It was a sweltering, hot day, and Horace was despondent since his Reid vapor pressure (RVP) analyzer was apparently at fault for an off-spec batch of premium fuel. In his refinery, the batch was “finaled” per procedure before being released to the pipeline. “Giveaway” is when a batch exceeds specification, meaning the blender could have increased the percentage of a less-valued blending stock–especially butane.
Butane, sold as automotive fuel, is an enormous upgrade for the component, worth literally millions of dollars in margins. It’s valuable to blend fuel right up to the very precipice of becoming off-spec, so online octane (that typically knocks motors) and vapor pressure analyzers are used throughout a batch to maximize consumption of feedstocks such as butane.
Decades ago, online analyzers were invented as facsimiles of lab methods that analyzer results aimed to duplicate. ASTM International’s RVP method was developed in the 1930s as a standard for internal combustion engine fuels. It calls for a pressure measurement with 4:1 vapor-to-liquid ratio in a 100 °F bath. Horace’s online machine drove its two pumps with cylinders sized to precisely pump one part liquid and four parts vapor, and measure the (absolute) pressure in a 100 °F temperature bath.
Packing wear is inevitable and often causes disparities to arise. Horace standardized against his water-displaced “proto” fuel (a golden batch that was thoroughly analyzed by the lab) to prevent any light ends from escaping into vapor space. Packing wear, bath temperature excursions and pressure transmitter drift can occur during a three- to six-hour batch—some refineries blend straight to pipeline—so a recent standardization wasn’t beyond reproach.
Traditional analyzers have multiple variables in play that can affect the measurement result, and digital integration of such online analytical equipment over various buses—primarily Modbus or Ethernet—is indispensable for measurement validation. Imagine if Horace knew his pumps were pumping precisely what they should have been, his temperature bath was on setpoint, and his pressure measurement was showing no evidence of drift. The highly mechanical, wear-prone device might have been more accurate and reliable than the lab.
I can’t recall a time when photometric or optical absorption instruments, for example non-dispersive infrared (BDIR) optical instruments, were regular appliances in the lab. However, lab instruments achieved amazing capabilities through by applying lasers, affordable computational power, and precise mechanics to provide high-resolution absorption spectra.
In the 1990s, a Fourier transform infrared (FTIR) of an interferometer signal could be resolved repeatably to produce a detailed spectrum in an online instrument. Sometimes, the raw spectra could be related to a component’s concentration, but in many interesting cases—gasoline blending among them—properties such as RVP or octane were derived by chemometrics, which uses partial least squares regression, among other statistical tools, to calibrate the spectra with lab results.
Like process modeling, which applies similar regression mathematics to derive process models, it’s often garbage in, garbage out. False or aliased measurements can produce models that are useless. As U.K. statistician George Box famously said, “All models are wrong, some are useful.” But like validating process models, relationships derived from spectra can have reassuring correlations. Science has cataloged specific wavelengths, where compounds and characteristic molecular bonds absorb in the infrared and near-infrared.
Still, mathematically derived models, proven effective and useful in FTIR and Raman spectroscopy, rely on repeatable and consistent lab results. Perhaps some outliers fall out during statistical analysis. Ultimately, there’s much reliance on lab standards, sampling and sample handling, and care and feeding of the lab instruments.
Horace and his old-school RVP analyzer were vindicated, as the second shift lab tech was meticulous about her use of a chilled container and ice bath to obtain a representative sample. In the commercial realm, the lab rules, but online analyzers can help adjudicate whether the lab’s infallibility needs another look.
