By Robert Sherman and Jim Montague
Process analyzer sample conditioning systems (SCSs) do not always get the attention they deserve. This is in part because analyzers were not able to do low-parts-per-million (PPM) analysis 40 years ago. So in past decades, process users could get away with crude, less highly developed measurements of what their applications were doing, and many of these traditional habits still persist today.
However, today's analyzers and sample conditioning systems can routinely do low-PPM analysis. Also, since 1990, new rules and regulations are requiring analyses of lower and lower PPM levels of H2S and total reduced sulfur (TRS) in continuous emission monitoring system (CEMS) applications regulated by the U.S. Environmental Protection Agency (EPA), so users must now achieve defined levels of precision and accuracy in their processes. In fact, the first rigorous sample conditioning and reporting requirements were part of the EPA's original Clean Air Act circa 1969-1974. It was revised in 1990 and had its third-generation revision released last fall, including a staged application with compliance deadlines in December 2011 and December 2012.
Lack Coordination? Be Meticulous
Unfortunately, in recent years, there have been an increasing number of sample conditioning systems built by integrators that were non-functional on shipment. This can happen when a user buys an analyzer from a vendor and then has a system integrator (SI) design the SCS. Another common problem is when the user keeps the existing SCS, but it doesn't match the requirements of the new analyzer. This means the overall system may not work, so the user may have to pay daily fines if the application is a CEMS.
This problem often happens because of the lack of cooperation between analyzer suppliers, unaligned SIs and end users. The SI may not have the right field application experience. The vendor might have the field application experience, but the vendor does not feel it should have to fix a competitor's sample conditioning product. Users can be caught in between, sometimes incurring fines of $100,000 per day per erroneous CEMS value.
The solution is to be meticulous! Spend the funds wisely to get your SCS properly designed for the specific application, especially in northern climates. If a user and an SI were to expend the same attention to all details to design their process analyzer sample conditioning systems as they do when designing an operating process unit, then most sample conditioning systems would work reasonably well.
Most users do not complete application data sheets (ADS), especially for their CEMS. It is extremely important that the analyzer manufacturers have these sheets completed with all sample-plus-matrix components and MIN/NORM/MAX values for each identified component. A user will often indicate they require a 40 CFR 60 subpart Ja system for a flare, and know roughly what that requires, but then tell the analyzer vendor to just sell him something designed for this general application. There will be time made to fix it later, but no time taken to properly specify it prior to order placement. This means the user is also asking his SI to have enough knowledge to give him something that "just works."
Unfortunately, the process details for a given process application are often very specific: Products at one company are different than at others; different types of crude oil come in; and different towers create different sample conditions, including many different temperatures, pressures and other parameters.
As a result, users often believe they do not have the time to do sample conditioning systems right the first time. They perceive other higher priorities, so they hope that the equipment they specify will operate properly. If not, they will be able to deal with any problems later. This adage comes to mind: "Seldom time to do the job right; always time to do the job over!"
Diagnose, Fix—and Gimme Shelter
For example, in past years, a Gulf Coast refinery had a flare system with a 150-ft to 175-ft long sample line. This line was operating at nominal saturation due to water washes of the sample, but the sample system was not working properly. When asked about the dew point of the sample, no one knew what it was! This application had a new, insulated, heat-traced, sample-transfer line that was not specified or routed properly. The vendor checked the order for this sample line. It was specified for a low temperature of 300 °F and a high of 360 °F. However, it was still cool to the touch on a day when the ambient temperature was 65 °F. Do you recognize a problem here?
You need to consider heat-traced enclosures for your calibration gases in low-PPM level CEMS applications. This is important! While ambient temperatures in northern Texas can be -15°F in late winter, they can reach 115 °F in summer. Add incident sunlight in southern Texas and you can have 160 °F wall temperature for your calibration gas cylinder stored outside on the south side of an analyzer shelter. This ambient temperature difference can cause stratification inside the cylinder containing the calibration gas. The analyzer depends on the uniform composition of this gas to develop its results. Previously, bottles of calibration gas were put inside of the analyzer shelter, but today's they're installed outside for safety (Figure 1). They should have a separate, climate-controlled shelter for low-PPM CEMS and process analyzer applications. Some refineries are installing 2-ft x 8-ft footprint, climate-controlled calibration gas enclosure adjacent to their analyzer shelter, and installing their calibration gas cylinders inside with heat-traced-and-insulated delivery tubing to carry the calibration gas from the cylinder to the analyzer.
