Sensor Health Index for Real-Time Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor monitoring technologies fail to accurately and efficiently detect sensor problems in real-time, particularly in industrial settings where sensors malfunction or exhibit variance, leading to false alarms and unnecessary maintenance.
Innovation Solution
The development of novel techniques that use a Sensor Health Index (SHI) to monitor sensor health in real-time, detecting issues without assuming standard signal distributions, and applying tests like Sequential Probability Ratio Test (SPRT) and asymmetric random walks to identify erratic or drifting sensor behavior, even in low-frequency data streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensor monitoring methods are used, then sensor problems can be detected, but false alarms and unnecessary maintenance actions occur due to inaccurate detection
Solution Approach 1:
The patent segments the sensor monitoring process into multiple independent statistical tests (flat-line test, outlier test, erratic test, drift test), each targeting specific sensor failure modes. This segmentation allows precise detection of different problem types without cross-interference, improving detection accuracy while reducing false alarms through specialized test design for each sensor condition
Solution Approach 2:
The patent transforms sensor readings into rate-of-change values and applies multiple statistical parameters (mean, variance, standard deviation) to detect different failure modes. By changing the parameter representation from raw sensor values to derived statistical features, the system achieves more accurate and reliable sensor health assessment across varying operating conditions
2Loss of time
If real-time sensor monitoring is implemented, then sensor problems can be detected early, but the system complexity increases
Solution Approach 1:
The patent pre-defines multiple statistical tests (flat-line, outlier, erratic, drift tests) with their respective thresholds and criteria before actual sensor monitoring begins. This preliminary setup allows the system to perform real-time detection using pre-configured logic without complex runtime calculations, reducing both detection response time and system complexity
Solution Approach 2:
The monitoring system automatically performs statistical calculations, compares readings against thresholds, and generates alerts without requiring manual intervention or complex external processing. The system self-evaluates sensor health using built-in statistical logic, simplifying the overall system architecture while enabling real-time detection
3Productivity
If standard statistical tests are used for sensor monitoring, then detection can be performed, but the tests require assumptions about signal distribution that may not hold in practice
Solution Approach 1:
The patent implements a universal monitoring framework that applies the same statistical test structure (rate-of-change calculation, threshold comparison) to multiple sensor types and failure modes. The methodology is adaptable to temperature, pressure, flow, and other industrial sensors without requiring type-specific assumptions, enhancing both efficiency and versatility across different sensor applications
Solution Approach 2:
The patent transforms sensor readings into unitless rate-of-change values, eliminating dependence on measurement units and distribution assumptions. This parameter transformation allows the same statistical tests to be applied universally across different sensor types and operating conditions without requiring distribution-specific calibration or adjustment
Data Source
AI summary
Streaming data at a sensor is sensed and received. The streaming data includes a plurality of observations. For a current observation in the plurality of observations, a health of the current observation is determined. Based upon the health of the current observation, a penalty is determined. A Sensor Health Index (SHI) for the current observation is obtained by aggregating the penalty with at least one SHI of one or more previous observations from the plurality of observations. An alarm is selectively generated based upon the SHI of the current observation.