New methods generate synthetic bearing vibration signals with target fault probabilities

Researchers have developed two novel methods to generate synthetic bearing vibration signals with specific fault probabilities, addressing the scarcity of borderline samples in existing datasets. The first method, Probability-Regularized Generative Adversarial Network (PR-GAN), modifies real signals using a residual generator while guiding a classifier towards a target probability. The second, a Wachter-style counterfactual (CF) procedure, directly optimizes input signals to achieve desired probabilities with minimal deviation from the original. Evaluations on bearing datasets showed that the CF method is more reliable in steering probabilities and requires smaller signal changes, while PR-GAN offers faster runtime in most scenarios. AI

IMPACT These methods could improve the training of AI models for predictive maintenance by providing more diverse and targeted synthetic data.

RANK_REASON The cluster contains an academic paper detailing novel methods for generating synthetic data. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Case Western Reserve University
• Hugging Face
• Paderborn
• PR-GAN
• Seyed Mohammadreza Alavi
• Wachter-style counterfactual
• WGAN-GP

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →