Machine Operation Anomaly Detection Using Localized Matrix Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional anomaly detection methods for time-varying system operations suffer from false positives and false negatives due to the unpredictability and volatility of time-varying signals, making it difficult to determine optimal maintenance schedules for machines.
Innovation Solution
The introduction of a statistical time series data mining primitive called Localized Matrix Profile (LMP), which represents local dissimilarities of time-varying signals, allowing for time-dependent statistical analysis and reducing false positives and negatives by comparing test signals with baseline signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional anomaly detection methods are used for time-varying system operations, then the system can detect anomalies, but it suffers from false positives and false negatives due to the unpredictability and volatility of time-varying signals
Solution Approach 1:
The patent segments the time-varying signal into multiple local windows or segments, and computes the matrix profile for each segment independently. This allows the system to capture local dissimilarities at different time points, improving the precision of anomaly detection by focusing on specific local patterns rather than treating the entire signal as a single unit.
Solution Approach 2:
The patent transforms the time-series anomaly detection problem into a two-dimensional matrix profile space, where rows represent different signal segments and columns represent template matches. This dimensional transformation enables the system to visualize and analyze local dissimilarities from multiple perspectives, reducing false positives and negatives by providing a more comprehensive view of signal variations.
2Ease of operation
If a single maintenance cycle is applied to all machines in a group, then the maintenance logistics are simplified, but it cannot be optimal for all machines since some are new and require maintenance less often while older machines require maintenance more often
Solution Approach 1:
The patent implements dynamic maintenance scheduling by continuously monitoring the matrix profile values of each machine's operational signals. When the local dissimilarity exceeds a threshold, the system automatically generates maintenance alerts for that specific machine, transitioning from static fixed-interval maintenance to dynamic condition-based maintenance. This optimizes maintenance timing for each machine based on its actual operational state.
Solution Approach 2:
The system enables machines to effectively self-diagnose their maintenance needs by comparing their operational signals against learned templates. The matrix profile computation automatically identifies when a machine's behavior deviates from normal patterns, allowing the system to determine maintenance requirements without external intervention, thus optimizing maintenance schedules based on actual machine conditions.
Data Source
AI summary
A system for detecting an anomaly in an execution of an operation of a machine determines a local matrix profile (LMP) of a test signal with respect to the baseline signals. LMP is a time series of values, each LMP value for a time instance is determined for a segment of the test signal based on a minimum distance between the segment of the test signal with corresponding segments of the baseline signals, such that each LMP value is a value of a local dissimilarity of the execution of the operation of the machine with respect to the baseline executions of the operation of the machine. The system determines an accumulation of the LMP values above an LMP threshold and detects an anomaly when the accumulation above an anomaly detection threshold.