SDD-CNN: Small Data-Driven Convolution Neural Networks for Subtle Roller Defect Inspection

Abstract

Roller bearings are some of the most critical and widely used components in rotating machinery. Appearance defect inspection plays a key role in bearing quality control. However, in real industries, bearing defects are usually extremely subtle and have a low probability of occurrence. This leads to distribution discrepancies between the number of positive and negative samples, which makes intelligent data-driven inspection methods difficult to develop and deploy. This paper presents a small data-driven convolution neural network (SDD-CNN) for roller subtle defect inspection via an ensemble method for small data preprocessing. First, label dilation (LD) is applied to solve the problem of an imbalance in class distribution. Second, a semi-supervised data augmentation (SSDA) method is proposed to extend the dataset in a more efficient and controlled way. In this method, a coarse CNN model is trained to generate ground truth class activation and guide the random cropping of images. Third, four variants of the CNN model, namely, SqueezeNet v1.1, Inception v3, VGG-16, and ResNet-18, are introduced and employed to inspect and classify the surface defects of rollers. Finally, a rich set of experiments and assessments is conducted, indicating that these SDD-CNN models, particularly the SDD-Inception v3 model, perform exceedingly well in the roller defect classification task with a top-1 accuracy reaching 99.56%. In addition, the convergence time and classification accuracy for an SDD-CNN model achieve significant improvement compared to that for the original CNN. Overall, using an SDD-CNN architecture, this paper provides a clear path toward a higher precision and efficiency for roller defect inspection in smart manufacturing.

A Review on Deep Learning Applications in Prognostics and Health Management

“…This type of transfer learning can markedly reduce the required number of labeled samples. Using the same strategy, Xu et al adopted and compared four pretrained networks (SqueezeNet v1.1, Inception v3, VGG-16 and ResNet-18) in the context of roller bearing surface defect diagnosis [138]. They validated a gain in both convergence speed and accuracy using a pretrained network.…”

“…We can use data augmentation techniques to obtain more training samples to improve the performance of PHM applications. For example, when methods like random crop, rotation, translation, zoom, shear and elastic transformation were adopted on natural images to generate more training samples for fault diagnosis, improved accuracy was reported [62], [124], [138]- [140], [142]. The success of this type of data augmentation on natural images is explained by the human visual perception mechanism; for instance, a rotated cat can still be recognized as a cat by the human brain.…”

Research on Detecting Bearing-Cover Defects Based on Improved YOLOv3

“…Y. Gao et al [30] also proposed a neural network-based semisupervised learning method to solve the problem of detecting steel defects. X. Xu's team [31] proposed a small data-driven convolutional neural network (SDD-CNN) and semisupervised data augmentation (SSDA) to effectively solve the problem of roll surface defect detection. However, data randomness and model quality instability exist in the above semisupervised learning.…”

FAR-Net: Feature-Wise Attention-Based Relation Network for Multilabel Jujube Defect Classification

“…In the data preprocessing stage, all samples were resized to 299 × 299 to meet the input requirement of Inception v3. Then, the semi-supervised data augmentation method (SSDA) [22] was used for data augmentation. Eventually, after the ARL, the label-wise fusion feature…”