本文選題：激光誘導(dǎo)擊穿光譜技術(shù) + 土壤　； 參考：《浙江大學(xué)》2016年博士論文

【摘要】：數(shù)字農(nóng)業(yè)是實(shí)現(xiàn)農(nóng)業(yè)精準(zhǔn)化管理和科學(xué)化種植的一條重要途徑,是現(xiàn)代農(nóng)業(yè)最前沿的發(fā)展領(lǐng)域之一,也是當(dāng)今農(nóng)業(yè)高效、生態(tài)、安全和可持續(xù)發(fā)展的關(guān)鍵和核心。數(shù)字化精準(zhǔn)農(nóng)業(yè)的實(shí)施中最為基本和關(guān)鍵的因素是農(nóng)作物-環(huán)境信息的準(zhǔn)確感知、快速獲取和智能控制。數(shù)字化和信息化技術(shù)可為農(nóng)業(yè)綠色生產(chǎn)和高效管理提供快速、準(zhǔn)確的信息獲取、科學(xué)的輔助決策和高效的作業(yè)控制,已成為農(nóng)業(yè)科技領(lǐng)域研究的熱點(diǎn)。土壤作為人類賴以生存的重要自然資源之一,是農(nóng)業(yè)生產(chǎn)的基礎(chǔ)和根本所在,對(duì)土壤的信息獲取和檢測的技術(shù)和方法研究是農(nóng)業(yè)環(huán)境信息領(lǐng)域的熱點(diǎn)。對(duì)土壤類型的分析研究可以為建立土壤的肥力和質(zhì)量評(píng)價(jià)系統(tǒng),為土壤的整治、規(guī)劃和合理利用提供科學(xué)依據(jù)；土壤的元素信息的檢測能夠?yàn)檗r(nóng)業(yè)田間作物營養(yǎng)診斷,農(nóng)田信息實(shí)時(shí)獲取和科學(xué)的肥水管理奠定理論基礎(chǔ)；對(duì)土壤的重金屬檢測可以有效防止農(nóng)田的重金屬污染,為農(nóng)業(yè)的高品質(zhì)安全生產(chǎn)提供理論指導(dǎo)作用。本研究在系統(tǒng)深入了解激光誘導(dǎo)擊穿光譜(Laser-induced breakdown spectroscopy, LIBS)技術(shù)的工作過程和原理基礎(chǔ),展開闡述了激光誘導(dǎo)等離子體的形成機(jī)理和作用特性的基礎(chǔ)上,以土壤為研究對(duì)象,研究分析LIBS系統(tǒng)單變量參數(shù)、土壤狀態(tài)參數(shù)以及系統(tǒng)多變量參數(shù)對(duì)土壤的等離子體特性影響；建立了基于LIBS技術(shù)結(jié)合化學(xué)計(jì)量學(xué)方法對(duì)不同地域類型的土壤類型判別模型；對(duì)比探索了基于單波段和多變量回歸的土壤中主要金屬元素含量的定量分析模型和方法；探討了基于LIBS技術(shù)結(jié)合單波段和多變量的定標(biāo)方法對(duì)土壤重金屬鉛和鎘元素含量快速檢測的方法,為后期開發(fā)土壤理化信息(類型信息、元素的種類和含量信息等)檢測儀器提供理論依據(jù)。具體的研究內(nèi)容如下：(1)研究了LIBS系統(tǒng)參數(shù)與土壤狀態(tài)參數(shù)對(duì)土壤等離子體特性影響規(guī)律。通過單因素分析,討論了LIBS試驗(yàn)系統(tǒng)的主要參數(shù)(激光脈沖能量、重復(fù)頻率、延時(shí)時(shí)間、采集方式以及聚焦透鏡到樣品的距離)以及土壤的狀態(tài)參數(shù)(水分含量、顆粒大小和緊實(shí)度等)對(duì)土壤等離子體特性的影響,得到優(yōu)化的試驗(yàn)參數(shù)：激光脈沖能量為100 mJ左右,重復(fù)頻率為1 Hz,延時(shí)時(shí)間要依具體元素來定,采集方式為20次取平均,以及聚焦透鏡到樣品的距離為98 mm(透鏡焦距為100 mm)；土壤含水率越低越好,土壤顆粒大小應(yīng)小于0.15 mm(過100篩子),土壤壓片的壓力(即緊實(shí)度)在10 MPa較好。(2)設(shè)計(jì)了三因素(激光脈沖能量LE、延遲時(shí)間DT和聚焦透鏡到樣品的距離LTSD)二次回歸旋轉(zhuǎn)正交組合的實(shí)驗(yàn),優(yōu)化了LIBS系統(tǒng)對(duì)土壤檢測的最佳實(shí)驗(yàn)條件。以土壤中主要元素的譜線綜合信背比(signal-background-ratio, SBR) YSBR為目標(biāo)函數(shù),分析了三因素之間交互作用對(duì)YSBR的影響,結(jié)果表明：DT對(duì)YSBR的線性效果顯著,而LE和LTSD對(duì)YSBR的線性效果均不顯著；三者的交互影響對(duì)YSBR的交互效果都不顯著；對(duì)于因素LE2、DT2和LTSD2對(duì)YSBR的曲面效應(yīng)均顯著。通過分析優(yōu)化得到最佳的試驗(yàn)條件是：激光能量LE為103.09 mJ,延遲時(shí)間為2.92μs,透鏡到樣品的距離LTSD為97.69 mm時(shí),得到最大綜合信背比YsBR為198.602。(3)建立了基于LIBS技術(shù)結(jié)合化學(xué)計(jì)量學(xué)方法對(duì)不同類型的土壤的判別分析模型,并且驗(yàn)證了該模型方法的可靠性。通過對(duì)6種標(biāo)準(zhǔn)土壤樣品的LIBS譜線特征進(jìn)行主成分分析和元素含量對(duì)比,選取了7條特征譜線,并建立了基于7條特征譜線的偏最小二乘判別分析(Partial least squares discriminant analysis, PLS-DA)、簇類的獨(dú)立軟模式法(Soft independent modeling of class analogy, SIMCA)和最小二乘支持向量機(jī)(Least-squares support vector machine, LS-SVM)判別模型,判別的精度分別為98%、90%和100%,并用受試者工作特征曲線(Receiver operating characteristic curve, ROC)評(píng)價(jià)模型的性能,表明基于激光誘導(dǎo)擊穿光譜的最小二乘支持向量機(jī)(LIBS-LS-SVM)判別模型的性能最好。針對(duì)選取的7條特征譜線,對(duì)選取另外8個(gè)不同類型的土壤樣品進(jìn)行分析驗(yàn)證,PCA得到8種土壤有明顯聚類,建立的LS-SVM判別模型準(zhǔn)確率為100%,ROC曲線也證明其預(yù)測性能的可靠性,這為研究土壤分類系統(tǒng)和農(nóng)田土地的管理和合理利用提供理論依據(jù)；(4)應(yīng)用LIBS技術(shù)結(jié)合定標(biāo)曲線法以及化學(xué)計(jì)量方法,實(shí)現(xiàn)了對(duì)于土壤中多種元素(Al、Ca、 K、Mg、Na和Fe)同時(shí)定量檢測。將LIBS數(shù)據(jù)經(jīng)預(yù)處理(數(shù)據(jù)歸一化,剔除異常光譜和平均處理)后,分別對(duì)比了基于譜線峰值強(qiáng)度、譜峰的積分信息(峰面積)和Si元素內(nèi)標(biāo)的定標(biāo)方法。