Pipeline Inspection Gauge's Velocity Simulation Based on Pressure Differential Using Artificial Neural Networks.

Area of Science:

• Engineering
• Petroleum Engineering
• Data Science

Background:

• Industrial pipelines require regular inspection for failures like obstructions and deformations.
• Pigging, a non-destructive technique using a Pipeline Inspection Gauge (PIG), is standard for inspecting buried petroleum pipelines.
• Current PIG velocity measurements rely on odometer systems, which can be prone to failure and inaccuracies.

Purpose of the Study:

• To investigate the use of artificial neural networks for calculating PIG velocity based on pressure differential.
• To develop a novel method for PIG velocity estimation that enhances data reliability.

Main Methods:

• A prototype PIG was deployed in a test pipeline to collect velocity data via an odometer system.
• A supervisory system simultaneously recorded pressure data from the testing pipeline.
• Multilayer Perceptron (MLP) and Nonlinear Autoregressive Network with eXogenous Inputs (NARX) models were trained using the collected pressure and velocity data.

Main Results:

• The trained neural networks demonstrated the capability to model PIG velocity effectively from pressure differential measurements.
• The results indicate a strong correlation between pressure differentials and PIG velocity when using artificial neural networks.

Conclusions:

• Artificial neural networks provide a viable and potentially more reliable method for estimating PIG velocity compared to traditional odometer systems.
• This neural network approach can serve as a redundant system, improving the overall reliability of data acquired during pipeline inspection tests.