International Journal for Numerical Methods in Biomedical Engineering · 2008 · 10 citations · 24 references
EngineeringMechanical EngineeringComputational Heat TransferHeat TransferThermal AnalysisDigital Signal ProcessingFrequency-domain AnalysisHeat Transfer ModelingInverse Heat TransferSpectral AnalysisTransient Heat TransferFrequency Spectrum AnalysisInverse Conduction ProblemInverse Heat ConductionInverse Heat Conduction ProblemsPower Spectral DensitySingular Value Decomposition
Abstract Sequential methods that use future times in order to solve an inverse heat conduction problem can be re‐formulated as a sequential digital filter algorithm. The aim of this paper is to explore the filter effect of the function specification method and the truncated singular value decomposition method, in a sequential form. For this purpose, the power spectral densities of the corresponding filter coefficients are evaluated and it is shown that they act as band‐pass filter. Three numerical tests with different frequency spectra are considered in order to solve a linear one‐dimensional transient inverse problem. Furthermore, three models of crescent complexity are used in each method. It is shown that both methods provide similar results and in most cases, the optimum estimations require the same number of future times. As it is expected, similar estimations correspond to a similar band‐pass filter. The effect of the number of future times, the influence of the frequency spectrum of the input and the complexity of the model used can be clearly interpreted in the spectral space. The most complex models act as a highly selective band‐pass filter, but they require greater number of future times. Copyright © 2008 John Wiley & Sons, Ltd.
24