Battling process information overload

Falconeer IV’s Sensor Validation and Proactive Fault Analysis (SV&PFA) methodology, called the Method of Minimal Evidence (MOME), continuously evaluates engineering models of normal process operation with current process measurements to identify process faults, such as faulty instrumentation. MOME is a generalized algorithm based on Fuzzy Logic1. The underlying logic of the MOME is similar to the approach an engineer or experienced operator would use to evaluate and troubleshoot the information—if they had the time and knew where and what to look for. 

LIGHTENING THE LOAD

Data analysis identifies process needs in FMC's operations.

A concurrent and equally powerful analysis occurs at the tag level. All process measurements and key performance indicators (KPI’s) are monitored using exponentially moving weighted averages (EWMAs), and then automatically alerted whenever those points go out of control. This virtual statistical process control (VSPC) feature of the application typically warns users of process shifts (normally non-fault events) well before they become DCS alarms or result in severe, non-optimal operating conditions. 

Benefits and Successes
FMC has used Falconeer IV on the ESP process since the summer of 2003, and on the LAP process since the fall of 2004. To date, we have used these two applications mostly to head off operating problems before they caused reject product, and to ensure the process is being operated in an optimal manner. Since operating conditions in the plant are seasonable, the proper process configuration changes over time.

In one situation that Falconeer IV caught, steam to a booster jet was found to be on when it wasn’t required. This one alert saved approximately $30,000 in operating expenses. Another alert on a low total ammonia inventory KPI occurred one full day before the problem would otherwise have been noticed. This proactive alert prevented a significant process chemistry upset that would have created product reject, and saved thousands of dollars. 

Falconeer IV embodies a complex but extremely useful and transparent diagnostic logic, which performs a comprehensive, continuous analysis of current process data, and then immediately reports its results in a variety of user-friendly formats. It’s designed to directly allow us at FMC to easily configure and change these two applications ourselves. The majority of the configuration effort involves developing the engineering models of normal process operation (they were generally already available in various spreadsheets, simulations and reports), and then evaluating them with sufficient normal process data to determine the necessary statistics. Such ease of configuration and maintenance makes this technology extremely viable and cost effective, helps ensure that actual applications don’t become obsolete over time, and that its full benefits are always available to its intended users.

Employing such complex analysis continuously enables us to more readily deal with the relentless deluge of process information gathered by our DCS. Continuously analyzing this information to better identify its decisive content, and then immediately conveying these results to the appropriate personnel, so that timely actions can be taken, has proven to be extremely beneficial. It acts as a virtual supervisor, being proactive and immediately responsive in its ability to provide its users better information for better decision-making in a timelier manner to save money and continuously improve current process operations.

Other examples of operational improvements directly derived from using Falconeer’s software suite include:

• Sensor validation and proactive fault analysis (SVPFA) module found several process sensors routinely used to control the process to be in error.
• Virtual statistical process control (VSPC) module detected Coriolis meter density and flow components failure 12 to 18 hours prior to TDC alarms. The meter’s temperature component was unaffected and remained validated.
• SVPFA module traced a crystallizer vacuum leak to a condenser unit.
• SVPFA module also correctly diagnosed crystallizer overflow pump failures twice over a six-month period, and pinpointed a catholyte pump flow failure.
• VSPC module detected numerous controlled variables operating unintentionally in manual control.
• SVPFA module identified a pH probe that was not properly located in its unit.
• VSPC module detected an incorrect operator adjustment to an air bleed for the current steam feed.
• SVPFA module identified numerous failed or miscalibrated flow, density, temperature and pressure sensors routinely used to monitor the process.
• SVPFA module discovered a higher-than-expected unmeasured flow in one unit operation.
• Some of the SV&PFA module engineering model residuals provided supporting evidence for expected, but unmeasured, process phenomena such as fines destruction and crystallizer boiling point elevation.

Consequently, these two performance auditing systems have already demonstrated a very acceptable return on investment (ROI) from improved process operations.

Bibliography    
Fickelscherer, Richard J., Douglas H. Lenz, and Daniel L. Chester, “Fuzzy Logic Clarifies Operations,” InTech, pp. 53 – 57, October, 2005.