By Ara Barsamian
ASTM provides the only internationally-recognized methodology to determine the precision of a measurement of hydrocarbon properties. It also provides the framework by which one evaluates the numerical performance of both laboratory and on-line analyzer precision.
The precision of an NIR analyzer depends on the design of the analyzer optics and signal processing electronics, and the comprehensiveness of the property prediction model in capturing the significant variability.
The precision is defined by ASTM in two ways: repeatability, which is indicative of short term stability, and reproducibility, which is indicative of comparative statistical stability.
Practically, in order to meet the ASTM precision (either r or R), this means that out of 20 consecutive measurements, there can only be one value outside the ASTM precision number, thus creating a 95% confidence level.
Assuming that the NIR modeling is done following the “Extreme Recipes Envelope” technique, the expected NIR analyzer precision will meet or exceed the ASTM precision. The ASTM precision numbers were established by round-robin blind testing of samples by hundreds of laboratories. In all cases, in order to demonstrate that the NIR analyzer performance guarantees have been met, it is essential to “track” the performance using Statistical Quality Control charts, either in MS Excel, or preferably in the refinery LIMS system.
By using protofuels for comparative measurements with on-line analyzers, it is possible to improve the precision by a factor of approximately four (the analyzer measurement precision “inherits” the protofuel accuracy).