結(jié)果表明：基于峰值強(qiáng)度信息和譜峰的峰面積的定標(biāo)曲線對(duì)多數(shù)元素都有較好的線性關(guān)系(Fe元素除外)；以Si元素內(nèi)標(biāo)的定標(biāo)曲線的線性相關(guān)系數(shù)優(yōu)于前兩種定標(biāo)方法；另外,利用自由定標(biāo)法(Calibration free-LIBS, CF-LIBS)對(duì)土壤中主要元素Al、Ca、Si、Fe、Mg、Na和K的含量進(jìn)行計(jì)算,結(jié)果有待于提高；建立了基于多變量偏最小二乘回歸(partial least squares regression, PLSR)的土壤中Al、 Ca、K、Mg、Na和Fe預(yù)測模型,結(jié)果明顯要優(yōu)于定標(biāo)曲線的分析精度,其各個(gè)的預(yù)測相關(guān)系數(shù)RP分別為：Ak,0.8455、Ca,0.9769、Fe,0.9744、K,0.8468、Mg,0.8260、 Na,0.9705,整體的預(yù)測精度要明顯優(yōu)于前幾種定標(biāo)方法,在應(yīng)用LIBS技術(shù)對(duì)物質(zhì)含量的定量分析中,多元的PLSR方法能夠展現(xiàn)其較好的分析精度,也有更好的應(yīng)用前景。(5)應(yīng)用LIBS技術(shù)結(jié)合定標(biāo)曲線法以及化學(xué)計(jì)量方法,實(shí)現(xiàn)了土壤中重金屬鉛和鎘元素的快速定量檢測。選取Pb Ⅰ 405.78 nm和Cd Ⅰ 361.05 nm為分析譜線,建立基于譜線峰強(qiáng)度,歸一化后洛倫茲擬合強(qiáng)度以及譜峰面積與對(duì)應(yīng)元素的濃度之間的關(guān)系模型。對(duì)于Pb元素,基于譜線峰強(qiáng)度、歸一化后洛倫茲擬合強(qiáng)度以及譜峰面積與對(duì)應(yīng)元素的濃度之間的線性關(guān)系分別為0.9839、0.9710、0.9932；而Cd元素,定標(biāo)曲線法沒有明顯的線性關(guān)系,其分析精度有待提高;同時(shí)建立了基于PLSR方法的土壤Pb和Cd元素的定量分析模型,Pb元素的定標(biāo)曲線法結(jié)果和PLSR模型的結(jié)果類似,其預(yù)測的相關(guān)系數(shù)RP為0.9485,預(yù)測均方根誤差RMSEP為2.044 mg·g-1；而Cd元素的PLSR模型的結(jié)果提升較大,預(yù)測的相關(guān)系數(shù)RP為0.9949,預(yù)測均方根誤差RMSEP為97.05 gg·g-1。
[Abstract]:Digital agriculture is an important way to realize the precision management and scientific planting of agriculture. It is one of the most advanced development fields of modern agriculture. It is also the key and core of agricultural efficiency, ecology, safety and sustainable development. The most fundamental and key factor in the implementation of digital precision agriculture is the quasi agricultural environment information. True perception, rapid acquisition and intelligent control. Digital and information technology can provide rapid, accurate information acquisition, scientific decision making and efficient operation control for agricultural green production and efficient management. It has become a hot spot in the field of agricultural science and technology. As one of the important natural resources for human survival, soil is an agricultural student. The research on soil information acquisition and testing is a hot spot in the field of agricultural environmental information. The analysis and study of soil types can be used to establish soil fertility and quality evaluation system for soil remediation, planning and rational use for scientific basis; the detection energy of soil element information can be found. It lays a theoretical foundation for agricultural field crop nutrition diagnosis, real-time acquisition of farmland information and scientific fertilizer management, which can effectively prevent heavy metal pollution in farmland and provide theoretical guidance for high quality and safe production of agriculture. In this study, laser induced breakdown spectroscopy (Laser-indu On the basis of the working process and principle of CED breakdown spectroscopy (LIBS) technology, the formation mechanism and function characteristics of laser induced plasma are expounded. Soil is taken as the research object, and the effects of the single variable parameters of the LIBS system, the soil state parameters and the multivariable parameters of the system on the plasma characteristics of the soil are studied and analyzed. A discriminant model of soil types based on LIBS technology combined with chemometrics method was established, and the quantitative analysis model and method for the content of major metal elements in soil based on single band and multivariable regression were compared and explored, and the soil weight based on LIBS technology combined with single band and multivariable calibration method to