Approaches to calibrating an NIR analyzer on-line

Ideally, an on-line analyzer will be calibrated before it is installed in the process. It may be possible to accomplish this by calibrating it off-line with process grab samples and/or synthetic samples. It may be possible to install the analyzer in a lab-scale reactor, or in a semi-works or pilot plant. It may be possible to transfer to the on-line analyzer a method developed on an off-line analyzer or on another on-line analyzer (e.g. at a different plant site). However, sometimes none of these are possible and the analyzer will need to be calibrated on-line. The challenges of on-line model development (calibration) and validation, as well as approaches to dealing with them, are discussed below. For information related to calibration transfer issues, please see Chapter 12 of this book. 

Limited variation in chemical composition This is often an issue at the point in the process where the analyzer is installed. A manufacturing plant s goal is to make only good product, with as little variability 

Note Monitoring an analyte indirectly, by relying on its coincidental covariance with another more easily detected analyte, is risky and should be avoided unless the covariance is 100% ironclad (which is rare). 

Nonexistent samples or reference values Samples may not exist because the plant does not exist yet (it is under construction and has not started up yet), or because the plant does not do any sampling at the process point of interest. Reference values may not exist because the plant lab is not set up to perform that particular analysis, or (worse) there is no established reference method for the analyte of interest. In any of these cases, one is left with no data to use for calibration. There are a number of ways to approach this challenge. If there are samples but no reference values, the plant samples can be sent off-site to be analyzed. The analyte concentration of interest can sometimes be estimated based on process conditions and/or the concentrations of other analytes. (This is one place where fixed covariance can come in handy.) If there is no plant yet, it may be possible to calibrate the analyzer elsewhere (different plant, semi-works, etc.). It may also be possible (or even necessary) to attempt lab value-less calibration, in which one assumes that the concentration of the analyte varies linearly with the height 

The factors which limit on-line process analyzer accuracy include  

Note that there are more sources of variability for accuracy than for precision. This is the basis of the common rule of thumb which states that the precision of an on-line analyzer will typically be about ten times better than its accuracy. 

---