A versatile software tool making best use of sparse data for closed loop process control

Abstract

This paper presents the design of a software supported sliding mode controller for a biochemical process. The state of the process is characterized by cell mass and nutrient amount. The controller is designed for tracking of a desired profile in cell mass and it is shown that the nutrient amount in the controlled bioreactor evolves bounded. A smart software tool named Support Vector Machine (SVM), which minimizes the upper bound of an empirical risk function, is proposed to approximate the nonlinear function seen in the control law by using very limited number of numerical data. This removes the necessity of knowing the functional form of the nominal nonlinearity in the control law. It is shown that the controller is robust against noisy measurements, considerable amount of parameter variations, discontinuities in the command signal and large initial errors. The contribution of the present work is the achievement of robustness and tracking performance on a benchmarking process, under the presence of limited prior knowledge.