Figure 2 - uploaded by Damodar Maity
Content may be subject to copyright.
Source publication
Methods to assess damage in a structure and to evaluate their life are very important to ensure the structural integrity of operating plants and structures. The difficulties faced in implementing traditional procedures and the need to develop computer based automated evaluation process motivates the application of soft-computing tools like artifici...
Context in source publication
Context 1
... analysis of the beam is carried out by generalized eigen value problem analysis. Mode shapes of undamaged cantilever beam are shown in Figure 2. The damage is introduced by reducing element stiffness, i.e., by giving SRF equal to 0.0622 to the element 9 in the FE model while calculating the CDFs in the beam. ...
Citations
... Several hybrid soft computing algorithms like neurogenetic algorithm [38], hybrid real GA [39] and radial basis neural network [40] have been used for effective damage identification and localization. Principle component analysis and pattern recognition were used by Bandara et al. [41] for FRF based damage identification. ...
Optimization algorithms are primarily responsible for efficiency in vibration-based damage detection particularly when utilizing the inverse approach. A complex problem of damage detection tends to converge into local minima, generated by a false damaged state which produces a response that is almost similar to the actual damaged state. Hence, there is a need for an efficient and accurate soft computing technique that can find the global minima or the actual damaged state. Recently, the teaching-learning based optimization (TLBO) algorithm has become quite popular due to its superior performance especially when compared to other metaheuristic algorithms. In this paper, damage estimation capability of the TLBO for frame structures and a benchmark problem of cantilever beam is studied and comparisons are made with some established soft computing techniques. TLBO is observed to produce better results relative to the other artificial intelligence-based techniques used for structural health monitoring.
... A BP neural network model was established with the cable force, cable length and damage degree as input vectors and the damage index RWE as output vectors [33]. It has been shown in the literatures that the 3-layer BP neural network can arbitrarily approximate the nonlinear function. ...
The prestressed cables used for external strengthening and the suspender cables used for arch bridges may suffer damage resulting from the corrosion or fracture of steel wires. Under these scenarios, the effective areas of the cables will decrease, but the cable forces will remain almost constant, which limits the ability to detect this damage with traditional frequency domain analysis. Due to the lack of understanding of the time and frequency domain characteristics, damage indexes and damage quantification for this kind of cable, hidden cable damage may not be detected in time, which can threaten bridge safety. To solve these problems, a series of performance experiments for prestressed cables were designed. The dynamic response signals of these cables to various damage levels, cable forces and cable lengths were obtained and analysed in the time domain, frequency domain and energy domain. Depending on the test, the change rate of wavelet packet total energy (RWE) was determined to be sensitive to cable damage and was chosen as the damage index for the cables. The damage level was quantified by a neural network algorithm with RWE, and a prediction procedure for cable damage was finally established. The damage detection method for external cables proposed in this paper will aid in the damage assessment and long-term monitoring of cable-supported bridges.
... In the radial basis neural network, the Gaussian function is the activation function, which implies that the center and width of the activation function are the two parameters which heavily influence the neural network performance. The weights of the neural network are adjusted based on minimizing the mean-squared error using the gradient descent algorithm [44,45] ...
The continuous increase in energy consumption has brought worldwide attention to its significant environmental effect, which is triggered by the increase in greenhouse gas emissions, global warming, and rapid climate change. As such, more energy efficient buildings are required to minimize the energy consumption of heating and cooling. The present study introduces a set of machine learning-based models to predict the heating and cooling loads in buildings. This includes backpropagation artificial neural network, generalized regression neural network, radial basis neural network, radial kernel support vector machines and ANOVA kernel support vector machines. The comparisons were conducted as per mean absolute percentage error (MAPE), mean absolute error
(MAE), relative absolute error (RAE), root relative squared error (RRSE) and root-mean squared error (RMSE). The significances of the capacities of the machine learning models are evaluated using two tailed student’s t-tests. Eventually, a holistic evaluation of the machine learning models is conducted using average ranking algorithm. Results demonstrate that the radial basis function network outperformed the afore-mentioned machine learning models significantly
... In it, the width and center significantly affect its performance. The weights of the radial basis neural network are obtained stepping on gradient descent algorithm through minimizing the mean-squared error between the actual and simulated values [23][24]. ...
