Abstract:
The paper analyses the relationship between model errors, data errors and illposedness of solutions for inverse problems of operator identification,establishes associated inversion techniques based on model optimization and data optimization,proposes monotonic denoise method of quantization data, which can deal with both digital data and nondigital data and establishes reliability concepts of numerical inversion, including reliability of direct operator, reliability of direct calculation, reliability of measurement design, reliability of inverse algorithm, reliability of inverse calculation, reliability of measuring data, and proposes their corresponding quantitative evaluation methods for numerical inversion reliability. By a numerical experiment of operator identification problem in geotechnical engineering, two conclusions are obtained: first, the associated inversion technique is a systematic optimization technique, and can increase reliability and degree of accuracy of numerical inversion obviously; second, through applying reliability evaluating methods of numerical inversion, the quality evaluation of solution estimation of inverse problem can be achieved objectively and quantitatively.