论文部分内容阅读
本文介绍了人工神经网络的基本原理、结构。针对化工中经常遇到的高度非线性问题，以塔板研究中的泄漏模型为例，采用人工神经网络中的ＢＰ（反向传播）算法进行处理。并对网络的拓扑结构、阈值函数、学习速率、动量因子等进行了初步的探讨。结果显示，ＢＰ算法和经典的多元因子分析相比，对数据具有更强的过滤能力和自适应能力，建模效果更佳。
This paper introduces the basic principle and structure of artificial neural network. Aiming at the highly nonlinear problems often encountered in the chemical industry, taking the leakage model in the plate research as an example, this paper uses the BP (Back Propagation) algorithm in artificial neural networks. And the network topology, threshold function, learning rate, momentum factor were discussed. The results show that compared with the classical multivariate analysis, the BP algorithm has better filtering ability and self-adaptability to the data, and the modeling effect is better.