Tackling correlated responses during process optimisation of rapeseed meal protein extraction

• Multi-response optimisation with correlated responses is discussed.
• New hybrid statistical method (RSM-PCA-desirability function) used.
• Use of PCA improved models’ prediction accuracy, better than RSM alone.
• Original responses can be predicted from PCs using Multiple Linear Regression.
• Experimental confirmation was carried out at predicted optimal condition.

Setting of process variables to meet the required specifications of quality characteristics is a crucial task in the extraction technology or process quality control. Simultaneous optimisation of several conflicting characteristics poses a problem, especially when correlation exists. To remedy this shortfall, we present multi-response optimisation based on Response Surface Methodology (RSM)-Principal Component Analysis (PCA)-desirability function approach, combined with Multiple Linear Regression (MLR). Experimental manifestation of the proposed methodology was executed using a multi-responses-based protein extraction process from an industrial waste, rapeseed press-cake. The proposed optimal factor combination reflects a compromise between the partially conflicting natures of the original responses. Prediction accuracy of this new hybrid method was found to be better than RSM alone, verifying the adequacy and superiority of the said approach. Furthermore, this study suggests the feasibility of the exploitation of the waste rapeseed oil-cake for extraction of valuable protein, with improved colour properties using simple, viable process.