Research And Development Of The Main Components Detector Of Fresh Milk Based On Dielectric Properties

As an ancient nutural drink,Milk is favored by the consumers because of its rich and comprehensive nutrition.Fresh milk is the main raw material of dairy products,and its quality detection is of great significance to ensure the quality and safety of dairy products.The content of protein and fat and the moisture content are often used as important indexes to evaluate the quality of milk.The national standard method for the detection of main components of milk has the advantages of high detection accuracy,but the operation is complex,time-consuming and laborious,and can not be used for on-site detection.The multi-functional dairy analyzer is a widely used dairy ingredient detector in large dairy enterprises at present.It has the advantages of high accuracy and rapid detection,but its price is very expensive and its volume is huge,so it is only suitable for laboratory testing.It is of great significance to develop a low-cost,easy to operate,suitable for on-site detection of the main components of fresh milk detector for guiding milk sales,processing and consumption.The existing research shows that the dielectric properties of milk are affected by protein,fat and moisture content.The dielectric spectrum technology has the advantages of fast detection speed,no pollution,suitable for online detection and so on.Therefore,based on the principle of parallel plate method(1-100 MHz)and coaxial probe method(40-2500 MHz)to measure the dielectric properties,this paper developed the main components detector of fresh milk,and tests the performance of the detector,so as to provide a scheme for the on-site and rapid detection of milk quality.The main contents and conclusions of this paper are as follows:(1)Hardware system design.With STM32F103C8T6 chip of arm Cortex-M3 core as control processor,the hardware system of parallel plate type milk and coaxial probe type milk main component detector is designed with parallel plate and open coaxial line as detection probes.The hardware system includes crystal clock,hardware reset,power supply,sweep signal source,reflection bridge,peak value detection,phase shift detection and analog-to-digital conversion circuit.(2)Software system design.Configure the software development environment and complete the program design with C language in keil MDK 5.13 software.STM32 firmware library mode was used to develop the program.Realized the initialization of single chip microcomputer,sweep signal output,analog-to-digital conversion,keyboard scanning,display driver and serial port transmission of data.(3)The establishment of milk protein content,fat content and moisture content prediction model.390 milk samples(360 raw milk and 60 fresh milk mixed with water)were collected at 1-100 MHz and 40-2500 MHz respectively by using the developed parallel plate type and coaxial probe type milk main component detector.The mobile window smoothing filter method was used to preprocess the collected data,SPXY method was used to divide the samples,and partial least squares(PLS),limit learning machine(ELM)and support vector machine(SVM)models were established to predict the content of each main component based on the full wavelength(FF)and the characteristic wavelength optimized by continuous projection algorithm(SPA).The results showed that the model based on FF has the best prediction accuracy whether it is parallel plate or coaxial probe.For the parallel plate type,the optimal prediction models of protein content,fat content and water content are FF-SVM,FF-PLS and FF-SVM respectively;for the coaxial probe type,the optimal prediction models of protein quality,fat content and water content are FF-PLS,FF-SVM and FF-SVM respectively.(4)The accuracy of the two instruments is verified.The prediction model was written into the software program of the tester,and 40 fresh milk samples were taken to verify the accuracy of the instrument.The results showed that the average errors of protein content,fat content and moisture content detected by parallel plate instrument are 0.17%,0.35% and0.49% respectively,and the average errors of protein content,fat content and moisture content detected by coaxial probe type are 0.29%,0.33% and 0.84% respectively.The results showed that the accuracy of the parallel plate type instrument is slightly better than that of the coaxial probe type.(5)The effect of milk homogenization on the performance of the detector was studied.The results showed that the homogenization treatment is helpful to improve the detection performance of the detector for fat,but it has little improvement for the detection performance of protein content and moisture content.The main components detector developed in this paper had the advantages of fast detection speed,simple operation and low cost,which provided a solution for the field and rapid detection of milk.