Algorithm For Accurate Inversion Of Low-field Nuclear Magnetic Resonance Signals And The Development Of Its Software Platform

Low-field Nuclear Magenetic Resonance(LF-NMR)detection technology is widely used in the fields of medicine,biology,chemistry and petroleum due to its advantages of non-destructiveness,accuracy and stability.As an important research branch of low-field nuclear magnetic resonance detection technology,the inversion algorithm analyzes and detects the amplitude signal distribution of each relaxing component in the sample to quantitatively learn the physical or chemical properties of the sample.At present,the existing inversion methods can be mainly divided into two kinds : iterative algorithms and regularization algorithms.Among them,iterative algorithms have defects in determining the number of iterations,while regularization algorithms encounter difficulties in determining smoothing coefficients.Besides these methods still have much room for improvement in the accuracy of multi-relaxation component inversion,so it is of great research significance to focus research on an accurate inversion algorithm of nuclear magnetic resonance signals.In this paper,based on the basic principles of low-field nuclear magnetic resonance,a mathematical model of low-field nuclear magnetic resonance signal inversion is established.By comparing the filter factors of the singular value decomposition method and the regularization method in the one-dimensional inversion algorithm,a joint Onedimensional iterative inversion algorithm for filtering factors of two algorithms.Then,for the NMR signal preprocessing process,a novel mapping function compression algorithm and signal noise evaluation method are proposed-through the mapping function and hash technology to achieve the compression of inversion data,and by fitting the noise histogram Calculate the standard deviation to evaluate the noise level.Then,the curvature method is used to automatically search for the best smoothing coefficient in the L curve,which improves the limitation of manually setting the smoothing coefficient.A new two-dimensional inversion objective function is proposed,and the BFGS algorithm is used to solve the inversion problem.The accuracy of the improved inversion algorithm is proved by numerical simulation and measured data experiments.Finally,a software platform for LF-NMR signal inversion was built,and the application research of inversion was carried out based on the software platform.Based on all the research content,the research of this paper has achieved the following innovative research results:1.By analyzing the filtering factors of the singular value decomposition method and the regularized inversion algorithm,a one-dimensional inversion algorithm combining these two filtering factors is proposed.This method corrects the eigenvalue matrix after singular value decomposition.When the eigenvalue is large,the filter factor of the singular value decomposition method is selected,and when the eigenvalue is small,the regularized filter factor is selected.The accuracy of the inversion is ensured during sex.Finally,the inversion simulation data and the measured data are used,and compared with the existing methods,it is verified that the improved inversion algorithm is the most accurate in the identification of samples containing multiple relaxation components.2.Based on the mathematical model of two-dimensional inversion,a new twodimensional inversion method is proposed.The two-dimensional inversion method mainly converts the two-dimensional inversion into a one-dimensional inversion problem through tensor product,and then uses the one-dimensional inversion method to solve.However,this method can directly transform the inversion problem without the tensor product,and transform the original non-negative minimum inversion problem into an unconstrained maximization problem.Finally,the new inversion method proposed in this paper is used to invert simulated data and measured data,and compare with the existing inversion effect to verify the accuracy of the method.3.Aiming at the problem of low-field nuclear magnetic resonance inversion,a low-field nuclear magnetic resonance inversion software platform was built,and inversion application research was carried out based on the software platform.The software platform integrates the current mainstream inversion methods,including onedimensional heterogeneous decomposition method,regularization method and joint inversion algorithm,two-dimensional BRD,LSQR and BFGS algorithms,etc.,so that researchers can choose the most accurate for different application scenarios Inversion algorithm.The software platform uses a graphical interactive way to configure the inversion parameters by default,which is convenient for the operator to use.In addition,the software platform was used to carry out LF-NMR inversion application research,which verified the accuracy of the software platform inversion,and also promoted the application and development of LF-NMR inversion.