Improved Monitoring of Flow Velocity along a Pipeline Using Acoustic Signals Measured by 1C Acoustic Sensors

28 Submissions
243 Views
$30,000 USD
Challenge under evaluation

Challenge overview

OVERVIEW

This Challenge is looking for ways to improve processing of Distributed Acoustic Sensing (DAS) data to efficiently monitor flow velocity along pipelines.

In recent years, there has been a growing interest in exploring DAS for monitoring oil and gas flow in surface and downhole pipelines, not the least because of DAS’s resistance to harsh environments. However, there is an urgent need to develop physics-based and/or machine learning-assisted signal processing workflows for continuous real-time interpretation of DAS distributed measurements.

This is a Prize Challenge which requires a written proposal to be submitted. The total payout may be up to $30,000 and there will be a guaranteed award of $20,000, with at least one award being no smaller than $10,000 and no award being smaller than $5,000. By submitting a proposal, the Solver grants the Seeker a right to use any information included in their proposal.

Submissions to this Challenge must be received by 11:59 PM (US Eastern Time) on January 28, 2024.

Login or register your interest to start solving!

Find out more about participation in Wazoku Crowd Challenges.

THE CHALLENGE

Real-time, non-intrusive pipeline surveillance allows continuous optimization of the fluid flow without frequent and labor-intensive performance tests; it can also potentially identify production anomalies (e.g., underperforming inlets, excessive local water production) hours before they are noticed downstream. However, the efficiency of online surveillance crucially depends on sensitivity and reliability of a monitoring system as well as the accuracy and robustness of processing of the collected data.

In recent years, there has been a growing interest in exploring Distributed Acoustic Sensing (DAS) for monitoring oil and gas flow in surface and downhole pipelines, owing to the DAS’s high spatial and temporal resolution measurements.

DAS is a sensing system based on light and consisting of laser and optical cables. In a typical case, the laser sends light pulses into the optical cable and then analyzes naturally scattered light returning to the receiver. The fiber itself is the sensing element enabling spatially continuous measurements comparable to those obtained by single-component (1C) accelerometers or geophones. This allows to detect acoustic signals passing through the cable over long distances with a spatial resolution of a few meters.

Our focus is on the application of DAS in monitoring pipelines where the fiber cable is attached to the pipe. In such scenarios, a widely used approach to process acoustic signals to gain insights about fluid flow velocity inside the pipe is by measuring the Speed of Sound (SoS) of the propagating pressure waves. The acoustic waves produced by the turbulent flow propagate at sonic velocity, but the presence of pipe flow creates a Doppler shift that modifies the upgoing and downgoing speed-of-sounds (SoS+ and SoS- respectively). As depicted in Figure 1 for a sample dataset, once the DAS measurements in a specific time window (and for a series of channels) are transferred to the frequency-wavenumber domain via the 2D Fast Fourier Transform FFT, the values of SoS+ and SoS- can then be determined from the FK-plot. By precisely measuring the up- and downgoing speed of sound, it becomes possible to estimate the flow velocity. Importantly, the Doppler-based velocity tracking allows for the quantitative estimation of the flow velocities for both single- and multi-phase flows.

Figure 1. An example of a F-K analysis for a pipe section to measure SoS in the flow (SoS+) and its opposite (SoS-) directions.

However, measuring flow velocity profile using the Doppler-based approach has limitations that must be addressed. The accuracy of this approach in determining flow velocity from DAS data relies on highly precise SoS values, which requires enhancing the strength of F-K analysis by incorporating more data. This can be achieved by either increasing the density of sensors (reducing the spacing between sensors/channels) while keeping the length of the spatial window small, or by extending the window size to encompass more sensors. The former approach is not always feasible as sensor spacing is set based on the best installation practices and DAS interrogator hardware constraints and is not entirely flexible. The latter translates into losing local information leading to a smoother, averaged solution over a larger pipe section. Essentially, the decision to enhance certainty by adding more sensors negatively influences the spatial resolution of the velocity profile along the pipe, revealing the trade-off between minimizing uncertainty and preserving spatial detail.

Figure 2 gives an example of such a trade-off, where to estimate the local velocity, the number of sensors used in the F-K analysis decreases from 500 on the left, to 300 at the center, and 100 on the right. Although lowering the number of sensors/channels is preferred as it helps produce more local estimates, as one can see that in this figure, it lowers the quality and clarity of the F-K plot, increasing therefore the uncertainty of the measured SoS. It should be noted that the FK approach is not the only technique that can be applied to estimate the SoS and/or the flow velocity, as there are some physics-based solutions that can potentially provide velocity estimations using the data measured by as few as three sensors.

Figure 2. An example of FK analysis for a pipe section using different spatial windows. From left to right the window size reduces from 500 to 100 sensors/channels.

Our team utilizes an experimental test facility which allows us to measure the pipe flow noise in a pipe using a commercial DAS interrogator.  In our test setup, the pipe is equipped with a fiber optics cable as well as three inflow points (via inflow control devices) at known locations along the pipe. The facility is also well instrumented with multiple pressure and temperature sensors along the pipe. The injection strategy can be changed by having a number of active inlets and injection rate as variables.

Figure 3. Schematic presentation of the flow facility used to measure the flow-induced acoustic signals.

The data the analysis of which we’d like to improve is hosted on Amazon Web Services (AWS) Simple Storage Service (S3). The tutorial Jupyter notebook (the link is provided in the supplementary information file) allows to either access and visualize the raw binary data directly from the AWS bucket or download the data first and then read and visualize it locally.

