Image Processing Reference
In-Depth Information
23.5.1 Networks for Control
Control signals can be divided into two categories: real-time and event-based. For real-time con-
trol, signals must be received within a specified amount of time for correct operation of the system.
Examples of real-time signals include continuous feedback values (e.g., position, velocity, acceler-
ation) for servo systems, temperature and flow in a process system, and limit switches and optical
sensors in material flow applications (e.g., conveyors). To support real-time control, networks often
must have a high level of determinism, i.e., they must be able to guarantee end-to-end communica-
tion of a signal within a specified amount of time. Further, QoS of NCSs can be very dependent upon
the amount of jitter in the network; thus, for example, fixed determinism is usually preferred over
bounded determinism.
Event-basedcontrolsignalsareusedbythecontrollertomakedecisions,butdonothaveatime
deadline. The system will wait until the signal is received (or a timeout is reached) and then the
decision is made. An example of an event-based signal is the completion of a machining operating
in a CNC; the part can stay in the machine without any harm to the system until a command is sent
to the material handler to retrieve it.
In addition to dividing control signals by their time requirements, the data size that must be trans-
mitted is important. Some control signals are a single bit (e.g., a limit switch) whereas others are very
large (e.g., machine vision). Generally speaking, however, and especially with real-time control, data
sizes on control networks tend to be relatively small and high levels of determinism are preferred.
Control networks in a factory are typically divided into multiple levels to correspond to the fac-
tory control distributed in a multitier hierarchical fashion. At the lowest level of networked control
are device networks, which are usually characterized by smaller numbers of nodes (e.g., less than
), communicating small data packets at high sample frequencies and with a higher level of deter-
minism. An example of networked control at this level is servo control; here network delay and jitter
requirements are very strict. Deterministic networks that support small data packet transmissions,
such as CAN-based networks, are very common at this level. Although seemingly nonoptimal for
this level of control, Ethernet is becoming more common, due to the desire to push Ethernet to all
levels in the factory and the increasing determinism possible with switched Ethernet. Regardless of
the network type, determinism and jitter capabilities for lower-level networked control are enhanced
oftentimes by utilizing techniques that minimize the potential for jitter through network contention,
such as MS operation, polling techniques, and deadbanding [].
An intermediate level of network is the cell or subsystem, which includes SCADA. At this level,
multiple controllers are connected to the network (instead of devices directly connected to the net-
work). he controllers exchange both information and control signals, but as the cells or subsystems
are typically physically decoupled, the timing requirements are not as strict as they are at the lowest
levels, or nonexistent if event-driven control is enforced []. [].These networks are also used to down-
load new part programs and updates to the lower-level controllers. TP and Ethernet-based networks
are commonly used at this level, with ability to communicate larger amounts of data and support for
network services generally taking precedence over strict determinism.
Networks at the factory or enterprise level coordinate multiple cells and link the factory-floor con-
trol infrastructure to the enterprise level systems (e.g., part ordering, supply chain integration, etc.)
Large amounts of data travel over these networks, but the real-time requirements are usually nonex-
istent. Ethernet is the most popular choice here primarily because Internet support at this level is
usuallycritical,andEthernetalsobringsattractivefeaturestothisenvironmentsuchassupportfor
high data volumes, network services, availability of tools, capability for wide area distribution, and
low cost.
Currently, wireless networks are rarely utilized for control. When they are utilized, it is often
because a wired system is impossible or impractical (e.g., a stationary host controller for an
autonomous guided vehicle [AGV]). Wireless systems are more applicable to discrete control; issues
 
Search WWH ::




Custom Search