Limitations of Joint and Dual Nonlinear Kalman Estimators in Low-Cost Bioprocess Monitoring

The biopharmaceutical industry constantly presses for fast and low-cost bioprocess monitoring strategies. However, a recent study has shown that the Joint Extended Kalman Filter (JEKF) is inefficient in this monitoring type under biomanufacturing conditions. This work investigates the Dual Extended Kalman Filter (DEKF), Joint Unscented Kalman Filter (JUKF), and Joint Cubature Kalman Filter (JCKF) under these challenging conditions. Our theoretical analysis also reveals inefficiencies in DEKF, while our empirical tests using a synthetic dataset indicate that JUKF and JCKF only perform well with specific initial conditions for the state error covariance matrix, although with unconventional Kalman gain behavior. These results suggest nonlinear Kalman estimators in biomanufacturing still merit further investigation.