Oil & gas
the value that lies within that data, advanced modelling techniques such as machine learning models have become increasingly important, where information such as the condition of instruments, fault detection and diagnosis, and future performance forecasting can be obtained. However, previous studies have shown
E
that the performance of flow meters which belong to the same technology subgroup (e.g., Coriolis, Electromagnetic, Ultrasonic), can still vary under the same operating conditions due to differing build and manufacturing materials, therefore making it less than straightforward to pinpoint the reasons behind sub-optimal performance. In addition, the performance of flow meters is dependent on operating conditions such as flow profile and pipe configurations. TÜV SÜD National Engineering Laboratory’s
data acquisition systems have logged and archived 20 years’ worth of data, detailing various flow meters’ performance, test facility configuration and operating conditions. The data contains many factors that could influence the performance of flow meters, for example pipe configuration (straight vs. u shaped), the distance from the flow disturbance, as well as the meter size. To offer insight into the reasons for sub-
optimal performance, TÜV SÜD NEL conducted experiments on four meters of different sizes within two technology subgroups, where they were deliberately set up differently and exposed to various conditions and disturbances. The study revisited the historical data
gathered from two electromagnetic meters and two ultrasonic meters (USMs) of two sizes which were exposed to different set ups, where factors such as different pipelines (straight vs. u-shaped), having disturbance in the system (normal vs. 75 per cent closed gate valve) and varying distances from disturbance were considered. Simple plots were initially produced to compare
the performance of these meters, where only a limited amount of insight was obtained and where it was challenging to interpret the degree of impact of changing operating conditions. Motivated by this, the study team used various statistical techniques and machine learning models in an attempt to better compare and visualise the performance of the meters under each set up, as well as detecting the presence of a closed gate valve within the system. These advanced modelling techniques allowed for more comprehensive insights to be extracted from the data and provided a more conclusive representation of the performance of each meter. It was found that the electromagnetic
meters outperformed the USMs by showing 26
Optimising flOw meter data fOr OperatiOnal and strategic decisiOn making
Dr Yanfeng Liang, mathematician at TÜV SÜD National Engineering Laboratory, highlights the challenges in creating ‘universally applicable’ data- driven models for flow meters.
the least degree of drift from the baseline value measured by the percentage error in volume flow. It was also observed that both of the full-bore electromagnetic meters and USMs had a similar drift pattern when changing from one operating condition to another. The effects of changing pipeline and having a closed gate valve in the system were more predominant in full-bore meters, where the worst performance was observed in the u-shaped pipeline. The presence of a closed gate valve can
disturb the reliability and the accuracy of outputs from flow meters. Therefore, having the ability to detect such disturbance will ensure any issues are rectified promptly and minimise the risk of errors propagating and affecting other related measurement outputs. Consequently, the study team trained and built seven machine learning models under three scenarios, where a mixture of data coming from different meters was used to detect a closed gate valve within the system, based entirely on the underlying patterns and trends in data.
March 2022 Instrumentation Monthly
very day, vast amounts of data are generated across different sectors, containing valuable information that could aid businesses in their operational and strategic decision-making. In order to unlock and extract
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74
