Page 331 - Design of Simple and Robust Process Plants

8.3 Automation of Operation 317
8.3.4.2 Software
In having to execute all the program tasks for which it is designed, software is an
important facet of a totally automated plant. Thus, on listing the required qualifica-
tions, the following question(s) should be asked as to whether the current software
meets these requirements:
. Software development is based on a well-documented operational and control
strategy and operational procedures.
. Soft ware should be easy to program, with an accepted high-level language.
. Standard tool kits for common parts need to reduce programming time, and
make it easier to adapt the program.
. Common software configuration should be used on the different processes
operated from the same control room.
. Software code should be understandable but protected from the operator, and
different protection levels must be foreseen.
. A database should be present to enable working with recipe tables for differ-
ent operational options.
. Different control levels need to be visible in a clear overview, so that the
operator knows which level is active and what is its objected functionality.
. Control software needs to handle from simple up to more complex predictive
model-based controllers within the same working environment.
. Software testing should be performed for logic by means of a quasi simula-
tion (Fujino et al). Detailed realistic dynamic simulations are required for
testing critical dynamics such as model-based alarming or complex model-
based controllers. Control parameters can be determined on a preliminary
basis using closed-loop dynamic simulations.
. During operation, an operator needs to have immediate access to the soft-
ware statements that determine its current (non)action. A hold-up in a pro-
gressing step should be explained to the operator.
. All operator actions need to be registered.
In conclusion, the selection of instrumentation is based on analyzing control, opera-
tion, safeguarding, and observation. The tendency is to over-instrument a process
plant, which in turn increases the likelihood of failures resulting in nuisance trips.
Nonetheless, there remains a drive for more instrumentation, and everybody adds
instruments during the development of P&ID. It is important to re-emphasize the
point that, based on the above findings, redundant instruments should not be
installed unless determined by safety or environmental requirements. Neither
should the performance of all units be measured routinely ± only those that are cri-
tical to the operation in terms of economics and reliability. As a yardstick for mea-
suring the degree of instrumentation, the ratios of input to output can be used.
The ratio between analog inputs and outputs (AI/AO) and digital inputs and out-
puts (DI/DO) can be used as a basis for comparison in instrument design. For con-
tinuous processes, an AI/AO ratio of 1 is the ultimate minimum, but in practice a
value of less than 3 is achievable. In cases where a reaction occurs in a packed bed,
and where local temperatures may lead to hot spots, the ratio may become higher.