TL;DR
This paper introduces a novel Wigner-Ville Transform (WVT)-based method for accelerating vehicle durability tests by efficiently editing load spectrums. By analyzing the instantaneous energy spectrum, the method effectively identifies and removes low-damage segments, significantly reducing test duration while preserving critical fatigue characteristics, as validated through simulations.
What Problem Does This Paper Address?
Vehicle durability testing is crucial for ensuring product reliability, but it's often time-consuming and expensive. The load spectrum applied to components during testing directly impacts fatigue life analysis. Traditional methods for accelerating these tests, like time-domain or frequency-domain editing, can inadvertently alter the signal's frequency characteristics or require extensive parameter tuning, leading to potential inaccuracies. This paper aims to develop a more efficient and robust accelerated editing method that preserves the essential damage-causing information of the original load spectrum while drastically reducing its duration.
Methods
The core of this paper is the Wigner-Ville Transform (WVT)-based accelerated editing method. Here's a breakdown:
- Wigner-Ville Transform (WVT): This time-frequency analysis technique is used to compute the Instantaneous Energy Spectrum (IES) of the original load signal. Unlike other time-frequency methods, WVT doesn't require parameter selection (like window functions), avoiding common limitations.
- Instantaneous Energy Spectrum (IES): The IES quantifies the energy contribution of each time point in the signal. The paper posits that higher instantaneous energy correlates with higher damage potential.
- Genetic Algorithm (GA) for Threshold Optimization: To automatically determine the optimal threshold for identifying "no/minor damage" segments in the IES, a GA is employed. This ensures that the editing process is objective and considers both time compression and the preservation of critical fatigue parameters.
- Editing Process:
- The original signal is pre-processed and transformed into an analytic signal.
- The WVT is applied to obtain the time-frequency distribution.
- The IES is calculated by integrating the WVT along the frequency axis.
- The GA determines an optimal threshold for the IES.
- Time-domain points corresponding to IES values below the threshold (deemed low-damage) are removed.
- The remaining segments are spliced to form the accelerated load spectrum.
The paper also compares this WVT-based method against traditional Time-Domain Editing, Short-Time Fourier Transform (STFT)-based, and S-Transform (ST)-based methods.
Datasets
The study utilizes a real-world dataset:
- SUV Suspension Coil Spring Load Signal: This signal was collected from a vehicle undergoing testing on a professional proving ground using strain gauges. The original signal had a duration of 320 seconds.
Key Insights and Ideas
- WVT's Advantage in Parameter-Free Analysis: The WVT's inherent lack of parameter selection (like window size or type) makes it more robust and less prone to user-induced errors compared to STFT or Wavelet Transforms, which often require extensive tuning.
- Instantaneous Energy as a Damage Proxy: The paper's central hypothesis is that instantaneous energy, as derived from the WVT, is a reliable indicator of damage contribution. This allows for targeted removal of low-impact segments.
- GA for Objective Thresholding: Using a GA to set the editing threshold ensures that the optimization is systematic and considers multiple criteria (time duration, pseudo-damage retention, statistical parameter errors), leading to a more balanced and effective editing process.
- Preservation of Fatigue Characteristics: The WVT-based method, even with significant time compression, effectively maintains key statistical parameters (mean, RMS, kurtosis) and frequency-domain characteristics (PSD), crucial for accurate fatigue life prediction.
Key Results
- Significant Time Compression: The WVT-based method achieved a time duration of 244.19 seconds, resulting in a time duration ratio of 76.30%. This is a substantial reduction compared to the original 320-second signal.
- Competitive Damage Retention: The pseudo-damage retention ratio was maintained at 97.00%, comparable to or better than other methods.
- Statistical Parameter Accuracy: Errors in mean, kurtosis, and RMS values for the WVT-edited signal were all below 15% (1.4%, 13.8%, and 1.7% respectively), demonstrating good agreement with the original signal.
- PSD Consistency: The Pearson correlation coefficient between the original and WVT-edited PSD was 0.990, indicating excellent preservation of frequency-domain characteristics.
- Fatigue Simulation Validation: Fatigue simulations using the edited signals showed that the WVT-based method produced results (damage value, failure point) virtually identical to the original, with a significantly reduced simulation duration (95.03s vs. 137.15s).
Limitations and Future Work
- Limitations: While the WVT-based method shows superior performance, the paper notes that the error in kurtosis for the WVT-edited signal (13.8%) was slightly higher than some other methods, though still within the acceptable 15% limit. The paper also acknowledges that the WVT can sometimes suffer from cross-term interference, although this is less of an issue when analyzing signals with clear energy concentrations.
- Future Work: The authors suggest that the proposed WVT-based accelerated editing method can be extended to process other types of signals, such as wheel center sextant force and strain, indicating its potential broad applicability in vehicle durability testing and analysis.
Summary updated