Adaptive Vehicle Prognostics for Fault Detection Model Updating
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle diagnostics and prognostics systems require significant time and resources to refine and update, especially to handle vehicle-to-vehicle variations, and often rely on a single set of logic and calibrations that may not effectively address unique vehicle conditions.
Innovation Solution
An adaptive prognostics system with a self-assessment module and training module that uses machine learning-based distance calculations and edge computing to automatically evaluate vehicle events, update models, and customize diagnostics and prognostics systems for individual vehicles or groups, excluding noise and corner cases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single set of logic and calibrations is used for one vehicle program, then the system is simple to implement, but it cannot achieve desired performance to handle vehicle to vehicle variations
Solution Approach 1:
The system dynamically adapts its logic and calibrations based on vehicle-to-vehicle variations. Instead of using a static single set of parameters, the system continuously learns from operational data and adjusts its diagnostic thresholds and parameters in real-time to accommodate different vehicle conditions, manufacturing variations, and operational environments.
Solution Approach 2:
The system changes its operational parameters based on detected vehicle characteristics and conditions. By monitoring deviations from reference data and automatically adjusting diagnostic thresholds and calibration parameters, the system adapts to handle variations across different vehicles without requiring manual recalibration for each vehicle program.
2Reliability
If traditional D&P systems are deployed in vehicles, then initial diagnostics functionality is provided, but it takes considerable amount of time and resources to refine and update the system to handle corner cases
Solution Approach 1:
The system performs self-updating by automatically learning from operational data and corner cases encountered during vehicle operation. The machine learning models continuously train on new data, enabling the system to refine its diagnostic capabilities and handle previously unseen corner cases without requiring external intervention, manual updates, or extensive recalibration time.
Solution Approach 2:
The system prepares for future diagnostic challenges by continuously pre-training on diverse operational data and corner cases before they are encountered in production vehicles. This preliminary learning ensures that when corner cases occur in the field, the system is already equipped with the knowledge to handle them accurately and immediately.
3Measurement precision
If machine learning-based distance calculations are used to evaluate vehicle events, then accurate fault identification is achieved, but computational resources and processing time increase
Solution Approach 1:
The system applies machine learning-based distance calculations selectively to only those events and parameters that are most indicative of potential faults. Instead of continuously processing all vehicle data with complex ML algorithms, the system identifies and focuses computational resources on critical deviation detection, achieving high accuracy while minimizing unnecessary computational energy consumption.
Data Source
AI summary
A system includes an assessment module and a training module. The assessment module is configured to receive event data about an event associated with a subsystem of a vehicle. The assessment module is configured to determine deviations between reference data for the subsystem indicating normal operation of the subsystem and portions of the event data that precede and follow the event. The assessment module is configured to determine whether the event data indicates a fault associated with the subsystem by comparing the deviations to a threshold deviation. The training module is configured to update a model trained to identify faults in vehicles to identify the event as a fault associated with the subsystem of the vehicle based on the event data in response to the deviations indicating a fault associated with the subsystem.