Power Converter Failure Prediction via Sensor Difference Histograms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional failure prediction systems for power converters in traction motors lack accuracy in predicting failures, which can lead to unexpected breakdowns and maintenance challenges.
Innovation Solution
A failure prediction system that includes sensors and a controller to calculate differences in measurement values over time, create a histogram of these differences, apply variable conversion using statistical processing or machine learning, and output a warning signal when a damage threshold is exceeded, indicating impending failure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional failure detection methods are used, then the system can detect failure signs, but the prediction accuracy is insufficient leading to unexpected breakdowns
Solution Approach 1:
The patent transforms raw sensor measurement values into damage levels through multiple parameter transformations: calculating differences between consecutive measurements, creating histograms of these differences, applying variable conversion with weight factors to histogram bin counts, and accumulating weighted values over time. This multi-stage parameter transformation process converts ordinary measurement data into a reliable damage level indicator that accurately predicts failures before they occur.
Solution Approach 2:
The patent introduces intermediate data structures as mediators between raw measurements and failure prediction. Specifically, it creates histograms as intermediate representations of measurement differences, then applies variable conversion to transform histogram bin counts into weighted values. These intermediate transformations serve as bridges that convert noisy sensor data into reliable damage level predictions, resolving the contradiction between measurement precision and reliability.
2Measurement precision
If simple threshold-based detection is used, then the system is easy to implement, but it cannot provide accurate timing predictions for failures
Solution Approach 1:
The patent segments the failure prediction process into distinct stages: (1) calculating differences between consecutive measurement values, (2) creating histograms that segment the difference values into bins, (3) applying variable conversion to segment the histogram bin counts into weighted contributions, and (4) accumulating segmented weighted values over time. This segmentation approach transforms a complex prediction problem into manageable discrete steps, achieving high prediction accuracy while maintaining systematic implementation.
Solution Approach 2:
The patent transitions from one-dimensional raw measurement values to two-dimensional histogram representations, where the x-axis represents measurement difference values and the y-axis represents frequency distributions. This dimensional transformation allows the system to capture temporal patterns and variations that simple threshold-based methods miss, enabling accurate failure timing predictions through the additional informational dimension provided by histogram analysis.
Data Source
AI summary
A failure prediction system disclosed herein is configured to predict a failure of a power converter which converts output power of a power source to power for driving a traction motor. The system may include: a sensor provided at the power converter; and a controller configured to predict a failure of the power converter based on a measurement value of the sensor. The controller may be configured to: calculate a difference between previous and present measurement values of the sensor, wherein the controller repeatedly calculates the difference at predetermined time intervals; obtain intermediate data by applying variable conversion to a plurality of the past differences; calculate a damage level of the power converter based on the intermediate data; and output a warning signal in a case where the damage level exceeds a damage threshold, wherein the warning signal indicates that a timing when the failure occurs is approaching.