Proxy Modeling for Reactive Transport Sensitivity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Reactive transport modeling (RTM) techniques are computationally intensive and time-consuming, particularly for sensitivity analysis and model calibration, due to the need for manual iteration and extensive computational resources, which limits their efficiency and accuracy.
Innovation Solution
A machine learning-based proxy model using neural networks is developed to surrogate process-based RTM, allowing for automated parameter selection, batch model running, and parallel computations, significantly reducing computational time and introducing accuracy while avoiding manual bias.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If process-based reactive transport modeling (RTM) is used for sensitivity analysis and model calibration, then accurate predictions of chemical reactions and fluid transportation can be obtained, but computational time and resource requirements increase significantly
Solution Approach 1:
The patent creates a proxy model that copies the essential behavior of the complex RTM system. A neural network is trained on RTM simulation data to create a surrogate model that reproduces RTM predictions with high accuracy (within 5% error) but executes thousands of times faster, enabling sensitivity analysis and model calibration that were previously computationally prohibitive
Solution Approach 2:
The patent performs preliminary action by pre-computing a training dataset using RTM simulations across a range of parameter values before the actual sensitivity analysis or calibration. This training data is then used to build the proxy model in advance, so that subsequent predictions and optimizations can be made rapidly without repeatedly running full RTM simulations
2Manufacturing precision
If manual iteration is performed for model calibration in RTM, then model parameters can be adjusted to fit observed data, but the process becomes time-consuming and prone to human bias
Solution Approach 1:
The patent implements self-service by creating an automated calibration workflow where the proxy model evaluates parameter sets and the optimization algorithm iteratively adjusts parameters without human intervention. The system automatically compares predictions against observed data, identifies parameter improvements, and converges on optimal parameter values, eliminating manual iteration and human bias while maintaining or improving calibration accuracy
Solution Approach 2:
The patent incorporates feedback mechanisms where the proxy model continuously evaluates parameter sets against observed data and provides performance feedback to the optimization algorithm. This automated feedback loop enables systematic exploration of parameter space and guides the calibration process toward optimal solutions without manual intervention, significantly improving both efficiency and objectivity
3Adaptability or versatility
If extensive computational resources are allocated for RTM sensitivity analysis, then comprehensive parameter exploration can be performed, but the cost and time requirements become prohibitive
Solution Approach 1:
The patent uses the proxy model to copy RTM behavior for sensitivity analysis, enabling comprehensive exploration of parameter space without the computational burden of full RTM simulations. The proxy model maintains sufficient accuracy to identify sensitive parameters and their effects, allowing extensive parameter exploration with minimal computational resources
Solution Approach 2:
The patent applies partial action by using the proxy model for the extensive iterations needed in sensitivity analysis while reserving full RTM simulations for final verification of key results. This approach performs sufficient analysis to identify sensitive parameters and their ranges without the excessive computational cost of running full RTM for every parameter combination explored
Data Source
AI summary
Systems and methods include a computer-implemented method for random selection and use of observation cells. Observation cells are randomly selected from a model of process-based reactive transport modeling (RTM). The observation cells are incorporated into a neural network for proxy modeling. A set of parameter-specific proxy models represented by a neural network is trained. Each parameter-specific proxy model corresponds to a specific RTM parameter from a set of RTM parameters. Blind tests are performed using the set of parameter-specific proxy models, where each blind test tests a specific one of the parameter-specific proxy models. Predictions are generated using the set of parameter-specific proxy models. 3-dimensional interpolation the observation cells is performed.