Similarly, because users are experiencing alteration of sample composition due to ambient exposure of sample transfer lines, some process vendors are designing gas chromatographs for field installation near the user's field application sites for hydrocarbon processing industry (HPI) and chemical processing industry (CPI) applications.
Implementing climate-controlled shelters for SCSs and analyzers also means that care is needed when designing and installing these shelters. For example, it's important to use steel I-beams, rather than C-channel, for the base structure of analyzer shelters. Lift extensions must be added to base beams for lifting the completed shelter into its position, often within the boundaries of an operating process unit. It's also important to use a lift cable spreader, so lifting cables will not crush the shelter's roof edges. In one recent case, it did not become evident that the weight of a shelter was more than its base beams could handle until a crew started to lift it using base beam lift extensions at its four corners, and the shelter base beams buckled! These are design and infrastructure problems, but they also require significant and continual attention to detail.
APC and Process Physical Properties
Though low-PPM sample conditioning for quality is a more recent trend than repeatable accuracy for regulatory compliance, process analyzers provide signals for a growing advanced process control (APC) engineering initiative.
Real-time process control and optimization techniques often need measurements of process physical properties in order to be fully effective. Ideally, these measurements would be available on-process in real time, but generally they're only available infrequently and as a result of an off-line laboratory measurement, which is often subject to high levels of method variability and significant delays. To overcome these limitations of lab-based measurements, inferred measurements called inferential sensors (ISs) are used to provide real-time predictions of process physical properties based on mathematical models and standard on-line process measurements such as temperature, pressure and flow.
To be effective, ISs must be based on well-maintained process measurements that can be related to the physical properties in a manner that can be captured in a mathematical relationship. Often, this relationship is developed using correlation models based on numerous lab samples and statistical techniques. Generally, this relationship is only valid in a certain operating region, so the process must be controlled to stay within this region of validity, and/or the process must be carefully monitored to alert when the operating regime no longer fits the range of model validity. Furthermore, the correlation model is often subject to a slowly developing bias or offset that comes from unmeasured disturbances and/or modeling error. This bias must be guarded against and corrected for by periodic validation of the on-line mathematical model predictions versus off-line lab results.
Alternatively, physical properties can often be measured with on-line process analyzers that have the advantages of being more accurate than the ISs, as well as having faster and more frequent measurements compared to lab results, with a (potentially) higher degree of repeatability. However, process analyzers are usually expensive to purchase and install, and they require on-going maintenance to maintain calibration and resolve sample system issues that periodically arise.
The choice of which technique to use in any given application can be difficult to answer. Generally, one has to trade the economic value of the optimization and control effectiveness with the cost and effort of the measurement technique. Clearly, optimization and control can generally provide more value with more accurate, more frequent and more quickly available physical property measurements. Online process analyzers do require an upfront capital outlay and an on-going maintenance effort, but can often provide superior measurements when done properly. Lab-based measurements often do not require the capital outlay, but do carry an on-going lab cost associated with the labor rates of the technicians running the samples, and often have higher levels of sampling and technician/method variability compared to process analyzers. Inferential sensors can be a good solution with a low cost of entry if the correlation model's accuracy is sufficient, and the appropriate monitoring and validity checks are put into place.
Nearer to NeSSI
A large distributed control system (DCS) can use percent analytical process measurements and low-PPM data to influence and improve product quality. Consequently, while we continue to generate analyzer system data to produce correct routine values, APC engineers will use these precise values to make better products!
Likewise, we're beginning to see more sample systems built by SIs that use the New Sampling and Sensor Initiative's (www.cpac.washington.edu/NeSSI) modular substrate platform and components. Some of these components are advertised to perform up to 10 million cycles before needing maintenance. This capability is especially useful in dealing with environmental measurements where we have to measure processes every 15 minutes and do zero and validation procedures every 24 hours. To determine where an analyzer is going to measure, zero is defined as the bottom of the measurement scale, and span is at the top of the measurement scale. Validation is a known measurement value between zero and span. These calibrations are required for every analyzer. Also, when switching from process operation measurements to EPA's CEMS measurements, the NeSSI substrate allows better precision for EPA reporting.
Because users must maintain product quality, avoid plant shutdowns and report accurate data to the EPA, they must require unaligned SI firms to provide better support and build more functional sample probes and sample conditioning systems. They must also demand that analyzer manufacturers cooperate on working with each other's devices when required by a user trying to achieve the absolute best system solution.
