Anomaly Detection Using Singular Value Decomposition for Multi-Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detecting anomalies in mechanical systems with unknown locations is inefficient, especially when using the Mahalanobis-Taguchi method, as it reduces high-dimensional data to a scalar value, leading to decreased information and inferior accuracy in identifying anomalous parts.
Innovation Solution
An anomaly detecting device that uses a singular value decomposition unit to process data from multiple sensors, determining anomalies by analyzing the variance-covariance matrix and identifying anomalous parts based on diagonal elements of a matrix representing the relationship between normal and anomalous data matrices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If high-dimensional measurement data is reduced to Mahalanobis distance scalar value for anomaly detection, then detection efficiency is improved, but information loss occurs and accuracy in identifying anomalous parts deteriorates
Solution Approach 1:
The patent transforms the anomaly detection approach by introducing a new dimensional representation. Instead of reducing N-dimensional data to a single scalar Mahalanobis distance, the invention computes multiple Mahalanobis distances corresponding to different dimensional components, thereby preserving the dimensional structure and information content while enabling efficient anomaly detection through comparative analysis of these multiple distance metrics.
2Measurement precision
If N-dimensional measurement data is fully utilized for anomaly detection, then accuracy in identifying anomalous parts is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent segments the N-dimensional measurement data into multiple dimensional components, computing Mahalanobis distances for each component separately. This segmentation approach allows the system to process high-dimensional data by breaking it down into manageable parts, maintaining measurement precision through component-wise analysis while reducing overall computational complexity compared to processing the full N-dimensional data as a single unit.
3Speed
If anomaly detection is performed without location identification, then detection speed is improved, but efficiency in locating and addressing anomalies deteriorates
Solution Approach 1:
The patent applies local quality analysis by examining Mahalanobis distances at different dimensional locations within the measurement data. By computing and comparing distances for individual dimensional components rather than treating the data as a homogeneous whole, the invention enables both rapid anomaly detection and precise location identification, as anomalies manifest as localized deviations in specific dimensional components.
Data Source
AI summary
An anomaly detecting device includes: a singular value decomposition unit configured to perform singular value decomposition of a variance-covariance matrix of a measured value matrix y0 composed of measured values acquired by a plurality of sensors in a time period considered to be normal, to thereby calculate a singular vector U and a singular value matrix S; an anomaly determination unit configured to apply the singular vector U and the singular value matrix S to a measured value matrix yt to be evaluated and which is acquired in an arbitrary time period to determine whether an anomaly is present from a result of application; and an anomalous part identification unit configured to, when the measured value matrix yt is determined to be anomalous, identify an anomalous part based on a diagonal element of a matrix obtained in association with the measured value matrix yt.