Abstract

Direct monitoring of bed conditions in large scale solid-state fermentation (SSF) bioreactors is difficult due to lack of reliable and affordable instrumentation. Although, relevant variables such as average bed temperature and water content can be inferred from energy and water balances, these estimations are prone to significant error since the process is complex and measured variables are extremely noisy. Hence, to obtain better estimation of bed conditions, simple and robust models, and effective noise suppression methods should be developed. A kinetic model was applied to obtain noise free CO2 production rate (CPR) values, which in turn was used in water and energy balances. The evolution of other relevant variables including total and active biomass, O2 consumption rate, secondary metabolite production and dry mass degradation were also calculated. The proposed model was calibrated with data obtained in an agitated aseptic SSF pilot-scale bioreactor of 200 kg, growing Gibberella fujikuroi on wheat bran and starch. The calibration was achieved by means of a non-linear optimization routine. With this model better average bed temperature estimates, compared with a model that uses noisy (CPR) measurements, were attained. Hence, the proposed model can be used on-line to estimate more reliable average bed conditions and to develop model based control strategies. In addition, a predictive model for bioreactor design can be derived from the proposed model, after good outlet gas temperature estimation has been achieved. © 2003 Elsevier Ltd. All rights reserved.