Research Of Machine Learning In Selection Of Spectral Features And Identification Of Grease Categories

With the rapid development of science and technology,there is a large amount of data and information in every field.The valuable information contained in these data does bring great convenience to our lives,it is very difficult to get meaningful and valuable information from large amount of data at the same time.Moreover,with the increasing trend of scale and type of information and data,more and more data was showed high dimension and small sample form.There is a lot of redundant and useless information in the feature space of the sample.These "noises" increased the difficulty of learning and recognizing.The feature selection method came into being in this background to meet such demand.By screening useful feature subsets in the original feature space,the useless and meaningless information features are excluded in order to reduce the difficulty of processing and analyzing data.As a result,feature selection has been widely valued and applied in real life.The application range of grease is very extensive.In order to meet the endless lubrication requirements of equipment,the production process,technology,varieties,output.Analysis and evaluation methods of grease need to be evaluated in detail.In the field of identification of grease categories,it is impossible to figure the grease quickly and truly through the conventional subjective experience,poor consistency.to determine the grease and brand type is not enough in the analysis of grease species identification.There will be no correspondence between grease data attributes and grease categories and lead to classification errors finally.Therefore,a rapid and nondestructive testing method is needed to realize the effective identification of grease.Based on the infrared spectrum data of grease,we select the infrared spectral band that represent the information of grease category through optimization and network algorithm.We also established a hierarchical classification identification model to identify the grease category.The main contents of this paper as follows:(1)5.3%of all infrared spectra is chosen after the optimization compared with 1868 spectral points of full spectrum(2)The accuracy of the model to distinguish the polyurea grease via the hierarchical training was raised from 88.8%to 95.2%.(3)The accuracy of using the recommended spectrum to distinguish lithium grease and calcium grease is 100%,100%..(4)Figure the three dimensional spatial distribution of different grease types at the spectral points,and the conclusion is proved that the characteristic spectrum holds the information of grease category.