Stuck Sensor Detection Using Correlated Signal Variation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional rationality diagnostics for detecting stuck-in-range sensors in vehicles are limited in sensitivity, often resulting in false negatives due to their broad fault limits, making it difficult to detect sensors that are stuck in a range close to nominal values.
Innovation Solution
A method involving the computation of variation values for correlated sensor signals, where a stuck sensor is identified by a significantly lower variation value compared to a non-stuck sensor, allowing for more sensitive detection and differentiation from steady-state conditions, and can be combined with conventional diagnostics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a conventional rationality diagnostic with a large fault limit is used to minimize false indications, then false positives are reduced, but detection sensitivity decreases and sensors stuck near nominal values cannot be detected
Solution Approach 1:
The diagnostic approach is segmented into two independent components: a conventional rationality diagnostic (comparing sensor values) and a variation diagnostic (comparing signal variations). Each component operates with its own threshold, allowing the variation diagnostic to use a sensitive threshold without causing false positives in the overall system.
Solution Approach 2:
Instead of only comparing sensor values in the value dimension, the invention introduces a new dimension of signal variation. By computing variations of sensor signals over time and comparing these variations, the system gains additional diagnostic information that is independent of the absolute sensor values, enabling detection of stuck sensors near nominal values.
2Measurement precision
If a small fault limit is used in rationality diagnostic to increase detection sensitivity, then more stuck sensors are detected, but false indications increase
Solution Approach 1:
The diagnostic is divided into separate value-based and variation-based components. The variation diagnostic specifically targets detection sensitivity by monitoring signal variations, while the conventional rationality diagnostic handles overall reliability. This segmentation allows each component to be optimized independently.
Solution Approach 2:
The variation computation acts as an intermediary that transforms raw sensor signals into variation values. This intermediary step filters out steady-state conditions and highlights dynamic changes, allowing the diagnostic to focus on meaningful variations rather than absolute values, thus reducing false indications.
3Measurement precision
If the fault limit is set to detect sensors stuck near nominal values, then detection sensitivity improves, but the diagnostic can no longer distinguish between stuck sensors and sensors in steady state
Solution Approach 1:
The diagnostic transitions from a static value comparison to a dynamic variation analysis. By computing variations of sensor signals over time, the system captures the dynamic behavior of sensors. A stuck sensor will show zero or minimal variation, while a sensor in steady state will show normal variation patterns, enabling clear distinction between the two conditions.
Solution Approach 2:
The variation computation is performed preliminarily on the sensor signals before the final diagnostic comparison. This preliminary processing of signals into variation values prepares the data in a form that inherently distinguishes between stuck and steady-state conditions, making the subsequent diagnostic comparison more effective.
Data Source
AI summary
A method for detecting a sensor that is stuck in range includes computing a variation value for each of a first signal and a second signal generated by a respective first and second sensor, wherein the first and second signals are correlated signals. The variation value for each of the first and second signals is computed for the same diagnostic period, such that when one of the first and second sensors is generating a signal which is stuck in range during the diagnostic period, the signal stuck in range will be characterized by a variation value which is much less than the variation value of the correlated signal. The magnitude of difference between the variation values of the first and second signals can be compared to a predetermined fault threshold to diagnose a sensor stuck in range.