【摘要】 针对深度置信网络(Deep Belief Network, DBN)用于轴承故障诊断时,网络层结构调试比较费时等问题,提出一种基于粒子群优化(Particle Swarm Optimization, PSO)的DBN算法,以及基于该算法的轴承故障诊断模型。该模型利用PSO算法优选DBN网络结构,并通过自适应时刻估计法微调模型参数,随后运用具有最优结构的DBN模型直接从原始振动信号中提取低维故障特征,并将其输入到Soft-max分类器中识别轴承的故障模式。该算法与支持向量机、BP神经网络、DBN、堆叠降噪自编码等方法进行对比分析,实验结果表明,PSO改进的DBN算法具有更高的准确率以及更好的鲁棒性。更多还原

【Abstract】 Aiming at the problem of debugging network layer structure being time-consuming during deep belief network(DBN) being applied in bearing fault diagnosis, the DBN algorithm improved with particle swarm optimization(PSO) and the bearing fault diagnosis model based on the DBN algorithm improved with PSO were proposed. In the proposed model, PSO algorithm was used to optimize DBN network structure, and the adaptive time instant estimation algorithm was used to finely tune the model parameters. Then, the DBN model with the optimal structure was used to extract low-dimensional fault features in the original vibration signals. The extracted fault features were input into a Soft-max classifier to identify bearing fault modes. The results using the proposed model were compared with those using SVM, BP neutral network, DBN and stacked de-noising auto-encoders, respectively. The comparison results showed that the DBN algorithm improved with PSO has higher accuracy and better robustness.更多还原