Bayesian Sensor Validation with Credible-Interval Control Lines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sensor validation techniques set control limits based on individual sensor data, making it difficult to generalize and apply these limits to all sensors within the same population, leading to unreliable validation and potential incorrect determination of sensor invalidity due to facility performance degradation.
Innovation Solution
A system that uses Bayesian inference to determine the posterior distribution of sensor parameters, sets credible intervals, and establishes control lines for sensor data validation, allowing for real-time monitoring and adjustment of control limits based on performance reduction indices to ensure accurate sensor validity assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If control limits are set based on individual sensor data, 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 establishes control limits that serve multiple sensors simultaneously by using population-level statistical parameters. Instead of creating separate control limits for each sensor, the system defines universal control limits based on the combined data from all sensors in the population, making the validation process adaptable across all sensors while maintaining a unified approach.
Solution Approach 2:
The patent transforms the approach by changing from individual sensor parameters to population-level statistical parameters. By calculating control limits using aggregate data from multiple sensors and expressing them as statistical ranges (e.g., mean ± standard deviation), the system achieves generalizability across the sensor population while maintaining mathematical rigor.
2Ease of operation
If control limits are set using conventional techniques, then the validation process is straightforward, but incorrect determination of sensor invalidity occurs due to facility performance degradation
Solution Approach 1:
The patent introduces statistical parameters (mean, standard deviation, control limit ranges) to transform fixed threshold validation into a probabilistic assessment. By expressing control limits as statistical ranges rather than fixed values, the system can distinguish between normal variation and actual sensor failure, improving reliability while maintaining operational simplicity through standardized statistical methods.
Solution Approach 2:
The patent implements a feedback mechanism where control limits are continuously refined based on historical sensor data and population statistics. The system uses past performance data to update and adjust control limits over time, creating a self-improving validation process that becomes more accurate while maintaining straightforward operation through automated statistical calculations.
3Adaptability or versatility
If statistical control limits are established for sensor population, then generalizability is improved, but the complexity of setting and maintaining these limits increases
Solution Approach 1:
The patent enables the sensor population itself to generate the control limits through automated statistical calculations. The system uses the sensors' own historical data to compute population parameters and establish control limits, eliminating the need for external manual calibration or complex configuration. This self-service approach reduces management complexity while achieving population-level generalizability.
Solution Approach 2:
The patent merges the control limit setting process with the normal sensor operation and data collection process. By integrating statistical calculations into the existing data flow and using the same infrastructure for both data collection and validation, the system achieves population-level control without adding separate complex management systems or procedures.
Data Source
Figure 1~2
Figure 3~4A
Figure 4B
AI summary
The present disclosure relates to a system for validating the validity of a sensor. The present disclosure provides a system for validating the validity of a sensor using a control limit, including an operation unit configured to infer 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, a setting unit configured to set a credible interval for the posterior distribution of the parameter and to set a control line of the sensor data, and a control unit configured to validate the validity of the sensor by monitoring whether the actual measurement data of the sensor deviates the control line. According to the present disclosure, it is possible to set the control limit based on the Bayesian inference and validate the validity of the sensor from the actual sensor data reliably