Study On Oil Detection Of Agricultural Machinery Based On Near Infrared Spectroscopy

The use of lubricating oil is a necessary condition for the normal operation of agricultural machinery.The power,safety,economy and life of agricultural machinery engine are closely related to the condition of lubricating oil,which plays a role in lubrication,anti-wear and anti-oxidation of agricultural machinery engine.In recent years,with the rapid development of the agricultural field,the demand for agricultural machinery is also increasing,accompanied by an increase in the demand for agricultural machinery lubricating oil,and more and more inferior lubricating oil in the market.Therefore,it is increasingly important to study a fast,accurate and non-destructive testing method for agricultural machinery lubricating oil.In this paper,the agricultural machinery lubricating oil as the research object,using the near infrared spectroscopy technology to collect the spectral data of the lubricating oil sample configuration,using the stoichiometry method,through the spectral data to establish the quality prediction model of agricultural machinery lubricating oil,realize the pollution concentration and adulteration concentration detection of agricultural machinery lubricating oil.The main contents and achievements of this study are as follows:(1)Aiming at the shortcomings of Random Frog(RF)feature wavelength selection algorithm,such as large number of iterations and low reproducibility,an iteratively retains informative variables-Random Frog(IRIV-RF)feature wavelength selection algorithm is proposed.On the one hand,the iterative retaining information variables(IRIV)algorithm is used to preliminatively determine the strong information variables and weak information variables that are beneficial to modeling among all the variables contained in the original spectrum,which are taken as the initial variable set in RF algorithm to eliminate the influence of the randomness of the initial variable set on the reproducibility of the results.On the other hand through the variable according to the selected probability value from big to small is sorted,starting from the first wavelength,in turn,increase a wavelength and partial least squares regression(PLSR)model is set up,when the PLSR model results in the smallest root mean square error of cross-validation(RMSECV),will be at this time as the corresponding variable subset selected characteristic wavelength,Eliminate the uncertainty of the number of characteristic wavelengths extracted in RF algorithm.(2)NIR spectrometer is used to collect 100 self-made original spectral data of agricultural machinery lubricating oil with different pollution concentrations.Three different pretreatment methods are used to process the original spectrum,and the best pretreatment method is determined as Standard Normal Variate(SNV).On this basis,RF,IRIV and IRIV-RF algorithms were used to select the characteristic wavelength of the whole spectrum,and the PLSR model was established.By comparing the prediction accuracy of full-spectrum PLSR,RF-PLSR,IRIV-PLSR and IRIV-RF-PLSR models,the results show that the PLSR model established by the IRIV-RF algorithm after extracting characteristic wavelengths has the highest prediction accuracy.The Correlation Coefficient of Prediction(Rp)is 0.9613,and the Root Mean Square Error of Prediction(RMSEP)is 5.6516.It significantly improves the prediction accuracy and operation efficiency,and reduces the complexity of the model.It provides a new detection method for pollution concentration of agricultural machinery lubricating oil.(3)Near-infrared spectrometer is used to collect the original spectral data of 100 samples of agricultural machinery lubricating oil with different adulteration concentrations prepared by ourselves.Three different pretreatment methods were used to process the original spectra respectively,and the PLSR model was established.Standard Normal Variate(SNV)is the best pretreatment method.On this basis,RF,IRIV and IRIV-RF algorithms were used to select the characteristic wavelength of the whole spectrum,and the PLSR model was established.By comparing the prediction accuracy of full-spectrum PLSR,RF-PLSR,IRIV-PLSR and IRIV-RF-PLSR models,the results show that the PLSR model established by iri V-RF algorithm after extracting characteristic wavelengths has the highest prediction accuracy,and the prediction set Rp=0.9759,RMSEP=3.7885.The rapid and accurate detection of adulteration concentration of agricultural machinery lubricating oil was realized.In this study,the rationality and effectiveness of the proposed iterative random leapfrog(IRIV-RF)feature wavelength selection algorithm was proved,and the feasibility of the improved IRIV-RF algorithm combined with n IR spectroscopy for detecting the pollution and adulteration concentration of agricultural machinery lubricating oil was proved,which provided a new idea for identifying the quality of lubricating oil.