The AWS bucket includes ten labeled and two unlabeled datasets. In all the cases, water is being injected for three minutes and the system is in a steady state condition. The attached PDF document, supplementary_info.pdf, lists the experiment’s operational conditions:

  • In the first three datasets (1-3), there are no active control devices (inlets), so the bulk flow velocity remains constant along the pipe length.
  • In the next three examples (4-6), only one inlet with a known location is active and contributes to the flow, thus there is a step change in flow rate when passing the inlet location.
  • In the next four datasets (7-10), as more complex but realistic scenarios, there are multiple inlets that contribute to the flow.
  • In the unlabeled datasets, the total flow rate as well as the contribution of each inlet is unidentified.

Please refer to the attached PDF document (supplementary_info.pdf), in order to properly label the first ten datasets. For potential machine learning applications, 20-30 percent of any combinations of the labeled data together with the unlabeled data can be used for testing.

SOLUTION REQUIREMENTS

This Challenge is looking for new or improved solutions to process acoustic data and determine flow velocity reducing the uncertainty and spatial limitations of the conventional prior approaches.

We envision that the following approaches can be tried:

  • A better algorithm (including possible alternatives to the Doppler-based methods) or an AI model (supported by appropriate code and training data) can be designed to process the measurements available in the dataset and estimate the velocity profile along the pipe and indicate active inlets and their contribution to the flow  
  • Finding optimized applications of moving windows in both time and space domains (e.g., optimizing number of channels and/or time intervals)
  • The proposed technology should apply to a single-phase flow scenario, but, ideally, be expandable to two- and three-phase flow situations.

However, regardless of the specific approach, the proposed solution should meet the following Solution Requirements:

  • Bulk Velocity: If physics-based (e.g., using Doppler- or coherence-based formulations), the proposed technology should be able to determine the outflow velocity with less than 20% error (MPE). If AI-based, the proposed model should be able to estimate the outflow velocity with less than 10% error (MPE). Keep in mind that velocity is a continuous variable.
  • Inlet Status: The inlet locations are known, but the proposed technology should be able to determine which inlets are automatically, particularly for the unlabeled data.
  • Inlet Contribution (optional): A proposed solution will be acceptable if it allows estimate the total outflow and detect active inlets. However, quantifying the inlet’s contribution to the flow thus providing a velocity profile along the pipe would be highly desirable. 
  • Gauge data: It should be noted that the pressure and temperature data shared in the supplementary data file are only for informational purposes and should not be used as model inputs/features.

Solutions with Technology Readiness Levels (TRLs) 3-5 are invited.

This is a Prize Challenge, which has the following features:

  1. The best solution in this Prize Challenge has the opportunity to win the award of up to $30,000 for meeting all solution requirements, as solely determined by the Seeker.
  2. Awards will be contingent upon the theoretical evaluation (and code validation, if submitted) of the proposal by the Seeker.
  3. By submitting a proposal, the Solver grants the Seeker a right to use any information included in their proposal. There will be a guaranteed award of $20,000, with at least one award being no smaller than $10,000 and no award being smaller than $5,000.
  4. The Seeker may also issue “Honourable Mention” recognitions for notable submissions that are not selected for monetary awards.

YOUR SUBMISSION

Please login and register your interest, to complete the submission form.

The submitted proposals must be written in English and should include:

  1. Participation type – you will first be asked to inform us how you are participating in this challenge, as a Solver (Individual) or Solver (Organization).
  2. Solution Stage - the Technology Readiness Level (TRL) of your solution, TRL1-3 ideation stage, TRL4-6 proof of concept stage, TRL7-9 production ready stage.
  3. Problem & Opportunity - highlight the innovation in your approach to the Problem, its point of difference, and the specific advantages/benefits this brings  (up to 500 words).
  4. Solution Overview - detail the features of your solution and how they address the Solution Requirements (up to 500 words, there is space to add more, and to add any appropriate supporting data, diagrams, code, etc).
  5. Experience - Expertise, use cases and skills you or your organization have in relation to your proposed solution. The Seeker may wish to partner at the conclusion of the Challenge; please include a statement indicating your interest in partnering (up to 500 words).
  6. Solution Risks - any risks you see with your solution and how you would plan for this (up to 500 words).
  7. Timeline, capability and costs - describe what you think is required to deliver the solution, estimated time and cost (up to 500 words).
  8. References - provide links to any publications or press releases of relevance (up to 500 words).

Submissions developed solely with generative AI are not of interest.

Find out more about participation in Wazoku Crowd Challenges.

Submissions to this Challenge must be received by 11:59 PM (US Eastern Time) on January 28, 2024.

Late submissions will not be considered.

Your submission will be evaluated by the evaluation team first reviewing the information and content you have submitted at the submission form, with attachments used as additional context to your form submission. Submissions relying solely on attachments will receive less attention from the evaluation team.

After the Challenge submission due date, the Seeker will complete the review process and make a decision with regards to the winning solution(s) according to the timeline in the Challenge header. All Solvers who submit a proposal will be notified about the status of their submissions.

What is the Wazoku Crowd?

Imagine a world where any problem you have can be looked at by millions of expert, motivated eyes. The Wazoku Crowd makes this possible. With twenty years of experience in crowdsourcing, the Wazoku Crowd is the gold standard for open talent and external innovation solutions. Commercial, governmental and humanitarian organizations engage with the Wazoku Crowd to solve problems that can impact humankind in areas ranging from the environment to medical advancements.

Use the slider to explore how the Challenge process works:

Register

Review & Accept

Submit

Win

To start solving this Challenge, log in to the Challenge Center or register as a Solver

Visit the Wazoku Crowd Community to start the conversation