Hong-Fu Yuan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hong-Fu Yuan has a total of 69 co-authors. Here are the top co-authors by number of shared publications:
Chun-Feng Song
18 shared publicationsXiao-Yu Li
17 shared publicationsJin-Chun Xie
12 shared publicationsAi-Qin Hu
5 shared publicationsXiaoyu Li
Key Laboratory of Carbon Fiber and Functional Polymers, Ministry of Education, College of Materials Science and Engineering, Beijing University of Chemical Technology, Beijing, China
4 shared publications- Chunfeng Song
Key Laboratory of Carbon Fiber and Functional Polymers, Ministry of Education, College of Materials Science and Engineering, Beijing University of Chemical Technology, Beijing, China
3 shared publications
Latest Publications
- Open Access
Determining Olefin Content of Gasoline by Adaptive Partial Least Squares Regression Combined with Near-Infrared Spectroscopy
The accurate and rapid determination of olefin content in gasoline is crucial for fuel quality control. While near-infrared spectroscopy (NIR) offers a rapid analytical solution, multiple parameters in the conventional partial least squares regression (PLSR) modeling process rely on the modeler’s subjective judgment. Consequently, the quantitative accuracy of the model is often influenced by the modeler’s experience. To address this limitation, this study developed an integrated adaptive PLSR framework. The methodology incorporates four core adaptive components: automated selection of latent variables based on the rate of decrease in PRESS values, dynamic formation of calibration subsets using Spectral Angle Distance and sample number thresholds, optimization of informative wavelength regions via correlation coefficients, and systematic database cleaning through iterative residual analysis. Applied to 248 gasoline samples, this strategy dramatically enhanced model performance, increasing the coefficient of determination (R2) from 0.7391 to 0.9102 and reducing the root mean square error (RMSE) from 1.51% to 0.866% compared to the global PLSR model. This work demonstrates that the adaptive PLSR framework effectively mitigates spectral nonlinearity and improves predictive robustness, thereby providing a reliable and practical solution for the on-site, rapid monitoring of gasoline quality using handheld NIR spectrometers.
Open AccessDetermining Olefin Content of Gasoline by Adaptive Partial Least Squares Regression Combined with Near-Infrared Spectroscopy
The accurate and rapid determination of olefin content in gasoline is crucial for fuel quality control. While near-infrared spectroscopy (NIR) offers a rapid analytical solution, multiple parameters in the conventional partial least squares regression (PLSR) modeling process rely on the modeler’s subjective judgment. Consequently, the quantitative accuracy of the model is often influenced by the modeler’s experience. To address this limitation, this study developed an integrated adaptive PLSR framework. The methodology incorporates four core adaptive components: automated selection of latent variables based on the rate of decrease in PRESS values, dynamic formation of calibration subsets using Spectral Angle Distance and sample number thresholds, optimization of informative wavelength regions via correlation coefficients, and systematic database cleaning through iterative residual analysis. Applied to 248 gasoline samples, this strategy dramatically enhanced model performance, increasing the coefficient of determination (R2) from 0.7391 to 0.9102 and reducing the root mean square error (RMSE) from 1.51% to 0.866% compared to the global PLSR model. This work demonstrates that the adaptive PLSR framework effectively mitigates spectral nonlinearity and improves predictive robustness, thereby providing a reliable and practical solution for the on-site, rapid monitoring of gasoline quality using handheld NIR spectrometers.
