Research On The Fast Multivariate Optimization Method For Separation Of Complex Samples

Accompanying the more and more applications of high performance liquid chromatography in the analysis of complex samples such as environmental samples, biological samples and medical samples, optimization of the separation conditions becomes more and more important. The development of optimization method undergoes a process from "Black box" method, empirical method to theoretical method, hi chapter one, the applicable range, advantages and disadvantages of different optimization method was described. Also, the development of optimization method (optimization strategy, optimization criterion) in recent years was emphasized.In HPLC, the most commonly used separation mode is reversed-phase mode, which can be used to solve most of the separation problems. Theoretical basis for the optimization of separation conditions hi RP-HPLC is the retention equation, which describes the retention behavior of solutes. However, it is difficult to obtain these factors theoretically, so it is the key point of optimization to quickly obtain the above factors by experiment.In chapter two, a new method for the precise prediction of retention tune of solutes under linear gradient elution by simultaneously calibrating the effect of delay time of instrument and the distribution of mobile phase in the column on the migration of solutes was first proposed and validated by the prediction of retention time of 15 kinds of derivatised amino acids and eight benzene. On this basis, a method for the quick calculation of the retention equation of solutes through two to three linear gradient elution was established. Combining this method and "moving overlapping separation range mapping" method, a strategy for the fast multi-step gradient optimization of complex samples under binary mobile phase was first proposed and validated by its application to the separation of 36 components hi Chinese medicine chuanxiong rhizome extract.In chapter three, a systemic strategy for the quick optimization of ternary gradient elution conditions (FMGOS) was proposed. By discussing and revising the expression of a value under multivariate mobile phase, a relatively simple retention equation under ternary mobile phase was obtained. Suitable gradient elution conditions were searched according to the hierarchical chromatography response function. A method to obtain the retention equation using least experiments and calibrated by gradient experiment results, which can further improve the precision of predicting. Based on this strategy, a software package which can be used to the fast optimization ofmultivariate gradient elution conditions (stepwise, linear and mixed gradient) for the separation of complex samples was developed.In chapter four, separation conditions of different types of complex samples were optimized based on the proposed fast multivariate gradient optimization strategy and corresponding software package. According to the relationship between analysis purpose of complex samples and the optimization target, the concrete separation mode and optimization criterion for separating complex samples was comprehensively evaluated. An idea that the separation mode and optimization target should match with the analysis purpose was proposed, which provides an overall frame for the method development of complex samples. According to the analysis method and properties of samples, different optimization method was applied to the different kinds of samples (known complex samples, target components in complex samples, total complex samples).In chapter five, artificial neural network (ANN) was applied to the simultaneous optimization of the concentration of organic modifier and the nonlinear chromatographic parameter-pH value of mobile phase. First, limited experiments were used to simulate the relationship between the retention value of solutes and pH values as well as composition of mobile phase by combining experimental design and ANN; then the optimal separation conditions was searched according to a response function. This method was validated by it...