurate outcomes when generalized. To this end, this paper presents the data-driven Bayesian probabilistic and machine learning prediction models (i.e., back-propagation artificial neural network, multi-gene genetic programming and support vector machine) with high accuracy. First, a comprehensive database containing 471 test results on the ultimate conditions of FRP-confined concrete cylinders was elaborately compiled from the open literature, and the quality of the database was examined and evaluated in detail. Then, an updating procedure characterized by the Bayesian parameter estimation technique was developed to evaluate the critical parameters in the existing models and refine the selected existing models accordingly. The database was also employed for deriving machine learning models. The computational efficiency, transferability and precision of the proposed models are verified. Results show that the proposed Bayesian posterior models, back-propagation artificial neural network, multi-gene genetic programming and support vector machine models achieved outstanding predictive performance, with the support vector machine yielding the highest prediction accuracy. The superior accuracy of the proposed models should assist various stakeholders in optimal use of FRP-confined concrete columns in diverse construction applications.

This paper focuses on the mechanical behaviors of carbon fiber-reinforced PEEK-titanium hybrid laminates (TiGr) under different temperatures. The tensile strength, flexural strength and inter-laminar shear strength (ILSS) of TiGr laminates with different stacking structures were investigated at room temperature, 80, 150 and 220 °C, respectively. The results show that the tensile strength, flexural strength and ILSS gradually decrease as the temperature increases, and reduced by 12.3%, 45.4% and 29.1% for Type I, 12.9%, 48.8% and 30.1% for Type II, 16.1%, 45.1% and 39.4% for Type III at 220 °C. Moreover, a three-layers back-propagation (BP) neural network was developed to predict the corresponding mechanical behaviors of TiGr laminates, then the genetic algorithm (GA) was introduced to optimize the established BP neural network and the accuracies of the two prediction models were compared. It is found that both the trained BP and GA-BP neural networks could predict the mechanical performances of TiGr laminates well while the GA optimized BP neural network has better prediction results, with average absolute errors of 1.78%, 1.92% and 2.41% for tensile, flexural and inter-laminar shear strengths, respectively. Furthermore, the predictability of the trained models was validated with data from published literature, providing a new concept for the application of BP neural networks to fiber metal laminates.

In this paper, we propose an efficient Artificial Neural Network (ANN) based on the global search capacity of evolutionary algorithms (EAs) to identify damages in laminated composite structures . With remarkable advances, ANN has taken off over the last decades. However, ANN also has major drawbacks relating to local minima issues because it applies backpropagation algorithms based on gradient descent (GD) techniques. This leads to a substantial reduction in the effectiveness and accuracy of ANN. Some researchers have been come up with some solutions to tackle the local minimal problems of ANN by looking for starting beneficial points to eliminate initial local minima based on the global search capacity of stochastic algorithms. Nevertheless, it is commonly acknowledged that those solutions are no longer useful or even counterproductive in some cases if the network contains too many local minima distributed deeply in the search space. Hence, we propose a novel approach applying the fast convergence speed of GD techniques of ANN and the global search capacity of EAs to train the network. The core idea is that EAs are employed to work parallel with ANN during the process of training the network. This guarantees that the network possibly determines the best solution fast and avoids getting stuck in local minima. To enhance the efficiency of the global search capacity, in this work, a hybrid metaheuristic optimization algorithm (HGACS) of EAs is also proposed, which possibly gains the advantages of both Genetic Algorithm (GA) and Cuckoo Search (CS). GA is applied to generate initial populations with the best quality derived from the ability of crossover and mutation operators , whereas CS with global search capacity is used to seek the best solution. Moreover, to deal with the large amount of data utilized to train the network, a vectorization technique is applied for the data of the objective function, which considerably decreases the computational cost. The obtained results prove that the proposed method is superior to traditional ANN, other hybrid-ANNs, and HGACS in terms of accuracy, and significantly reduces computational time compared with HGACS.

