Modern ultrasonic flow meters (USMs) output vast quantities of digital data in the form of individual process values commonly referred to as diagnostics. In theory, this information can be used to determine meter performance as well as infer process and installation conditions. For example, any potential flow measurement errors resulting from incorrect installation, particle deposition (e.g. wax build up on the transducer ports) or the presence of an unexpected second fluid phase, usually manifest as measurement errors.
However, without extensive field experience, it can be challenging for end-users to interpret these values and relate them to specific sources of error due to the fact that the same diagnostic drift of a variable may be observed for multiple scenarios. We therefore cannot rely on simply trending and alarming on one specific variable. To obtain reliable and actionable information, a combination of digital variables and their associated response to given scenarios must therefore be evaluated, which in turn renders basic human observations largely subjective and unreliable.
Using our flow laboratories, we have conducted experimentation on multiple brands of USM, in which we intentionally exposed them to undesirable operating conditions while logging the available diagnostic values output via the meter’s fieldbus port. Through data driven modelling techniques, our Digital Services team were able to identify hidden correlations between the various digital process values and were able to build a tool which can identify specific fault conditions and alert end-users via simple graphical interfaces.
The output of this project has demonstrated the potential for end-users to increase operational efficiency through harnessing data, which is already available to them through standard USM transmitters. Equipped with the ability to identify a specific fault, there is no longer a need for end-users to shut down production, remove a meter from the pipeline and conduct multiple diagnostic tests before a solution can be found. This in turn has significant long-term cost-saving implications and the potential to increase plant operation efficiency. The tool can be applied to both historical data sets and live streamed data.