On-Board Component Failure Detection Using Virtual Sensor Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for monitoring machine components, such as those described in U.S. Patent Application Pub. No. 2017/0092021, can detect abnormal vehicle conditions but fail to specifically identify failing components, requiring further investigation to determine the cause of the abnormality.
Innovation Solution
An electronic control module (ECM) on-board a machine uses a reduced-order model and a machine learning model to generate virtual sensor data and offset data, adjusting sensor data based on component age and operating conditions, to predict component failures and estimate remaining usable life by comparing against failure condition thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If PCA analysis is used to detect abnormal vehicle conditions, then abnormal conditions can be revealed, but specific failing components cannot be directly identified
Solution Approach 1:
The patent segments the monitoring system into multiple specialized models, each dedicated to monitoring a specific machine component (engine, transmission, brakes, etc.). Each component monitor processes sensor data independently and identifies failures for its specific component, thereby retaining component identification information that would be lost in a holistic PCA analysis.
Solution Approach 2:
The patent introduces component monitors as intermediary processing layers between the sensor data and the final failure identification output. These intermediaries translate general vibration patterns into specific component failure diagnoses, preserving the information about which component is failing.
2Reliability
If holistic vehicle monitoring is used, then overall abnormal conditions can be detected, but further investigation is needed to determine the cause
Solution Approach 1:
The patent performs preliminary analysis by having each component monitor continuously assess its specific component's health status in advance. This preliminary action identifies potential failures before they manifest as general vehicle abnormalities, eliminating the need for subsequent investigation to determine the cause.
Solution Approach 2:
The system implements feedback loops where each component monitor continuously receives sensor data, compares it against expected patterns, and provides immediate feedback about component health. This real-time feedback mechanism enables rapid identification of failure causes without requiring additional investigation time.
3Measurement precision
If multiple component-specific monitors are implemented, then specific component failures can be identified, but system complexity increases
Solution Approach 1:
The patent implements universal monitoring algorithms that can be applied across multiple component types. Each component monitor uses the same general approach of comparing sensor data against expected vibration patterns, making the system scalable and reducing complexity compared to designing unique monitoring systems for each component.
Solution Approach 2:
The system manages complexity by changing parameters (thresholds, frequency ranges, sensor configurations) rather than changing the fundamental monitoring approach for each component. This allows the same monitoring logic to be reused across different components with simple parameter adjustments.
Data Source
AI summary
An electronic control module (ECM) on-board a machine executes a reduced-order model configured to generate virtual sensor data associated with a machine component based on sensor data measured by actual sensors. The ECM can also execute a machine learning model configured to generate offset data for the sensor data and/or the virtual sensor data, based on operating conditions, an age of the machine component, and/or other factors that the reduced-order model may not consider. The sensor data and/or the virtual sensor data, adjusted based on the offset data, can indicate whether the machine component has failed or is predicted to fail at a future point in time, and/or a remaining usable life of the machine component.