联合3D建模与改进CycleGAN的故障数据集扩增方法  

Equipment fault dataset amplification method combine 3D model with improved CycleGAN

作　　者：李宝平 戚恒熠 王满利 魏坡 LI Baoping;QI Hengyi;WANG Manli;WEI Po(College of Physics and Electronic Information,Henan Polytechnic University,Jiaozuo 454000,China)

机构地区：[1]河南理工大学物理与电子信息学院,河南焦作454000

出　　处：《光学精密工程》2023年第16期2406-2417,共12页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.52074305);河南理工大学博士基金资助项目(No.B2021-64)。

摘　　要：基于深度学习的设备故障检测系统性能很大程度上依赖于样本集的规模及类别多样性。由于工业生产中难以全面采集各类故障样本,由此就有样本集扩增需求。本文提出联合3D模型和改进CycleGAN的故障数据集扩增方法。首先,提出利用3D建模软件模拟生成各类故障图片,将其作为CycleGAN迁移网络输入,约束引导生成真实故障图像,以解决样本不足及分布不均衡问题;其次,对CycleGAN网络生成器进行改进,提出U-ResNet生成器,用以解决数据集扩增过程中的边缘模糊和梯度消失问题。将该方法应用于带式输送机跑偏检测任务,结果表明相较于其他扩增方法,该方法训练过程中轮廓结构收敛快,时效性好,应用于目标检测网络准确率达到98.1%,较原真实数据集提升4.5%。说明该数据集扩增方法可以满足故障数据集类别分布均衡,图像质量高的要求。The performance of deep-learning-based equipment fault detection systems relies heavily on the size and class diversity of the sample set.Because it is difficult to collect all types of fault sample compre⁃hensively in industrial production,there is a demand for sample set augmentation.A fault dataset amplifi⁃cation method combining 3D modeling with an improved cycle generative adversarial network(Cycle⁃GAN)is proposed.First,various equipment malfunction images generated by 3D modeling software are applied to the CycleGAN network training to guide it in generating pseudo-real images to address the prob⁃lem of insufficient samples and an uneven distribution.Second,a U-ResNet generator is used in the Cycle⁃GAN network to solve the problem of edge blurring and gradient vanishing during network training.The method was applied to the task of belt conveyor deviation detection.The experimental results show that the contour structure of the method converges quickly in the training process and has good timeliness in comparison with other amplification methods.The accuracy rate of the method is 98.1%when applying to the target detection network,which is 4.5%higher than that of the original real dataset.It meets the basic requirements of a balanced distribution of amplified datasets and high image quality.

关 键 词：数据集扩增 3D模型 CycleGAN U-ResNet 带式输送机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]