基于增量型极限学习机的材料力学性能预测
|
摘要
高Co-Ni二次硬化钢是一种多元系合金,其力学性能受材料成分及热处理工艺等条件影响。传统的数学模型很难对材料力学性能进行预测。采用增量型极限学习机(I-ELM)建立高Co-Ni二次硬化钢材料模型,根据模型预测的结果研究微量元素Co和时效温度对高Co-Ni二次硬化钢材料力学性能的影响,并且在多项性能指标方面对高Co-Ni二次硬化钢数据的I-ELM模型与BP神经网络模型进行比较。实验结果表明,在对较少样本数据的模型训练时,高Co-Ni二次硬化钢的I-ELM模型预测结果与实验数据基本吻合,I-ELM模型的拟合精度和训练速度均优于BP神经网络模型,为今后高Co-Ni二次硬化钢的材料研究提供参考。
High Co-Ni secondary hardening steel is a muti-component alloy, whose mechanical properties are affected by material composition and heat treatment process. Traditional mathematical models are difficult to predict the mechanical properties of materials. This paper uses the incremental extreme learning machine(I-ELM) algorithm to establish a high Co-Ni secondary hardening steel material model, studies the effect of trace element Co and aging temperature on mechanical properties of high Co-Ni secondary hardening steel according to the model predicting results, and comparing the I-ELM model and BP neural network model of high Co-Ni secondary hardened steel data on multiple performance indicators. The experimental result in this paper shows that the I-ELM model predicting results of the high Co-Ni secondary hardening steel are basically consistent with the experimental data when training the model with less sample data, both the fitting accuracy and training speed of the I-ELM model are better than the BP neural network, thereby providing a reference for material research of high Co-Ni secondary hardening steel in future.
High Co-Ni secondary hardening steel is a muti-component alloy, whose mechanical properties are affected by material composition and heat treatment process. Traditional mathematical models are difficult to predict the mechanical properties of materials. This paper uses the incremental extreme learning machine(I-ELM) algorithm to establish a high Co-Ni secondary hardening steel material model, studies the effect of trace element Co and aging temperature on mechanical properties of high Co-Ni secondary hardening steel according to the model predicting results, and comparing the I-ELM model and BP neural network model of high Co-Ni secondary hardened steel data on multiple performance indicators. The experimental result in this paper shows that the I-ELM model predicting results of the high Co-Ni secondary hardening steel are basically consistent with the experimental data when training the model with less sample data, both the fitting accuracy and training speed of the I-ELM model are better than the BP neural network, thereby providing a reference for material research of high Co-Ni secondary hardening steel in future.
引文
[ 1 ]刘贵立,张国英,曾梅光.用人工神经网络模型研究微量元素对钢力学性能的影响[J].钢铁研究,2000(28):48-50.
[ 2 ]HUANG G B,ZHU Q Y,SIEW C K.Extreme learning machine:Theory and applications[J].Neurocomputing,2006,70(1):489-501.
[ 3 ]HUANG G B,LEI C,SIEW C K.Universal approximation using incremental constructive feed forward networks with random hidden nodes[J].IEE Trans Neural Networks,2006,17(4):879-892.
[ 4 ]HUANG G B.Learning capability and storage capacity of two-hidden-layer feedforward networks[J].IEE Trans Neural Networks,2003,14(2):274-281.
[ 5 ]TAMURA S,TATEISHI M.Capabilities of a four-layered feedforward neural network:four layers versus three.[J].IEE Trans Neural Networks,1997,8:251-255.
[ 6 ]宋绍剑,向伟康,林小峰.增量型极限学习机改进算法[J].信息与控制,2016,45(6):735-741,758.
[ 7 ]崔建国,张善好,于明月,等.基于增量型极限学习机的飞机复合材料结构损伤识别[J].科学技术与工程,2018,18(4):191-196.
[ 8 ]沈龙凤.基于极限学习机的聚变堆结构材料力学性能预测方法及其应用研究[D].合肥:中国科学技术大学,2016.
[ 9 ]张静.基于人工神经网络的材料力学性能预测[J].上海电机学院学报,2006,9(5):20-23.
[10]张国英,刘贵立,曾梅光,等.高Co-Ni二次硬化钢性能预测及多指标优化[J].钢铁,2000(9):56-59.
[11]刘贵立,张国英.高Co-Ni二次硬化钢力学性能研究[J].材料科学与工程,2000,18(1):49-51,26.
[12]LUO Xiong,YANG Xiaona,JIANG Changwei,et al.Timeliness online regularized extreme learning machine[J].Int Jmach Learn Cyb,2018,9(3):465-476.
[ 2 ]HUANG G B,ZHU Q Y,SIEW C K.Extreme learning machine:Theory and applications[J].Neurocomputing,2006,70(1):489-501.
[ 3 ]HUANG G B,LEI C,SIEW C K.Universal approximation using incremental constructive feed forward networks with random hidden nodes[J].IEE Trans Neural Networks,2006,17(4):879-892.
[ 4 ]HUANG G B.Learning capability and storage capacity of two-hidden-layer feedforward networks[J].IEE Trans Neural Networks,2003,14(2):274-281.
[ 5 ]TAMURA S,TATEISHI M.Capabilities of a four-layered feedforward neural network:four layers versus three.[J].IEE Trans Neural Networks,1997,8:251-255.
[ 6 ]宋绍剑,向伟康,林小峰.增量型极限学习机改进算法[J].信息与控制,2016,45(6):735-741,758.
[ 7 ]崔建国,张善好,于明月,等.基于增量型极限学习机的飞机复合材料结构损伤识别[J].科学技术与工程,2018,18(4):191-196.
[ 8 ]沈龙凤.基于极限学习机的聚变堆结构材料力学性能预测方法及其应用研究[D].合肥:中国科学技术大学,2016.
[ 9 ]张静.基于人工神经网络的材料力学性能预测[J].上海电机学院学报,2006,9(5):20-23.
[10]张国英,刘贵立,曾梅光,等.高Co-Ni二次硬化钢性能预测及多指标优化[J].钢铁,2000(9):56-59.
[11]刘贵立,张国英.高Co-Ni二次硬化钢力学性能研究[J].材料科学与工程,2000,18(1):49-51,26.
[12]LUO Xiong,YANG Xiaona,JIANG Changwei,et al.Timeliness online regularized extreme learning machine[J].Int Jmach Learn Cyb,2018,9(3):465-476.