(24c) Evaluating Individual Instrument Readings and Potential Errors in the Analysis of Monitoring Data

Evaluating Individual Instrument Readings and Potential Errors in the
Analysis of Monitoring Data

E. Tamakloe¹ , G.T.Polley¹ &
Y.V.Kapoor²

1. Dept of Chemical
Engineering, University of Guanajuato, Mexico

2. CERT, Total s.a., France

Data
reconciliation techniques have been applied to measurements obtained from the
monitoring of pre-heat train performance for many years. In fact, the pre-heat
train appears to be a standard problem addressed in text books on the subject
of data reconciliation [1]. These techniques are usually based on statistical
analysis.

In 2011 Ishiyama et al [2] introduced an alternative to data
reconciliation based upon statistical analysis. This approach involved ?data
adjustment?. It involved the application of a fouling model to the monitoring
data in order to minimize differences in hot and cold heat balances, generate
missing temperature information and the grouping of data in order to identify
trends over specific operating periods.

The technique
can be used to identify ?best values? of missing measurements and to identify
which measurements are likely to be unreliable and generate estimates of what
the true values are likely to be.

In this paper
the authors follow the lead of Ishiyama et al but
look at how errors in individual instrument readings can be addressed.

Instrument errors

Instrument
errors are assumed to be made up of two components: an off-set and a drift.

Off-set is an
absolute temperature difference between actual and apparent reading. It can be
positive or negative.

Drift is a
percentage of the reading relating to difference between actual and apparent
reading. Again, it can be positive or negative.

Identification of Onset of Instrument Failure

Experience in
the analysis of monitoring data has shown that the behavior of temperature
instruments deteriorates several days before it fails. It is therefore possible
to detect the onset of failure and provide a warning to the plant operator.

Example will be
described.

References

1. Narasimhan, S., Jordache, C., Data Reconciliation & Gross Error
Detection, Gulf Publishing Company, Houston, TX, 2000.

2. Ishiyama E, Pugh
S.J., Wilson D.I., Paterson W.R. & Polley G.T., ?Importance of data
reconciliation on improving performances of crude refinery preheat trains?, AIChE Spring Meeting, 2011