soil weight was discussed. The rapid detection method of metal lead and cadmium content provides a theoretical basis for the later development of soil physical and chemical information (type information, element type and content information etc.). The specific research contents are as follows: (1) the effects of LIBS system parameters and soil state parameters on soil plasma characteristics are studied. The main parameters of the LIBS test system (laser pulse energy, repetition rate, delay time, acquisition mode and the distance of focusing lens to the sample) and the influence of the state parameters of soil (moisture content, particle size and compaction) on soil plasma characteristics are discussed. The optimized experimental parameters are obtained: laser pulse energy is 100 Around mJ, the repetition rate is 1 Hz, the time delay time must depend on the specific elements, the acquisition mode is 20 times averaging, and the distance of the focusing lens to the sample is 98 mm (lens focal distance is 100 mm); the lower the soil water content the better, the soil particle size should be less than 0.15 mm (100 sieves), and the pressure of the soil press (i.e. compaction) is better in 10 MPa. (2) The experiment of three factors (laser pulse energy LE, delay time DT and the distance LTSD of focusing lens to sample) was designed for the experiment of two regression rotation orthogonal combination. The optimum experimental conditions for soil detection were optimized. The target function of the spectrum line integrated signal back ratio (signal-background-ratio, SBR) YSBR of the main elements in the soil was analyzed by three. The effect of interaction between factors on YSBR shows that DT has a significant linear effect on YSBR, while LE and LTSD have no significant linear effect on YSBR; the interaction effect of the three is not significant to the interaction effect of YSBR; for factors LE2, DT2 and LTSD2 are both obvious to the YSBR surface effect. The optimal test conditions are obtained by analysis and optimization. It is: the laser energy LE is 103.09 mJ, the delay time is 2.92 Mu s, and the distance LTSD of the sample is 97.69 mm, the maximum integrated signal back ratio YsBR is 198.602. (3), and the discriminant analysis model of different types of soil based on LIBS technology combined with chemometrics method is established, and the reliability of the model method is verified. 6 kinds of methods are verified. The LIBS line characteristics of the standard soil samples are analyzed by principal component analysis and element content comparison, and 7 characteristic spectral lines are selected, and the partial least squares discriminant analysis (Partial least squares discriminant analysis, PLS-DA) based on the spectral lines of the soil is established, and the independent soft mode method of Soft independent modeling of class (Soft independent modeling) is established. CA) and the least squares support vector machine (Least-squares support vector machine, LS-SVM) discriminant model, the accuracy of discrimination is 98%, 90% and 100% respectively, and the performance of the subjects' working characteristic curves (Receiver operating characteristic curve, ROC) is used to show the least square support vector machine based on the laser induced breakdown spectroscopy. LIBS-LS-SVM) the performance of the discriminant model is the best. According to the selected 7 characteristic spectral lines, 8 different types of soil samples are selected to be analyzed and verified by 8 different types of soil samples. 