论文部分内容阅读
神经网络具有优良的非线性映射逼近能力,广泛应用于化工过程建模,但神经网络建模方法属于黑箱法,所获得的模型缺乏透明性,各变量的解释性差,限制其指导化工企业优化技术决策。结合神经网络释义图、连接权法和改进的随机化测验三种方法,对复杂化工过程神经网络模型进行透明化研究。首先利用神经网络释义图可视化模型,再用连接权法对决策参数贡献率定量分析,最后利用改进的随机化测验,对模型的连接权、决策参数的综合贡献度和相对贡献率进行显著性检验,进而修剪模型。通过对复杂化工过程氢氰酸生产模型验证研究,结果表明该方法获取了过程变量的内部信息,极大地提高了模型的“可理解”能力。因此,本研究为复杂化工过程神经网路模型的透明化提供了一条很好的途径。
Neural networks have excellent nonlinear mapping approximation ability and are widely used in chemical process modeling. However, the neural network modeling method belongs to the black box method. The obtained model lacks transparency and the explanatory power of each variable is poor, which restricts its guidance to chemical enterprise optimization techniques decision making. Combined with three kinds of methods: neural network definition map, connection rights method and improved randomization test, the transparent chemical process neural network model is studied. Firstly, the visual model of ANN paraphrase graph is used, then the weight of decision-making parameter is quantitatively analyzed by using the connection weight method. Finally, the improved randomization test is used to test the model’s connection weight, the comprehensive contribution rate of the decision-making parameters and the relative contribution rate , And then trim the model. Through the validation of hydrocyanic acid production model in complex chemical processes, the results show that the method obtains the internal information of process variables and greatly enhances the model’s “understandable” ability. Therefore, this study provides a good way for the transparency of the complex chemical process neural network model.