The paper discusses the results of laboratory experiments in which three independent measurement techniques were compared: a digital oscilloscope, phased array acquisition system, a laser vibrometer 3D. These techniques take advantage of elastic wave signals actuated and sensed by a surface-mounted piezoelectric transducers as well as non-contact measurements. In these experiments two samples of aluminum strips were investigated while the damage was modeled by drilling a hole. The structure responses recorded were then subjected to a procedure of signal processing, and features' extraction was done by Principal Components Analysis. A pattern database defined was used to train artificial neural networks for the purpose of damage detection.


Keywords: Artificial neural networks, damage detection, structural health monitoring, elastic waves, non-destructive testing.

In this paper, a time series-based damage detection algorithm is proposed using Gaussian mixture model (GMM) and expectation maximization (EM) framework. The vibration time series from the structure are modelled as the autoregressive (AR) processes. The first AR coefficients are used as a feature vector for novelty detection. To test the efficacy of the damage detection algorithm, it has been tested on the pseudo-experimental data obtained from the FEM model of the ASCE benchmark frame structure. Results suggest that the presented approach is able to detect mainly major and moderate damage patterns.


Keywords: dynamics, inverse problems, structural monitoring, damage detection, mixture model, novelty detection.

This paper presents neural networks prediction of load capacity for eccentrically loaded reinforced concrete (RC) columns. The direct modelling of the load capacity of RC columns by means of the finite element method presents several difficulties, mainly in geometry representation and handling of several nonlinearities. Properly trained neural network can provide a useful surrogate model for such columns. The paper discusses architecture and training methods of the both multi-layer perceptron (MLP) and fuzzy weights neural networks (FWNN) for this application. It also presents the performance analysis of the networks trained on data from three independent databases available in the literature.


Keywords: concrete reinforced columns, load capacity, neural networks, fuzzification.

The present paper focuses on the identification of delamination size and location in homogeneous and composite laminates. The modal analysis methods are employed to calculate the data patterns. An aggregated approach combining Haar wavelets, support vector machines (SVMs) and artificial neural networks (ANNs) is used to solve identification problems. The usability and effectiveness of the proposed technique are tested by several numerical experiments. The advantages of the proposed method lie in the ability to make fast and accurate calculations.


Keywords: delamination identification, free vibrations, Euler-Bernoulli beam theory, Haar wavelets, machine learning methods.

In the paper a proposal of using selected swarm intelligence algorithms for solving the inverse heat conduction problem is presented. The analyzed problem consists in reconstructing temperature distribution in the given domain and the form of heat transfer coefficient appearing in the boundary condition of the third kind. The investigated approaches are based on the Artificial Bee Colony algorithm and the Ant Colony Optimization algorithm, the efficiency of which are examined and compared.


Keywords: Swarm Intelligence, ACO algorithm, ABC algorithm, Inverse Heat Conduction Problem.

Validation of an experimental approach requires that both model and data errors are proved to be within acceptable ranges. In case of destructive testing none of the classic, statistically based methods can be applied for that task due to the lack of independent data series required for building data statistics. The aim of the paper is to present a non-statistical methodology for performing such validation, developed within the framework of physically based approximation (PBA). It has been developed to validate a neutron diffraction based experimental-numerical approach applied for studying 3D rail residual stress. It is for the PBA technique's capability to provide high quality physically reasonable data fits for one data set only, treated here as higher order reference fields that made it possible to develop this methodology and perform error analysis/validation. In many ways this approach is analogical to Zienkiewicz-Zhu type of error estimators, and its performance will be demonstrated for a defective RE136 rail sample that was installed in a US DOT test track.


Keywords: physically based approximation, experimental data error estimation, validation of experimental technique, residual stress in railroad rails, neutronography.

A thin metal film subjected to a laser pulse is considered. The problem is described by the system of energy equations describing the electron gas and lattice temperatures. The thermal interactions between electrons and lattice are determined by the parameter G called the electron-phonon coupling factor. To estimate the unknown parameter G the identification problem is formulated. The additional information necessary to solve an inverse problem is the knowledge of transient measurements of the reflectivity or transmissivity variation which is proportional to the variation of the electron temperature. So, at the stage of inverse problem solution, it is possible to assume the knowledge of electrons temperature on the irradiated surface of the system (x = 0). To solve the identification problem the gradient method basing on the least squares criterion and sensitivity coefficients is used. In the final part of the paper the results of computations are shown.


Keywords: microscale heat transfer, laser heating, two-temperature model, inverse problem, finite difference method.

The paper presents a solution of an inverse problem consisting in determination of boundary conditions in the process of binary alloy solidification when temperature measurements in selected points of the cast are known. In the investigated model the distribution of temperature is described using the Stefan model with the liquidus temperature varying in dependance on concentration of the alloy component. For description of the concentration we apply the model in which the immediate equalization of chemical composition of the alloy is assumed (lever arm model). Experimental verification of the developed algorithm is also presented.


Keywords: solidification, segregation, binary alloy, genetic algorithm.