Challenges and Opportunities on Nonlinear State Estimation of Chemical and Biochemical Processes

Abstract

This paper provides an overview of nonlinear state estimation techniques along with a discussion on the challenges and opportunities for future work in the field. Emphasis is given on Bayesian methods such as moving horizon estimation (MHE) and extended Kalman filter (EKF). A discussion on Bayesian, deterministic, and hybrid methods is provided and examples of each of these methods are listed. An approach for nonlinear state estimation design is included to guide the selection of the nonlinear estimator by the user/practitioner. Some of the current challenges in the field are discussed involving covariance estimation, uncertainty quantification, time-scale multiplicity, bioprocess monitoring, and online implementation. A case study in which MHE and EKF are applied to a batch reactor system is addressed to highlight the challenges of these technologies in terms of performance and computational time. This case study is followed by some possible opportunities for state estimation in the future including the incorporation of more efficient optimization techniques and development of heuristics to streamline the further adoption of MHE.

Model predictive control and moving horizon estimation for adaptive optimal bolus feeding in high-throughput cultivation of \textit{E. coli}

“…State and parameter estimation methods can be classified into recursive methods and optimization-based methods, namely Kalman filter (KF) and MHE, respectively (Alexander et al, 2020). Most of the estimators follow the assumption that online measurements are available for every time instance and desired state variable.…”

Cascade PID Control Systems Based on DNA Strand Displacement With Application in Polarization of Tumor-Associated Macrophages

“…Due to some natural properties (e.g., variability and nonlinearity) [1], biochemical systems have difficulty in gaining a predetermined behavior, which hinders the further application of synthetic biology [2] in various fields. In order to obtain a stable and robust biochemical system, it is essential to build control systems in the form of biomolecular circuits [3], [4].…”

Seeded Cooling Crystallization Process Optimization of β Form l-Glutamic Acid Based on Variable Moving Horizon State Estimation

“…Based on a comparative study on the estimation performance of different nonlinear observers for an evaporative crystallization process, it was confirmed that UKF could provide good state estimation for implementing satisfactory closed-loop control performance . To accommodate for the operational constraints involved with nonlinear process dynamics, e.g., the supersaturation level and cooling rate during crystallization, a moving horizon estimator (MHE) was studied in refs , , which is an optimization-based estimator that minimizes the sum of squared errors between past measurements and state predictions over an N -length sliding window, weighted by the covariance matrices associated with the state and measurement uncertainties. , Mangold et al designed an MHE based on a population balance model (PBM) of barium sulfate precipitation to estimate the evolution of particle size distribution, thus improving batch control of product quality. Moreover, MHE was efficiently used for estimating batch-to-batch parametric drift, so as to conduct run-to-run process optimization .…”