By Ara Barsamian
The purpose of the survey was to determine the usage of NIR-type on-line blending analyzers, and what it takes to have a successful installation. The reason for using an NIR type analyzer is that it replaces 6 to 8 conventional analyzers with correspondingly lower capital investment, lower maintenance costs, and overall much higher reliability and fast response.
The results of the survey can be used to determine if the NIR type of online blending analyzer is for you, what parameters it can measure, how fast, modeling requirements, lessons learned, etc.
We requested information from approximately 400 refineries world-wide (out of a total of ~700), and received 41 replies. This was a bit disappointing considering that there are approximately 500 NIR analyzers used in refining as of July 2007, out of which 215 are used in gasoline and diesel blending (Table 1). This tells us that many NIR owners are not entirely satisfied with their performance; we can surmise the reasons why later, when we look to the survey data.
The 41 responders are quite aggressive in their use of on-line analyzers, NIR in particular, for blend control, optimization, and product release. They have taken the time to understand how they work and how to install and support them properly; as a result, they are happy with their choice, and would use NIR analyzers again.
The summary of survey results are in Table 3. Basically, the 59% using NIR blending analyzers are pretty happy with them: they meet or exceed ASTM precision, measure a wide range of properties very fast (2 minutes on the average), are used both for gasoline and diesel blending, are robust, reliable, and low maintenance, and some use them also for blend component property measurements.
The biggest user complaint was with property prediction modeling: the fact that many NIR analyzer vendors do not provide process property measurements performance guarantees (i.e., that it meets or exceeds ASTM precision), and that in general, the vendor modeling support is unsatisfactory because the vendors themselves understand poorly how to model the blending process.
When the modeling is done right, one does not need to adjust it every couple of months; once or twice a year is sufficient, if warranted by the number of outliers. Before plunging into modeling, one requires examining blend component variability, rundown segregation strategy, and the blend recipe “space,” all detailed in excruciating detail in ASTM E1655 NIR modeling practice … but very few read it and follow it.
A minor complaint is with filters clogging and film deposition on optical parts; again, the successful installations use a properly designed sample conditioning system and automatic wash and validation either daily or twice a day (as recommended in ASTM D6122).
Table 4 shows the gasoline properties measured by an NIR analyzer; Table 5 shows the Diesel properties measured by NIR.
Overall, the use of NIR analyzer technology was used more aggressively and successfully in the EU, where newer refineries or recent modernization promoted the use of NIRs. In the USA, the use of NIR, although numerically larger, was less robust because of the tendency to “do-it-yourself” cheaply without understanding the subtleties of modeling. Currently, a number of US refineries are in the process of upgrading models through extended support agreements with NIR vendors to provide chemometric specialists assistance.
In conclusion, the major factors for success are:
Success is defined as meeting or exceeding ASTM PRM test method precision for 90% of the time, and not requiring model update more frequently than once every six months during the normal course of refinery operation.
We plan to update the survey in 2009; please feel free to suggest additional survey questions that should be included.
I would like to thank all survey respondents for their support in providing the information.
 “How to Develop NIR Property Models That Work,” Barsamian, A.,
Carmagen Newsletter, Nov. 2005
 ASTM E1655, Standard Practice for Infrared Multivariate Quantitative Analysis
 ASTM D6122, Standard