Time Series Clustering for Machine Behavior Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI-based monitoring systems for machines, particularly in oil or gas production, face challenges in efficiently and accurately classifying time series data due to complex rule formulation, the black-box nature of machine learning, and the need for large training datasets, making it difficult to address new behavior types and resolve conflicting classifications.
Innovation Solution
A device and method that segment time series data into clusters based on probabilistic models, utilize existing label information for classification, and involve user input to generate new label information, allowing for intuitive management of new behavior types and resolving conflicts through probabilistic models and user interaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pre-defined rules are used for classification, then classification accuracy can be maintained, but the complexity of formulating and maintaining rules increases
Solution Approach 1:
The system automatically generates classification rules by clustering time series data and extracting patterns, eliminating the need for manual rule formulation. The clustering algorithm self-organizes the data into groups, and the system automatically creates classification logic based on these clusters, making the system self-sufficient in rule generation.
Solution Approach 2:
Manual rule formulation is replaced with an automated machine learning approach. The system uses clustering algorithms and pattern recognition to automatically generate classification rules, substituting the mechanical process of manual rule creation with an automated computational process.
2Adaptability or versatility
If machine learning is used for classification, then adaptability to new behavior types improves, but the black-box nature reduces interpretability
Solution Approach 1:
The system provides feedback by presenting clustered results to users for verification and refinement. Users can review the automatically generated clusters and classifications, providing feedback that improves both the adaptability of the system to new patterns and the interpretability through user validation and adjustment.
Solution Approach 2:
The clustering process acts as an intermediary between raw data and final classification. By visualizing and analyzing the intermediate clustering results, users gain interpretability while the system maintains adaptability to new behavior types through the flexible clustering approach.
3Measurement precision
If machine learning is used for classification, then classification accuracy can be improved, but the requirement for large training datasets increases preparation time
Solution Approach 1:
The system performs preliminary clustering on the available data before final classification. This preliminary action organizes the data into meaningful groups, reducing the need for extensive manual training data preparation and enabling faster deployment while maintaining classification accuracy.
Solution Approach 2:
The system automatically clusters and classifies data without requiring large manually prepared training datasets. The clustering algorithm self-organizes the data and generates classification rules autonomously, eliminating the time-consuming process of manual training data preparation.
4Extent of automation
If machine learning is used for classification, then automated classification is achieved, but the ability to immediately address new behavior types is limited
Solution Approach 1:
The system dynamically adapts to new behavior types by continuously clustering incoming time series data. When new patterns are detected, the clustering algorithm automatically adjusts and creates new clusters, enabling the automated system to immediately address and adapt to new behavior types without retraining.
Data Source
AI summary
Device and method for analyzing time series data monitored on a machine, wherein the device segments the time series data into multiple time segments, determines a cluster of time segments estimated to have the same dynamics of the time series data, then classifies the cluster based on label information associated with at least one of the time segments, presents at least a part of the time series data of the cluster to a user if none of the time segments of the cluster has associated label information, classifies the cluster and generates label information associated with the time segments of the cluster based on a user input received in response to presentation of the time series data, where the generated label information indicates a result of classifying the cluster.