Q & A – Understanding DGA Techniques and Interpretations

Understanding DGA Techniques and Interpretations Webinar Q & A


During our recent Dissolved Gas Analysis webinar “Understanding DGA Techniques and Interpretations,” Dr. Gregor Hsiao answered a number of excellent questions from attendees. Read on to learn more about these questions and find out how you can watch the recorded webinar!

Why do I get different results depending on which lab I send the oil to?

This is a very common question and something we’ve experienced ourselves! It can happen even when sending identical samples to the same lab at the same time. In offline analysis, there are several different factors coming into play. First, the manual sampling process may be completed differently by each operator. Another factor is timing, where one sample is taken at the end of the day and shipped out immediately while another was obtained and then sat in the truck for a few days while other samples were collected at other locations before shipping them out several days later. Shipment may also impact the sample as sometimes gas bubbles form in the sample. Once it arrives at the lab, there is an operator preparing the sample manually. So, as you can see, by the time the sample enters the analytical instrument, there are multiple sources of variation that increase the odds that the sample is not the same as when it was taken.

Are absolute values more important or the change values more important?

It is important to remember that this data is developed empirically and that not every transformer is the same. There are transformers for various reasons that may start out at different points. They may have been operating at gas levels or ratios which would typically indicate a problem but they’ve been operating in those conditions for many years and are actually stable. So, what’s most important is the change over time and not the absolute value. For Duval Triangle, you are not looking at an isolated data point, you are looking at the evolution and the behavior of that transformer. For example, with the way the Duval Triangle is set up, there is no indeterminate diagnosis, as everything falls into zones. You don’t look at just a single data point or level but you’re looking at how that transformer is changing over time.

If there is a difference or conflict of data with several methods, how do you determine what is most accurate?

If you are using different analysis methods and they are giving you indication of different fault types, we suggest looking at the different methods and the rate of false or indeterminate diagnosis. If you are getting agreement with multiple methods and one with a high rate of false diagnosis, perhaps that one should be discounted. If you want to do as much as possible, you can also look at case studies by IEEE IEC where they’ve compiled these statistics to check your conditions against correct and false results and really do a comparison relative to that. Basically, by comparing the case studies, you can gain insight as to how valid your results are.

Also, asking what is the nature of one technique if one is indicating a fault and the others are not as in what is the severity and can it be looked into further? Can it be looked at during the next planned scheduled maintenance or is a serious fault that requires immediate action? Another great resource can be the transformer manufacturer to talk with them about their experience with the type of faults that have been diagnosed.

Watch the webinar on demand by registering here!