Contribuição à modelagem de sistemas de controle via rede.

Abstract

Networked control systems (NCSs) are ones where control loops are closed by means of a commumication network. These systems can be found in many complex control applications such as the automotive, aircraft, robotics, manufacturing industry among others. The use of network architectures enable greater reliability and flexibility of operation, reducing the amount of cabling and costs, ease of installation, maintenance and system diagnostics. However, network architectures introduce new challenges that make the design and implementation of control systems a complex activity. Basically, communication networks can cause varying delays and loss of messages during the operating cycles of the control loops as well as others particuiarities that can degrade performance and even cause the instability of the physical processes. In the last two decades, several studies have been developed in an attempt to overcome these challenges. However, most of these works. the design methodologies are complex and generally the applicability is questionable due to unrealistic assumptions and especially the large computational effort required. Furthermore, many of these methodologies apply only in simpler architectures and is also eomrnon to consider NCSs with a single control loop, which is not consistem with reality applications. On the other hand, some studies have allowed the extension of control techniques simpler and already established in practice in the design of NCSs. However, some behaviors eaused by network architecture are still disregarded, which in practice compromise the functioning of the system. This work presents a procedure to design networked control systems (NCSs) using CAN Network. considering the integration of related theories, in order to check and mitigate the effecrs eaused by the network architectures in the performance control loops. Firstly, were developed four models of the CAN network in timed automata and priced timed automata considering different neíwork behavior characteristics. Two of those models have been extended to model CAN NCSs and determine the temporal behavior of the control loops. Then, the procedure was developed, that consists of three activities: to design PIDs controllers for processes with one input and one output, considering variants control delays; to develop a framework using formal modeling in timed automata and priced timed automata, and approaches classical and statistical model-cheking to determine the temporal behavior of the control loops in order to verify the conditions under which the time requirements will be satisfied, and develop an experimental plataform to simulate in real-time the CAN NCSs using technique hardware in the loop, including as real parts a CAN network and a random fault injector. A case study to a CAN NCS with five control loops was realized and the impact eaused by network in performance control loopd was checked.