论文部分内容阅读
溶解度作为一项重要的物化指标,一直是化学学科的研究重点。然而,通过实验测量获得数据耗时费力,因此,科研人员建立了多种理论方法来进行估算,其中,人工神经网络因其能够关联复杂的多变量情况而受到广泛关注。本文综述了人工神经网络在物质溶解度预测方面的应用,介绍了应用最广泛的3种神经网络(BP神经网络、小波神经网络、径向基神经网络)的模型结构、预测方法和预测优势,探讨了神经网络的不足以及改进方法。文章最后对神经网络在物质溶解度预测方面的发展前景进行了展望。与其他方法相比,人工神经网络技术在物质溶解度预测方面具有预测结果精确度高、操作简单等特点,具有广阔的应用前景,但输入变量选择、隐含层节点数确定、避免局部最优等问题还需逐步建立系统的理论指导。
As an important material and chemical index, solubility has always been the focus of chemistry research. However, it takes time and effort to obtain data through experimental measurements. Therefore, researchers have established a variety of theoretical methods to make estimates. Among them, artificial neural network has drawn much attention due to its ability to relate to complex multivariable conditions. This article summarizes the application of artificial neural network in the prediction of material solubility, and introduces the model structure, prediction method and prediction advantages of the three most widely used neural networks (BP neural network, wavelet neural network, RBF neural network) The lack of neural networks and ways to improve. In the end, the prospect of neural network in the prediction of material solubility is prospected. Compared with other methods, artificial neural network technology has the advantages of high accuracy and simple operation in the prediction of material solubility, which has broad application prospects. However, the selection of input variables, the determination of hidden layer nodes and the avoidance of local optimal problems It is also necessary to gradually establish systematic theoretical guidance.