Key Engineering Materials Vols. 293-294

Abstract: Two alternative software tools for damage identification are presented. The first tool, developed on the basis of the Virtual Distortion Method (VDM), takes advantage of an analytical formulation of the damage identification problem. Consequently, gradient-based optimization method is applied to solve the resulting dynamic inverse problem in time domain. Finite element model of the structure is necessary for the VDM approach. The second tool utilizes the Case-Based Reasoning (CBR) for damage identification. This method consists in i) extracting principal features of the response signal by wavelet transform, ii) creating a base of representative damage cases, iii) organizing and training the base by neural networks, and finally iv) retrieving and adapting a new case (possible damage) by similarity criteria. Basic description of both approaches is given. A comparison of numerical effectiveness, in terms of accuracy and computational time, is provided for a simple beam structure. Advantages and weaknesses of each approach are highlighted.

Abstract: In mechanical structures, the impact force is related to the structural damage. To identify the location where impact force occurs, the triangle method has long been used. This method requires three acceleration signals or strain signals to be measured on the mechanical structure. Time delay among these signals is useful information to estimate the location of the impact force. It is very difficult to estimate time delay by using the raw data of three signals because the propagation wave of the structure is a dispersive wave. Therefore, three signals should be analyzed in the time and frequency domain in order to estimate the time delay at each frequency. For the time-frequency analysis of highly non-stationary signals like impulse signals, time-frequency methods or time scale methods have been used. These methods use the first or second order statistical characteristics of the signal. This paper outlines the higher order Wigner method to obtain time and frequency information of a signal. Since it uses the high order statistics of signals, this method is useful for identifying the impact signal embedded in the background. It has a better time-frequency resolution for a non-linear signal than other time-frequency and time scale methods. This method can be applied to estimate the location of an impact force, which becomes a cause of damage of mechanical plants. Finally, in order to prove this method, experimental work was conducted on an aluminum plate in the laboratory.

Abstract: This paper presents experimental results of source identification for a non-minimum phase system. Generally, a causal linear system may be described by matrix form. The inverse problem is considered as a matrix inversion. Direct inverse method cannot be applied for a non-minimum phase system, because the system has ill-conditioning. Therefore, in this study the SVD inverse technique is introduced to execute an effective inversion. In a non-minimum phase system, its system matrix may be singular or near-singular and has very small singular values. These very small singular values have information about a phase of the system and ill-conditioning. Using this property we could solve the ill-conditioned problem of the system and then verify it for the practical system (cantilever beam). The experimental results show that the SVD inverse technique works well for a non-minimum phase system. This inverse technique can be applied to the estimation of the magnitude of impact force, which becomes often a cause of damage to a mechanical system.

Abstract: Noise is the biggest obstacle that makes the incipient fault prognosis results uncorrected. According to the theories of correlation analysis and threshold de-noising by wavelets, wavelet transform domain filter (WTDF) is constructed. WTDF is an iterative process. By selecting the process parameters adaptively, WTDF can de-noise signal efficiently. More important, the faint component in the signal will become stronger. WTDF method is used to analyze the signals collected from a bearing that has incipient unbalance and misalignment faults. Results show that WTDF is effective for bearing incipient fault prognosis.

Abstract: This paper considers the performance of radial basis function neural networks for the purpose of data classification. The methods are illustrated using a simple two class problem. Two techniques for reducing the rate of misclassifications, via the introduction of an “unable to classify” label, are presented. The first of these considers the imposition of a threshold value on the classifier outputs whilst the second considers the replacement of the crisp network weights with interval ranges. Two network training techniques are investigated and it is found that, although thresholding and uncertain weights give similar results, the level of variability of network performance is dependent upon the training approach

Abstract: Ritz vectors are parameters, which describe object dynamics in similar way as modal vectors. Their advantage with respect to modal vectors is a fact, that less number of them is required to estimate object response. It is also proofed, that Ritz vectors are more sensitive for damage detection then modal vectors. The problem which arises, when one wants to use Ritz vectors for damage assessment is that there are no many methods to extract Ritz vectors from operational data. In addition, the methods require knowledge of analytical mass matrix or state a space model matrix, which makes them rather complicated. In the paper two Ritz vectors extraction methods are taken into account. First bases on flexibility matrix and second is related with state space model. Both of them are evaluated with respect to analytical procedure of Ritz vectors calculation. For the vectors evaluation, a simulation data from theoretical model were used. In next step Ritz vectors were extracted from data measured on a real structure. On the structure damage was simulated. The Ritz vectors extracted with use of both tested methods were used to detect damage.

Abstract: The application of a multilayer perceptron (MLP) neural network to a damage location problem on a GNAT aircraft wing is considered. The problems associated with effective network training and evaluation are discussed, focussing on ensuring good generalisation performance of the network to the classification of new data. Both conventional Maximum Likelihood and Bayesian Evidence based training techniques are considered and a simple thresholding technique is presented to aid in the rejection of poorly regularised network structures. Examples are presented for an artificial simple 2 class problem (drawn from a Gaussian distribution) and a real 9 class problem on the GNAT aircraft wing.

Abstract: A new methodology for load identification is proposed. The global dynamic structural response is modeled using only pre-computed, time dependent, dynamic influence matrix, describing structural response to locally generated unit impulses. Then, the impact load identification procedure is based on distance minimization between the modeled and measured local dynamic responses in sensor locations. The theoretical background as well as numerical examples is presented.

Abstract: Bispectrum is a powerful tool for non-Gaussian signal processing and nonlinearity detection. However, it is difficult to use in practical applications due to that it is a 2-dimensional function. Bispectral slices are widely used reduction methods, and they can only represent a small part of the whole bispectral information. Integrated bispectrum contains more signal features than that of the bispectral slices, whereas the integration will lose the focus of some signal features. To overcome these problems, a new approach called maximal bispectrum is proposed to extract signal features. Maximal bispectrum is obtained by selecting the maximal values of every row of the magnitude bispectrum in the whole bispectral plane and it is a 1-dimensional function. Feature extraction based on maximal bispectrum is investigated and the maximal bispectrum is used to extract features of gear fault. Experimental results indicate that the maximal bispectrum is effective for diagnosing gear crack fault.

Abstract: Data mining is used not only for database analyses, but also for machine learning. The data mining technique described in this paper was used for steam turbine fault diagnostics based on continuous data measurements. The classification rules are based on standardized vibration frequency data for steam turbines and field experts’ analyses of turbine vibration problems. The expert knowledge enables the steam turbine fault diagnosis system to be more powerful and accurate. The system can identify twenty types of standard steam turbine faults. The system was developed using 2000 simulated data sets. The data mining methods were then used to identify 20 explicit rules for the turbine faults. The method was also used with actual power plant data to successfully diagnose real faults. The results indicate that data mining can be effectively applied to diagnosis of rotating machinery by giving useful rules to interpret the data.