Studies On Nondestructive Quantitative Pharmaceutical Analysis Using ANN On SW-NIR Spectroscopy Technique

With extensive and in-depth application of computer science , applied mathematics and statistical methodology in analytical chemistry,near-infrared (NIR) spectroscopy and chemometrics have been spread rapidly and became an independent analytical technique.NIR region involves the electromagnetic spectrum from 780nm to 2500nm. Normally,NIR region is divided into two regions,spectrum from 780nm to 1100nm is called Short Wavelength Near-infrared (SW-NIR) and spectrum from 1100nm to 2500nm is called Long Wavelength Near-infrared (LW-NIR). Which NIR obtains are the spectra of multiple and combined frequency vibrations of O-H,C-H and N-H,and most compounds have absorption in this region,their spectra are correlated to molecular inner structure,functional group and molecular state,therefore the quantitative and qualitative information may be obtained from NIR spectroscopy. Compared with the traditional analytical methods,NIR spectroscopy technology has a lot of advantages and is a simple rapid method. The main advantage of NIR is that the samples are not restricted in appearance,which can be determined no matter what they are gas,liquid and solid. We may determine the samples using diffuse reflectance spectroscopy directly because of its strong penetrable ability. The solid samples may be any shapes such as fruits,corns and solid tablets. However,NIR spectroscopy has some disadvantages resulting in weak,partly overlapped,non-specific bands that require the use of a multivariate calibration technique in order to correlate the spectral information they contain to the concentration of the target analyte. Hence,it needs to be accurately analyzed combined with method of chemometrics.NIR analytical method,a fast and efficient measurement technique which is suitable for online process analysis,has been widely used in the fields of agriculture,food,petrochemical,textile and pharmaceutical industries,as it is featured in simple and rapid operation,not destroying samples,and no chemical pollution.Pharmaceutical analysis is an important component of analytical chemistry and pharmaceutical science as well. With the development of pharmaceutics,new requests were raised in pharmaceutical analysis. In recent years,due to rapid scientific development and mutual infiltration of various subjects,pharmaceutical analysis has developed into an equipment-based modern pharmaceutical analysis. The urgent task the pharmaceutical analyser facing is no longer just static conventional medication test,but to go into the process of medication production,and to achieve real-time,in situ,and online quality control to medication raw materials,intermediates,semi-finished and finished products. The main task of modern pharmaceutical analysis is to ensure medication is safe and effective by controlling the quality of raw materials and preparation based on accurate and rapid analytical methods during the development,production,distribution,and use of medication.The normally used pharmaceutical quantitative analyses are Gas Chromatography (GC),High Performance Liquid Chromatography (HPLC),infrared (IR) and ultraviolet (UV) spectrophotometry. However,those methods both need to dissolve samples and separate active components to measure their proportions,which are time-consuming,laborious,destroying medications and even causing a certain amount of chemical pollution. In line with the tenet of green chemistry,a nonlinear correction model of NIR spectrum was established by using Artificial Neural Network (ANN). In the case of larger amount of samples,ANN models can get much smaller calibration and prediction errors.In this dissertation,a non-destructive and rapid quantitative analysis of active components of three medications is performed by using of ANN combined with SW-NIR spectroscopy. The result shows that the best mathematical model established by using ANN and SW-NIR spectroscopy performed well and can monitor pharmaceutical quality easily and rapidly.With this proposed model , both single-component medications and multi-component medications can be measured and analyzed simultaneously. This research focused on processing the original data obtained from experiments,in order to extract more valuable chemical information after choosing the most appropriate analytical and measurement method and the most optimized information models.The main content of this dissertation is as follows:1. A method for non-destructive quantitative analysis of the active components of three medications was carried out successfully by using of SW-NIR spectroscopy and Back Propagation ANN (BP-ANN) algorithm. The optimal ANN models that suitable for the three medications were established with good results.2. The application of four spectral preprocessing methods (the first derivative,the second derivative,standard normal variate (SNV),and multiplicative scatter correction (MSC)) brought positive effects on the removal of noise of SW-NIR spectra,the filteration of data,the improvement of operation efficiency,the optimization of spectral information and the removal of side effects of other factors,which set up solid foundation for the further establishment of calibration model and the accurate prediction of unknown samples.3. This article went deeply into the establishment of the optimal ANN models of corresponding active components in metronidazole , ciprofloxacin and acetylspiramycin. In addition,the partial least squares (PLS) model of ciprofloxacin samples was also designed in order to compare with the optimal ANN model,and the results indicated that ANN model had more superiority in dealing with complex non-linear systems.Through the analysis and research of three powder medications,calibration models of the above three medications were established respectively by using of ANN method,and very satisfactory results were obtained. The research of medication analysis model has achieved the purpose of systematism and scientificity,via improving the method and correcting model parameters in the experiment. Experimental results show that the concentrations of corresponding active components in the medications could be quantitatively predicted with high accuracy by using of SW-NIR spectroscopy combined with ANN or PLS method,however, ANN is more powerful to process complex non-linear information than PLS.Therefore,ANN as a chemometrics method possesses the directive significance in the scientific exploration of pharmaceutical analysis,and ANN have broad development prospects and application value in the field of pharmaceutical analysis.