Validation of process data quality

Process data is, to large extent, measurement values created by analogical meters and, therefore, measurement uncertainty is always present. Meters are always more or less unstable, and thus, measurement values suffer almost always from crawling to certain extent. If measurement values are used for invoicing (between two parties), their inaccuracy and crawling may destroy invoicing reliability. If process data is used for decision making regarding the process, measurement inaccuracy and crawling may result as wrong analysis results and decisions.

Acceptable quality of measurement data means that measurement values can be traced to international measurement standards, measurement uncertainty is managed continuously and that it is small enough to serve its practical purpose. Also, it needs to be verified that handling and transfer chains of measurement data are covered by measurement quality validation.

1. Measurement uncertainty

Measurement uncertainty is a parameter related to measurement results that describes the expected variation of measurement unit. Normally, 95% measurement uncertainty is used and its size is typically described as percentage of measured value.

If, for instance, measurement uncertainty is 3% and measured value of a unit is 100, the real value is of 95% certainty between 97...103.

2. Measurement data traceability to international standards

This is ensured by covering adequate amount of important measurements in balance by calibration program. Traceability to real measurement circumstances is something that is required from calibrations, and often this calls for field calibrations.

3. Continuous validation of measurement quality

Measurement accuracy of a single position can be extended to cover entire balance (energy, material, water), using balance analysis. When this is done on a continuous basis, measurement accuracy of a balance is secured continuously.

4. Sufficient measurement accuracy

Sufficient accuracy depends always on the case. For instance, operating efficiency of energy production can typically be improved by 1…5 percentages. Uncertainty of major measurements needs to be clearly smaller than this range, in order to have reliable analysis results for improving operating efficiency properly. Another example is big occasional emissions. Relative measurement accuracy is often enough to identify and reduce these.

5. Validation of measurement chain quality

Measurement message created by field equipment is transferred and modified in automation system before it is saved to history database. The same measurement unit may be measured via many meters, or different versions of it might exist – and a part of these is not covered by measurement quality control. Therefore, handling and transfer chains of measurements need to be validated to cover process data used for various analyses. Additionally, it needs to be ensured that analyses use only data that is covered by quality validation.  

Many changes and updates are done on automation system during its life-cycle and these may cause uncontrollable changes to measurement chains, and through that, to process data. Based on our experience, especially automation system renewals have caused unjustified changes on measurement chains.