| Microbial pollution has always been the most important factor endangering food safety.It is of great significance to realize online detection of microbial biomass concentration in food production process to ensure product quality and food safety.Meanwhile,quantitative detection of microorganisms in food production process can also provide corresponding data support for standardized production and management.However,at present,there are many disadvantages of microbial detection methods,such as time-consuming,complex operation process,the need to add additional reagents to destroy samples,etc.these limitations make the detection technology of microbial concentration relatively lagging behind,unable to achieve the online detection in the food production process.The spectral method has the advantages of fast,simple operation and no direct contact with the tested samples.Therefore,it is introduced into the on-line quantitative detection of microbial concentration in the food production process in this study.Milk is rich in nutrition,complex in composition,and the scale of the consumer market is huge,which is very representative.Therefore,it is selected as the research background of this paper.Escherichia coli is widely distributed in nature,and dairy enterprises suffer huge losses due to the pollution of Escherichia coli every year.Escherichia coli is selected as the target detection object in the milk production process.The main contents of this paper are as follows:In order to explain the cause of absorption spectrum of milk infected with bacteria and assist the realization of on-line detection,the photochemical absorption characteristics of milk dry matter and Escherichia coli were analyzed in detail.The existing biomacromolecule model was analyzed to explain the cause of sample scattering.According to the characteristics of biomacromolecule multi particle sizecombined scattering,the combination of multiple scattering correction and normal variable transformation was selected The algorithm of de scattering.According to the absorption characteristics of the spectral curve,the wavelet transform is selected to remove the noise in the spectral curve.According to the characteristics of milk,the cause of red shift of spectrum curve of Escherichia coli was analyzed.Combined with the analysis method of two-dimensional correlation spectrum,40 groups of spectrum obtained in the experiment were analyzed.It was determined that the characteristic absorption peak position of Escherichia coli after red shift was 310nm-385 nm,and 300nm-400 nm was selected as the analysis range to ensure the spectrum continuity.At the same time,in order to ensure the reasonable division of the data set,40 groups of original spectra were divided into 30 groups of training sets and 10 groups of prediction sets by K-S algorithm.The plate culture method was used as the calibration experiment to compare the prediction results of the prediction model with the quantitative results of Escherichia coli in the samples estimated by the plate method.Finally,this study compares the BP artificial neural network,GRNN neural network and elm neural network,analyzes the prediction accuracy of the three models,the stability of the prediction time and results,and determines that the limit learning machine neural network is the most suitable prediction model for this research background.The limit learning machine method can guarantee the prediction accuracy and have the shortest running time,and finally predict the overall model The accuracy is 92.74%,the prediction time is 6s,and the whole time of spectral method is 3min.It is proved that the method can meet the requirements of on-line detection of microbial biomass concentration in the production process. |
PDF Full Text Request