- 8 June 2021
- Published by Taylor & Francis in Analytical Letters
- Vol. 55 (3) , 388-398
Determination of Cellulose, Water, and N-Methylmorpholine N-Oxide during the Production of Lyocell by Attenuated Total Reflection (ATR) – Fourier Transform Infrared (FTIR) Spectroscopy
Attenuated total reflection (ATR) – Fourier transform infrared (FTIR) spectroscopy with the pretreatment of converting the absorption spectrum into angle spectrum (CASAS) and partial least squares (PLS) was employed to rapidly and simultaneously determine the ingredients of cellulose slurries and solutions during lyocell production. ATR was employed to rapidly and conveniently collect the spectra of the swollen slurries and viscous solutions during production. CASAS was used to eliminate the influence on modeling by the optical path differences due to uneven ingredient distributions in the samples. The PLS models for the cellulose, water, and N-methylmorpholine N-oxide (NMMO) concentrations of the cellulose slurry/solution were constructed using the raw spectra, pretreated spectra by CASAS, and common pretreatment methods, respectively. The results show that CASAS eliminates interferences due to the differences in optical path and improves the performance of PLS model for each analyte. The standard errors of prediction (SEP) of models using CASAS for cellulose, NMMO, and water concentration of cellulose slurry were 0.368, 0.196 and 0.171, while the values for the cellulose solution were 0.442, 0.353, and 0.066, respectively. The standard errors of prediction (SEPs) of CASAS models are significantly lower than the reproducibility requirement of the reference method, 0.5. The reported approach using ATR-FTIR spectroscopy combined with CASAS and PLS is rapid and accurate for determining the ingredients of the cellulose slurries and solutions and offers potential for identifying similar samples.
- 8 June 2021
- Published by Taylor & Francis in Analytical Letters
- Vol. 55 (3) , 388-398
Determination of Cellulose, Water, and N-Methylmorpholine N-Oxide during the Production of Lyocell by Attenuated Total Reflection (ATR) – Fourier Transform Infrared (FTIR) Spectroscopy
Attenuated total reflection (ATR) – Fourier transform infrared (FTIR) spectroscopy with the pretreatment of converting the absorption spectrum into angle spectrum (CASAS) and partial least squares (PLS) was employed to rapidly and simultaneously determine the ingredients of cellulose slurries and solutions during lyocell production. ATR was employed to rapidly and conveniently collect the spectra of the swollen slurries and viscous solutions during production. CASAS was used to eliminate the influence on modeling by the optical path differences due to uneven ingredient distributions in the samples. The PLS models for the cellulose, water, and N-methylmorpholine N-oxide (NMMO) concentrations of the cellulose slurry/solution were constructed using the raw spectra, pretreated spectra by CASAS, and common pretreatment methods, respectively. The results show that CASAS eliminates interferences due to the differences in optical path and improves the performance of PLS model for each analyte. The standard errors of prediction (SEP) of models using CASAS for cellulose, NMMO, and water concentration of cellulose slurry were 0.368, 0.196 and 0.171, while the values for the cellulose solution were 0.442, 0.353, and 0.066, respectively. The standard errors of prediction (SEPs) of CASAS models are significantly lower than the reproducibility requirement of the reference method, 0.5. The reported approach using ATR-FTIR spectroscopy combined with CASAS and PLS is rapid and accurate for determining the ingredients of the cellulose slurries and solutions and offers potential for identifying similar samples.
- 10 February 2021
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 13 (12) , 1479-1488
Quantitative detection of dithiocarbamate pesticides by surface-enhanced Raman spectroscopy combined with an exhaustive peak-seeking method
Surface-enhanced Raman spectroscopy (SERS) based on nanosilver colloid substrates has great potential for rapid detection of pesticide residues because of its advantages of sensitivity, rapidity, simplicity, low cost, etc. However, its poor repeatability and narrow linear quantitative range limit its practical application. In this paper, a silver colloid SERS analysis method combined with an exhaustive peak-seeking method was introduced for quantitative determination of thiram and ziram. This method can establish a linear quantitative relationship in a wide range by use of an own characteristic peak of analysis as an internal standard (IS) which is found via judging the linear correlation between the intensity ratios of two SERS peaks of analytes and the concentrations. Combined with improving the preparation method of silver colloids, adding suitable activators and optimizing the detection process, a liquid detection system with good repeatability and a wide linear quantitation range was obtained. The relative standard deviation (RSD) of the strongest SERS peak is no more than 8.98%, which is better than the general case of the silver colloid SERS substrate. The ratio of I1384/I1148 has a good linear relationship with the concentration of thiram solution, and the 1148 cm−1 characteristic peak was utilized as the IS to establish the standard curve equation for the determination of thiram concentration. The equation is I1384/I1148 = −1.7930 × lg[cthiram (ppm)] + 6.0078 with a linear range of 10−2 to 102 ppm (4.16 × 10−8 to 4.16 × 10−4 mol L−1) and a limit of detection (LOD) of 10−2 ppm. The peak of IS for the determination of ziram concentration is at 938 cm−1, and the equation is I1384/I938 = 4.5531 × lg[cziram (ppm)] + 6.4792 with a linear range of 10−1 to 102 ppm (3.27 × 10−7 to 3.27 × 10−4 mol L−1) and a LOD of 10−4 ppm. Thiram or ziram in apple juice was successfully detected by using this liquid detection system. This analysis system effectively solves the problem of poor repeatability and a narrow linear quantification range in SERS analysis based on silver colloid substrates, and the linear quantification range meets the requirements of the national standard (GB-2763-2019).
