Time Series Partitioning via Distance Matrix Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive maintenance methods for equipment health monitoring require specific knowledge about the database and failure modes, limiting their applicability to various systems and databases.
Innovation Solution
A partitioning method that systematically extracts group information from time series data using a distance matrix representation as images, segmented by a U-NET-type convolutional neural network, allowing for automatic and unsupervised classification of severity classes without prior knowledge of failure modes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a partitioning method using joint optimisation of temporal coherence criterion and energy criterion is used, then the convergence level of energy criterion is improved, but specific knowledge about failure modes is required
Solution Approach 1:
The patent segments the time series data into multiple severity classes by dividing the observation matrix into sub-matrices corresponding to different failure modes. This segmentation allows the system to automatically identify and separate different degradation patterns without requiring prior knowledge of failure modes, thereby improving both measurement precision and adaptability.
Solution Approach 2:
The patent transforms the one-dimensional time series data into two-dimensional images by calculating distance matrices and visualizing them as images where pixels represent distance values. This dimensional transformation enables the application of image processing techniques and automatic segmentation algorithms, eliminating the need for manual definition of severity classes and failure mode knowledge.
2Device complexity
If deep learning algorithms are used for diagnostic and prognostic steps, then the signature extraction step can be avoided, but the need for specific knowledge about failure modes remains
Solution Approach 1:
The patent implements a self-service mechanism where the system automatically discovers and defines severity classes and failure modes through unsupervised learning from the data itself. The distance matrix calculation and image-based segmentation enable the system to identify patterns and groupings without external guidance or prior knowledge, making the deep learning approach truly autonomous and adaptable to any equipment type.
Data Source
AI summary
A partitioning method includes the steps of acquiring an observation matrix including time series (x1 x2, . . . , xp); for each time series calculating a distance matrix comprising distance values between the elements of the time series, then generating the primary image on the basis of the distance matrix; implementing a learning algorithm for segmenting the primary image so as to obtain a segmented image; defining, on the basis of the segmented image, a primary boundary signal representative of the boundaries; and merging the primary boundary signals in order to obtain a global boundary signal, and defining classes on the basis of the global boundary signal.