Soft sensors play a crucial role in monitoring bioprocesses by providing real-time data on variables that are otherwise challenging to measure directly. In a recent study by Dominik Geier et al., an innovative approach was introduced to develop an adaptive soft sensor that reliably predicts biomass concentration in bioprocesses, especially in the presence of faulty sensor inputs. The research aimed to address the inherent challenges associated with sensor faults in bioreactor systems, focusing on fault-tolerant biomass monitoring using an ensemble-based algorithm.
The study involved the development of three independent soft sensor submodels based on different model inputs, including base addition, CO2 production, and mid-infrared spectrum readings. These submodels were then combined using an ensemble-based algorithm to create an adaptive soft sensor capable of fault-tolerant prediction. By assigning weights based on the reliability of each submodel, the ensemble model demonstrated high robustness and accuracy in predicting biomass concentration, even under simulated and real sensor fault conditions.
Key takeaways:
– Soft sensors are essential for real-time monitoring of bioprocesses, allowing indirect determination of critical variables.
– The ensemble-based adaptive soft sensor model developed in the study effectively predicted biomass concentration in bioprocesses despite sensor faults.
– By combining three independent submodels using a fault-tolerant ensemble approach, the model showed resilience to single and multiple fault scenarios.
– The study’s findings highlight the potential of ensemble-based algorithms for improving fault tolerance in biomass monitoring, a critical aspect of bioprocess optimization.
Tags: bioreactor, bioprocess, fed batch
Read more on pmc.ncbi.nlm.nih.gov