Robust online identification method for biofabrication processes with multiple unknown disturbances

Abstract

This paper addresses the challenge of online parameter identification for biofabrication processes with multiple sensors, particularly under unknown disturbances. A robust recursive multitask expectation maximization (RMTEM) algorithm is proposed within Bayesian framework. The algorithm integrates data of multi-sensor to recursively estimate both unknown noise variances and system parameters, ensuring adaptability to plug-and-play sensors and real-time applications. By leveraging information from heterogeneous noise sources, the RMTEM algorithm exhibits enhanced robustness and adaptability to fluctuating disturbances. Numerical simulations demonstrate its superior identification accuracy compared to existing methods, while a continuous fermenter case further validates its effectiveness and practical relevance in complex biofabrication scenarios.