Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Temporal Data and Short-Time Averages Improve Multiphase Mass Flow Metering

Amanda Nyholm, Yessica Arellano, Jinyu Liu, Damian Krakowiak, Pierluigi Salvo Rossi

TL;DR

This work tackles the inaccuracy of Coriolis mass flowmeters in multiphase (air–water–oil) flows by leveraging temporal information through ML corrections. It compares a baseline MLP, a windowed MLP, and a 1D-CNN on a large three-phase dataset, showing that short-time averaging within experiments and preserving temporal structure yield substantial accuracy gains, with the 0.25 Hz CNN achieving about 95% of relative errors below 13% and a MAPE around 4.3%. The study demonstrates robustness across data splits and random seeds, and finds that downsampling to 4 seconds provides the best performance, while latency remains suitable for real-time deployment. Overall, the results highlight the practical value of time-series-aware corrections for improving CMF accuracy in multiphase conditions using only user-accessible inputs.

Abstract

Reliable flow measurements are essential in many industries, but current instruments often fail to accurately estimate multiphase flows, which are frequently encountered in real-world operations. Combining machine learning (ML) algorithms with accurate single-phase flowmeters has therefore received extensive research attention in recent years. The Coriolis mass flowmeter is a widely used single-phase meter that provides direct mass flow measurements, which ML models can be trained to correct, thereby reducing measurement errors in multiphase conditions. This paper demonstrates that preserving temporal information significantly improves model performance in such scenarios. We compare a multilayer perceptron, a windowed multilayer perceptron, and a convolutional neural network (CNN) on three-phase air-water-oil flow data from 342 experiments. Whereas prior work typically compresses each experiment into a single averaged sample, we instead compute short-time averages from within each experiment and train models that preserve temporal information at several downsampling intervals. The CNN performed best at 0.25 Hz with approximately 95 % of relative errors below 13 %, a normalized root mean squared error of 0.03, and a mean absolute percentage error of approximately 4.3 %, clearly outperforming the best single-averaged model and demonstrating that short-time averaging within individual experiments is preferable. Results are consistent across multiple data splits and random seeds, demonstrating robustness.

Temporal Data and Short-Time Averages Improve Multiphase Mass Flow Metering

TL;DR

This work tackles the inaccuracy of Coriolis mass flowmeters in multiphase (air–water–oil) flows by leveraging temporal information through ML corrections. It compares a baseline MLP, a windowed MLP, and a 1D-CNN on a large three-phase dataset, showing that short-time averaging within experiments and preserving temporal structure yield substantial accuracy gains, with the 0.25 Hz CNN achieving about 95% of relative errors below 13% and a MAPE around 4.3%. The study demonstrates robustness across data splits and random seeds, and finds that downsampling to 4 seconds provides the best performance, while latency remains suitable for real-time deployment. Overall, the results highlight the practical value of time-series-aware corrections for improving CMF accuracy in multiphase conditions using only user-accessible inputs.

Abstract

Reliable flow measurements are essential in many industries, but current instruments often fail to accurately estimate multiphase flows, which are frequently encountered in real-world operations. Combining machine learning (ML) algorithms with accurate single-phase flowmeters has therefore received extensive research attention in recent years. The Coriolis mass flowmeter is a widely used single-phase meter that provides direct mass flow measurements, which ML models can be trained to correct, thereby reducing measurement errors in multiphase conditions. This paper demonstrates that preserving temporal information significantly improves model performance in such scenarios. We compare a multilayer perceptron, a windowed multilayer perceptron, and a convolutional neural network (CNN) on three-phase air-water-oil flow data from 342 experiments. Whereas prior work typically compresses each experiment into a single averaged sample, we instead compute short-time averages from within each experiment and train models that preserve temporal information at several downsampling intervals. The CNN performed best at 0.25 Hz with approximately 95 % of relative errors below 13 %, a normalized root mean squared error of 0.03, and a mean absolute percentage error of approximately 4.3 %, clearly outperforming the best single-averaged model and demonstrating that short-time averaging within individual experiments is preferable. Results are consistent across multiple data splits and random seeds, demonstrating robustness.
Paper Structure (20 sections, 6 equations, 6 figures, 7 tables)

This paper contains 20 sections, 6 equations, 6 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Previous Work
  3. Conventional Coriolis Mass Flow Metering Models
  4. Sequence Models in Flow Measurements
  5. Contributions
  6. Experimental Setup and Data
  7. Preliminaries
  8. Multivariate Time Series
  9. Multilayer Perceptron Models
  10. Windowed Multilayer Perceptron
  11. Convolutional Neural Networks
  12. Evaluation Metrics
  13. Implementation
  14. Results and Discussion
  15. Model Performance
  16. ...and 5 more sections

Figures (6)

  • Figure 1: Representative time series showing (from top to bottom) the true mass flow rate, apparent mass flow rate, GVF, and pressure for four consecutive experiments (columns).
  • Figure 2: Illustration of the experimental setup.
  • Figure 3: Illustration of the multiphase skid, rotated 90 degrees clockwise for compactness.
  • Figure 4: Relative errors versus overall mass flow rate for the best odd model, compared with the original errors on the same test points.
  • Figure 5: Coverage percentiles averaged per seed for both data splits. Results are shown for the 4s CNN (top) and 60s CNN (bottom) models. Shading indicates $\pm 1$ standard deviation across seeds.
  • ...and 1 more figures