论文部分内容阅读
使用概率神经网络(PNN)对制冷系统7种常见故障进行诊断,包括系统故障和局部故障。详细介绍了应用PNN建立故障诊断模型以及平滑因子寻优过程,并探索了样本规模对最佳平滑因子和诊断正确率的影响。将PNN与人工神经网络中最常用的误差反向传播(BP)神经网络进行比较,结果表明,PNN网络的诊断正确率比BP网络诊断正确率高3.48%,且诊断耗时更短,并且PNN网络的单次训练结果更可靠。尽管2种网络的训练结果均显示系统故障比局部故障更难以被识别,但使用PNN网络进行诊断时,系统故障的诊断正确率明显高于BP网络的诊断正确率。
Probabilistic neural network (PNN) is used to diagnose seven kinds of common faults of refrigeration system, including system faults and local faults. The application of PNN to establish the fault diagnosis model and the smoothing factor optimization process are introduced in detail, and the influence of the sample size on the optimal smoothing factor and diagnosis accuracy is explored. Comparing the PNN with BP neural network, which is the most commonly used artificial neural network, the results show that the diagnostic accuracy of PNN is 3.48% higher than that of BP neural network, and the diagnosis takes less time and the PNN The network’s single training results are more reliable. Although training results of both networks show that system faults are more difficult to identify than local faults, the diagnostic accuracy of system faults is significantly higher than that of BP networks when using PNN networks.