【摘要】 以磷酸盐石墨铸型作为研究对象,用2组不同的正交实验方案结果作为样本集,用BP神经网络对这2个样本进行了互预测,结果表明:对样本集进行恰当的预处理,包含信息量大的样本集能够对包含信息量小的样本集以极高的精度进行预测,而包含信息量小的样本集不能对包含信息量大的样本集进行预测。这给科研人员提供了新的实验设计思路,能大大节省时间和劳力。更多还原

【Abstract】 The studied object was the phosphategraphite molds. The samples were the data of two groups of the orthogonal tests. Two samples were mutually forecast in the artificial neutral network of BP. The results were as follows. If the samples were preprocessed, the sample collection contained more information could accurately forecast the little information’s?sample collection. But the little information sample collection could not forecast the more information sample collection. And thus, it provided a new testdesign route for the learners, and the time and labor could be greatly saved.更多还原