- 10 February 2021
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 13 (12) , 1479-1488
Quantitative detection of dithiocarbamate pesticides by surface-enhanced Raman spectroscopy combined with an exhaustive peak-seeking method
Surface-enhanced Raman spectroscopy (SERS) based on nanosilver colloid substrates has great potential for rapid detection of pesticide residues because of its advantages of sensitivity, rapidity, simplicity, low cost, etc. However, its poor repeatability and narrow linear quantitative range limit its practical application. In this paper, a silver colloid SERS analysis method combined with an exhaustive peak-seeking method was introduced for quantitative determination of thiram and ziram. This method can establish a linear quantitative relationship in a wide range by use of an own characteristic peak of analysis as an internal standard (IS) which is found via judging the linear correlation between the intensity ratios of two SERS peaks of analytes and the concentrations. Combined with improving the preparation method of silver colloids, adding suitable activators and optimizing the detection process, a liquid detection system with good repeatability and a wide linear quantitation range was obtained. The relative standard deviation (RSD) of the strongest SERS peak is no more than 8.98%, which is better than the general case of the silver colloid SERS substrate. The ratio of I1384/I1148 has a good linear relationship with the concentration of thiram solution, and the 1148 cm−1 characteristic peak was utilized as the IS to establish the standard curve equation for the determination of thiram concentration. The equation is I1384/I1148 = −1.7930 × lg[cthiram (ppm)] + 6.0078 with a linear range of 10−2 to 102 ppm (4.16 × 10−8 to 4.16 × 10−4 mol L−1) and a limit of detection (LOD) of 10−2 ppm. The peak of IS for the determination of ziram concentration is at 938 cm−1, and the equation is I1384/I938 = 4.5531 × lg[cziram (ppm)] + 6.4792 with a linear range of 10−1 to 102 ppm (3.27 × 10−7 to 3.27 × 10−4 mol L−1) and a LOD of 10−4 ppm. Thiram or ziram in apple juice was successfully detected by using this liquid detection system. This analysis system effectively solves the problem of poor repeatability and a narrow linear quantification range in SERS analysis based on silver colloid substrates, and the linear quantification range meets the requirements of the national standard (GB-2763-2019).
- 5 January 2021
- Published by Elsevier in Vibrational Spectroscopy
- Vol. 113, 103206
Investigation on spectral standardization among multi-channel of an on-line near-infrared spectrometer
- 5 January 2021
- Published by Elsevier in Vibrational Spectroscopy
- Vol. 113, 103206
Investigation on spectral standardization among multi-channel of an on-line near-infrared spectrometer
- 13 February 2020
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 12 (10) , 1311-1315
A novel deep learning-based chemical image identification method of infrared spectroscopy using external perturbation
Perturbation-induced infrared spectroscopy combined with a deep learning-based chemical image identification method enables rapid and non-destructive identification of different samples with a quite similar composition and morphological complexity.
