基于模糊聚类分析和神经网络的显微硬度预测

Micro-hardness prediction based on fuzzy clustering analysis and neural network

数学思想加以神经网络算法建立的数学模型,能够解决实际问题,神经网络作为辅助计算工具已经在相关力学性能预测中得到了广泛的应用。基于显微硬度数据样本的非均匀性特征,提出将模糊聚类分析与神经网络结合对显微硬度测量值进行预测。将模糊聚类分析方法对训练数据进行相似性分类的结果作为BP神经网络的训练样本训练各聚类预测模型。以TiAl复合材料的显微硬度实验观测值为研究对象,根据样本与各聚类之间的相似程度作为测试标准进行实验验证,预测结果的平均绝对误差为4.8157,均方误差为5.0904,该结果表明实验建立的预测模型的精度优于神经网络预测模型。

Supplemented by computer tools,the mathematical model was established by appropriate mathematical ideas with the neural network algorithm,which can solve the practical problems.Neural network as an auxiliary computing tools have be widely used in some mechanics performance prediction.Because of heterogeneity characteristics in micro hardness data samples,a fuzzy cluster method and neural network were combined to forecast micro hardness measurement in this paper.Fuzzy cluster analysis was applied to carry on the similarity training data classification,and BP neural network prediction model was established based on clustering samples.Similar data was chosen as training samples for each forecast model.The experimental values of micro hardness of TiAl composites were taken as the research object,and the similarity between samples and clusters was used as the test standard for experimental verification.The value of MAE is4.8157and MSE is5.0904,which show that the model is superior to the general prediction model.