The use of predictive models to optimize sugar recovery obtained after the steam pre‐treatment of softwoods

Acid catalyzed steam pre-treatment is recognized as an effective method for defibrillating plant cell walls while providing good hemicellulose sugar recovery and a more readily accessible cellulosic component for subsequent enzymatic hydrolysis. However, much of the past work to try to optimize the overall sugar recovery after pre-treatment and enzymatic hydrolysis has been limited to more qualitative comparisons which offer little insight into the steam pre-treatment process itself. Better prediction of sugar recoveries from steam pre-treated biomass will likely prove to be invaluable in helping us design more effective steam pre-treatment reactors. The work discussed here attempted to determine which of three options – namely the severity factor Ro, the combined severity factor CS, and response surface methodology RSM – was best suited for the development of predictive empirical equations that would be used to optimize the acid catalyzed steam pre-treatment of softwood chips of industrially relevant size and provide maximum soluble sugar recovery. It was apparent that the combined severity factor CS resulted in predictions that were slightly more accurate than those of the severity factor Ro, and that RSM and the severity factor Ro possessed similar predictive capability. A comparison of several RSM models that were used to evaluate the SO2 catalyzed steam pre-treatment of softwood indicated that a hybrid design, when used in conjunction with a narrow process space, provided the most robust model. The applicability of an RSM model which was developed when pretreating radiata pine was assessed against pre-treated lodgepole pine and was found to provide good predictability.