Quantitative Evaluation of Blocking-Recovery Discrete Modular Agents
Palomera Palacios, Francisco
MetadataShow full item record
Modern manufacturing production environments, made up of a set of independent machines, execute high production plans trying to develop the full potential of the machines, but still keeping a safety operation. The synthesis of control entities (supervisors, controllers, agents) requieres a quantitative evaluation rather than just a qualitative one in order to measure its performance. Quantitative measurements of evaluation are defined in terms of production parameters or production conditions to achieve a desired production plan. Since control entities rule the processing machines for a high production system, an undesirable condition named blocking (dead-lock) may occur. If so, instead of rejecting such behavior, that at first sight seems to be problematic, we should evaluate how convenient is to use a blocking recovery procedure (use of an un-blocking mechanism) and move to a free-blocking production condition in order to re-initialiaze and resume a production process. This thesis work provides three complementary methods to deal with blocking- occurrence and evaluates, quantitatively, its blocking-recovery. To illustrate and show the proposed methods, we control a manufacturing transfer line that includes re-processing work, which feeds-back the system. The methods are developed to evaluate the performance of the transfer line in transient and steady-state analysis. The first method helps to cope up with dead-lock occurrence by synthesizing a set of blockingrecovery discrete-modular agents (BRDMA) along with an un-blocking mechanism (UBM). The second method provides a procedure to compute the time to block, tb, as a quantitative evaluation metric of the controlled transfer line, following two different approaches. The third method proposes a way to evaluate quantitatively and make a comparison of the profit rate between the best blocking recovery policy and its counter part of the free-blocking solution, both in steady-state. To show our experimental results, we provide state-graphs, plots and numerical results to synthesize and evaluate quantitatively the performance of the control entities named as Blocking-Recovery Discrete Modular Agents (BRDMA), under transient and steady-state analysis. Theory from discrete-event systems, agents and multiagents, discrete events and supervisory control theory, automated manufacturing systems and continuous time Markov chains provide the support of this thesis work.