基于UD-PLS对牛粪堆制Cu和Zn钝化预测模型的研究

Research of copper and zinc passivation prediction model during cattle manure composting based on uniform design—partial least squares method

选取钝化材料木醋液和影响堆肥质量的关键因素:含水量和C/N,每个因素安排6个水平,利用均匀设计进行多因素多水平试验,对试验结果进行偏最小二乘回归分析,建立对重金属的钝化预测模型.结果表明:木醋液添加比例为0.50%、含水量为40%和C/N为40时,Cu和Zn的钝化效果均达到最大值,分别为13.5%和30.2%;其中重金属Cu的钝化预测模型为CuA B Cy=15.4748+0.3524x-0.1100x+0.0131x,其中交叉有效性为22Q=-2.0767<0.0985,模型达到精度要求;重金属Zn的钝化预测模型为ZnA By=34.3512+11.0905x-0.2561x-C0.0531x,其中交叉有效性为22Q=-3.0863<0.0985,模型达到精度要求.针对多因素多水平的复杂堆肥系统中,将均匀设计与偏最小二乘法有机耦合,有效地解决了试验次数多、因素间多重相关性的问题,从而使模型精度和实用性都得到提高.

The wood vinegar was elected the passivation materials and the key factors affecting the quality of compost for moisture content and C/N ratio. Each factor had six levels, uniform design was used for multiple factors level test, the test results were analyzed by using partial least squares regression, and the heavy metal passivation prediction model was established. Results showed that the test combination with 0.50% wood vinegar, 40% water content and 40C/N ratio had the maximum passivation effects for Cu and Zn （13.5% and 30.2%, respectively）. Partial least squares regression was also applied to the test results. The prediction model for heavy metal Cu passivation effect was YCu ==15.4748＋0.3524x-0.1100x＋0.0131x, where XA, XB, and Xc were wood vinegar content, water content, and C/N ratio, respectively. The cross-validation was： Q2 =-2.0767 〈 0.0985. The model reached precision requirement. The prediction model for heavy metal Zn passivation effect was YZn =34.3512＋11.0905x-0.2561x-C0.0531x. The cross-validation was： Q2 =-3.0863 〈 0.0985. The model reaches precision requirement. In view of the multiple factors level complex composting system, the uniform experimental design combined with partial least squares analysis to effectively solve the experiment many times, and the problems of multiple correlation between factors, so that the precision and practicability of model was improved.