故障诊断问答指令数据集
收藏资源简介:
该数据集是由华南理工大学智能工程系构建的大规模多模态指令数据集,包含振动时频图像-文本标签对和人类指令-真实标签对。数据集用于大型视觉语言模型(LVLM)的指令微调,目的是提高模型在生成高质量故障诊断报告方面的能力。数据集整合了不同模态的数据,如声学、振动和自然语言,旨在为机械健康管理提供更全面的视角,解决机械系统中的故障诊断问题。
This large-scale multimodal instruction dataset was developed by the Department of Intelligent Engineering, South China University of Technology. It contains two types of paired data: vibration time-frequency image-text label pairs and human instruction-ground truth label pairs. This dataset is intended for instruction tuning of large vision-language models (LVLM), with the goal of improving the models' ability to generate high-quality fault diagnosis reports. By integrating multi-modal data including acoustic, vibration and natural language data, the dataset aims to provide a more comprehensive perspective for mechanical health management and address fault diagnosis issues in mechanical systems.




