Sensor Anomaly Detection via Status Probability Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting anomalous sensors are computationally expensive and do not provide an absolute measure of anomaly, nor do they consider temporal information effectively.
Innovation Solution
An apparatus comprising a processor and computer-readable medium that generates a status probability distribution based on healthy and anomalous data distributions, utilizing temporal information to automate the detection of anomalous sensors, with features such as Beta prior distributions and Bernoulli distribution approximations to reduce computational resources and time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing anomaly localization methods are used to determine the degree of anomaly of sensors, then anomaly detection capability is provided, but computational cost and time consumption increase significantly
Solution Approach 1:
The patent transforms the anomaly detection problem from computing absolute anomaly scores to computing relative status probabilities. By changing the parameter from absolute anomaly score to relative status probability ratio, the computational complexity is reduced while maintaining detection capability. The status probability distribution is computed using efficient algorithms that leverage the relative nature of the problem.
Solution Approach 2:
The patent segments the computational task by first computing the status probability distribution for all sensors, then using this distribution to determine anomaly status. This segmentation allows for efficient computation by separating the probability computation phase from the anomaly determination phase, reducing overall computational burden.
2Measurement precision
If existing anomaly localization methods are used to determine the degree of anomaly of sensors, then anomaly detection capability is provided, but computational resources increase significantly
Solution Approach 1:
The patent changes the computational parameter from absolute anomaly scores requiring intensive computation to relative status probabilities that can be computed more efficiently. The status probability distribution computation uses algorithms that are less resource-intensive than traditional anomaly scoring methods.
Solution Approach 2:
The patent uses the healthy data distribution as a reference model (copy) to compare against actual sensor data. By creating and utilizing this reference distribution, the system avoids computing anomaly scores from scratch for each sensor, instead leveraging the pre-computed healthy distribution to efficiently determine anomaly status.
3Measurement precision
If existing anomaly localization methods are used, then anomaly degree is determined, but temporal information of sensor data is not considered
Solution Approach 1:
The patent performs preliminary computation of the status probability distribution using historical and current sensor data before determining anomaly status. This preliminary action incorporates temporal information by using past data to establish the probability distribution, which then informs the anomaly determination process.
Solution Approach 2:
The patent incorporates temporal information through feedback mechanisms where the status probability distribution is continuously updated based on incoming sensor data. The system uses past anomaly determinations and sensor readings to refine the probability distribution, creating a feedback loop that maintains temporal awareness.
Data Source
AI summary
Anomalous sensors are detected using an apparatus including a processor and one or more computer readable mediums collectively including instructions that, when executed by the processor, cause the processor to: obtain a plurality of healthy sensor data, wherein each of the healthy sensor data includes a plurality of sensed values of a corresponding sensor among a plurality of sensors in normal operation, generate a healthy data distribution of at least two sensors among the plurality of sensors based on the plurality of healthy sensor data, and generate a function of a status probability distribution of the plurality of sensors with respect to time under the condition of sensor data with respect to time based on the healthy data distribution.