Continuous Annealing Line Predictive Control for Dead Time
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current process control systems in continuous processing lines face significant dead time issues, leading to suboptimal control actions and potential downgrades in product quality due to inadequate consideration of forthcoming process dynamics and constraints, especially when producing diverse strip dimensions and materials with transient mechanical, surface, and geometry properties.
Innovation Solution
The Dynamic Property Predictive Control (DPPC) method integrates Materials Property Models (MPM) and Dynamic Process Models (DPM) to predict and optimize process settings by iteratively examining property and process predictions against targets, incorporating forthcoming process dynamics and constraints, allowing for real-time adjustment of process settings to achieve desired mechanical, surface, and geometry properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor measurement at the end of process sections is used for property control, then the control can detect deviations, but the large dead time in the control loop leads to substantial material losses before control can remedy the deviations
Solution Approach 1:
The patent applies preliminary action by using a property model to predict intermediate and final strip properties before the actual processing is complete. This allows the control system to identify potential quality deviations in advance and adjust process settings proactively, rather than waiting for sensor measurements from the end of the line. The prediction is performed based on process histories and scheduled strip transitions, enabling early intervention to prevent material losses.
2Loss of time
If property model prediction is used to compensate dead time, then control can adjust processes in advance, but the control actions do not consider forthcoming process dynamics and constraints, leading to infeasible and unstable operation
Solution Approach 1:
The patent implements feedback by iteratively examining the property model prediction results against operational constraints and forthcoming process dynamics. The control system continuously compares predicted properties with target specifications and adjusts process settings based on this feedback loop. This ensures that control actions are not only proactive but also feasible and stable, considering the actual capabilities and constraints of the processing line.
Solution Approach 2:
The patent applies dynamics by considering scheduled strip transitions and forthcoming process dynamics in the control strategy. The system adapts process settings dynamically based on the specific characteristics of upcoming strips, such as thickness changes that require line speed adjustments to prevent heating buckles or coating thickness variations. This dynamic adaptation ensures control actions remain feasible under varying operational conditions.
3Productivity
If classical set-up and feed-back models are used, then control is adequate for large batch production, but the long dead time results in downgraded product when producing unique strips with transient properties
Solution Approach 1:
The patent applies parameter changes by transitioning from classical set-up and feed-back models to a dynamic property model-based control system. The property model uses process histories and scheduled strip transitions to predict intermediate and final properties, enabling the system to adapt to unique strips with transient properties. This parameter change in the control methodology allows the system to maintain high manufacturing precision while producing diverse, small-batch orders with just-in-time delivery requirements.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
This invention relates to a method for operating a continuous processing line comprising an annealing step for the production of continuously processed rolled steel strip using a computer aided dynamic property predictive control model.