Cluster Analysis Software

At the heart of our software is an Excel addin that automates data retrieval. We combine this with a scripting file that automates our addin. Other files include a web page populated at runtime and a help file detailing how to use the software. Not all data in a historian is good data - often the process is unstable or even down. Including this data will result in ridiculous conclusions. We solve this problem by identifying time slices in the historian that meet certain criteria. These time slices form the data set to be examined. Every time the application runs, it discards the oldest time slice and adds a new one (assuming it passes the filter criteria).

Once we identify the data set, we load the application with the variables, or tags, in that process area. We have run as many as 5000, but more often the list is 500 to 3000 tags.

The user then identifies the Key Performance Indicators (KPIs) to study. For each KPI, the user creates a folder in a file server. At run time, the application as well as the following:

The tags that changed at the same time(s) the KPI did and the magnitude of the change
The 25 tags that have the highest correlation (R2) to the KPI
A list of tags that are dead in other words, have not changed
A list of tags that changed significantly since the last run, even if they don't correlate with the KPI.

Imagine a bearing temperature that rises 20 degrees but does not generate a DCS alarm - we identify that bearing.

At runtime, we create two Excel output files that together with our addin, allow users to pinpoint in seconds the drivers of the identified KPIs or for that matter, ANY other variable in that process area. The application can even create a PowerPoint file showing you every tag that correlates with a given KPI with just the click of a button!


Envoy Development software is protected by US patents and patents pending