Vehicle Fault Diagnostics Using Data Segmentation and Trend Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle fault diagnostics and prognostics systems face challenges in predicting failures during rapid subsystem degradation and lack of data, especially when fault signatures do not exceed predefined thresholds, and struggle with accurately evaluating repairs using historical data, particularly with OEM components and warranty claims.
Innovation Solution
An improved fault diagnostics and prognostics system using automatic data segmentation and trending, which smooths data with moving averages, identifies transition points, detects trends using regression, and extrapolates to predict failures, while also validating repairs by analyzing data post-service.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional threshold-based diagnostics are used, then system-level faults can be detected, but component-level root cause analysis and early failure prediction are not achieved
Solution Approach 1:
The patent segments historical sensor data into distinct phases (normal operation, degradation, failure) using change point detection algorithms. This segmentation enables precise identification of when and how subsystems transition from healthy to faulty states, allowing component-level root cause analysis rather than just system-level fault detection.
Solution Approach 2:
The system performs preliminary trend analysis and extrapolation on segmented data to predict future failures before they occur. By analyzing degradation trends in segmented phases and extrapolating forward, the system can predict component failures in advance, enabling proactive maintenance and reducing time to identify root causes.
2Quantity of substance
If all historical data is used for trend analysis, then more data is available for prediction, but healthy data dilutes the abnormal trends reducing prediction accuracy
Solution Approach 1:
The patent divides historical data into separate segments (normal operation phase and degradation phase) using change point detection. This allows the system to select only the relevant degradation segment for trend analysis, excluding healthy data that would dilute the abnormal trends and reduce prediction accuracy.
Solution Approach 2:
The system extracts and isolates the degradation phase data from the complete historical dataset. By taking out only the abnormal segment containing the actual degradation trend, the system performs trend analysis on concentrated faulty data rather than diluted mixed data, significantly improving prediction accuracy.
3Ease of repair
If repair validation is performed using traditional methods, then service completion is recorded, but the actual effectiveness of repairs especially for OEM components cannot be accurately evaluated
Solution Approach 1:
The patent establishes baseline degradation trends and failure thresholds before service occurs. After repair, the system compares post-service data against these pre-established baselines to objectively verify whether the repair was effective, rather than simply recording service completion. This preliminary setup enables reliable evaluation of repair effectiveness.
Solution Approach 2:
The system continuously monitors post-service data and provides feedback on whether the subsystem returns to normal operation or continues degrading. This feedback mechanism validates whether repairs were effective, especially for OEM components, by comparing actual post-service performance against expected outcomes based on historical patterns.
Data Source
AI summary
A controller processes data from one or more sensors of a subsystem of a vehicle. The processing includes smoothing the data and calculating a mean of the data. The controller identifies a transition point in the processed data where a moving average of the data is less than the mean by a predetermined amount indicating a trend. The controller selects a segment of the processed data subsequent to the transition point, detects the trend in the segment using regression, and extrapolates the detected trend to reach a predetermined fault threshold. The controller predicts a failure of the subsystem based on a slope of the extrapolated trend and provides an indication of the prediction based on the slope to schedule a service for the subsystem.