一种云样本的控制产生及在电路故障诊断中的应用

A method of cloud-sample control and generation with application to circuit fault diagnosis

为了选择电路故障诊断中的特征样本,提出了产生云样本的方法,并用于神经网络的训练和识别.首先采用逆向云理论对初始特征样本进行统计以获取数字特征,其次采用正向云理论产生扩展训练样本,并用新产生的样本训练两种神经网络.仿真结果表明,采用云样本训练的神经网络要比采用常规样本训练的性能稳健,具有较好的抗噪声性能,在模拟电路故障诊断中达到了较好的诊断效果.

To select feature samples in circuit fault diagnosis, we propose a method of cloud-sample generation, and apply it to artificial-neural-network training and recognition. First, the inverse cloud model theory is employed to obtain the statistical digital feature of the samples, and then the extended training data set is produced by positive cloud theory. Second, two kinds of networks are trained with the newly produced data set. Simulation results reveal that the performance of the neural network trained by the cloud samples is better than that trained by the conventional methods. The results also proved that the network is robust to random noise, and the proposed method is valid in the faults diagnosis of analog circuit.