Industrial Change Point Detection With Hybrid ML Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current change point detection (CPD) methods in industrial processes are hindered by the high human effort required for labeling data in supervised approaches and the complexity of selecting suitable algorithms and parameters in unsupervised methods, leading to limited practical application due to tedious manual steps and unavailability of necessary data.
Innovation Solution
A computer-implemented method combining unsupervised and supervised machine learning, where candidate change points are initially detected offline without labels, refined by human experts, and then used to train an online system for efficient detection with reduced labeling effort, enabling effective change point identification in industrial processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised CPD algorithms are used to improve detection accuracy, then measurement precision is improved, but human effort for labeling data increases significantly
Solution Approach 1:
The system performs preliminary unsupervised change point detection to generate candidate change points before the supervised learning phase. This preliminary action creates a structured dataset that reduces the manual labeling effort required for training, as experts only need to verify or correct algorithm-generated candidates rather than label all data points from scratch.
Solution Approach 2:
The system enables semi-automated change point detection where the algorithm independently generates candidate change points and the system automatically trains on these candidates. This self-service approach minimizes human intervention while maintaining high detection accuracy, as the system serves itself by generating its own training data through unsupervised detection.
2Ease of operation
If unsupervised CPD algorithms are used to reduce labeling effort, then ease of operation is improved, but detection accuracy deteriorates due to parameter selection complexity
Solution Approach 1:
The system merges unsupervised and supervised learning approaches into a hybrid framework. The unsupervised component generates candidate change points with minimal human effort, while the supervised component refines detection accuracy using expert-labeled candidates. This combination achieves both ease of operation and high detection accuracy by leveraging the strengths of both methodologies.
Solution Approach 2:
The unsupervised detection algorithm performs preliminary analysis to identify candidate change points before expert verification. This preliminary action filters the data to only those points requiring expert attention, reducing overall labeling effort while maintaining accuracy through subsequent supervised refinement of the candidates.
3Measurement precision
If more expert verification is performed to improve change point selection accuracy, then measurement precision is improved, but productivity decreases due to increased manual steps
Solution Approach 1:
The system applies partial expert verification only to candidate change points generated by the unsupervised algorithm, rather than requiring expert review of all data points. This partial action approach maintains high detection accuracy for critical candidates while preserving productivity by avoiding exhaustive manual verification of every potential change point.
Solution Approach 2:
The system creates a copy of the signal data and applies unsupervised detection to generate candidate change points, which are then verified by experts. This copying approach allows parallel processing where algorithmic detection and expert verification occur on separate data copies, improving productivity by eliminating sequential bottlenecks.
Data Source
AI summary
A method for detecting change points, CPs, in a signal of a process automation system, includes, in an offline learning phase, unsupervised, candidate CPs on at least one offline signal using unsupervised detection method are detected, CPs are selected from the candidate CPs; the selected CPs are provided to a supervised process; in the supervised process, an offline machine-learning (ML) system is trained to refine CPs from the selected CPs using a supervised machine learning method; a training data set for an online ML system is created using the offline ML system by projecting the refined CPs on the signal; the online ML system is trained in a supervised manner, using the created training data set; and after the offline learning phase, CPs are detected using the trained online ML system.