Local intelligent control in biological systems and industrial processes

Regulatory mechanisms in large scale systems typically involve a hierarchy of regulatory systems, be it in a biological organism or in an industrial process. This thesis focuses on two aspects of distributed or multilevel hierarchical control: (1) Analysis of the “local function” of each controller in the hierarchy. (2) Design of distributed controller network that provides required “global function”.; In this thesis, the first aspect is explored to delineate the previously unknown functional role of a local cardiac reflex in cardiovascular regulation in the rat. The local cardiac reflex hypothesis is that (i) small intensely fluorescent interneurons receive cardiac sensory inputs and then project to Principal Neurons (PNs); and (ii) selected PNs in specific locations and subserving specific functions receive discrete and extremely dense vagal input from the dorsal motor nucleus of the vagus and nucleus ambiguus. A nonlinear mathematical model for the local reflex has been developed from the anatomical experimental results and physiological data in the literature. Simulation analysis of the coherence between vagal input and arterial pressure indicates that robust nonlinear attenuation is an underlying principle in the local cardiac reflex function. Based on the parametric sensitivity studies, it is proposed that the robust modulation of specific phase-related characteristics of the cardiac cycle is the underlying mechanism of the nonlinear compensation.; In the chemical process industry, the second aspect of hierarchical control mentioned above is critical. The algorithmic scalability and the computational load on the central processor are key issues that render the centralized approach impractical for plantwide process systems. In contrast, a distributed control system can be formulated such that the resulting coordinating control network provides the specified “global function”, while overcoming the disadvantages of the centralized control. In this thesis, a scalable distributed state estimation and control algorithm has been developed for multi-rate plantwide systems. This methodology has been demonstrated in two separate case studies involving a simulated large scale industrial reaction separation system and a pulp mill process. The issues involved in the model decomposition for employing the distributed control algorithm are examined. The distributed algorithm is found to be scalable for application to plantwide processes.