Optimization Of Electronic Nose And Its Application To The Detection Of Nuts Quality

Recently,the detection of nuts quality has increasingly rasied people's concentration.However,due to the block of hard shell,internal information of pecan is difficult to access directly and accurately.As an olfactory bionic instrument,electronic nose(e-nose)judge nuts'quality based on "fingerprinter figure" of sample's volatile,which makes it owns a good prospect in the field of nuts' quality detection.At present,the research and development of e-nose mainly focus on the derection of automation and intelligence,and according to the type of e-nose,it can be divided into portable type and desk type.Thereinto,portable e-nose mainly has the problesm of poor human-computer interation and low data processing efficiency,while the desk type lacking of automatic injection mechanism and performs poor in automiaztion.Therefore,in view of the above problems and demands of different application condition,this paper designed an embedded portable e-nose and an automatic desk type e-nose based on DSP and DELTA parallel manipulator respectively.Moreover,pecans were taken as object for experiment to verify the actual detection abiliy of e-nose developed.The main research contents and conclusion of this paper are as follows:(1)An embedded portable e-nose based on DSP is developed,which includes gas collection and response module,signal acquisition and condition module,DSP control processing module and information display module.In this research,FLUENT fluid analysis software is applied for the optimal design of air charmber,and human-computer interaction interface is designed based on the PS-LCD to achieve the visualization of detection process.Meanwhile,programs of data preprocessing,feature extraction,pattern recognition are implanted into DSP to make the e-nose with independent computing ability.(2)An automatic desk type e-nose based on DELTA parallel manipulator is developed,which includes pluggable sensor response module,automatic sampling module,signal acquisition and sampling control module and PC computer.In this research,an sensor mounting pedestal is designed for the quick replacement of different types of sensors,and the best scheme of sesors distribution and air inlet height are determined through the analysis of FLUENT.While automatic sampling module is designed based on the structure of DELTA 3 degree of freedom parallel manipulator.Meanwhile,according to the function module of e-nose,human-machine interface of PC is designed based on Matlab GUI.(3)Specific to different detection object,three feature extraction methods are applied to each sensor's abstraction to generate the initial feature matrix,thus are mean-differential coefficient value,stable value and response area.Then this paper presents an optimization method basded on non-search feature selection strategy,and through the procedure of mean analysis,variation coefficient analysis,cluster analysis,correlation coefficient analysis and multiple co-linear analysis to obtain the optimal feature matrix.Eventually,sensors corresponding to the featrues are selected to generate the optimal sensor array.(4)In order to verify the performance of e-nose developed,automatic desk type e-nose is used for the detection of aging time of raw pecan,strorage time and doping ratio of processed pecan.The result shows that detection data of raw pecans with diferent aging time can be distinguished well based on PCA;LDA performs better than PCA in the classifcaiton of processed pecans with different storage time and doping ratio.Meanwhile,detection data through PCA dimension reduction are used to generate prediction model based on PLSR and BPNN respectively.Thereinto,in the experiment of strorage time detection of processed pecan,the prediction model based on PLSR owns lower precision,while the one based on BPNN performs better.And the prediction model based on PLSR and BPNN of the other two experiments both performs well in prediction of aging time and doping ratio.