Process Evaluation of Key Parameters during Plant-field Composting using Genetic Algorithms and Near-infrared Spectroscopy

摘要

The nondestructive estimation of key parameters during plant-field chicken manure composting is of great importance for quality evaluation. In the process of developing regression models using nearinfrared spectroscopy (NIRS), methods used for wavelength selection significantly influence on the efficiency of the calibration. This study explored the method of genetic algorithms (GAs) for selecting highly related wavelengths to improve NIRS models for moisture (Miost), pH and electronic conductivity (EC), total carbon (TC), total nitrogen (TN) and C/N' ratio determination in chicken manure during composting. Based on the values of coefficient of determination in the validation set (R2) and root mean square error of prediction (RMSEP), the prediction results were evaluated as excellent for Miost, TC and TN, good for pH and EC, and approximate for C/N ratio. But GAs had better performance than using full spectrum for near-infrared spectroscopy model construction in the process of evaluating key parameters during plant-field chicken manure composting.