[Extraction Optimization of Rhizome of Curcuma longa by Response Surface Methodology and Support Vector Regression]

To optimize the optimal microwave-assisted extraction method of curcuminoids from Curcuma longa.

On the base of single factor experiment, the ethanol concentration, the ratio of liquid to solid and the microwave time were selected for further optimization. Support Vector Regression (SVR) and Central Composite Design-Response Surface Methodology (CCD) algorithm were utilized to design and establish models respectively, while Particle Swarm Optimization (PSO) was introduced to optimize the parameters of SVR models and to search optimal points of models. The evaluation indicator, the sum of curcumin, demethoxycurcumin and bisdemethoxycurcumin by HPLC, were used.

The optimal parameters of microwave-assisted extraction were as follows: ethanol concentration of 69%, ratio of liquid to solid of 21 : 1, microwave time of 55 s. On those conditions, the sum of three curcuminoids was 28.97 mg/g (per gram of rhizomes powder).

Both the CCD model and the SVR model were credible, for they have predicted the similar process condition and the deviation of yield were less than 1.2%.