Utilising advanced pressure diagnostics to enhance process insight

Advanced diagnostics embedded in a pressure transmitter can proactively detect abnormal process conditions.

Pressure transmitters measure pressure, flow or level and may be used to control or monitor changes in a process. Although plants make every effort to instrument and control their process effectively, there may be abnormal conditions that could cause a loss of production, plant shutdown or safety hazard. New types of pressure transmitters that include advanced statistical diagnostics can now provide a means for early detection of abnormal situations in a process environment. This technology can also detect integrity issues in the electrical loop that connects the transmitter to the host system, and enables the user to proactively respond to changes in the process, to troubleshoot and to prevent future shutdowns.

Process intelligence

Virtually all dynamic processes have a unique noise or variation signature when operating normally. Changes in the signatures may signal that an abnormal event has occurred or will occur soon. Examples of abnormal events that are detectable using process variation include plugged impulse lines, furnace flame instability, distillation column flooding, pump cavitation and entrained gas in liquid flow measurements. Since pressure measurement applications require a very stable reading, damping is typically applied to both the transmitter and the host system to reduce or eliminate variation. Most control systems sample the pressure once per second or slower and process historians may sample the process variable as slow as once per minute. By the time the pressure measurement reaches the control system, most of the noise and variation characteristics have been filtered out, making it difficult or impossible to detect these abnormal process conditions.

Figure 1: Effect of changing noise levels on the mean and standard deviation values for a process signal.

Statistical process monitoring (SPM) technology provides the means to detect abnormal process conditions within the limitations of typical control systems and process historians. A smart pressure transmitter measures the pressure and calculates statistical parameters at a high speed, allowing it to monitor the process noise prior to any analog or control system damping. Statistical parameters (mean, standard deviation and coefficient of variation) are computed within the transmitter’s microprocessor and sent to the host system via HART or fieldbus. These statistical parameters are then used to determine the presence of an abnormal condition. Figure 1 shows one example of how the standard deviation value is affected by changes in process noise while the process variable (PV), as typically seen by the control system, remains unchanged.

Figure 2 illustrates statistical process monitoring technology in greater detail. The process pressure is measured by the pressure sensor and sent to the host system via a 4-20 mA analog signal, just as with any other typical pressure measurement. At the same time, a statistical calculations module computes the mean, standard deviation and coefficient of variation (CV). The statistical parameters (or SPM variable) are made available as outputs to the host system as HART or fieldbus digital variables, where they may be trended in a process historian and viewed by a plant operator.

Figure 2: Pressure transmitter implementing statistical process monitoring technology.

If a process upset occurs, plant staff can look at the process historian and examine the SPM variables to determine if changes in these values gave any prior indication. After understanding how changes in the SPM variables are affected by process conditions, a control engineer can create DCS alerts or alarms based on the SPM data.

Alternatively, one may utilise intelligence resident in the pressure transmitter to detect abnormal process conditions. The Learning Module of SPM technology learns the operating conditions and statistical patterns (baseline values for mean, standard deviation and CV) when the process is operating normally. The Decision Module compares the current statistical values against the baseline values and detects an abnormal event if the current value deviates from the baseline value by more than some limit. The host system is notified of this detection either via device alert or analog alarm.

Figure 3 illustrates in more detail how the Learning and Decision intelligence of SPM technology works. There are two primary modes: the learning/verifying mode (performed in the Learning Module) and the monitoring mode (performed in the Decision Module).

Figure 3: Simplified view of statistical process monitoring technology.

During the learning/verifying mode, the mean and standard deviation are computed over a user-configurable length of time (default is three minutes). First, the module determines if there is sufficient process variation to perform the SPM diagnostic. If there is not enough variation it gives the message “Insufficient Dynamics”. If there is enough variation, then the SPM computes mean and standard deviation over the next three minutes to determine if the process is stable. If the process is stable, the SPM takes the current statistical values to be the baseline values and moves into the monitoring mode. If the process is not stable, the SPM remains in the verifying mode. This stability check is an important part of the SPM technology because if the transmitter learns the baseline values while the process is not stable there is a much greater chance of generating a false detection during the monitoring phase.

