Vehicle Component Health Monitoring With On-Board Predictive Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In complex vehicle systems, predictive analysis of component degradation is challenging due to the high number of signals and interactions, leading to inefficient data processing and incorrect health monitoring results, especially when centralized servers handle data from thousands of vehicles with limited on-board calculation capabilities.
Innovation Solution
A method that preprocesses sensor data on board the vehicle using neural networks and multivariate statistical techniques to estimate health states and residual useful life, reducing data flow to central servers and improving predictive diagnosis accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all sensor signals and control data are transmitted to the central server for post-processing, then comprehensive predictive analysis can be performed, but the data transmission load and processing complexity increase significantly
Solution Approach 1:
The system performs preliminary processing of sensor data on-board the vehicle before transmission. Health indicators are calculated locally using processed sensor signals, so that only essential health indicator data needs to be transmitted to the central server, not the raw sensor data itself. This preliminary calculation reduces the data transmission load while maintaining diagnostic accuracy.
Solution Approach 2:
The invention extracts only the essential health indicator information from the complex sensor data and transmits only this extracted information to the central server. The on-board system filters and processes raw sensor signals to extract meaningful health indicators, leaving the detailed raw data processing to be performed locally rather than centrally.
2Measurement precision
If specific algorithms are developed for monitoring each component's health, then component-specific monitoring accuracy improves, but the system complexity and difficulty of managing heterogeneous subsystems increases
Solution Approach 1:
The system uses a universal on-board processing unit that can handle multiple different component types using the same neural network architecture. While specific algorithms are developed for different component types (engine, transmission, brakes, etc.), they all follow the same general processing framework and are integrated through a common health indicator calculation approach, reducing overall system complexity.
Solution Approach 2:
The system segments the monitoring task by creating separate neural network models for different component types (engine, transmission, brakes, suspension, etc.), allowing each component to be monitored with specialized algorithms while maintaining a unified overall system architecture. This segmentation allows component-specific accuracy without requiring a completely custom system for each part.
3Speed
If heavy computational analyses are performed on-board the vehicle, then real-time predictive diagnosis is improved, but the required calculation capabilities and energy consumption increase
Solution Approach 1:
The system performs partial computational analysis on-board (sufficient for real-time health indicator calculation) and leaves the more intensive analytical work for the central server. The on-board unit calculates health indicators using neural networks with limited computational requirements, while the central server performs more extensive predictive analysis and long-term trend analysis, dividing the computational workload to balance real-time response with energy consumption.
Data Source
Figure 1~2
Figure 3a~3b
Figure 4
AI summary
Method for monitoring a state of health of a vehicle component or subsystem such as a particulate filter, an air filter, a sensor for measuring the NOx produced by a Diesel engine, etc., the method comprising (i) first acquisition of a variable measured target, representative of a state of health of the component or subsystem, (ii) second acquisition of a set of input variables interdependent with said target variable, (iii) calculation of an estimated target variable based on the second acquisition, (iv) comparison of said measured target variable with said estimated target variable and (v) calculation of an index representative of a state of health of the component or subsystem, function of said comparison, (vi) acquisition of said index and monitoring of its temporal variability and /o mileage of the vehicle, (vii) calculation of a residual life of the component or subsystem based on said monitoring.