Vibration-based damage detection is a promising structural health monitoring technique for identifying delaminations in composite structures. However, its accuracy may be adversely affected by measurement errors and the discrepancy between the real composite structures and their numerical models. Therefore, it is of critical importance to study the effect of such noise on the prediction accuracy. The ideal way to do this is to analyze the noise effects based on real measurements on large number of specimens. However, evidently, this is cost and time intensive in terms of materials, manufacturing and labor. An alternative way is to add artificial noise to the structural dynamic parameters obtained from numerical model and study diverse cases using this modified numerical data. In this paper, different levels of random artificial noise (1%, 2%, 3% and 5%) are added to the numerical frequencies. In order to prevent any undesirable bias, 10 different sets of random noise, both uniformly and normally distributed, are employed at each noise level for the same frequencies. The frequency shifts from these 10 test cases (at each noise level) were then input into three inverse algorithms, namely graphical technique (GT), surrogate-assisted optimization (SAO) and artificial neural network (ANN), to predict the location, size and interface of delaminations in a composite beam. The predictions were subsequently compared to the actual damage parameters to evaluate the effect of the noise on the prediction accuracies. SAO and ANN were also tested with 100 cases of noise and then further with Monte Carlo simulation with 100,000 samples in ANN to verify the results. Additional noise sensitivity tests were carried out on beams with different delamination sizes and interfaces, to investigate the dependence of noise effect on the damage parameters themselves. Results show that the trends in prediction errors with increasing noise levels become more consistent as the number of samples increased. With 100,000 samples, the predictions by ANN with uniformly distributed noise and Gaussian noise were nearly indistinguishable. Furthermore, prediction accuracies are also affected by the location and size of the delamination. Among the three inverse algorithms, the performance of ANN was found to be less reliable in presence of noise compared to SAO and GT.

One of the fundamental criteria in the design of air structures is to achieve required strength at the highest possible reduction of structure weight w. However, it is necessary to keep the second design parameter i.e. stiffness of the air structural element on the proper level in order to satisfy durability and reliability of aircrafts. This paper presents the application of an integrated system based on artificial neural networks and calculations by the finite element method (FEM) for the optimization of geometry of a thin-walled element of an air structure. The main criterion of optimization was to reduce the structure’s weight w at the lowest possible deformation (high stress level) of the tested object. The objective of the analyses – using artificial neural networks (ANN) – was to investigate the effect of 4 individual variables defining geometry of the model (including: system of ribs and their inclination, system of holes in ribs and side walls) on its deformation and final value of the reduced weight w. Numerical analysis showed that the most important variable is the diameter of holes in the side walls of the model.

Among recent artificial intelligence techniques, machine learning (ML) has gained significant attention during the past decade as an emerging topic in civil and structural engineering. This paper presents an efficient and powerful machine learning-based framework for strength predicting of concrete filled steel tubular (CFST) columns under concentric loading. The proposed framework was based on the gradient tree boosting (GTB) algorithm which is one of the most powerful ML techniques for developing predictive models. A comprehensive database of over 1,000 tests on circular CFST columns was also collected from the open literature to serve as training and testing purposes of the developed framework. The efficiency of the proposed framework was demonstrated by comparing its performance with that obtained from other ML methods such as random forest (RF), support vector machines (SVM), decision tree (DT) and deep learning (DL). The accuracy of the developed predictive model was also verified with the current design equations from modern codes of practice as well as existing ML-based predictive models.

In this paper, artificial neural network (ANN) models are developed to predict the load-displacement curves for better understanding the behavior of cotton fiber/polyvinyl chloride (PVC) composites. Series of experiments were undertaken in the laboratory for a varying percentage of composite fiber to characteristic loading. Based on those experimental data, the ANN models were trained and tested on the TensorFlow backend using Keras library in Python by implementing the back-propagation method. For better prediction and accuracy of the load-displacement curves, the grid search hyperparameter tuning method was used, followed by k-fold cross-validation. The developed approach proved to be very efficient and reduced the time and effort of the behavioral study for numerous samples, and it will help materials designers to design their future experiments effectively. A similar approach to predict load-displacement curves using ANN can be extended for any kind of composite material if the necessary experimental data are available.

In this paper an application of Artificial Neural Networks and Probability Ellipse methods for damage detection using Lamb waves is presented. The wave propagation data are used to determine the location and degree of damage in metallic and composite plates. The structural components are to be instrumented with array of actuators and sensors (PCB) able to excite and record its dynamic response. A damage index, calculated from the measured wave propagation data in a reference state (baseline) and the current state, is introduced as a determinant of structural damage. The index is a relative measure comparing the two states of the structure under the same ambient conditions. The damage index increase its intensity when the damage size increase too. Moreover, its calculation in different paths associated with the Artificial Neural Networks or Probability Ellipse methods allow to identify position and entity of the damage. Two panels made by aluminum and fabric composite material have been used for tests. A damage entities simulated as through-thickness hole with diameter equal to 6 mm has been investigated for aluminium panel while three damage entities (through-thickness hole with size equal to 2.5 mm, 4.5 mm and 9 mm) have been investigated for that in composite material. In order to get a training, needful for the application of the ANN method, finite elements analysis have been performed for the aluminum panel taking into account several damage paths. Simulations have been performed in order to characterize the propagation path of the Lamb waves in composite plate changing the damage entities from hole diameter equal to 2.5 mm up to 9 mm. Damage index for each damage size for both simulations and tests will be compared. The potential applications of these methods in developing health monitoring systems in defects-critical structures and in composite material structures are discussed.

