Research Of Several Key Technologies For Automatic Experiment And Data Processing Of High-Resolution NMR

Nuclear Magnetic Resonance technology has been developing rapidly since1945and has become one of the most successful analytical techniques. High resolution is the basis of the fact that the NMR spectra can provide the information of chemical structure and dynamics. In recent years, the rapid development of NMR in the field of metabolomics has raised a question of how to get the spectra automatically, which is an important challenge and opportunity for NMR spectrometer manufacturers.In this thesis, the question of how to get the high resolution spectra automatically is discussed through both experiment and data processing. A variety of methods of automatic shimming and automatic data processing methods are implemented in the spectrometer with independent intellectual property rights.The research of automatic shimming includes shimming based on search algorithm, gradient shimming and gradient shimming with lineshape optimization. In the gradient shimming, an improved pulse sequence is utilized to make three dimensional gradients shimming available in the probe without XY gradient coils. Moreover, some improvements are also made to address two issues of gradient shimming:One is that the shimming current will be out of range, and the other one is that a long time is needed in three dimensional automated shimming.In the research of automatic data processing, a robust automatic phase correction method and a new automatic baseline correction method are proposed. In the automatic phase correction method, a new strategy combining’coarse tuning’ with ’fine tuning’ is introduced to correct various spectra accurately. In the’coarse tuning’ procedure, the preliminary phased spectra are obtained by minimizing the objective function based on the height difference of’critical signal points’in peaks’tails. After that, the peaks in the preliminary corrected spectra can be categorized into three classes:positive, negative, and distorted. Based on the classification result, a new custom negative penalty function used in the step of ’fine tuning’ is constructed to avoid the negative points in the spectra excluded in the negative peaks and distorted peaks. Finally, the fine phased spectra can be obtained by minimizing the custom negative penalty function. This automatic method is proven to be very robust for it is tolerant to low signal-to-noise ratio, large baseline distortion and independent of the starting search points of phasing parameters. The new baseline correction method is based on an improved baseline recognition method and a new iterative baseline modeling method. The presented baseline recognition method takes advantages of three baseline recognition algorithms in order to recognize all signals in spectra. While in the iterative baseline modeling method, except for the well-recognized baseline points in signal-free regions, the ’quasi-baseline points’ in the signal-crowded regions are also identified and then utilized to improve robustness by preventing the negative regions. The experimental results on both simulated data and real metabolomics spectra with over-crowded peaks show the efficiency of this automatic method.