Sudhirkumar V. Barai

Multilayer perceptron in damage detection of bridge structures

Abstract Recent developments in artificial neural networks (ANN) have opened up new possibilities in the domain of structural engineering. For inverse problems like structural identification of large civil engineering structures such as bridges and buildings where the in situ measured data are expected to be imprecise and often incomplete, the ANN holds greater promise. The detection of structural damage and identification of damaged element in a large complex structure is a challenging task indeed. This paper presents an application of multilayer perceptron in the damage detection of steel bridge structures. The issues relating to the design of network and learning paradigm are addressed and network architectures have been developed with reference to trussed bridge structures. The training patterns are generated for multiple damaged zones in a structure and performance of the networks with one and two hidden layers are examined. It has been observed that the performance of the network with two hidden layers was better than that of a single-layer architecture in general. The engineering importance of the whole exercise is demonstrated from the fact that measured input at only a few locations in the structure is needed in the identification process using the ANN.

Evaluating machine learning models for engineering problems

The use of machine learning (ML), and in particular, artificial neural networks (ANN), in engineering applications has increased dramatically over the last years. However, by and large, the development of such applications or their report lack proper evaluation. Deficient evaluation practice was observed in the general neural networks community and again in engineering applications through a survey we conducted of articles published in AI in Engineering and elsewhere. This status hinders understanding and prevents progress. This article goal is to remedy this situation. First, several evaluation methods are discussed with their relative qualities. Second, these qualities are illustrated by using the methods to evaluate ANN performance in two engineering problems. Third, a systematic evaluation procedure for ML is discussed. This procedure will lead to better evaluation of studies, and consequently to improved research and practice in the area of ML in engineering applications.

Studies on recycled aggregates-based concrete

Reduced extraction of raw materials, reduced transportation cost, improved profits, reduced environmental impact and fast-depleting reserves of conventional natural aggregates has necessitated the use of recycling, in order to be able to conserve conventional natural aggregate. In this study various physical and mechanical properties of recycled concrete aggregates were examined. Recycled concrete aggregates are different from natural aggregates and concrete made from them has specific properties. The percentages of recycled concrete aggregates were varied and it was observed that properties such as compressive strength showed a decrease of up to 10% as the percentage of recycled concrete aggregates increased. Water absorption of recycled aggregates was found to be greater than natural aggregates, and this needs to be compensated during mix design.

Concrete fracture models and applications

Foreword.- Preface.- List of Symbols and Abbreviations.- Introduction to Fracture Mechanics of Concrete.- Fracture Mechanics of Concrete - State-of-the-Art Review.- Fracture Behavior of Concrete using Cohesive Crack and Size Effect Models.- Crack Propagation Study using Double-K and Double-G Fracture Parameters.- Fracture Properties of Concrete based on the KRcurve Associated with Cohesive Stress Distribution.- Comparison of Fracture Parameters of Concrete using Nonlinear Fracture Models.- Appendix. Subject index.

Time-delay neural networks in damage detection of railway bridges

The recent developments in multilayer perceptron using the backpropagation algorithm, has opened up new possibilities in structural identification. Limitation of traditional neural networks (TNN) in dealing with patterns that may vary in time domain has given birth to time-delay neural networks (TDNN). In the present paper the TNN and the TDNN have been implemented in detecting the damage in bridge structure using vibration signature analysis. A comparative study has been carried out for the various cases of complete as well as incomplete measurement data. It has been observed that TDNNs have performed better than Tows in this application.

A methodology for building neural networks models from empirical engineering data

Abstract Neural networks (NN) are general tools for modeling functional relationships in engineering. They are used to model the behavior of products and the properties of processes. Nevertheless, their use is often ad hoc. This paper provides a sound basis for using NN as tools for modeling functional relationships implicit in empirical engineering data. First, a clear definition of a modeling task is given, followed by reviewing the theoretical modeling capabilities of NN and NN model estimation. Subsequently, a procedure for using NN in engineering practice is described and illustrated with an example of modeling marine propeller behavior. Particular attention is devoted to better estimation of model quality, insight into the influence of measurement errors on model quality, and the use of advanced methods such as stacked generalization and ensemble modeling to further improve model quality. Using a new method of ensemble of SG(k-NN), one could improve the quality of models even if they are close to being optimal.

Size-effect Prediction from the Double-K Fracture Model for Notched Concrete Beam

The size-effect predictions from double-K fracture criterion, characterized by two parameters: the initiation toughness and the unstable toughness, are compared with the fictitious crack model or cohesive crack model for practical (laboratory) size range of three-point bend test notched specimens. Both the fracture models, although, adopt different crack propagation criteria, they yield indistinguishable crack initiation and unstable fracture loads for usual laboratory size specimens. Notable difference in the predicted crack initiation and unstable fracture loads are observed for asymptotic large size specimens and these loads are more conservative than those obtained using the fictitious crack model by ∼20 and 22%, respectively.

Neural networks modeling of shear strength of SFRC corbels without stirrups

Based on developed semi-empirical characteristic equations an artificial neural network (ANN) model is presented to measure the ultimate shear strength of steel fibrous reinforced concrete (SFRC) corbels without shear reinforcement and tested under vertical loading. Backpropagation networks with Lavenberg-Marquardt algorithm is chosen for the proposed network, which is implemented using the programming package MATLAB. The model gives satisfactory predictions of the ultimate shear strength when compared with available test results and some existing models. Using the proposed networks results, a parametric study is also carried out to determine the influence of each parameter affecting the failure shear strength of SFRC corbels with wide range of variables. This shows the versatility of ANNs in constructing relationship among multiple variables of complex physical relationship.

Neural Network Models for Air Quality Prediction: A Comparative Study

The present paper aims to find neural network based air quality predictors, which can work with limited number of data sets and are robust enough to handle data with noise and errors. A number of available variations of neural network models such as Recurrent Network Model (RNM), Change Point detection Model with RNM (CPDM), Sequential Network Construction Model (SNCM), and Self Organizing Feature Maps (SOFM) are implemented for predicting air quality. Developed models are applied to simulate and forecast based on the long-term (annual) and short-term (daily) data. The models, in general, could predict air quality patterns with modest accuracy. However, SOFM model performed extremely well in comparison to other models for predicting long-term (annual) data as well as short-term (daily) data.

Performance of the generalized delta rule in structural damage detection

Abstract The paper examines the suitability of the generalized data rule in training artificial neural networks (ANN) for damage identification in structures. Several multilayer perceptron architectures are investigated for a typical bridge truss structure with simulated damage states generated randomly. The training samples have been generated in terms of measurable structural parameters (displacements and strains) at suitable selected locations in the structure. Issues related to the performance of the network with reference to hidden layers and hidden neurons are examined. Some heuristics are proposed for the design of neural networks for damage identification in structures. These are further supported by an investigation conducted on five other bridge truss configurations.