Recently, more attention has been paid to Artificial Neural Network (ANN) in the field of damage identification of engineering structures based on modal analysis. This paper proposes a new modified damage indicator, using transmissibility technique to improve Local Frequency Response Ratio (LFCR), combined with ANN. The main objective of the proposed damage indicator is to reduce the number of collected data for fast prediction and with higher accuracy instead of collecting all modal analysis data, i.e. natural frequencies, damping ratios, and mode shapes, or using inverse analysis for damage quantification. The suggested approach is tested using three layers laminated cross-ply [0°/90°/0°] composite beam and plate having single and multiple damage(s). The reliability and accuracy of the proposed application are demonstrated by predicting the severity of damages in the considered composite structures after analysing four damage scenarios.

This paper presents the use Machine Learning (ML) techniques to study the behavior of shear-deficient reinforced concrete (RC) beams strengthened in shear with side-bonded and U-wrapped fiber-reinforced polymers (FRP) laminates. An extensive database consisting of 120 tested specimen and 15 parameters was collected. The resilient back-propagating neural network (RBPNN) was used as a regression tool and the recursive feature elimination (RFE) algorithm and neural interpretation diagram (NID) were employed within the validated RBPNN to identify the parameters that greatly influence the prediction of FRP shear capacity. The results indicated that the RBPNN with the selected parameters was capable of predicting the FRP shear capacity more accurately (r 2 = 0.885; RMSE = 8.1 kN) than that of the RBPNN with the original 15 parameters (r 2 = 0.668; RMSE = 16.6 kN). The model also outperformed previously established standard predictions of ACI 440.R-17, fib 14 and CNRDT200. A comprehensive parametric study was conducted and it concluded that the implementation of RBPNN with RFE and NID, separately, is a viable tool for assessing the strength and behavior of FRP in shear strengthened beams.

A simple, yet powerful, new technique based on Artificial Neural Network (ANN) combined with Particle Swarm Optimization (PSO) for damage quantification in laminated composite plates using Cornwell indicator (CI) is proposed. The analysis is performed in two stages. In the first stage, IsoGeometric Analysis (IGA) is formulated for square laminated composite plates having three layers [ 0 ° / 90 ° / 0 ° ] . In the second stage, IGA model is coupled with PSO for damage quantification using an inverse problem approach and CI as an objective function to minimize the difference between calculated and measured values. This paper aims to assess the application of ANN-PSO for damage quantification in composite structures in order to achieve efficient computational time. IGA is used as modelling technique, CI is used as input data, whereas damage locations and severities are used as output data. The result indicates that high accuracy of damage quantification is achieved using ANN-PSO-IGA-CI. Furthermore, it is demonstrated that huge saving in computational time is achieved when using ANN-PSO-IGA-CI compared with PSO-IGA-CI.

The main objective of this paper is to derive a new empirical formula for predicting the axial compression capacity (ACC) of square concrete-filled steel tubular (SCFST) columns using the artificial neural network (ANN). A total of 300 experimental data of SCFST columns extracted from the literature were used for training, testing, and validating the ANN models. The trial and error method was used to determine the best ANN model, which had the highest correlation coefficient (R) and the lowest mean square error (MSE). In addition, several existing and design code formulae were adopted to evaluate the performance of the current study. The comparative results revealed that the ANN model was more stable and accurate than any other existing formula. Using the validated ANN, a number of master curves were generated to establish a new formula to predict the ACC of the SCFST column. The comparisons with the existing formulae showed a higher accuracy of the proposed empirical formula.

This paper presents a neural network modelling method for damage behaviour of composite materials in conjunction with full-field strain measurements. The proposed method utilises the overall structural response of a laminate composite specimen to develop the constitutive model of a single ply unidirectional laminate. Based on an energy principle, a performance function for training the neural networks is derived in terms of the applied external work and the induced strain energy. This allows the proposed method to develop the neural networks without the presence of stress information that is not necessarily obtainable in experiments with non-uniform deformation. The use of neural networks also enables the proposed method to model the damage behaviour without the constraints on the parameter space, such that a more representative model is developed for the actual material behaviour. An example of tailoring the proposed method to model the in-plane shear damage behaviour of a carbon fibre reinforced plastic (CFRP) laminate is demonstrated as well as its numerical validation. The practical application of the proposed method to multi-axial damage-related nonlinear behaviour of composite is presented using the experimental data obtained from a tensile test with an open-hole specimen.

