Page 67 - Chemical Process Equipment - Selection and Design

3

PROCESS CONTROL

I! processes are subject to disturbances that tend to of the corrective action is provided by feed forward and 10%
change operating conditions, compositions, and by feedback with the result that the integrated error is
physical properties of the streams. In order to reduced by a factor of 10.
minimize the ill effects that could result from such A major feature of many modern control systems is
disturbances, chemical plants are implemented with composition control which has become possible with the
substantial amounts of instrumentation and automatic control development of fast and accurate on-line analyzers. Figure 3.2
equipment. In criti,cal cases and in especially large plants, shows that 10 analyzers are used for control of ethylene
moreover, the instrumentation is computer monitored for cornposition in this plant within the purities shown. High
convenience, safety, and optimization. speed on-line gas chromatographs have analysis times of
For example, a typical billion lb/yr ethylene plant may 30- 120 sec and are capable of measuring several
have 600 control loops with control valves and 400 interacting components simultaneously with a sensitivity in the
loops with a ccst of about $6 million. (Skrokov, 1980, pp. 13, part.s/million range. Mass spectrometers are faster, more
49; see Sec. 3.7); the computer implementation of this stable, and easier to maintain but are not sensitive in the ppm
control system wi~Y cost another $3 million. Figure 3.1 shows range. Any one instrument can be hooked up to a half-dozen
the control system of an ethylene fractionator which has 12 or so sample ports, but, of course, at the expense of time lag
input signals to the cornputer and four outgoing reset signals for controller response. Infrared and NMR spectrometers also
to flow controllers. are feasible for on-line analysis. Less costly but also less
In order for a ,process to be controllable by machine, it specific analyzers are available for measuring physical
must represented by a mathematical model. Ideally, each properties such as refractive index and others that have been
eiement of a dynamic process, for example, a reflux drum or calibrated against mixture composition or product purity.
an individual tray of a fractionator, is represented by The development of a mathematical model, even a
differential equations based on material and energy balances, simplified one that is feasible for control purposes, takes a
transfer rates, stage efficiencies, phase equilibrium relations, major effort and is well beyond the scope of the brief
etc., as well as the pammeters of sensing devices, control treatment of process control that can be attempted here.
valves, and control instruments. The process as a whole then What will be given is examples of control loops for the
is equivalent to a system of ordinary and partial differential common kinds of equipment and operations. Primarily these
equations involving certain independent and dependent are feedback arrangements, but, as mentioned earlier,
variables. When the values of the independent variables are feedback devices usually are necessary supplements in
specified or measured, corresponding values of the others are primarily feedforward situations.
found by computation, and the information is transmitted to When processes are subject only to slow and small
the control instruments. For example, if the temperature, perturbations, conventional feedback ?ID controllers usually
composition, and flow rate of the feed to a fractionator are are adequate with set points and instrument characteristics
perturbed, the computer will determine the other flows and fine-tuned in the field. As an example, two modes of control
the heat balance required to maintain constant overhead of a heat exchange process are shown in Figure 3.8 where
purity. Economic factors also can be incorporated in process the objective is to maintain constant outlet temperature by
models; then the computer can be made to optimize the exchanging process heat with a heat transfer medium. Part (a)
operation con tin ually. has a feedback controller which goes into action when a
For control purposes, somewhat simplified mathematical deviation from the preset temperature occurs and attempts to
models usually are adequate. In distillation, for instance, the restore the set point. Inevitably some oscillation of the outlet
Undenuood-Fenske-Gilliland model with constant relative temperature will be generated that will persist for some time
volatilities and a simplified enthalpy balance may be preferred and may never die down if perturbations of the inlet condition
to a full-fledged tray-bytray calculation every time there is a occur often enough. In the operation of the feedforward
perturbation. In contro,l situations, the demand for speed of control of part (bi, the flow rate and temperature of the
response may not be realizable with an overly elaborate process input are continually signalled to a computer which
mathematical system. Moreover, in practice not all then finds the flow rate of heat transfer medium required to
disturbances are measurable, and the process characteristics maintain constant process outlet temperature and adjusts the
are not known exactly. Accordingly feedforward control is flow control valve appropriately. Temperature oscillation
supplemented in ,most instances with feedback. In a amplitude and duration will be much less in this mode.
well-designed system (Shinskey, 1984, p. 186) typically 90%

3.3. FEEDBACK CONTROL mode of action of the controller. The usual controllers provide one,
two, or three of these modes of corrective action:
In feedback control, after an offset of the controlled variable from a
preset value has been generated, the controller acts to eliminate or 1. Proportional, in which the corrective action is proportional to
reduce the offset. Usually there is produced an oscillation in the the error signal.
value of the controlled variable whose amplitude, period, damping 2. Integral, in which the corrective action at time t is proportional
and permanent offset depend on the nature of the system and the to the integral of the error up to that time.
39