基于改进卷积受限玻尔兹曼机的滚动轴承故障诊断

Fault Diagnosis of Bearing Based on Improved Convolution Restricted Boltzmann Machine

针对滚动轴承初期故障信号所存在的信息微弱,背景噪声复杂,早期故障易被覆盖,特征提取难以提取和识别的难题。结合卷积玻尔兹曼机的特性,文章提出了一种基于改进卷积玻尔兹曼机的滚动轴承故障诊断方法。该方法通过对原始输入数据的边缘数据进行补零操作,得到重构模型,此外,在目标函数中引入了最大似然,在隐单元中,结合交叉熵的概念,利用稀疏惩罚的特性,进而抑制其激活概率。实验证明,改进后的卷积玻尔兹曼机误差更小,解决了传统的受限玻尔兹曼机不能够处理大量数据、诊断种类单一、用时较长、计算复杂以及灵活性差等问题,同时利用该方法能够有效提高滚动轴承的故障诊断的准确率。

Problems with the initial fault signal of the rolling bearing are many. The information is weak, the background noise is complex, the early faults are easily covered, and the feature extraction is difficult to extract and identify. Combined with the characteristics of the Convolution Boltzmann Machine, this paper proposes a fault diagnosis method for rolling bearing based on improved Restricted Convolution Boltzmann Machine. The method obtains a reconstructed model by performing a zero-padding operation on the edge data of the original input data. In addition, maximum likelihood is introduced in the objective function. In the implicit unit, combined with the concept of cross entropy, the characteristics of sparse punishment are utilized to suppress the activation probability. Experiments have shown that the improved Convolution Boltzmann Machine has less error. It solves the problem that traditional Restricted Boltzmann Machines cannot solve. These problems are dealing with large amounts of data, a single type of diagnosis, a long time spent, complex calculations, and poor flexibility. At the same time, the method can effectively improve the accuracy of fault diagnosis of rolling bearings.