Errors such as vertical misalignment and wax depositions in transducers ports often result in drifts in diagnostic variables. However, there arises the challenge of being able to distinguish between errors, as some errors could induce the same drifts. This report aims to overcome this challenge via the use of machine learning models. Thus, helping end-users pinpoint, with high accuracy, the type of error that is responsible for causing drifts in flow meter data.
This report adopts a modelling approach to extract valuable information and correlations from flow meters’ diagnostic variables to enable condition-based monitoring. This report contains statistical analysis results on experimental data used. Machine learning models are constructed to predict and classify different types of error that have occurred within flow meters based on the relationship of different variables. The results and techniques, shown in the report, will be of interest to end-users who wish to see the benefits of using data-driven models and how modelling techniques can be used to enhance their understanding on the condition of flow meters. The high prediction results obtained from machine learning models can improve the process of fault diagnosis and aid end-user’s decision-making. The models used in this report are not limited to one specific meter type and therefore can be extended and applied to other flow meters and instrumentations.
Select your location
Bosnia and Herzegovina