Estimation and control of non-linear variables in a continuous fermentation process using sliding mode techniques

Abstract

Biomass, substrate or metabolite concentrations are difficult to measure online in fermentation processes because of the lack of reliable, cheap and sterilizable transducers. Currently, many of the measurements required may be determined through offline analysis, which is costly and time consuming. Furthermore, the specific growth rate conditions involved in the fermentation are typically non-linear and uncertain. In this paper, a new variable, the substrate consumption rate, consisting of a combination of substrate concentration, biomass concentration, specific growth rate and yield production coefficient, is introduced to overcome these problems and simplify the non-linear differential equations of the fermentation process. A sliding mode observer, which only requires measurement of substrate concentration, is then developed to solve the estimation problem, providing a soft sensor to estimate the substrate consumption rate. It is shown that the sliding mode exhibited by the corresponding observer error dynamics is exponentially stable. This parameterization and the resulting estimate of biomass concentration are then utilized within a feedback control strategy. Non-linear simulation results in the presence of both parameter uncertainties and external disturbances illustrate the approach.

A Discrete-Time Sliding Mode Controller for the Fermentation Process

“…The fermentation process (FP) has a wide range of applications, from the production and storage of food [1] to the production of pharmaceutical products and wastewater treatment. The FP control design is challenging.…”

“…The FP control design is challenging. The main obstacle to precise control is the presence of microorganisms [2], as well as the dynamics of FP, which is often incomprehensible, strongly non-linear, and non-stationary [1]. The model parameters vary over a prolonged period due to metabolic variations and physiological modifications [3].…”

Soft-sensor modeling for l-lysine fermentation process based on hybrid ICS-MLSSVM

“…Measurement of these key variables is necessary to control and optimize the fermentation process in real-time to enhance the productivity. However, it is hard to measure cell concentration, substrate concentration, product concentration during fermentation process in real-time due to the highly time-varying, non-linear and uncertain nature of the fermentation process 5 – 7 .…”

An efficient sliding mode observer-based model predictive control: experimental implementation on a DC/DC power converter

“…A SMO was developed to solve the estimation problem by providing a soft sensor to estimate the substrate consumption rate. In addition, it was shown that the observer error dynamics was exponentially stable in Rahman et al (2012). A comparison of observation techniques such as the unscented Kalman filter (UKF), extended Kalman filter (EKF), and a new nonlinear version of SMO were presented for state estimation in Afshar et al (2015).…”