Distributed Real-Time Control Systems using CAN

Searching for more efficient ways to improve process quality and increase plant throughput has encouraged changes in the control system architecture toward greater distribution. A distributed control system, illustrated in Fig.15.1, can be defined as control systems in which functions around a plant are distributed into computing nodes (i.e., control computers and smart devices) that are physically separated; and all the computing nodes are interconnected by digital communication networks (e.g., fieldbuses). Distributed control systems are now found in many industrial fields, such as oil and gas production (Hars, 1994), automobiles (Hansson et al., 1997), and gas turbine engine control (Shaffer, 1999). Advances in distributed control system technology have been driven by the ever increasing performance of smart field devices and fieldbuses using smaller, less expensive and more functional microprocessors. In fact, the goal of a distributed control system is the same as its centralised counterpart; that is, to get input from the plant through its sensors and give output using its actuators. However, the use of field-buses and smart devices introduces many advantages into a distributed control system over a centralised one. Not only does a distributed architecture offer reduced wiring and simplified maintenance, it also provides the opportunity to implement different control strategies. Consequently, the emergence of distributed control systems is promoting changes in control systems design issues. One of the key changes from conventional control techniques is driven by the local processing power. In distributed architectures, smart devices employing a built-in microprocessor with a fieldbus interface offer the possibility of the faster and more reliable self-diagnosis and self-compensation. Several researchers (Henry 1993; Koscielny et al 1997) have discussed the benefits of local fault-diagnosis. Extra information about the element status reported through a fieldbus network is also very beneficial for reconfigurable control. For example, Lee et al. (2001) presented an on-line retuning strategy for PI control to tolerate actuator faults using this information.