Automated root-cause analysis earns patent for Metso

Metso Automation USA has received a patent from the US Patent Office for Automated Determination of Root Cause. The patented method, which is used in Metso’s PlantTriage Control Loop Monitoring software, determines the most likely root cause using a "big data" technique and eliminates several costly and time-consuming steps in the problem-solving process.

''Our clients see the benefits in energy savings, production increases and quality improvements. With the proper training and methodology, a single engineer can perform like a large team of engineers,'' says George Buckbee, General Manager, Automation, and the inventor of record.

The new tools are available in the current release of Metso ExperTune PlantTriage software which is an integral part of Metso's Control Performance Business Solution. First, it gathers real-time data from hundreds or thousands of controllers. Then, it looks at correlations between controllers and especially at the effect of time shifting on correlations. Since root causes usually appear as the first in a series of similar events, the software is able to identify the most likely root cause from this combination of correlation and time shift.

Because the new technology does not rely on process models, it is very adaptable and can be applied across many various unit operations. This is especially important in complex process plants where interactions are complex, and accurate dynamic process models may not be available.

Results of the automated analysis are displayed in a Process Interaction Map. The map shows the relationship graphically, using bands of colour to highlight the strongest correlations. For oscillatory interactions, the software also includes the Oscillation Details Problem-Solver.

The advantages of Root Cause Analysis Tools:

• Locates the root cause of routine variation in a manufacturing process
• Determines the root cause in quality upsets or alarms, as well as the variation under normal conditions
• The end user needs no prior process knowledge: the cause can be determined without physical description of the process
• Prioritises a list of likely root causes, displaying it in an easy-to-understand graphic