论文部分内容阅读
为提高柴油组分近红外法检测的精度,提出了一种基于训练字典稀疏表示下的建模方法并用于柴油十六烷值、沸点和芳烃总量的检测.该法先用柴油光谱结合K均值奇异值分解(K-SVD)算法完成对冗余字典的训练,再用正交匹配追踪算法(OMP)寻找柴油光谱在该训练字典下的稀疏表示系数,用该系数建立了柴油十六烷值、沸点和芳烃总量偏最小二乘预测模型.实验比对了训练字典、傅里叶字典和小波字典稀疏表示下的柴油组分预测模型性能,其中训练字典的表示系数建模性能最优且比其他两种字典的预测性能有较大幅度改进,验证了该法在近红外光谱检测建模领域推广能改善预测的准确度和稳健性.
In order to improve the accuracy of diesel component NIR detection, a modeling method based on training dictionary sparse representation was proposed and used for the detection of cetane number, boiling point and aromatics content in diesel. The algorithm of mean-value singular value decomposition (K-SVD) is used to train the redundant dictionary. Orthogonal Matching Pursuit (OMP) is then used to find the sparse representation coefficient of diesel spectrum under the training dictionary. Diesel cetane Value, boiling point and the total aromatics partial least square prediction model.Experiments were performed to compare the performances of the diesel components prediction models under the sparse representation of training dictionary, Fourier dictionary and wavelet dictionary, in which the performance of the training dictionary was best Compared with the other two kinds of dictionaries, the predictive performance of the two dictionaries has been greatly improved, which proves that this method can improve the prediction accuracy and robustness in the field of near infrared spectroscopy detection and modeling.