Current methods for structural damage identification, such as genetic algorithms and artificial neural networks, are often implemented based on a few measured data and a large number of simulation data. The tremendous time-consuming computational work needed for calculating the response data to establish the dynamic model of damaged structures is an important issue for dynamic damage detection. In this paper through using the advanced modeling method of element stiffness matrix modification, the order of the global stiffness matrix can be kept invariable in establishing the model of intact and damaged structures. Then, eigenvalue perturbation theory is introduced to obtain the eigenvalues and eigenvectors of the damaged structure for reducing the computation load. Two artificial neural networks (ANN) are trained based on the response data simulated using finite element method (FEM) and perturbation theory enhanced finite element method (PFEM), respectively. The damage identification capability of these two ANN’s are compared. Results show that the PFEM using the first order eigenvalue perturbation theory provides enough precision for detecting small structural damage and the computational requirement is greatly reduced. Typically, the eigensolution computational time for obtaining the train sample data using PFEM is only 1% of that using the traditional FEM.

Many current methods on structural damage identification such as GA algorithms and neural networks technology are often implemented based on a few measured data and a large number of simulation data from structural vibration responses. Therefore, to establish an accurate and efficient dynamics model for a structure with different damage is an important precondition, so that plentiful simulation data of structural vibration response can be acquired using the dynamics model of the structure with damage. There are two problems when directly meshing small structural damage in FEM modeling, i.e., excessive gridding number and unavoidable errors from differently meshing for the same damaged structure. In order to solve these two problems, this paper presents an improved modeling method based on modifying element stiffness matrix at structural damage position using a modification coefficient. The first step of this improved modeling method is to determine modification coefficient of element stiffness matrix based on the coherence of natural frequencies for two kinds of models, and the second step is to verify the coherence of the frequency-response functions. This study also introduces algorithm and calculating results of damaged element stiffness matrix. Influence of structural damage position and constraint conditions on the modification coefficient for small structural damage are also discussed.

The use of shearography for impact damage detection and characterization is often difficult to achieve in laminated composites due to lack of a formation of clear fringe patterns as in the case of delaminations. Furthermore, existing techniques for interpreting shearograms are often inadequate for impact damage assessment in composite materials. In this paper, a technique is reported that combines conventional image analysis with neural networks (NNs) classification and fuzzy logic inference to characterize shearograms. The objective is to train an automated system for recognition of pertinent characteristics and features from shearograms of composites damaged by impact. The results demonstrate the potential of the NN-based shearography technique in characterizing damage in laminated composites. Its applicability to structural health monitoring is also discussed.

The stiffness prediction of textile composites has been studied intensively over the last 20 years. It is the complex yarn architecture that adds exceptional properties but also requires computationally expensive methods for the accurate solution of the homogenization problem. Braided composites are of special interest for the aerospace and automotive industry and have thus drawn the attention of many researchers, studying and developing analytical and numerical methods for the extraction of the effective elastic properties. This paper intends to study the effect of uncertainties caused by the automated manufacturing procedure, to the elastic behavior of braided composites. In this direction, a fast FEM-based multiscale algorithm is proposed, allowing for uncertainty introduction and response variability calculation of the macro-scale properties of 3D braided composites, within a Monte Carlo framework. Artificial neural networks are used to reduce the computational effort even more, since they allow for rapid generation of large samples when trained. With this approach it is feasible to apply a variance-based global sensitivity analysis in order to identify the most crucial uncertain parameters through the costly Sobol indices. The proposed method is straightforward, quite accurate and highlights the importance of realistic uncertainty quantification.

This paper examines the crashworthiness performance of composite rectangular tubes using experimental and artificial neural network (ANN) techniques. Based on experimentally obtained values of different crashworthiness parameters under various loading conditions, ANN models are constructed to identify the optimum cross-sectional aspect ratio of cotton fiber/epoxy laminated composite to achieve the targeted mechanical properties such as load carrying and energy absorption capability. Experimental findings show that axially and laterally loaded rectangular tubes were significantly affected by their aspect ratio. Furthermore, the predictions obtained from the ANN models showed consistency with the experimental data. In addition, the developed ANN captured the complicated nonlinear relationship among crashworthiness parameters to obtain insight into the practical design of the composite materials.