Construction is a complex process which is associated with lots of changes. This results in issuing change orders. Change orders have a significant negative impact on time, cost and quality of construction projects. Labor productivity is regarded as one of the main performance metrics to judge the efficiency of the construction process. The implications of change orders on labor productivity are difficult to evaluate. Moreover, traditional-based models often fail to deal with such kind of input-output relationships. As such, the present study introduces a set of machine learning-based models to evaluate the implication of change order on labor productivity. This includes multiple linear regression, hybrid particle swarm optimization-liner regression, back-propagation artificial neural network, Elman neural network, radial basis neural network, generalized regression neural network and Cascade forward neural network. The comparisons were conducted as per mean absolute percentage error (MAPE), mean absolute error (MAE) and root-mean squared error (RMSE). Results demonstrate that the radial basis function network outperformed the afore-mentioned machine learning models such that it achieved MAPE , MAE and RMSE of 2.447%, 0.0141 and 0.0279, respectively. Finally, the significances of the capacities of the machine learning models are evaluated using two-tailed student's t-tests.
... This perspective is additionally motivated by studies about radial basis functions applied to beam and plate deformation and stability problems [25,26]. Structural integrity concerning damage assessment of structures was described in [27] causing a better damage detection by means of RBFNN compared to classical FFNN. A clustering algorithm, which calculated the output error of an RBFNN in each cluster, was described in [28]. ...
The aim of the present study is to develop a series of artificial neural networks (ANN) and to determine, by comparison to experiments, which type of neural network is able to predict the measured structural deformations most accurately. For this approach, three different ANNs are proposed. Firstly, the classical form of an ANN in the form of a feedforward neural network (FFNN). In the second approach a new modular radial basis function neural network (RBFNN) is proposed and the third network consists of a deep convolutional neural network (DCNN). By means of comparative calculations between neural network enhanced numerical predictions and measurements, the applicability of each type of network is studied.
... In the radial basis neural network, the Gaussian function is the activation function, which implies that the center and width of the activation function are the two parameters which heavily influence the neural network performance. The weights of the neural network are adjusted based on minimizing the meansquared error using the gradient descent algorithm (Vallabhaneni & Maity, 2011;Pinar et al., 2010). ...
The continuous increase in energy consumption has brought worldwide attention to its significant environmental effect, which is triggered by the increase in greenhouse gas emissions, global warming, and rapid climate change. As such, more energy efficient buildings are required to minimize the energy consumption of heating and cooling. The present study introduces a set of machine learning-based models to predict the heating and cooling loads in buildings. This includes back-propagation artificial neural network, generalized regression neural network, radial basis neural network, radial kernel support vector machines and ANOVA kernel support vector machines. The comparisons were conducted as per mean absolute percentage error (MAPE), mean absolute error (MAE) and root-mean squared error (RMSE). Finally, the significances of the capacities of the machine learning models are evaluated using two-tailed student’s t-tests. Results demonstrate that the radial basis function network outperformed the afore-mentioned machine learning models.
... The main idea of this technique is to average the variations of mode shape curvatures at a given coordinate j with respect to the number of considered modes. The use of several modes enables the detection of damages affecting mode shapes other than that of the fundamental mode and reduces the weight of misleading informations [27]. This method is computed as follows: ...
The Structural Health Monitoring (SHM) in civil engineering faces several challenges. The main issue lies in defining a reliable and precise methodology of damage detection and localization in order to allow preventive maintenance or to enable the definition of repair actions. In this paper, a new methodology of SHM is proposed. Using Vibration-Based Damage Detection Methods (VBDDM), a damage detection and localization algorithm is elaborated and tested on a Finite Element Model (FEM) of an existing building. In a first case, the damage is introduced artificially by a local reduction of stiffness, while in the second case, the damage is calculated according to a real seismic signal from the italian L'Aquila earthquake. The advantages and disadvantages of each dynamic monitoring technique are discussed and the usefulness of the algorithm is highlighted.