- 13 February 2020
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 12 (10) , 1311-1315
A novel deep learning-based chemical image identification method of infrared spectroscopy using external perturbation
Perturbation-induced infrared spectroscopy combined with a deep learning-based chemical image identification method enables rapid and non-destructive identification of different samples with a quite similar composition and morphological complexity.
- 4 December 2019
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 12 (1) , 112-121
A qualitative recognition method based on Karhunen–Loeve decomposition for near-infrared (NIR) and mid infrared (MIR) spectroscopy analysis
Qualitative recognition is an important research area of NIR and MIR spectroscopy analyses.
- 4 December 2019
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 12 (1) , 112-121
A qualitative recognition method based on Karhunen–Loeve decomposition for near-infrared (NIR) and mid infrared (MIR) spectroscopy analysis
Qualitative recognition is an important research area of NIR and MIR spectroscopy analyses.
- 3 April 2019
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 11 (18) , 2435-2442
Online determination of chemical and physical properties of poly(ethylene vinyl acetate) pellets using a novel method of near-infrared spectroscopy combined with angle transformation
Melt flow rate (MFR) and vinyl acetate content (VAC) are the target parameters for quality control of poly(ethylene vinyl acetate) (EVA) pellets in production processes. Standard test methods for these two properties are too time consuming to meet the actual instant needs in industry. In order to solve this problem, an online near-infrared spectroscopy (NIR) system was integrated in this study to determine the MFR and VAC of EVA pellets in process. When collecting spectra, the difference of sample loading density and optical path will cause noise and decrease the spectral and prediction precision of quantitative models of the MFR and VAC. Although routine spectral pre-treatment methods, including the first order (1st) derivative and multiplicative scatter correction (MSC), are effective in reducing noise, the standard error of prediction (SEP) values of partial least squares (PLS) models of MFR and VAC using the pretreated spectra are still much larger than the reproducibility requirement of standard methods. The key aim of this present study is to explore a novel method based on angle transform to improve the online models. Before establishing models, the original spectra are transformed to angular spectra that reflect the composition information of the sample and are not influenced by the spectral intensity. This method can effectively reduce the severe noise and improve the performance of the models. The SEPs of PLS models of the MFR and VAC using angular spectra are 0.08 and 0.21, respectively. These two values are lower than the reproducibility requirement of standard test methods for the MFR, 0.1 and VAC, 0.3. A new online method of NIR spectroscopy combined with PLS using angular spectra has been successfully established to simultaneously determine the chemical and physical properties of EVA pellets. This research solves a tough technical problem existing in online NIR analysis of pellet samples carried by conveyor belts or pipelines with diffuse reflectance mode.
- 3 April 2019
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 11 (18) , 2435-2442
Online determination of chemical and physical properties of poly(ethylene vinyl acetate) pellets using a novel method of near-infrared spectroscopy combined with angle transformation
Melt flow rate (MFR) and vinyl acetate content (VAC) are the target parameters for quality control of poly(ethylene vinyl acetate) (EVA) pellets in production processes. Standard test methods for these two properties are too time consuming to meet the actual instant needs in industry. In order to solve this problem, an online near-infrared spectroscopy (NIR) system was integrated in this study to determine the MFR and VAC of EVA pellets in process. When collecting spectra, the difference of sample loading density and optical path will cause noise and decrease the spectral and prediction precision of quantitative models of the MFR and VAC. Although routine spectral pre-treatment methods, including the first order (1st) derivative and multiplicative scatter correction (MSC), are effective in reducing noise, the standard error of prediction (SEP) values of partial least squares (PLS) models of MFR and VAC using the pretreated spectra are still much larger than the reproducibility requirement of standard methods. The key aim of this present study is to explore a novel method based on angle transform to improve the online models. Before establishing models, the original spectra are transformed to angular spectra that reflect the composition information of the sample and are not influenced by the spectral intensity. This method can effectively reduce the severe noise and improve the performance of the models. The SEPs of PLS models of the MFR and VAC using angular spectra are 0.08 and 0.21, respectively. These two values are lower than the reproducibility requirement of standard test methods for the MFR, 0.1 and VAC, 0.3. A new online method of NIR spectroscopy combined with PLS using angular spectra has been successfully established to simultaneously determine the chemical and physical properties of EVA pellets. This research solves a tough technical problem existing in online NIR analysis of pellet samples carried by conveyor belts or pipelines with diffuse reflectance mode.
