Vehicle Actuator Failure Detection Using Temporal Steering Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting actuator failures in vehicles, such as those in tractor trailers, face challenges in accurately determining failures due to the complexity of controlling these vehicles, as they do not effectively consider the temporal aspect of data, leading to inaccurate detection.
Innovation Solution
A device and method utilizing deep learning models, where a first model processes behavior data and steering compensation angles to output a steering compensation angle, and a second model processes this output along with lateral data and failure probability values to accurately detect actuator failures, enabling rapid and precise detection without complex calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional Kalman filter-based prediction methods are used to detect actuator failure, then the detection process can be implemented, but the detection accuracy is insufficient due to the complexity of vehicle control and inability to capture temporal data characteristics
Solution Approach 1:
The patent replaces conventional Kalman filter-based mechanical prediction methods with deep learning neural network models. The neural networks automatically learn temporal patterns and nonlinear relationships from historical data, substituting complex manual control algorithms with adaptive data-driven models that capture temporal characteristics without requiring explicit mathematical models of vehicle dynamics.
Solution Approach 2:
The patent transforms the detection approach by changing from fixed-parameter Kalman filter predictions to dynamic parameter learning through neural networks. The system learns optimal detection parameters and relationships from training data, allowing the detection accuracy to adapt to varying operating conditions while maintaining computational efficiency during actual detection operations.
2Measurement precision
If deep learning models with multiple layers are used to process temporal data, then the detection accuracy improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by training the deep learning models offline using historical training data before actual failure detection is needed. The neural networks learn and store optimal detection patterns during the training phase, so that during actual operation, the models can rapidly process new data and provide accurate failure detection without requiring extensive real-time computational resources.
Data Source
AI summary
A device for detecting a failure of an actuator of a vehicle includes: a training device that trains a model using training data comprising behavior data of the vehicle and a steering compensation angle, and a controller that detects the failure of the actuator in the vehicle based on the model.