Abstract: To monitor in real-time changes in the contents of multiple components during the counter-current extraction of curcumin, near-infrared spectra of samples collected at different times were acquired, preprocessed and optimized by variable selection methods. The contents of the target components in the samples were determined by high performance liquid chromatography (HPLC) and used as reference values. The key parameters of the support vector regression (SVR) model were optimized by hybrid strategy improved dung beetle optimizer (IDBO), artificial hummingbird algorithm (AHA), or whale optimization algorithm (WOA). The results indicated that the IDBO-SVR model performed best, with coefficients of determination of prediction (R2p) exceeding 0.93. It enabled simultaneous monitoring of changes in the contents of multiple components within 10 seconds, faster than traditional HPLC (40 minutes per sample). This model facilitates visual monitoring of curcumin counter-current extraction, providing technical support for real-time process control to enhance extraction efficiency and product quality stability.

Key words: near-infrared spectroscopy; support vector regression; countercurrent extraction; curcumin; hybrid strategy improved dung beetle optimizer