Non-Integrating Model Filter for Integrating Process Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional parametric internal model-based control (IMC) for integrating processes faces computational difficulties due to unbounded growth of predicted process variables and control outputs, leading to numerical instability and design complexities, especially with non-zero disturbances.
Innovation Solution
A non-integrating process model is used to approximate the behavior of the integrating process, applying a filter represented by (γs+1)m(εs+1) to minimize steady-state errors for step input disturbances, where γ is selected to optimize the lead time constant and ε is the filter time constant, and a control output is generated based on the predicted process variable.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional integrating process model is used in parametric IMC, then the model accurately represents the integrating process behavior, but the predicted process variable and control output grow without bound leading to numerical instability
Solution Approach 1:
The patent transforms the integrating process model parameters by introducing a filter with time constant ε and lead time constant γ. The filtered model uses modified parameters where the pole is moved from the origin to a location determined by ε, converting the integrating model into a stable first-order model that prevents unbounded growth while maintaining accuracy for practical control purposes.
Solution Approach 2:
The patent introduces a filter as an intermediary element between the conventional integrating model and the control implementation. This filter acts as a mediator that modifies the model output to prevent numerical instability while preserving the essential integrating behavior for control performance.
2Productivity
If an integrating process model is used with non-zero disturbances, then the model predicts process behavior, but the predicted process variable grows without bound due to the integrating nature
Solution Approach 1:
The patent modifies the model parameters by applying a filter that changes the integrating behavior into a stable first-order response. The filter time constant ε and lead time constant γ are selected to optimize the balance between maintaining prediction capability and preventing unbounded growth, thereby reducing computational complexity.
3Reliability
If a filter is applied to the inverse model to stabilize the system, then numerical stability is improved, but steady-state errors increase for step input disturbances
Solution Approach 1:
The patent optimizes the filter parameters ε and γ to achieve a balance between stability and accuracy. By carefully selecting these parameters, the filter stabilizes the system while minimizing steady-state errors for step input disturbances.
Solution Approach 2:
The patent introduces dynamic elements through the filter that allow the system to adapt its response characteristics. The filter provides dynamic compensation that reduces steady-state errors while maintaining numerical stability during transient and steady-state operations.
Data Source
AI summary
Various methods and systems for the parametric control of a process include representing the process with a process model used to generate future predictions of a process variable. In one embodiment, the process exhibits integrating behavior that is represented by a non-integrating process model. In another embodiment, an inverse of the model is filtered using a filter that includes a lead time constant that is selected to minimize a steady state error of the predicted process variable. In yet another embodiment, an array of output model values is revised or reindexed in response to a change in a time-varying parameter related to the process.