In the monitoring phase, the SPM again checks to see if the system is stable (that is, if the pressure value is near its baseline). If the mean pressure has changed significantly, the SPM goes back into the learning mode. If the system is stable, then the SPM checks if the process dynamics (standard deviation or CV) have increased or decreased significantly. If so, either a 'High Variation Detected' or 'Low Variation Detected' alert is generated.

Electronic Device Description Language (EDDL) and FDT/DTM technologies allow advanced pressure diagnostics to be utilised in any host system environment that supports one of these standards. Figure 4 shows an example of SPM technology in AMS Suite: Intelligent Device Manager. When the process is operating normally, the standard deviation (red) is near the baseline (blue). When an impulse line is plugged, the standard deviation decreases below the lower threshold (grey) and a low variation alert is triggered. At the same time, the mean (which follows the pressure value) remains unchanged.

Figure 4: SPM abnormal event detection visible in AMS Device Manger.

Coefficient of variation

In DP flow applications, the standard deviation is affected by the differential pressure. Thus, if the flow rate increases, the standard deviation also increases. If the flow decreases, the standard deviation decreases. A change in standard deviation caused by a changing flow rate should not be interpreted as an abnormal event. SPM has the ability to relearn new baseline operating conditions if the mean changes significantly, which can reduce the possibility of a false detection if the flow rate changes.

However, field test and laboratory data have shown that for most DP flow applications, the standard deviation is approximately proportionate to the DP mean. This relationship has been observed in a variety of fluids (liquid and gas), primary elements (orifice plate, venturi, flow nozzle, Annubar, etc) and beta ratios. The ratio of standard deviation to mean is defined as the CV.

CV will remain approximately constant even if the flow rate changes. Thus, if an operator observes a change in CV, it is much less likely due to a change in the flow rate and much more likely due to the presence of some abnormal process condition. The newest pressure transmitters with advanced diagnostics and statistical process monitoring provide CV directly as a digital variable and can generate alerts if the CV changes from baseline conditions.

Figure 5 shows a comparison of standard deviation with CV. During the abnormal event (eg, plugged impulse line, entrained air, etc) both the standard deviation and the CV increase. But when the flow rate is increased, the standard deviation increased while the CV stayed generally constant.

Figure 5: Standard deviation versus coefficient of variation with changing flow rates.

Note that CV is recommended to be used only with DP flow applications, because for other pressure applications (eg, line, absolute, DP level) the relationship of standard deviation being proportionate to pressure does not hold.

Diagnostic applications

Within every process there are many abnormal situations that can result in an untimely plant shutdown or loss in productivity. Over 20 different applications have been identified as being detectable using SPM technology. One example application is detection of plugged impulse lines. In many instances pressure transmitters are installed with impulse piping that can become plugged with solids or frozen in cold environments. The user typically has no idea that this has occurred because the pressure is effectively trapped between the plug and the transmitter, resulting in the pressure measurement not responding to changes in the process.

The trapped pressure causes a change in the process noise, which can be seen by the SPM technology and can be used to detect plugging in either one or both impulse lines. Using SPM technology to detect impulse line plugging may increase the safe failure fraction and safety rating of the measurement point.

Figure 6: An example of furnace flame instability under actual factory conditions.

Another application of the SPM technology is detecting the instability of furnace flames. In many pieces of fired equipment, NO_x emission standards are met by multistage burner design and exhaust gas recirculation. Either technique tends to give a burner design with a much narrower stable operating window, thus making flameouts more likely. Rather than detecting flameout after it has occurred, SPM can detect an incipient flameout. This makes it possible to proactively detect the instability and prevent the flameout. The example in Figure 6 shows how the process noise (standard deviation) registers instability in the furnace flame on all three of the transmitters monitoring this furnace, while the primary variable (draft pressure) measurement does not register any change in the process.

In conclusion, advanced diagnostics embedded in a pressure transmitter such as the Rosemount 3051S can provide a means for early detection of abnormal conditions in a process environment. Using this technology enables the user to proactively respond to changes in the process, troubleshoot and prevent future shutdowns.

By Erik Mathiason and John Miller, Emerson Process Management