Process Event Analyzer

The PM Event Analyzer is a visual root cause data analysis tool that helps process engineers interpret massive volumes of data collected by industrial databases such as PI System, OPC HDA, INSQL or by an internal SQL Server based high-speed servers. The Analyzer does this by providing event-driven analysis of process variables. In the papermaking industry, this provides a shortcut to determining the causes of web breaks and other process related events or upsets.

The database systems acquire data from plants or processes via interfaces to automated control systems and other sources. These systems record data from thousands of these process ‘tags’, at specified intervals or by manual entry. When an event takes place, it is normal for operators to review alarm status on the DCS or on drive control panels and to check selected process tags. But which tags, of the thousands being logged, should the operator check first?

The Event Analyzer automatically identifies the most likely process variable to examine. It does this by displaying a graphical representation of several hundred selected process variables and by ranking these according to an algorithm based on signal processing techniques. The result is that the tags associated with the most probable cause of the event identify themselves – without tedious searching by the engineer or the operator.

Multiple Applications
The Event Analyzer has built-in features for paper machine break analysis, spectrum analysis and general process root cause analysis. Its high level mathematical tools can be used to analyze many kinds of process upsets or events.

Break Analysis
Root cause analysis of web breaks is still heavily dependent on the skill, experience and intuition of engineers and mill operations staff. As a plant control engineer explains: "When we get a break, first we look at the video. Then we look at the DCS alarms, then the motor drive alarms, and then the PI tags. But with 40,000 tags being monitored, we’ve got to use our gut to figure out which ones to start looking at first." The Event Analyzer software provides help in sorting through this mass of data. It continuously monitors a selection of a few hundred tags, graphically showing long-term and short-term changes in values.

In case of a break, the software displays the tags that have shown the most significant variation in the time immediately preceding the break, on the basis that these are most likely to be correlated with the cause of the failure. "It’s like one of the tags raises its hand and says ‘Look at me first!’" explains one of the engineers. The value of this is not hard to see: time is saved in diagnosis, and the quality of the diagnosis is higher, avoiding the possibility of unnecessary repair work and – worse – the likelihood of another break due to the same cause.

He cites two recent examples of the value of the Event Analyzer tool. In the first, a maintenance team was about to change a solenoid, which the shift supervisor was convinced had caused a web break; but the control engineer’s review of a motor speed tag highlighted by the software caused him to direct the team instead to change the motor controller. This proved to have failed, and the early identification of the culprit saved several hours of downtime and unnecessary work. The second example was of a break that occurred while all systems were apparently performing perfectly – but the Event Analyzer quickly identified a small speed variation in a fan pump motor, which coincided with a grade change. Human error was identified as the root cause in this case.

Process Analysis
A major upset occurred in the MD Moisture. The moisture normally runs around 2%, however, there was a sharp peak at 6%. Luckily this upset did not create a break on the machine, which is common when this happens.
The actual process upset is highlighted in the graph below.

Below is a list of the correlated variables. BW is listed first and couch vacuum is second. Of course those correlations are directly related to the same problem. The actual root cause is the third variable in the list, #4 HD to #2 PM Pres Cntl.

The trend below shows the BW peaks as the green trend. The HD tower pressure controller is in yellow and shows drastic drops in pressure. Using the measurement tool provided with the trend tool we see that these pressure drops occurred almost 1½ hour earlier than the upset on the machine.

Spectral Analysis
The following three pictures show the steps to select a problem and identify the source using the spectral analyzer.
In this case the variability problem with the moisture. When the moisture profile is selected, the trend window shows how this variable has been moving for the last 48 hours. The first step is to select a time period that distinctly shows the variability. The red selection bars on the graph show this.

Spectral Correlation List shows that the Softwood Washed Stock Consistency has the same frequency in its variability domain with that frequency contributing over 50% to the consistency’s variability.

It is easy to see that the moisture is varying with the consistency.

Conclusion
These are just two examples showing how one customer is using the tool to perform root-cause analysis to identify the source of process upsets so that corrective actions can be taken.

Feature:

• Graphical user interface
• Automatic ranking of high variability tags
• Quick to install
• Database-driven analytical tools
• Ability to e-mail results
• Based on industry standards ‘Good run’ versus ‘bad run’ comparison
• On site access Laptop, H P4700
• User-selectable ranking criteria
  

Benefits:

• Easy to understand and to use
• Immediate analysis available
• Quick payback, minimum disruption to operations
• Facilitate process improvements
• Improves knowledge-sharing
• Integrates with existing systems
• Easy to check the effect of changes
• Can deal with gradual process changes
  
  

Fast Payback:

The Event Analyzer helps operations staff to quickly diagnose process failures such as web breaks, generator trips and other process events, reducing the likelihood of misdiagnosis.

• Minimizes lost production
• Saves labor
• Avoids unnecessary repairs on misdiagnosed componentss
• Avoids secondary failures due to misdiagnosis
• Leads to rapid process improvement
  

Detail Power Spectrums: