Neural Network Predicting Reservoir Simulation Convergence Failures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer simulations used to model fluid flow in subterranean reservoirs often fail to converge due to input data characteristics or simulation limitations, leading to premature termination and unnecessary resource expenditure.
Innovation Solution
A machine learning-based system that uses a neural network to predict convergence failures by analyzing input data from previous simulations, generating notifications for users, and modifying data to prevent failures, thereby preventing premature terminations and optimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computer simulations are performed using iterative calculation processes, then the characteristics of physical environments can be modeled, but the simulations may fail to converge and be prematurely terminated
Solution Approach 1:
The system performs preliminary analysis of input data using a trained machine learning model before executing the full simulation. This preliminary action identifies characteristics likely to cause convergence failures, allowing users to modify input data beforehand and avoid premature termination of simulations.
Solution Approach 2:
The system uses feedback from historical simulation outcomes (successful and failed simulations) to train the machine learning model. This feedback loop enables the model to learn patterns in input data that lead to convergence failures, improving its ability to predict and prevent future failures.
2Loss of time
If simulations are run with problematic input data, then computational resources are consumed, but the simulations terminate prematurely without producing results
Solution Approach 1:
The system performs preliminary analysis of input data using a trained machine learning model before executing the full simulation. This preliminary action identifies characteristics likely to cause convergence failures, allowing users to modify input data beforehand and avoid premature termination of simulations.
Solution Approach 2:
The system converts the harmful effect of convergence failures (wasted computational resources) into a benefit by using historical failure data to train a predictive model. This model then prevents future failures, turning past resource waste into improved efficiency.
3Reliability
If the machine learning model predicts low likelihood of successful simulation, then users can be warned beforehand, but the simulation may still be executed consuming resources
Solution Approach 1:
The system uses feedback from historical simulation outcomes (successful and failed simulations) to train the machine learning model. This feedback loop enables the model to learn patterns in input data that lead to convergence failures, improving its ability to predict and prevent future failures.
Solution Approach 2:
The system enables users to self-diagnose potential simulation failures by providing predictions and recommendations. Users can independently assess whether their input data is likely to cause convergence issues and make corrections before execution, reducing reliance on manual review processes.
Data Source
AI summary
In an example method, a system obtains first data indicating a plurality of properties of a first reservoir. The system determines, using a computerized neural network, a first metric representing a likelihood that a first computer simulation of the first reservoir can be performed to completion using a computer model and the first data. Further, the system determines that the first metric is less than a threshold level, and in response, generates a notification indicating the first metric for presentation to a user.