By Robert C. McCue, PE and Donald Keer, PE
MDC Systems® Consulting Engineers
Process Plant profitability depends in large part on operational continuity or up-time. MDC Systems® has recently become aware of a developing technology to detect and then prevent unplanned shutdowns due to up-set operating parameters in process plants. This capability results from the real time collection and analysis of all reportable operating data…“BIG DATA.”
Process Plants typically produce more data than can be efficiently collected and reviewed by the operators. A plant with 320 tags (equipment items with associated data collection points such as pressure, temperature, flow etc), recording at 5 second intervals, will produce 5 million data readings per day. That is a billion data points over six months. Buried in this cascade of information are subtle leading indicators of up-set operating conditions.
In order to unlock the information computer processing of the ongoing data stream is required. Essentially using the plants operating signature through the data history allows for the identification of negative trends which have previously been a precursor of the developing up-set operating condition.
This approach to prediction does not result from the typical process safety and systems approach to plant operation but instead relies on the data relationships imbedded in the actual operating data for the process plant. Ferreting out the subtle data relationships that can predict the developing conditions is the task of proprietary data mining hardware and software developed by Near-Miss Management LLC.
A traditional engineering approach to failure prediction rests on systems analysis using failure modes and effects analysis, or quantitative risk assessment approaches to see, via system process relationships, where potential failures originate and to understand the root cause events to prevent them from occurring.
Alternatively, taking a data centric view allows one to let the data relationships established in the normal operation of the process plant provide the insight into the key cause and effect events and to determine the acceptable pattern of variation in data trends. This fact based historical trend analysis provides insight into the subtle correlation of process plant variables needed to avoid unplanned shut-down situations.
Key features of this BIG DATA approach provide:
• Insights into process risk levels of a plant by harnessing BIG DATA
• Evaluation of risk trends on a periodic basis
• Actionable leading indicators to highlight risky conditions at their developing stages
• Drill down capabilities to discover underlying drivers
Process Plant Operators should consider and evaluate the power of BIG DATA collection and computer analysis to increase their bottom line and lead to enhanced safer plant operation.
Click here for a copy of the original white paper, Using Big Data to Predict Process Risks.
MDC Systems® provides Forensic Project Management and Forensic Engineering services to a wide range of clients around the world. Questions or comments on this article should be forwarded to the authors at MDC Systems®, Robert C. McCue and/or Donald Keer, PE.