- 1 January 2016
- Vol. 36 (1) , 70-4
[Determination of Alcoholysis Degree and Volatile Matter of Poly-Vinyl Alcohol Using Diffuse-Reflection Near Infrared Spectroscopy].
A new method using reflection NIR technology was developed to determine the alcoholysis degree and volatile matter of Poly-vinyl alcohol (PVA). 120 samples were used in this research. NIR spectra of the sample were scanned by the spectrometer from 1 000 to 1 800 nm. The alcoholysis degree and volatile matter were determined by the national standard method of volumetric and gravimetric method respectivily. Partial least squares (PLS1) was used to establish the quantitative correction model of alcoholysis degree and volatile matter of PVA. The corrected relationship (Rc) of alcoholysis degree and volatile matter was 0.976 and 0.981 respectively. The corrected standard deviation(SEC) was 0.176 and 0.197. The predicted relationship (R(p)) was 0.967 and 0.969. The predicted deviation(SEP) was 0.202 and 0.193. The test for actual samples showed that the NIR method was fitted for the requirement of PVA analysis.
- 1 January 2016
- Vol. 36 (1) , 70-4
[Determination of Alcoholysis Degree and Volatile Matter of Poly-Vinyl Alcohol Using Diffuse-Reflection Near Infrared Spectroscopy].
A new method using reflection NIR technology was developed to determine the alcoholysis degree and volatile matter of Poly-vinyl alcohol (PVA). 120 samples were used in this research. NIR spectra of the sample were scanned by the spectrometer from 1 000 to 1 800 nm. The alcoholysis degree and volatile matter were determined by the national standard method of volumetric and gravimetric method respectivily. Partial least squares (PLS1) was used to establish the quantitative correction model of alcoholysis degree and volatile matter of PVA. The corrected relationship (Rc) of alcoholysis degree and volatile matter was 0.976 and 0.981 respectively. The corrected standard deviation(SEC) was 0.176 and 0.197. The predicted relationship (R(p)) was 0.967 and 0.969. The predicted deviation(SEP) was 0.202 and 0.193. The test for actual samples showed that the NIR method was fitted for the requirement of PVA analysis.
- 1 October 2015
- Vol. 35 (10) , 2757-60
[Rapid Quantitative Analysis of Content of the Additive in Gasoline for Motor Vehicles by Near-Infrared Spectroscopy].
A new rapid quantitative method for the determination of oxygenates and the compounds not included in the national standard in gasoline using near-infrared spectroscopy is raised by this paper. This method combine near-infrared spectroscopy with oblique projection. This experiment choose four different types of gasoline, including reconcile gasoline, FCC refined gasoline, reformed gasoline and desulfurizing gasoline. Prepare series gasoline samples containing different concentrations and different types of compounds. Using FTIR spectrometer to measure those samples and got transmission spectrums. Oblique projection method could separate quantity spectral signal from mixed spectrum signal, and using projection to calculate and analyze the separated signal to obtain the content of measured component. The deviation between this method and the real content is low, the absolute error is less than 0.8 and the relative error is less than 8%. For the actual gasoline samples, compare results of this method with gas chromatography, the absolute error are less than 0.85 and the relative error are less than 6.85%. This method solves the problem of general multivariate calibration methods. It is very significant for the development of rapid detection technology using NIR suitable for on-site and the improvement of the quality of gasoline.
- 1 October 2015
- Vol. 35 (10) , 2757-60
[Rapid Quantitative Analysis of Content of the Additive in Gasoline for Motor Vehicles by Near-Infrared Spectroscopy].
