Harmonic Drive Anomaly Detection Using GMM and Time Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods for industrial robots fail to detect early stages of performance degradation, as they rely on significant deviations in parameter data that may not be apparent in repetitive but slightly varying operations, leading to component failures.
Innovation Solution
A Gaussian mixture model (GMM) is used to learn distribution parameters from time-series data collected during normal robot operations, aligning and differencing the data with a reference file, and computing log likelihood to detect anomalies through statistical variance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional anomaly detection methods are used to evaluate parameter data, then significant changes can be detected, but early stages of performance degradation cannot be detected
Solution Approach 1:
The system performs preliminary alignment of current parameter data with reference data using dynamic time warping before anomaly detection. This preprocessing step prepares the data in advance to reveal subtle deviations that would otherwise be hidden, enabling early detection of performance degradation before it becomes significant.
Solution Approach 2:
The invention transforms the anomaly detection approach by introducing time alignment as an additional dimension of analysis. Instead of directly comparing parameter values, the system aligns time series in the time domain using DTW, then performs detection in the aligned domain, effectively adding a temporal alignment dimension that reveals early degradation patterns.
2Ease of operation
If parameter data is evaluated directly without alignment, then simple comparison is possible, but minor changes in trajectory or acceleration cannot be properly assessed
Solution Approach 1:
The system performs preliminary alignment of current parameter data with reference data using dynamic time warping before anomaly detection. This preprocessing step prepares the data in advance to reveal subtle deviations that would otherwise be hidden, enabling early detection of performance degradation before it becomes significant.
Solution Approach 2:
Dynamic time warping serves as an intermediary transformation that bridges the gap between raw parameter data and meaningful anomaly detection. The DTW alignment process mediates the comparison by warping time axes to match corresponding phases of operation, allowing accurate assessment of minor trajectory and acceleration changes without direct raw data comparison.
3Stability of the object's composition
If robot operations are monitored for exact repetition, then consistency can be measured, but natural variations in repetitive motions cause false anomalies
Solution Approach 1:
The system dynamically adapts the reference model by continuously updating it with recently collected good data. This dynamic reference allows the system to accommodate natural variations in repetitive robot motions while maintaining sensitivity to actual anomalies, resolving the conflict between measuring consistency and avoiding false positives.
Solution Approach 2:
The invention changes the parameter comparison approach by applying dynamic time warping to align time series before comparison. This parameter transformation allows the system to measure operational consistency in a time-warped domain where natural variations are accounted for, while still detecting true anomalies that deviate from the aligned reference pattern.
Data Source
AI summary
A method and system for anomaly detection from time-series input data. A Gaussian mixture model (GMM) learns distribution parameters in an offline learning stage using sample data. The data used for the offline learning, and for a subsequent online anomaly detection stage, is time-series data collected for multiple parameters of a machine operation, such as a robot performing a repetitive set of operations. The method includes aligning the data to and taking a difference from a known good reference data file, before providing the data to the GMM. In the online anomaly detection stage, the GMM computes a probability that each time-series data point fits the distribution, and a log summing computation is performed on each data file to determine the likelihood that the file contains anomaly data. The file log likelihood is compared to previous values and an alarm is issued when statistically variant from the historical data.