水溶性有机物荧光指标LM神经网络法评价堆肥腐熟度研究

Levenberg-Marquardt Neural Network combining with the fluorescence spectra characteristics of DOM derived from organic waste composting for assessment of compost maturity

由于受到有机废弃物中各组分间的影响,有机废弃物腐熟程度的判断呈现模糊性,使传统的评价方法很难正确认识。水溶性有机物(DOM)的荧光特性可作为评价有机废弃物堆肥腐熟程度的重要手段。文章通过获取多种有机废弃物在堆肥各阶段DOM的荧光特性参数,统计分析结果表明,相互间呈现显著相关(p<0.01)的荧光参数有A_(FLR)、A_4/A_1、r_(A,C)、P_(Ⅱ,n)、P_(Ⅳ,n)、P_(Ⅴ,n)/P_(Ⅲ,n)、P_((Hs)/(Pr)),故可将其作为综合评价的指标。在此基础上,结合LM神经网络模型定量表征堆肥腐熟度等级,并将腐熟度划分为4个腐熟等级:未腐熟(I级))、基本腐熟(II级))、较腐熟(III级))、完全腐熟(IV级),并以4组确定腐熟度样本作为标准,进行LM神经网络训练,32组已知腐熟度样本进行预测,网络预测准确性为84.37%,因此,该方法对评价有机废弃物堆肥有重要意义。

The traditional methods are difficult to assessing the compost maturity in a correct way owing to the influence of the different organic components, which make the assessment fuzziness. However, the properties of fluores-cence spectra of dissolved organic matter（DOM）derived from composting could be an important means of assessing the compost maturity. In this study, the germination percentage（GI）and fluorescence spectra of DOM of organic waste composting（chicken manure, swine manure, kitchen waste, lawn waste, fruits and vegetables waste, straw, green waste, and municipal solid waste）were measured. Person correlation analysis between the fluorescence parameters of DOM and GI demonstrated that A（FLR）, A4/A1, r（A,C）, P（ Ⅱ,n）, P（ Ⅳ,n）, P（ Ⅴ,n）/P（ Ⅲ,n） and P（Hs/Pr） are more suitable（p〈0.01; 2-tailed）to assess compost maturity as comprehensive indexes in LM Neural Network. Four degrades of compost maturity were divided on the base of the GI value during composting and they were immaturity（Ⅰ, GI〈 50%）, maturity（Ⅱ,50%〈 GI〈 60%）, better degree of compost maturity（Ⅲ, 60%〈 GI〈 80%）and best degree of maturity（Ⅳ, GI〉80%）, respectively. Four groups of samples, which have known to the degree, were used as the standard for LM Neural Network training. Furthermore, thirty-two groups of samples, which also have known to the degree, were used for vali-dation. The accuracy of LM Neural Network was 84.37%, which suggested LM Neural Network showed a good performance in the the assessment of compost maturity.