A new rapid quantitative method for the determination of oxygenates and the compounds not included in the national standard in gasoline using near-infrared spectroscopy is raised by this paper. This method combine near-infrared spectroscopy with oblique projection. This experiment choose four different types of gasoline, including reconcile gasoline, FCC refined gasoline, reformed gasoline and desulfurizing gasoline. Prepare series gasoline samples containing different concentrations and different types of compounds. Using FTIR spectrometer to measure those samples and got transmission spectrums. Oblique projection method could separate quantity spectral signal from mixed spectrum signal, and using projection to calculate and analyze the separated signal to obtain the content of measured component. The deviation between this method and the real content is low, the absolute error is less than 0.8 and the relative error is less than 8%. For the actual gasoline samples, compare results of this method with gas chromatography, the absolute error are less than 0.85 and the relative error are less than 6.85%. This method solves the problem of general multivariate calibration methods. It is very significant for the development of rapid detection technology using NIR suitable for on-site and the improvement of the quality of gasoline.
- 1 July 2015
- Vol. 35 (7) , 1879-84
[Research on Rapid Discrimination of Edible Oil by ATR Infrared Spectroscopy].
A rapid discrimination method of edible oils, KL-BP model, was proposed by attenuated total reflectance infrared spectroscopy. The model extracts the characteristic of classification from source data by KL and reduces data dimension at the same time. Then the neural network model is constructed by the new data which as the input of the model. 84 edible oil samples which include sesame oil, corn oil, canola oil, blend oil, sunflower oil, peanut oil, olive oil, soybean oil and tea seed oil, were collected and their infrared spectra determined using an ATR FT-IR spectrometer. In order to compare the method performance, principal component analysis (PCA) direct-classification model, KL direct-classification model, PLS-DA model, PCA-BP model and KL-BP model are constructed in this paper. The results show that the recognition rates of PCA, PCA-BP, KL, PLS-DA and KL-BP are 59.1%, 68.2%, 77.3%, 77.3% and 90.9% for discriminating the 9 kinds of edible oils, respectively. KL extracts the eigenvector which make the distance between different class and distance of every class ratio is the largest. So the method can get much more classify information than PCA. BP neural network can effectively enhance the classification ability and accuracy. Taking full of the advantages of KL in extracting more category information in dimension reducing and the features of BP neural network in self-learning, adaptive, nonlinear, the KL-BP method has the best classification ability and recognition accuracy and great importance for rapidly recognizing edible oil in practice.
- 1 July 2015
- Vol. 35 (7) , 1879-84
[Research on Rapid Discrimination of Edible Oil by ATR Infrared Spectroscopy].
A rapid discrimination method of edible oils, KL-BP model, was proposed by attenuated total reflectance infrared spectroscopy. The model extracts the characteristic of classification from source data by KL and reduces data dimension at the same time. Then the neural network model is constructed by the new data which as the input of the model. 84 edible oil samples which include sesame oil, corn oil, canola oil, blend oil, sunflower oil, peanut oil, olive oil, soybean oil and tea seed oil, were collected and their infrared spectra determined using an ATR FT-IR spectrometer. In order to compare the method performance, principal component analysis (PCA) direct-classification model, KL direct-classification model, PLS-DA model, PCA-BP model and KL-BP model are constructed in this paper. The results show that the recognition rates of PCA, PCA-BP, KL, PLS-DA and KL-BP are 59.1%, 68.2%, 77.3%, 77.3% and 90.9% for discriminating the 9 kinds of edible oils, respectively. KL extracts the eigenvector which make the distance between different class and distance of every class ratio is the largest. So the method can get much more classify information than PCA. BP neural network can effectively enhance the classification ability and accuracy. Taking full of the advantages of KL in extracting more category information in dimension reducing and the features of BP neural network in self-learning, adaptive, nonlinear, the KL-BP method has the best classification ability and recognition accuracy and great importance for rapidly recognizing edible oil in practice.