Development Of LFNMR-NIR System

Low-field nuclear magnetic resonance(LF-NMR)uses hydrogen atom as a probe to detect the energy release speed of hydrogen atom and obtain the distribution and content of hydrogen-containing substance.Near-infrared spectroscopy(NIR)is based on the characteristic absorption information of C,H,O,N groups in the sample components in the near infrared spectrum to realize the rapid detection of substances.Low field nuclear magnetic resonance and near infrared spectroscopy are widely used in food agriculture,biomedicine and other fields because of their fast,efficient,green and non-destructive detection characteristics.However,it mainly focuses on the independent detection of the two instruments at present.In order to make up for the shortcomings such as limited information detected by a single low-field NMR analyzer or a single near-infrared spectroscopy instrument,tedious and time-consuming separation sampling operation,the project combined PQ001 low-field NMR analyzer and INSION miniature near-infrared spectrometer to develop a device capable of insitu and synchronous sampling.To realize the low field NMR and near infrared analysis system with complementary information,not only improve the efficiency and accuracy of joint analysis,but also obtain more abundant physical and chemical information.The main work of this paper is as follows:1.On the basis of extensive reference and analysis of domestic and foreign literature,according to the principle,characteristics and instrument composition of LFNMR and NIR,this paper reviews the research status of the two technologies from hardware to data fusion.Based on the modern co-use instrument analysis theory,this paper proposes that the co-use instrument should have the basic functions of joint sampling,information complementary and improved detection accuracy.The overall scheme of LFNMR-NIR combined instrument is designed.2.The hardware platform of the co-use system was built to realize the simultaneous and in-situ sampling function of LFNMR and NIR.Firstly,the transmission sampling structure of NIR was determined,and then the simultaneous sampling scheme of LFNMR and NIR probe and the selection of related components were determined.The radio-frequency coil,tuned matching circuit,NIR fiber and light source of nuclear magnetic probe are designed and manufactured.NIR and LFNMR software were combined to realize the sampling control and detection visualization of the joint system.C# language was used to design the human-computer interaction interface of the near infrared upper computer,and LFNMR sampling and analysis software was embedded and combined.The software included device connection,sampling,display,result prediction and other functions.The data processing software is written based on python3.9,which covers data preprocessing,feature extraction,model regression and other methods,and provides a simple and fast means for the fusion analysis of LFNMR and NIR information obtained by the combined system.3.The performance of the LFNMR-NIR combined system was tested.Firstly,the nuclear magnetic performance was tested,and the base test was carried out for the signals in the flat section at the tail end.The results showed that the ratio of the average value of the bottom noise signal from the empty mining to the average value of the 1%porosity sample was 3.63%.In the 25 mm effective region,the relative range is 5.22%.After the inversion calculation of T2 spectrum of the sample,the proportion of small peaks is 0.1% and 0.2%.The performance test results are in line with the production standards of the enterprise and meet the application indicators of the combined system.Then,the near-infrared part was tested.Since the near-infrared spectrometer used can be calibrated for life,the absorbance repeatability test was only needed in the combined system with the nuclear magnetic resonance instrument,and the results showed that the standard deviation of the spectrum was less than 0.003.Finally,it is proved that low field nuclear magnetic RF signal and near infrared light signal do not interfere with each other when they work.4.The practical capability of 84 diesel oil samples was tested on the combined system.Spyder prepared data processing software to LFNMR and NIR single data set respectively and sulfur content,polycyclic aromatic hydrocarbons content,distillation range,density and other 6 indicators to establish the correction model,which LFNMR and recovery temperature,NIR and density correction correlation coefficient and predicted correlation coefficient are above 0.85,indicating that the correlation is good;In order to further optimize the model,SPA and CARS algorithm were used to screen LFNMR and NIR variables to effectively find out the optimal variable combination,and then four regressors were used for training.The results showed that,compared with the single dataset model of LFNMR and NIR,the prediction accuracy of SPA feature extraction method was improved by 10.8% and 8.0%,respectively.Compared with the single dataset model,the prediction accuracy of CARS feature extraction method was improved by 8.9% and 2.0%,respectively,indicating that higher prediction results were obtained after data fusion.The results show that the LFNMR-NIR combined system has a trend of information complementation in diesel analysis,and has practical application significance and promotion value in the detection fields of petroleum products,food agriculture,biomedicine and so on.