Process Simulation AI Integration for In-System Model Training

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing process simulation systems face limitations due to the isolation of intelligent model execution, which restricts the capability and functionality of process simulation models, necessitating a solution for integrating intelligent models within these systems to enhance performance and efficiency.

Innovation Solution

A computer-implemented method and apparatus that tightly integrate intelligent operations with process simulation models by using specially configured algorithms to identify and flag qualified data for training and deploying intelligent models, storing data in a common repository, and performing pre- and post-processing operations to generate and evaluate model predictions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If intelligent models are executed in isolation from process simulation systems, then model execution simplicity is maintained, but system capability and functionality are restricted

Engineering Contradiction:
Improvesystem capabilityVSAvoidsystem integration
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges intelligent model execution directly within the process simulation system by integrating model training, execution, and re-training capabilities into the simulation environment. This allows intelligent models to leverage process simulation data and capabilities while maintaining unified system management, thereby enhancing system capability without requiring separate external processing systems.

Inventive Principle:
Principle #5Merging (Combining)

2Productivity

If data is stored in separate external repositories for model training, then data accessibility is maintained, but network traffic and processing efficiency are reduced

Engineering Contradiction:
Improvecomputational efficiencyVSAvoiddata management architecture
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent consolidates data storage and processing within the process simulation system by implementing unified data structures that store both process simulation data and intelligent model training data in the same system environment. This eliminates the need for external data repositories and reduces network traffic, as all data operations occur within the unified system architecture.

Inventive Principle:
Principle #5Merging (Combining)

3Ease of operation

If intelligent models require external processes for training and deployment, then model independence is maintained, but system integration and operational efficiency are reduced

Engineering Contradiction:
Improveoperational efficiencyVSAvoidsystem architecture
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The patent integrates intelligent model training, validation, and deployment operations directly within the process simulation system. The system provides unified management of model lifecycles, allowing models to be trained on process simulation data, validated against simulation results, and deployed for real-time operations without requiring external processing environments. This streamlined approach enhances operational efficiency while maintaining manageable system architecture through unified control.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4446837A1Systems, apparatuses, methods, and computer program products for artificial intelligence and machine learning integration within a process simulation system
Publication Date: 2024.10.16 HONEYWELL INTERNATIONAL INC
  • EP4446837A1 patent drawingFigure 1
  • EP4446837A1 patent drawingFigure 2
  • EP4446837A1 patent drawingFigure 3

AI summary

Embodiments of the disclosure provide for intelligent model integration within a process simulation system. Some embodiments receive data associated with the operation of a plant, determine, using at least one specially configured algorithm and based on the received data, at least one qualifying dataset determined qualified to train an intelligent model, train the intelligent model using the at least one qualifying dataset, and deploy the trained intelligent model for use.