Identificação de sistema em uma unidade de compressão de cloro.

Abstract

The influence of impurities is often mentioned as a cause o f accidents in chemical industry. Worldwide oil and chemical industries have been involved in an accident whose consequences economic, environmental and human often exceed the limits of its facilities. In the production industry CI2 (chlorine) due to the presence of ammonia (NH3) in salt, giving rise to trichloramine (NCI3), a compound that is extremely unstable and explosive under certain conditions of pressure and temperature. The automatic control of the composition of NCI3 is the form that allows operation of the process safely. In this context, we approached the process of producing chlorine by electrolysis of brine. The stage held until then to present the seminar, included an identification system by obtaining mathematical models. Mathematical modeling is the area of knowledge that studies the ways of representation the real systems, where the model is only an approximation o f some characteristics of the real system. Thus this work aims to make a System Identification chlorine production, and propose a control strategy, developed in Simulink / Matlab to process variables that need to control. Therefore, this Dissertation deals with the identification system of the plant model Chlorine Compression of Braskem in Maceio, where this system was simulated by Brito (2009) in Aspen Plus and Dynamics environment, and provided for the development of this work. The necessity of adding controllers in certain process variables, due to the need to keep them operating at certain benchmarks established for the better functioning of the plant. Excitation signals of type Step and PRBS (pseudo-random binary signal) were applied at a flow rate of chlorine gas coming from the electrolytic cells as well as the flow of liquid chlorine that entered the pre cooler, and the thermal load o f the reboiler. The open loop tests were performed in order to understand and draw conclusions about the transient behavior of the variables under study. The Transfer Function models, ARX (autoregressive with exogenous inputs) and state space, were identified as those that best represent the transient behavior of the system under study. Since the proposed control loop at the end of this study, includes a justification to implement a particular control variable under study.