Constrained Surrogate Modeling for Feasible Chemical Process Outputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Surrogate models for chemical production processes lack constraints based on real-world physical conditions, leading to unrealistic output predictions that exceed equipment capacity and are not feasible in actual production environments.
Innovation Solution
A machine-learned surrogate model is developed that uses training data from both simulations and actual production data, incorporating constraints to ensure output predictions align with feasible physical limitations of equipment, such as capacity constraints, to provide plausible and realistic output estimates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a surrogate model is used to predict chemical production outputs, then computational speed is improved, but output feasibility is worsened because the model suggests outputs that exceed equipment capacity
Solution Approach 1:
The patent applies feedback by using the simulation model to validate surrogate model predictions. The simulation model, which incorporates physical constraints and equipment capacities, provides feedback to correct or reject surrogate model outputs that are not feasible in the real world. This feedback loop ensures that the fast surrogate model predictions are adjusted to match actual production capabilities.
Solution Approach 2:
The patent introduces a constraint model as an intermediary between the surrogate model and the final output. This constraint model, built from simulation data and real production data, acts as a mediator that filters and adjusts the surrogate model's predictions to ensure they comply with physical constraints and equipment capacities before being presented as feasible solutions.
2Device complexity
If a surrogate model is trained without physical constraints, then model complexity is reduced, but manufacturing precision is worsened because the model cannot distinguish viable from non-viable solutions
Solution Approach 1:
The patent applies preliminary action by pre-processing simulation data to extract and encode physical constraints and equipment capacities into the constraint model before training the surrogate model. This preliminary preparation of constraint information allows the surrogate model to learn from data that already incorporates feasibility boundaries, improving its ability to distinguish viable solutions without significantly increasing model complexity.
Solution Approach 2:
The patent changes parameters by transforming the surrogate model's output parameters to align with real-world feasibility. Instead of directly outputting unconstrained predictions, the model adjusts its output parameters based on the constraint model, which encodes equipment capacities and physical limitations. This parameter transformation enables the model to produce viable solutions while maintaining relatively simple model architecture.
Data Source
AI summary
Aspects of the technology described herein comprise a surrogate model for a chemical production process. A surrogate model is a machine learned model that uses a collection of inputs and outputs from a simulation of the chemical production process and/or actual production data as training data. Once trained, the surrogate model can estimate an output of a chemical production process given an input to the process. Surrogate models are not directly constrained by physical conditions in a plant. This can cause them to suggest optimized outputs that the not possible to produce in the real world. It is a significant challenge to train a surrogate model to only produce outputs that are possible. The technology described herein improves upon previous surrogate models by constraining the output of the surrogate model to outputs that are possible in the real world.