Near-infrared (NIR) spectroscopy is widely used to estimate product quality and other key variables. Eliminating redundant variables is very important in constructing a high-quality NIR model. This article proposes a new wavelength-selection method for NIR spectroscopy based on joint mutual information and weighted bootstrap sampling (WBS). The method considers the combination effect of variables and involves the dynamic selection of wavelength in each iteration to increase the model prediction accuracy. The index based on joint mutual information is used to determine the importance of variables and thus accurately reflects the variable-combination effect. WBS is further used to dynamically adjust the importance of candidate variables, i.e., to increase the weights of samples with poor prediction results and decrease those of samples with accurate prediction. This process ensures that the subsequently selected wavelength focuses on inaccurately estimated samples. The performance of this method is demonstrated through three NIR datasets of gasoline, shootout, and diesel fuels. The proposed method is found to have better accuracy than the traditional partial-least-squares method, variable iterative space shrinkage approach, and several other wavelength-selection methods.
5 May 2021 in IEEE Transactions on Instrumentation and Measurement
This article presents a study about the rapid characterization of petroleum naphtha properties based on near-infrared (NIR) spectroscopy. The major challenge of this problem is the low prediction accuracy and poor robustness of predictive models caused by significant noise and insufficient sample data. To address this challenge, a machine learning (ML) method and a correlation analysis (CA) method are applied, respectively. The ML approach uses wavelet transform to reduce noise and kernel partial least square (kPLS) algorithm to deal with the nonlinearities. The CA method utilizes the correlation relationship between petroleum naphtha samples and real-time NIR data to solve robustness problem. In order to exploit the advantages of both methods, a novel integration approach is then proposed, which systematically integrates the ML and correlation methods for both good accuracy and robustness. Application studies on NIR spectroscopy data from industry have been conducted. The results confirm the issues with only use of the ML or CA method and demonstrate the advantages of the proposed integrated approach and its potential in industrial applications.