采用BP神经网络模型预测改性骨胶的初始黏度

Initial Viscosity Prediction of Bone Glue with BP Neural Network Model

建立了一个基于L-M(Levenberg-Marquardt)优化算法的BP神经网络预测模型,用于骨胶初始黏度的预测。对涉及骨胶接枝改性的几个主要因素,如氢氧化钠用量、碱解时间、接枝共聚温度、环氧氯丙烷用量和接枝共聚时间等为考察对象,以初始黏度为指标,对16个试验样本进行了训练建模,并对2个测试样本进行了预测。结果表明:骨胶初始黏度预测值和试验值符合良好,相对误差小于2%。该方法可行有效,为快捷、经济地开发研制新的胶粘剂提供了新的思路和有效手段。

A model using L - glue. Some factors about of alkaline hydrolysis, th grafting M optimization algorithm BP neural network was established to predict the initial viscosity of bone of bone glue were studied such as the amounts of sodium hydrate and epichlorohydrin, the time e polymerization temperature. The initial viscosity was used as inspecting marker, the model was estab- lished on training the 16 samples, and 2 samples were used as prediction. The resuhs show that the predicted dates are in good agreement with the experimental dates, with the relative error being less than 2%. This system has friendly interfaces, extensive application, good operating feasibility and reliability examined with the present of adhesive.