Timely failure detection for bearings is of great importance to prevent economic losses in the industry. In this article we propose a method based on Convolutional Neural Networks (CNN) to estimate the level of wear in roller bearings in each of its elements, inner race, outer race and rolling element. First of all, an automatic labeling of the raw vibration data for each element of the bearing is performed to obtain different levels of wear, by means of the Root-Mean-Square features along with the Shannon’s entropy to extract features from the raw data, which is then grouped in different classes using the K-means algorithm to obtain the labels for seventeen levels of degradation condition. Then, the raw vibration data is converted into small square images, each point of the data representing one pixel of the image. Following this, we propose a CNN model based on the AlexNet architecture to classify the wear level and diagnose the rotating system. To train the network and validate our proposal, we use a dataset from the center of Intelligent Maintenance Systems (IMS), and extensively compare it with other methods reported in the literature. The effectiveness of the proposed strategy proved to be excellent compared to other approaches in the state of the art.

及时检测轴承的故障对于防止行业中的经济损失非常重要。在本文中，我们提出了一种基于卷积神经网络（CNN）的方法，用于估计滚动轴承各要素，内座圈，外座圈和滚动元件的磨损程度。首先，借助Root-Mean-Square特征以及Shannon熵从原始数据中提取特征，对轴承每个元素的原始振动数据进行自动标记以获得不同程度的磨损，然后使用K-means算法将其分组为不同的类别，以获取17个级别的降解条件的标签。然后，原始振动数据被转换成小的正方形图像，数据的每个点代表图像的一个像素。按照此，我们提出基于AlexNet架构的CNN模型，以对磨损程度进行分类并诊断旋转系统。为了训练网络并验证我们的建议，我们使用了来自智能维护系统（IMS）中心的数据集，并将其与文献中报道的其他方法进行了广泛比较。与现有技术中的其他方法相比，所提出的策略的有效性被证明是极好的。