Nonlinear Process Modeling for Pulp and Paper Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current industrial processes lack effective methods to detect and model nonlinearity in chemical processes, leading to inefficiencies and inaccuracies in process control and optimization.
Innovation Solution
A computer-implemented method that automatically determines and configures a non-linear or quasi-linear prediction model based on sensor data, optimizing resource consumption and process control by training models using techniques like PLS regression and XGBoost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If data-driven empirical modeling approach is used, then model simplicity and interpretability are improved, but accuracy in modeling non-linear industrial processes deteriorates
Solution Approach 1:
The patent implements a dynamic model selection approach where the system automatically switches between linear and non-linear models based on real-time process conditions. The model selector monitors process characteristics and dynamically adjusts the modeling approach, allowing the system to maintain simplicity for linear regions while achieving high accuracy for non-linear regions through adaptive model switching.
Solution Approach 2:
The patent divides the process modeling into multiple segments by identifying different operating regimes and selecting appropriate models for each segment. The process is segmented into linear and non-linear regions, with separate models trained and applied to each segment, allowing the system to optimize for both simplicity and accuracy across different process conditions.
2Measurement precision
If non-linear models are used to accurately model industrial processes, then model accuracy is improved, but computational complexity and cost increase
Solution Approach 1:
The system dynamically adjusts computational complexity by selecting between simple linear models and complex non-linear models based on the specific process conditions and required accuracy. This dynamic model selection allows the system to use computational resources efficiently, applying complex models only when necessary while using simpler models for linear regions.
Solution Approach 2:
The patent changes the model parameters and structure based on process conditions. The system adjusts the degree of non-linearity, model complexity, and computational methods according to the specific operating regime, allowing accurate modeling of non-linear processes only when needed while maintaining computational efficiency for linear regions.
3Reliability
If first-principle-based modeling is used, then physical and chemical accuracy is improved, but modeling cost and complexity increase
Solution Approach 1:
The patent segments the modeling approach by applying first-principle-based models to specific critical processes where physical and chemical accuracy is essential, while using data-driven models for other processes where computational cost is a concern. This segmented approach allows the system to optimize for reliability where needed while controlling costs in other areas.
Solution Approach 2:
The system implements a universal modeling framework that can handle both first-principle-based and data-driven approaches within the same platform. The model selector and process simulator provide multi-functional capabilities, supporting both rigorous physical-chemical modeling and flexible data-driven modeling, allowing users to choose the appropriate approach based on their specific needs and resource constraints.
Data Source
AI summary
A method, system, and computer program product are described capable of controlling an industrial based chemical process through accessing sensor data, pre-processing accessed sensor data through an automated process, forming an initial prediction model of the chemical process, and automatically determining linearity or non-linearity of the initial prediction model. As a function of quantitatively measured linearity, being non-linear, quasilinear, or linear, automatically train the initial prediction model, then deploy the trained prediction model in a manner that controls the subject chemical process at the industrial plant, including optimizing consumption of a certain resource. Data-driven modeling of non-linear, continuous-like industrial or chemical processes results.