To foster better cooperation between unaligned SIs and vendors, users must fill out their process application data sheets and encourage everyone involved to perform the required calculations based on this process data. These ADSs must include the physical parameters that the sample system will be required to control, as well as the upper (MAX) and lower (MIN) characteristics of the overall process in a tabular format. In addition, a sample conditioning system's specifications should describe the specific physical and electrical hazard requirements that it must meet. The data sheets should also provide the system's environmental parameters: temperatures (ambient and process stream) and hazardous atmosphere monitoring (emergency alarm system).
NeSSI is Not a Monster
"I’ve been working with sampling systems since 1969, and NeSSI is the biggest advance in process analytics in the past 50 years," says Bob Sherman, an ISA fellow and former industry specialist for process analytical systems at CIRCOR Tech, which manufactures modular, NeSSI-compliant devices. "Whenever manufacturers move from legacy sampling systems to modular substrate systems, they can optimize components such as filters, for example, and achieve many new and more efficient capabilities."
Sherman adds that several hundred NeSSI-compliant components have been installed and tested and are serving in sampling systems in some West Coast refineries for more than five years. The maintenance time they require is 30% to 70% less than the maintenance time required by legacy components in the same services.
One of the first modular systems CIRCOR Tech put in was cleaning up samples for Reid vapor pressure analysis on a gasoline blender. The refiner’s technician said this application was just not going to work with these new modular components. Eight weeks later this same technician was enthusiastically asking, "How many more of these modular systems can we get?" These components became a de facto standard for this refinery.
The question potential users often have is, "Does NeSSI have staying power, or is it just another flash in the pan?" Sherman adds that one major refiner previously required its technicians to go out and read rotameters and pressure gauges each shift, and they couldn’t apply legacy flow controllers and pressure transmitters because these components were too large to fit into the sample conditioning system enclosures. This refiner now uses NeSSI-based mass flow controllers and pressure transmitters that report directly to the process unit’s control room, which means technicians only need to visit the analyzer system in the field if an exception report is generated because the analyzer system isn’t operating within its defined range.
Users, SIs and vendors must draft well-written specification documents that sufficiently describe each party's roles and define all expected results. Users should never say, "Just give me whatever equipment works; we'll fix it later."
In the past couple of years, many users have started requiring better-defined specifications and started filling out process ADSs for every sample conditioning system project. Even though it can be difficult to gather some of this process and ambient information, these users are requiring expected process flow rates, temperatures and pressures at the specific points where samples will be extracted by the in-line sample probe.
Glossary of Process Sampling System Acronyms
ADS Application Data Sheet
Similarly, analyzer technicians and engineers seem to be getting better at arguing that proper design of analyzer sample conditioning systems will reduce overall analyzer system lifecycle costs. Many refining companies encourage their technical personnel to join the American Petroleum Institute (www.API.org), American Society of Testing and Material (www.ASTM.org), get involved and contribute to the International Society of Automation's (www.ISA.org) Analytical Division or one of its other divisions relating to analysis and control aspects of the HPI and CPI, and join supplier user groups and customer advisory boards. The reality is that you can gather significant useful information at these professional forums and bring it back for use within your organization.
Ten or 15 years ago, most analyzer technicians could not get a climate-controlled shelter for their PPM analyzers because finance personnel would argue that if an analyzer in its own enclosed housing was rated to meet field location ambient conditions, then a separate shelter wasn't needed. However, while a manufacturer can say an analyzer in its enclosure is rated to -30 °F, it might also add that, if the internals get below 55 °F, then primary vacuum pump oil will be too viscous for the pump to operate, and it will need to be heated up for several hours.
How would you heat an enclosure filled with electronics and delicate mechanical devices in a northern winter location when this enclosure is located in the middle of an operating process unit? A plastic sheet tent and a steam hose are about all you'd have available, and this combination would be absolutely lethal to the electronics and could possibly also damage delicate mechanical devices. If there is blowing rain or drifting snow where an externally mounted analyzer has been installed, then analyzer and instrument technicians can't open its enclosure because they'll risk getting the internal circuit boards too cold or thoroughly damp. This is why users are currently installing more 8-ft x 8-ft or 10-ft x 10-ft base footprint shelters to protect their process control and environmental analyzers.