Application of adaptive chaotic dung beetle optimization algorithm to near-infrared spectral model transfer

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Transferring Near-Infrared (NIR) spectroscopy calibration models across different instruments is challenging due to spectral variations.
• Accurate determination of hexamethylenetetramine concentration is crucial in various applications.
• Existing model transfer methods have limitations in accuracy, robustness, and sample requirements.

Purpose of the Study:

• To develop a novel and robust approach for sharing NIR calibration models for hexamethylenetetramine analysis.
• To enhance the accuracy and efficiency of model transfer across diverse NIR spectrometers.
• To overcome the limitations of conventional model transfer techniques.

Main Methods:

• A new transfer approach combining Savitzky-Golay first derivative (S_G_1) and orthogonal signal correction (OSC) preprocessing.
• Feature variable optimization using an adaptive chaotic dung beetle optimization (ACDBO) algorithm with tent chaotic mapping and nonlinear decreasing strategy.
• Partial Least Squares (PLS) regression for model building and validation using CEC-2017 benchmark functions.

Main Results:

• The ACDBO algorithm demonstrated superior convergence, accuracy, and stability compared to traditional optimization methods.
• The proposed transfer strategy achieved excellent calibration and validation metrics (R_c²=0.99999, RMSEC=0.00195%, R_v²=0.99643, RMSEV=0.03818%, RPD=16.72574).
• The novel method yielded competitive prediction performance (R_p²=0.96228, RMSEP=0.12462%, RER=17.62331), comparable to direct standardization (DS) and piecewise direct standardization (PDS).

Conclusions:

• The developed transfer approach effectively minimizes spectral discrepancies across different NIR instruments.
• This method significantly enhances the accuracy and robustness of hexamethylenetetramine concentration predictions.
• The approach broadens applicability by eliminating the need for analyzing identical samples across instruments, making it highly beneficial for specific measurement sample transfers.