GitHub Project Classifies Mechanical Faults Using Contrastive Language‑Audio Pretraining

The repository showcases a method for detecting mechanical faults from sound. It employs Contrastive Language‑Audio Pretraining (CLAP) to learn audio representations. The approach treats fault detection as a classification problem. Audio recordings of equipment are fed into the pretrained model. The system outputs predicted fault categories for each sample. Code and model checkpoints are provided for reproducibility. The project highlights the potential of multimodal pretraining for industrial monitoring. Users can adapt the pipeline to other acoustic diagnostic tasks.