Study On The Novel Vibrational Spectral Fusion Technique For Diagnosis Of Adulterated Milk Powder

The milk powder plays an extremely important role in people's daily life because of its perfect nutritional composition and the strong similarity to breast's characteristics.However,driven by the loss of morality and the imperfection of supervising system,some manufactures adulterate the milk powder with other cheap materials,thus resulting in the accidents of melamine and other contaminants.The problem of dairy safety is thus greatly concerned by the public.In routine screening of dairy product,the complicated sample pretreatment,high labor cost and expensive instruments combined to limit the applications of Massspectrum,Chromatography methods.Moreover,these complex instrument methodologies required highly quantified persons to operate,which is hard to match the huge demand of screening dairy products.Alternatively,Raman spectroscopy and infrared spectroscopy stand out because of its unique advantages of rapid speed,sensitivity and high throughput.Both of vibrational spectroscopic technologies can detect a large number of samples in a reasonably short time.In order to detect the adulterations in milk powder much more effectively,a multispectral fusion technique(MFT)method consisted of Raman and infrared spectrometries was developed to ulitize their complementary molecular charactericstics localized in their spectra.In MFT,the qualitative analysis usually relies on chemometrics methods to detect the presence of analytes,requiring removel of different interference one by one.In this regard,several data fusion strategies was proposed to facilitate the extraction of essential information presented in MFT,where we combine the discrete wavelet transform(DWT)and typical feature selection methods,e.g.montecarlo elimination uninformative variables(MCUVE),competitive adaptive reweighted sampling(CARS),random frog(RF)with partial least squares discriminant analysis(PLSDA),etc.to eliminate redundancy spectral variables.Theoritically,these strategiesare capable of encoding the multi-spectra information in the presence of complex spectral interference analysis respectively.The objective of this paper is to make the qualitative detection of nine typical adulterations such as flour,starch,talcum,sucrose,dextrin,expired milk powder,nondairy creamer,soy protein isolate and soybean powder,etc.All these methods were validated and refined by performing MFT analysis of 9 typical milk powder adulteration systems to simulate the complicated situations encountered in practice.The results showed that the MFT outperformed solitary Raman spectroscopy or infrared spectroscopy in both of detection sensitivity and accuracy.In MFT,it is of great interest to investigate the performance of different chemometrics strategies.The experimental results indicated that the DWT-RF-PLSDA algorithm presented the best calibration performance in enhancing the discriminant accuracy and stability,and the reason is being that DWT-RF-PLSDA selected the globally optimized features according to the variance of enormous regression vectors.One might expect MFT is a promising tool for discriminating the milk powder adulteration,which may well extend to screen other complicated systems.