You have got the data, now extract the hidden information
Archives of plant data.
In most plants the important process
variables like temperatures, pressures, flows, are monitored continuously by
Plant Information and Control Systems. The numerical values of these process
variables are stored at regular time intervals. Thus huge archives of
historical plant data have been built up. The consultants of the group
Mathematics & Statistics can assist you in exploiting these data mines and in
extracting valuable information from raw data. Our resources are many years of
industrial experience, state-of-the-art software tools and statistical
expertise of the highest academic level.
Consistency with Standard Operating Procedures (SOP's).
By
statistically analysing the time profiles of the process variables, it is
possible both for continuous processes as for batch processes to:
-
quantify the variability in your process/plant and calculate the capability
indices;
-
detect systematic deviations from the SOP's;
-
assess the effectiveness and possibilities for improvement of the plant's
measurement systems (with respect to accuracy, frequency, delays, ...) and
control loops;
-
tighten control limits and specification limits.
Multivariate Statistical Process Control.
Historical plant data
can be used to build statistical models that capture both the systematic and
the random components in the normal operating modes of the plant. Once the SPC
models are built, they can monitor the process in real time and can detect
deviations from normal operating conditions, thus allowing for corrective
action in an early stage.
Black box models, predicting the values of quality variables.
Based on historical plant data, statistical black box models can be built.
Functional relationships are constructed that link process conditions and
quality variables (like: yield, viscosity, purity, water content, ...). Once
these relationships have been found and quantified, the values of the quality
variables can be predicted in real time, i.e. while the process is running.
Set points or trajectories can be chosen to maximize or minimize the values of
the quality variables.
