Neural-Network Air Knife Control for Uniform Strip Coating
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing manual operation methods for controlling coating weight in hot dipping processes are inaccurate, leading to deviations and poor surface quality due to limitations in directly deriving optimal air knife gaps and pressures for various steels and coating weights.
Innovation Solution
A coating weight control apparatus and method utilizing a prediction model trained with neural networks to derive absolute values of air knife gap and pressure based on input operation conditions, incorporating statistical methods and look-up tables to adjust and correct air knife conditions for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual operation methods are used to control coating weight by adjusting air knife gap and pressure, then operation simplicity is maintained, but coating weight control accuracy deteriorates
Solution Approach 1:
The system automatically derives optimal air knife gap and pressure values through the neural network prediction model without requiring manual operator intervention. The model self-adjusts parameters based on input conditions (steel type, line speed, target coating weight), eliminating the need for operator experience while maintaining high accuracy.
Solution Approach 2:
The patent replaces manual mechanical adjustment of air knife parameters with an intelligent prediction model based on neural networks. The system substitutes human operator judgment and manual calculation with automated computational derivation of optimal parameters, significantly improving accuracy while simplifying the operational process.
2Stability of the object's composition
If air knife gap and pressure are adjusted based on operator experience or setting tables, then ease of operation is maintained, but coating weight uniformity deteriorates
Solution Approach 1:
The system dynamically changes air knife parameters (gap and pressure) based on multiple input variables including steel type, line speed, and target coating weight. The neural network model processes these parameter changes automatically, ensuring optimal settings for each specific condition while maintaining coating uniformity across different steel varieties and operating speeds.
Solution Approach 2:
The control system is segmented into distinct functional modules: input condition reception, neural network prediction model processing, and output parameter derivation. This modular structure manages system complexity by separating concerns while integrating multiple control factors (gap and pressure) through a unified prediction framework.
3Manufacturing precision
If coating weight is controlled by adjusting air knife pressure alone according to line speed changes, then operation simplicity is maintained, but coating weight accuracy deteriorates
Solution Approach 1:
The prediction model serves multiple functions simultaneously: it adapts to different steel types, line speeds, and target coating weights while deriving both air knife gap and pressure parameters. This multi-functional approach replaces the limited single-parameter adjustment method, providing comprehensive control adaptability across various operating conditions and steel varieties.
Solution Approach 2:
The system transitions from one-dimensional control (pressure adjustment only) to two-dimensional control by simultaneously optimizing both air knife gap and pressure parameters. This dimensional expansion enables more precise coating weight control while maintaining adaptability to different steel types and operating conditions through the neural network's multi-parameter processing capability.
Data Source
AI summary
An apparatus for controlling coating weight coated on a strip by using an air knife disposed in a travelling direction of the strip in a continuous plating process in which the strip is dipped in a molten metal pot and is coated includes: a prediction model unit including a prediction model in which a neural network is trained with accumulated operation conditions; and an optimum air knife condition calculation unit configured to derive an absolute value of at least one of an air knife gap and an air knife pressure by using the prediction model based on an input operation condition.