8 kinds of soil have obvious clustering, the accuracy rate of the established LS-SVM discriminant model is 100%, and the ROC curve also proves the reliability of its prediction performance. This is the study of soil classification system and agriculture. The management and rational utilization of field land provide theoretical basis; (4) the simultaneous quantitative detection of various elements (Al, Ca, K, Mg, Na and Fe) in the soil is realized by using LIBS technology and the method of calibration curve and chemical metrology. After the LIBS data is pretreated (data normalization, elimination of abnormal spectra and average processing), the data of LIBS are compared. The line peak strength, the integral information (peak area) of the peak and the calibration method of the Si element internal standard. The results show that the calibration curves of peak area based on peak intensity information and peaks have a good linear relationship with most elements (except Fe elements), and the linear correlation coefficient of the calibration curve with the internal standard of Si is superior to the first two calibration methods. In addition, the content of the main elements of soil Al, Ca, Si, Fe, Mg, Na and K in the soil are calculated by Calibration free-LIBS (CF-LIBS), and the results of the soil, which are based on the multivariable partial least squares regression (partial least), are established. Compared with the accuracy of the calibration curve, the prediction correlation coefficients RP are Ak, 0.8455, Ca, 0.9769, Fe, 0.9744, K, 0.8468, Mg, 0.8260, Na, 0.9705, and the overall prediction accuracy is obviously superior to the previous methods. In the quantitative analysis of the material content in the application of LIBS technology, the multivariate PLSR method can show its better analysis. Precision and better application prospect. (5) the rapid quantitative detection of heavy metal lead and cadmium in soil was realized with LIBS technique combined with calibration curve method and chemical metrology method. Pb I 405.78 nm and Cd I 361.05 nm were selected as the analytical spectral lines, and the intensity of spectral line peak, normalized intensity and peak surface after normalization were established. For the Pb element, based on the spectral peak intensity, the linear relationship between the Lorenz fitting strength and the peak area of the spectral peak and the concentration of the corresponding elements is 0.9839,0.9710,0.9932, while the Cd element, the calibration curve method has no obvious linear relationship, and the accuracy of the analysis needs to be improved. At the same time, the quantitative analysis model of soil Pb and Cd elements based on PLSR method is established. The calibration curve method of Pb element is similar to the result of PLSR model. The correlation coefficient of the prediction is 0.9485 and the mean square root error RMSEP is 2.044 mg. G-1, while the PLSR model of the Cd element is higher and the predicted correlation coefficient RP is 0.9949. The mean square root error of RMSEP is 97.05 GG. G-1.
【學(xué)位授予單位】：浙江大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2016
【分類號(hào)】：S151.9;TN249
，

本文編號(hào)：2006654