Determining the Importance of Diagnostic Variables

Sector Challenge

As manufacturers of instrumentation continue to embrace digital technologies in place of traditional analogue electrical signals as a means of communicating data, end-users can now be faced with interpreting screens of data with multiple process variables which, as well as providing the primary variable of interest (flow rate, pressure, temperature), include device specific diagnostic values. Unless reviewed by a subject expert these can either be meaningless or misinterpreted. In many cases, the usefulness of these diagnostic variables can be very application specific and different stimuli can cause similar responses in any one variable.

How Can TÜV SÜD National Engineering Laboratory Help?

Using our flow laboratories, we have conducted experimentation on specific flow meters, during which we intentionally exposed them to undesirable operating conditions while logging all available diagnostic values from the device. Through detailed analysis and data-driven modelling our Digital Metrology team were able to identify subtle and hidden correlations between the various digital process variables and the physical effects experienced by the meter. The diagram below demonstrates the output of an interactive tool which was subsequently developed to allow end-users to interact with the data and identify which diagnostic variables, and combinations thereof, are important in identifying a specific operating condition.

Sector Impact

By combining the facilities of the UK’s National Standard for Flow Measurement with emerging data science techniques, we have been able to provide end-users with a means of interpreting meter-specific diagnostic data and relating the response of said data to specific operating conditions. The experiments and data science techniques utilised to produce the above results can be redeployed to address specific end-user requirements on a case by case basis, as well as form the foundation of a much larger condition-based monitoring system. These approaches can also be used to provide the back-end functionality to a much simpler interface that simply tells the user what the prevailing fault condition is.