Plant Operating Condition Modeling for Multistep Parameter Setting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing plant operating condition simulation models require extensive manual adjustment and are dependent on operator experience, making it difficult to accurately simulate and predict the impact of manipulation parameters in complex multistep processes, especially for plants producing chemical or industrial products.
Innovation Solution
A plant operating condition setting support system that includes a learning device to acquire and learn a regression model from recorded state and manipulation parameters, allowing for the calculation of predicted output values, and an operating condition setting support device to determine optimal manipulation parameters using the learned regression model, thereby reducing the need for advanced simulators and improving precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a process simulator is used to simulate unit operations, then simulation capability is provided, but manual adjustment requires large numbers of man hours and precision depends on operator experience
Solution Approach 1:
The patent creates a digital twin (virtual model) that copies the actual plant's behavior characteristics. This virtual model is trained using machine learning on historical operation data, allowing it to replicate plant responses without manual simulation adjustment. The digital twin enables accurate predictions of manipulation parameter impacts while eliminating time-consuming manual calibration.
Solution Approach 2:
The system uses self-learning through machine learning algorithms that automatically improve simulation accuracy by processing historical operation data. The model autonomously identifies patterns and relationships between manipulation parameters and process outcomes without requiring continuous manual intervention or expert adjustment, enabling the system to serve itself in improving precision.
2Adaptability or versatility
If process simulators are combined to simulate multistep processes, then comprehensive simulation is achieved, but interaction complexity makes impact prediction difficult
Solution Approach 1:
The patent merges multiple process simulators into a unified virtual model that handles multistep processes as an integrated system. Rather than managing separate simulators for each unit operation, the machine learning model consolidates their functions, capturing inter-unit interactions and parameter dependencies in a single coherent framework that simplifies impact prediction.
Solution Approach 2:
The system transforms the complex physical-chemical process descriptions into learned parameter relationships through machine learning. By training on historical data, the model identifies effective parameter interactions and representations that simplify the complexity of multistep process simulations while maintaining comprehensive simulation capability across all units.
3Reliability
If manual adjustment of simulation models is performed, then model calibration is achieved, but the process requires large numbers of man hours and operator experience
Solution Approach 1:
The patent replaces manual mechanical adjustment processes with automated machine learning algorithms. Instead of operators manually tuning simulation parameters based on experience, the system uses computational algorithms to automatically train the virtual model on historical data, objectively determining optimal parameters without human intervention and eliminating dependence on operator skill levels.
Solution Approach 2:
The system implements feedback loops where the virtual model's predictions are continuously validated against actual plant performance data. This feedback mechanism automatically refines and calibrates the model, improving accuracy over time without requiring manual intervention. The system learns from discrepancies between predicted and actual outcomes, autonomously adjusting to maintain high reliability.
Data Source
AI summary
A plant operating condition setting support system for supporting the setting of a plant operating condition includes: a learning device that learns a regression model for calculating, from values of a plurality of state parameters indicating an operating condition of a plant and values of a plurality of manipulation parameters set to control an operation of the plant, a predicted value of an output indicating a result of operating the plant when the values of the plurality of manipulation parameters are set in the operating condition indicated by the values of the plurality of state parameters; and an operating condition setting support device that calculates the values of the plurality of manipulation parameters that should be set to control the operation of the plant, by using the regression model learned by the learning device.