... Li and Yang [27] used the back propagation neural network for damage location identification and assessing the extent of the damage of continuous beams. The Radial Basis Function (RBF) neural network is a better choice for damage detection assessment of structures and takes less training time when compared to the back propagation learning of multilayer perceptron (MLP) neural network [28,29]. Reddy and Ganguli [30] used RBF neural networks for the damage detection of helicopter rotor blades using the rotating frequencies as inputs. ...
Dynamic characteristics such as natural frequencies and mode shapes are used to identify the location of damage and the damage level in a laminated composite beam with localized matrix cracks. Such cracks can be the result of low velocity impact damage and are hard to detect visually. Natural frequencies are used in conjunction with modular radial basis function neural networks for damage detection. The mode shapes are utilized to obtain a damage indicator called curvature damage factor (CDF). A matrix crack based damage model is integrated with a beam finite element model to simulate the damaged composite beam structure. In the matrix crack model, the stiffness of the beam is degraded by a reduction in A, B and D matrices to simulate the damage and the damage level is represented by matrix crack density. It is found that the combination of modular radial basis neural networks with natural frequencies and CDF can be used as robust damage detection tools for localized matrix cracks in composite beams.
... For example, the convergence speed of FABPN and GRNN in training is higher than that of MLBPN [39,65]. • To avoid sticking at local minima instead of global minima at the error surface, which is why RBFN and GRNN were used in several studies [45,48,65,67] as these networks use localized nonlinearities for approximation. • To avoid the 'black-box' optimization procedure of hidden layer architecture. ...
... Therefore, modification of ANN parameters such as introducing novel network architecture, activation function or convergence theory in order to enable ANNs for robust extrapolation is an important area of future studies. e) In several past studies [45,48,65,67] including the experimental validation of this study, RBFN and GRNN have been found to be more accurate for approximation tasks. As described in Sec. ...
Vibration behavior of any solid structure reveals certain dynamic characteristics and property parameters of that structure. Inverse problems dealing with vibration response utilize the response signals to find out input factors and/or certain structural properties. Due to certain drawbacks of traditional solutions to inverse problems, ANNs have gained a major popularity in this field. This paper reviews some earlier researches where ANNs were applied to solve different vibration-based inverse parametric identification problems. The adoption of different ANN algorithms, input-output schemes and required signal processing were denoted in considerable detail. In addition, a number of issues have been reported, including the factors that affect ANNs’ prediction, as well as the advantage and disadvantage of ANN approaches with respect to general inverse methods Based on the critical analysis, suggestions to potential researchers have also been provided for future scopes.
... As the mathematical relationship between structural vibration response and the location and extent of damage is quite complex, nowadays computational intelligence techniques, such as genetic algorithm Tripathi 2005, Mehrjoo et al. 2013), artificial neural network (Maity and Saha 2004, Bakhary et al. 2010, Sahoo and Maity 2007, Vallabhaneni and Maity 2011 and swarm based intelligence techniques such as ant colony optimization (Majumdar et al. 2012, Yu andXu 2011) and particle swarm optimization (Nanda et al. 2012, Sayedpoor 2012 are widely employed to solve such problems. To construct suitable objective function for this purpose, most of the authors used natural frequency, mode shape or their derivatives. ...
A two-step procedure to detect and quantify damages in structures from changes in curvature mode shapes is presented here. In the first step the maximum difference in curvature mode shapes of the undamaged and damaged structure are used for visual identification of the damaged internal-substructure. In the next step, the identified substructures are searched using unified particle swarm optimization technique for exact identification of damage location and amount. Efficiency of the developed procedure is demonstrated using beam like structures. This methodology may be extended for identifying damages in general frame structures.
