Neural Networks for the Simulation and Identification Analysis of Buildings Subjected to Paraseismic Excitations

The chapter deals with an application of neural networks to the analysis of vibrations of medium-height prefabricated buildings with load-bearing walls subjected to paraseismic excitations. Neural network technique was used for identification of dynamic properties of actual buildings, simulation of building responses to paraseismic excitations as well as for the analysis of response spectra. Mining tremors in strip mines and in the most seismically active mining regions in Poland with underground exploitation were the sources of these vibrations. On the basis of the experimental data obtained from the measurements of kinematic excitations and dynamic building responses of actual structures the training and testing patterns were formulated. It was stated that the application of neural networks enables us to predict the results with accuracy quite satisfactory for engineering practice. The results presented in this chapter lead to a conclusion that the neural technique gives new prospects of efficient analysis of structural dynamics problems related to paraseismic excitations.

Developments in Structural Optimization and Applications to Intelligent Structural Vibration Control

The chapter introduces developments in intelligent optimal control systems and their applications in structural engineering. It provides a good background on the subject starting with the shortcomings of conventional vibration control techniques and the need for intelligent control systems. Description of a few basic tools required for intelligent control such as evolutionary algorithms, fuzzy rule base, and so forth, is outlined. Examples on vibration control of benchmark building and bridge under seismic excitation are presented to provide better insight on the subject. The chapter provides necessary background for a reader to work in intelligent structural control systems with real-life examples. Current trends in the research area are given and challenges put forward for further research.

Evolutionary Seismic Design for Optimal Performance

The chapter provides an overview of optimal structural design procedures for seismic performance. Structural analysis and design for earthquake effects is an evolving area of science; many design philosophies and concepts have been proposed, investigated, and practiced in the past three decades. The chapter briefly introduces some of these advancements first, as their understanding is essential in a successful application of optimal seismic design for performance. An emerging trend in seismic design for optimal performance is speculated next. Finally, a state-of-the-art application of evolutionary algorithms in probabilistic performance-based seismic design of steel moment frame buildings is described through an example. In order to follow the concepts of this chapter, the reader is assumed equipped with a basic knowledge of structural mechanics, dynamics of structures, and design optimizations.

Neuro-Fuzzy Assessment of Building Damage and Safety After an Earthquake

This chapter describes the algorithmic basis of a computational intelligence technique, based on a neuro-fuzzy system, developed with the objective of assisting nonexpert professionals of building construction to evaluate the damage and safety of buildings after strong earthquakes, facilitating decision-making during the emergency response phase on their habitability and reparability. A hybrid neuro-fuzzy system is proposed, based on a special three-layer feedforward artificial neural network and fuzzy rule bases. The inputs to the system are fuzzy sets, taking into account that the damage levels of the structural components are linguistic variables, defined by means of qualifications such as slight, moderate or severe, which are very appropriate to handle subjective and incomplete information. The chapter is a contribution to the understanding of how soft computing applications, such as artificial neural networks and fuzzy sets, can be used to complex and urgent processes of engineering decision-making, like the building occupancy after a seismic disaster.

Neural Network-Based Identification of Structural Parameters in Multistory Buildings

A detailed study of the capabilities and powerfulness of soft computing techniques such as artificial neural network with respect to the identification of structural parameters and structural responses are presented. This chapter includes the definition of neural architectures and system identification of multistory structure. An efficient identification algorithm for the multistory structure subject to initial condition and ground displacement is presented. Response identification subject to real earthquake data has also been discussed. Several example problems are incorporated to show the efficiency and reliability of the proposed algorithm.

Genetic Algorithms in Structural Identification and Damage Detection

Genetic algorithms (GA) have proved to be a robust, efficient search technique for many problems. In this chapter, the latest developments by the authors in the area of structural identification and structural damage detection using genetic algorithms are presented. A GA strategy involving a search space reduction method (SSRM) using a modified genetic algorithm based on migration and artificial selection (MGAMAS) is first used to identify structural properties in multiple degree-of-freedom systems. The SSRM is then incorporated in a structural damage detection strategy using response measurements both before and after damage has taken place. Numerical studies on 10 and 20 degree-of-freedom systems show that a small damage of only 2.5% can be accurately and consistently identified from incomplete acceleration measurements in the presence of 5% input and output noise.

Soft Computing Techniques in Probabilistic Seismic Analysis of Structures

Earthquake-resistant design of structures using probabilistic analysis is an emerging field in structural engineering. The objective of this chapter is to investigate the efficiency of soft computing methods when incorporated into the solution of computationally intensive earthquake engineering problems. Two methodologies are proposed in this work where limit-state probabilities of exceedance for real world structures are determined. Neural networks based metamodels are used in order to replace a large number of time-consuming structural analyses required for the calculation of a limit-state probability. The Rprop algorithm is employed for the training of the neural networks; using data obtained from appropriately selected structural analyses.

Optimal Reliability-Based Design Using Support Vector Machines and Artificial Life Algorithms

Reliability-based optimization is considered by many authors as the most rigorous approach to structural design, because the search for the optimal solution is performed with consideration of the uncertainties present in the structural and load variables. The practical application of this idea, however, is hindered by the computational difficulties associated to the minimisation of cost functions with probabilistic constraints involving the computation of very small probabilities computed over implicit threshold functions, that is, those given by numerical models such as finite elements. In this chapter, a procedure intended to perform this task with a minimal amount of calls of the finite element code is proposed. It is based on the combination of a computational learning method (the support vector machines) and an artificial life technique (particle swarm optimisation). The former is selected because of its information encoding properties as well as for its elitist procedures that complement hose of the a-life optimisation method. The later has been chosen du to its advantages over classical genetic algorithms. The practical application of the procedure is demonstrated with earthquake engineering examples.

Optimum Design of Structures for Earthquake Induced Loading by Wavelet Neural Network

Optimum design of structures for earthquake induced loading is achieved by a modified genetic algorithm (MGA). Some features of the simulated annealing (SA) are used to control various parameters of the genetic algorithm (GA). To reduce the computational work, a fast wavelet transform is used. The record is decomposed into two parts. One part contains the low frequency of the record, and the other contains the high frequency of the record. The low-frequency content is used for dynamic analysis. Then using a wavelet neural network, the dynamic responses of the structures are approximated. By such approximation, the dynamic analysis of the structure becomes unnecessary in the process of optimisation. The wavelet neural networks have been employed as a general approximation tool for the time history dynamic analysis. A number of structures are designed for optimal weight and the results are compared to those corresponding to the exact dynamic analysis.

Structural Assessment of RC Constructions and Fuzzy Expert Systems

The chapter deals with the structural assessment of existing constructions, with a particular attention to seismic risk mitigation. Two aspects are involved: the appraisal of the actual conditions of the structure (material deterioration, preexisting damages) and the evaluation of the structural “vulnerability,” that is, the propensity to suffer damage because of the intrinsic geometric and structural arrangement, boundary conditions, specific structural details. Attention is first focused on the investigation protocol, which is organized through a multilevel, hierarchical scheme: the procedure includes visual inspections, surveys, experimental testing on site and in laboratory, and gradually proceeds into the details of the problem, progressively refining and verifying hypotheses and preliminary judgments. In a second part, the definition of effective tools for uncertainty management and decision making is performed, by presenting a genetic-fuzzy expert system which handles the procedure of the assessment properly accounting for uncertainty and errors, and is able to tune the parameters involved on the basis of experts’ knowledge, “training” the system. Finally, a case study is presented, applying the whole assessment procedure and the fuzzy genetic algorithm.