Time Series Fault Localization Using Piecewise Change-Point Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault detection and localization methods in systems are inadequate in accurately identifying and localizing changes in time series or spatial data, leading to unnecessary equipment failures, secondary damage, and unscheduled maintenance, which result in significant indirect costs and downtime.
Innovation Solution
A method and system for fault detection and localization that smooths data using kernel regression, identifies split points, and iteratively fits piecewise linear functions to minimize residuals below a domain-dependent threshold, effectively reducing noise and outlier impact, thereby improving accuracy and precision in detecting faults and their onset times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fault detection methods are used, then the system can identify changes in time series data, but the accuracy of fault detection and localization is poor, leading to spurious faults and imprecise fault onset timing
Solution Approach 1:
The patent segments the time series data into multiple segments by identifying change points where the statistical properties of the data change. This segmentation allows the system to detect faults more accurately by analyzing each segment separately, reducing false positives while maintaining high detection accuracy.
Solution Approach 2:
The patent applies preliminary smoothing of the time series data using a moving average or similar technique before detecting change points. This preliminary action reduces noise in the data, enabling more accurate fault detection and localization while minimizing spurious faults.
2Loss of time
If the system detects faults early and accurately, then equipment failures can be predicted and unscheduled downtime can be avoided, but the complexity of the detection system increases
Solution Approach 1:
The patent employs a self-service approach where the system automatically detects change points and segments data without requiring external intervention or complex manual configuration. The algorithm autonomously identifies fault conditions and determines when to alert operators, reducing system complexity while enabling early fault detection.
Solution Approach 2:
The patent monitors changes in statistical parameters of the time series data (such as mean, variance, or other domain-specific parameters) to detect faults. By focusing on parameter changes rather than complex pattern recognition, the system achieves early fault detection with relatively simple computational methods.
Data Source
AI summary
A method for fault detection and localization calls for obtaining a data set, smoothing the data set, identifying a plurality of split points within the data set, fitting a piecewise linear function to the plurality of split points; and determining a residual between the function and the smoothed data set. Related systems and computer program products are disclosed and claimed.