Active Learning Batch Data Alignment for Variable Industrial Processes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing batch process alignment methods, such as the online dynamic alignment method, face challenges with phase identity prediction errors, batch maturity prediction inconsistencies, and lack of robustness for industrial batch data with higher variability, requiring manual tuning of hyperparameters and lacking quantified measurement for alignment results, which hinders accurate batch modeling, monitoring, and control.
Innovation Solution
The implementation of intermediate derived pseudo time-series variables and alignment-guidance-only process variables, combined with purpose-built performance metrics and active learning workflows, to automate batch data alignment, providing systematic guidance for hyperparameter selection and ensuring consistent alignment results through an alignment environment library and supervised machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the online dynamic alignment method is used for batch data alignment, then alignment can be performed in real-time, but phase identity prediction errors and batch maturity prediction inconsistencies occur
Solution Approach 1:
The patent introduces an alignment score as an intermediary metric to evaluate alignment quality between batches. This alignment score serves as a mediator to quantify the similarity between batch trajectories, enabling more accurate phase identity prediction and batch maturity assessment by comparing alignment scores across different phase segments rather than relying solely on raw process variable comparisons
Solution Approach 2:
The patent replaces the traditional mechanical threshold-based phase transition detection with a machine learning classifier that uses alignment scores as input features. This substitution of mechanical detection methods with intelligent algorithms enables more accurate and consistent phase identity prediction, resolving the contradiction between real-time capability and prediction accuracy
2Measurement precision
If manual tuning of hyperparameters is performed to improve alignment accuracy, then alignment quality can be enhanced, but time consumption and operational complexity increase
Solution Approach 1:
The patent implements self-service through automated hyperparameter optimization using grid search and cross-validation. The system automatically selects optimal hyperparameters based on alignment score maximization, eliminating the need for manual tuning while maintaining high alignment quality. This automation directly addresses the contradiction by preserving measurement precision while eliminating time loss
Solution Approach 2:
The patent transforms the hyperparameter selection problem into an automated parameter optimization process. By systematically varying hyperparameters through grid search and evaluating their impact on alignment scores, the system automatically identifies optimal parameter combinations, resolving the contradiction between alignment quality and time consumption
3Device complexity
If traditional alignment methods are used without quantified measurement, then implementation is simpler, but alignment result evaluation and comparison are hindered
Solution Approach 1:
The patent introduces feedback through the alignment score metric, which provides quantitative measurement of alignment quality. This feedback mechanism enables evaluation and comparison of alignment results by calculating alignment scores for different batch pairs and phase segments. The alignment score serves as a feedback signal that guides phase identification and alignment optimization while maintaining implementation simplicity through a unified scoring framework
Data Source
AI summary
Computer-based methods and systems provide automated batch data alignment for a batch production industrial process. An example embodiment selects a reference batch from batch data for a subject industrial process and configures batch alignment settings. In turn, a seed model configured to predict alignment quality given settings for one or more alignment hyperparameters is constructed. Collectively the selected reference batch, the configured batch alignment settings, the constructed seed model, and a set of representative batches, representative of the batch data for the industrial process, are used to perform at least one of: (i) automated active learning, (ii) interactive active learning, and (iii) guided learning to determine settings for the one or more alignment hyperparameters. Then, a batch alignment is performed using the determined settings for the one or more alignment hyperparameters and the configured batch alignment settings. The resulting aligned batch data of the subject industrial process enables improved modeling and control of batch productions by the subject industrial process.