Analysis and control of large scale distributed networks with competing autocatalators

Distributed systems offer robust manufacturing environments as they have ultimate flexibility to reconfigure in response to external disturbances and demands. They are hard to analyze and control because of the complexity of interactions between simultaneously running sub-systems. The type of distributed system of interest is the paradigm of autocatalytic replicators populating in a Continuous Stirred Tank Reactor (CSTR) environment. The autocatalytic reaction scheme is an abstract scheme which has been extensively applied to study the behavior of real systems. When autocatalytic reactions are carried out in an interconnected network of CSTRs, there is an observed geometric proliferation of steady states stemming from combinations of multiple steady states of a CSTR in isolated operating conditions. The system displays additional features that are not possible when the components are considered on their own. The thesis focuses on developing various methods to analyze and control a distributed system of autocatalytic replicators.;The behavior of system is investigated using bifurcation diagrams, continuation studies and dynamics. It is shown that heterogeneous networks characterized by unequal reactor parameters shows an expanded phase space, stability characteristics and improved coexistence in case of multiple species. The results of the study highlight the importance of symmetry breaking in facilitating evolution, by limiting the dominance of the overbearing single CSTR steady states. Tracer experiments are another well established method for analysis of flow systems. Different performances measures are designed based on the distribution of tracer in different network topologies. These measures are valuable to assess the performance of non-linear systems. Control of distributed systems is challenging because of the non-linearity of reactions present and the vast number of controlled and manipulated variables. Agent based approaches offer a way to decompose a complex system wherein lower level agents performed local well-defined tasks and higher-level agents perform global tasks. An intelligent agent-based system is implemented to detect and control an external impurity in the network. The underlying mechanisms for agent knowledge are formulated by rigorous understanding of the process behaviors for varying parameters.