Fault Detection Using Sliding Window Deviation and Stableness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault detection systems are unable to dynamically and automatically detect anomalies in dynamic systems, fail to distinguish between improbable behavior and actual faults, and are computationally expensive, making them infeasible for real-time operation on a large scale.
Innovation Solution
A method and apparatus that compute deviation and stableness values from observation data, using a data collection module to receive values, a compute module to calculate these values, and a decision module to transmit fault indications when specific criteria are met, allowing for real-time detection of faults in processing systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If dynamic or adaptive thresholds are used to detect abnormal behaviors, then detection capability for dynamic systems is improved, but computational cost increases making real-time operation infeasible
Solution Approach 1:
The patent applies dynamics by using a moving window of fixed size that slides through the data stream, dynamically adapting to changing system behavior without requiring complex adaptive threshold calculations. The window moves forward with each new observation, automatically adjusting the reference frame to current system state while maintaining computational simplicity through fixed-size sliding window operations.
Solution Approach 2:
The patent changes parameters by transforming the detection problem from computing adaptive thresholds to computing statistical moments (mean and standard deviation) over a sliding window. This parameter transformation enables dynamic adaptation through simple arithmetic operations on window contents rather than complex threshold adaptation algorithms, significantly reducing computational cost while maintaining detection effectiveness.
2Adaptability or versatility
If dynamic or adaptive thresholds are used to detect abnormal behaviors, then detection capability is improved, but device complexity increases
Solution Approach 1:
The patent applies dynamics through a sliding window approach that dynamically adapts to changing system conditions by continuously updating the window position. This dynamic mechanism handles variability in system behavior without requiring complex adaptive threshold algorithms, maintaining simplicity while achieving adaptability through the moving reference frame.
Solution Approach 2:
The patent segments the data stream into fixed-size windows, dividing the continuous data flow into manageable segments for analysis. This segmentation allows the system to focus computational resources on a limited portion of data at a time, simplifying the overall system architecture while enabling dynamic adaptation through sequential window processing.
3Use of energy by moving object
If predefined static thresholds are used to detect abnormal behaviors, then computational cost is reduced, but ability to detect anomalies in dynamic systems is lost
Solution Approach 1:
The patent resolves this contradiction by implementing a dynamic sliding window that adapts to changing system behavior over time. The window continuously moves through the data stream, automatically adjusting the reference frame to match current system state. This dynamic mechanism enables detection of anomalies in dynamic systems while maintaining low computational cost through simple arithmetic operations on window contents.
Solution Approach 2:
The patent applies preliminary action by pre-establishing a fixed-size window that slides through the data stream. This pre-configured window structure enables the system to adapt to dynamic conditions without requiring complex real-time threshold calculations, as the window simply needs to move and compute basic statistics on its contents, maintaining both adaptability and computational efficiency.
4Reliability
If known fault detection systems are used, then fault detection is possible, but ability to distinguish faults from abnormal behavior is lost
Solution Approach 1:
The patent applies feedback by continuously monitoring the sliding window contents and comparing current observations against the window's statistical characteristics. The system uses feedback from the window's mean and standard deviation calculations to dynamically adjust the detection criteria, enabling it to distinguish between temporary abnormalities and actual faults based on the context provided by recent historical data.
Solution Approach 2:
The patent uses dynamics through the sliding window that continuously adapts to changing system behavior. By maintaining a moving reference frame that reflects current system state, the system can dynamically distinguish between abnormal behavior that is consistent with current operational patterns and actual faults that represent deviations from expected behavior, improving detection reliability without losing distinguishing capability.
Data Source
AI summary
In some embodiments, a method includes receiving, at a data collection module implemented in at least one of a memory or a processing device, from a processing system, an observation value for a variable. The observation value of the variable is associated with operation of the processing system at a time. The method further includes computing a deviation value of the variable from a baseline value at the time based on the observation value. The method further includes computing a stableness value of the variable at the time based on the baseline value and a variance of the variable during a time period including the time. The method further includes transmitting an indication of the processing system as operating with a fault in response to the deviation value meeting a first criterion and the stableness value meeting a second criterion.