Bayesian Sensor Validation Using Population-Based Control Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for validating sensor validity using control limits are limited, as they are often set based on sample data from a single sensor, making it difficult to generalize and apply these limits to all sensors within the same population.
Innovation Solution
A method and system for validating sensor validity by inferring posterior distributions of parameters using Bayesian techniques, setting a target credible interval, and establishing control lines for sensor data, which can be modified by considering the performance reduction index of the facility where the sensor is installed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If control limits are set based on sample data from a single sensor, then the validation process is simple to implement, but the control limits cannot be generalized to all sensors in the same population
Solution Approach 1:
The patent applies universality by establishing control limits based on the population distribution of sensor data rather than individual sensor samples. The control limit calculation uses the cumulative distribution function of the population, allowing a single set of control limits to be applied universally across all sensors in the population, thus achieving generalizability while maintaining systematic validation
Solution Approach 2:
The patent implements preliminary action by pre-establishing control limits based on population characteristics before actual sensor validation occurs. The control limits are calculated in advance using the population's cumulative distribution function, so that when validation is needed, the pre-computed limits can be directly applied without requiring complex real-time calculations for each sensor
2Reliability
If conventional rule-based algorithms are used for sensor validation, then the implementation is straightforward, but the validation reliability is insufficient
Solution Approach 1:
The patent replaces conventional rule-based algorithms with a statistical methodology based on cumulative distribution functions. Instead of using fixed thresholds or heuristic rules, the validation system uses probabilistic statistical models to determine whether sensor data deviates from expected population behavior, thereby improving validation reliability through mathematically rigorous statistical inference
Solution Approach 2:
The patent implements feedback by continuously monitoring sensor data against the population-based control limits and using the validation results to update the understanding of sensor population characteristics. The system compares actual sensor measurements with the statistically derived control limits and provides feedback on whether the sensor remains within acceptable parameters, enabling continuous validation and potential recalibration
Data Source
AI summary
The present disclosure relates to a system and a method for validating the validity of a sensor, in particular, validating the validity of a sensor using a control limit. The method for validating the validity of a sensor using a control limit includes inferring a posterior distribution of a parameter in a Bayesian technique using a prior distribution of the parameter of sensor data and historical data of the sensor, setting a target credible interval for the posterior distribution of the parameter and setting a control line of the sensor data using the set credible interval, and validating the validity of the sensor by monitoring whether the actual measurement data of the sensor deviates from the control line.