Quality Pattern Prediction for Closed-Loop Process Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data-driven models are inadequate to portray the dynamic change characteristics of product quality and stochastic characteristics of processes, are only applicable to open-loop systems, and cannot consider the influence of control inputs on changes in product quality.
Innovation Solution
A real-time prediction method based on a process dynamic pattern using a state space probability model, involving constructing a state space probability model, calculating probability density function distribution, performing optimized learning, and online prediction to obtain optimal model parameters, and introducing these into a closed-loop system for regulation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If data-driven models are used to predict product quality, then applicability and reflection of actual process changes are improved, but ability to portray dynamic change characteristics and stochastic characteristics deteriorates
Solution Approach 1:
The patent applies dynamics by transitioning from static data-driven models to dynamic models that capture temporal evolution. The state space model represents quality patterns as dynamic states that evolve over time according to differential equations, enabling the model to portray dynamic change characteristics while maintaining applicability to actual processes.
Solution Approach 2:
The patent changes parameters by introducing probabilistic parameters and stochastic elements into the modeling framework. The quality pattern is represented as a probability distribution rather than a deterministic value, allowing the model to capture stochastic characteristics while maintaining computational tractability through parameterized distributions.
2Ease of manufacture
If open-loop data-driven models are used, then model construction simplicity is improved, but consideration of control input influence on quality changes deteriorates
Solution Approach 1:
The patent applies feedback by incorporating control inputs into the state space model through feedback loops. The control inputs directly influence the evolution of quality patterns through the system dynamics equations, allowing the model to capture the influence of control actions on quality changes while maintaining a systematic modeling approach.
3Loss of information
If mechanistic models are used to predict product quality, then interpretability is improved, but model accuracy and reliability deteriorate
Solution Approach 1:
The patent uses an intermediary approach by introducing quality patterns as latent variables that mediate between process variables and product quality. These patterns are not directly observable but can be inferred from process data, providing both interpretability through the pattern concept and accuracy through probabilistic inference mechanisms.
Data Source
AI summary
The invention provides real-time prediction and regulation methods and systems of product quality based on a process dynamic pattern. The prediction method includes: constructing a state space probability model of quality pattern dynamic motion equation; calculating a probability density function distribution of a quality pattern according to the probability model; performing optimized learning on parameters of the state space probability model of quality pattern dynamic motion equation according to the probability density function distribution, to obtain optimal model parameters and an optimized state space probability model of quality-pattern dynamic motion equation; and performing online prediction on product quality indicators based on the optimized state space probability model. The invention can implement online prediction of product quality, and perform online regulation on product quality by using the analytical relationship between the control input and the quality pattern, to ensure that the product quality remains at an optimal level.