摘要
以三元乙丙橡胶(EPDM)胶料配方和天然橡胶(NR)胶料配方为例,将配方中各组分的用量作为输入,硫化橡胶的基本物理机械性能作为输出,建立了基于极限学习机(ELM,extreme learning machine)神经网络的配方性能预测模型,并给出两种配方的预测结果和相对误差。结果表明,ELM神经网络模型能够准确预测出EPDM配方和NR配方硫化橡胶的基本物理机械性能,且平均相对误差在7%以内,具有较高的预测精度。
EPDM(Ethylene propylene diene monomer)compound formula and NR(Natural rubber)compound formula are taken as an example. The amount of each component in the formula is taken as input,and the basic physical and mechanical properties of vulcanized rubber are taken as output. After a series of training and debugging,the formulation performance prediction model of the ELM(Extreme learning machine) neural network is established,and the prediction model gives the prediction results and relative errors of the two formulations. The results show that the ELM neural network model can accurately predict the basic physical and mechanical properties of EPDM formula and NR formula vulcanized rubber,and the average relative error is less than 7%,which has high prediction accuracy.
引文
[1] Bin Wang,Jian Hua Ma,You Ping Wu. Application of artificial neural network in prediction of abrasion of rubber composites[J]. Materials & Design,2013,49:802-807.
[2] Rodica-Mariana Diaconescu,Marinela Barbuta,Maria Harja.Prediction of properties of polymer concrete composite with tire rubber using neural networks[J]. Materials Science and Engineering:B,2013,178:1259-1267.
[3] Liqun Yang,Yuancheng Li,Zhoujun Li.Improved-ELM method for detecting false data attack in smart grid[J]. International Journal of Electrical Power & Energy Systems,2017,91:183-191.
[4] Nathan Lemahieu,Henri Greuner,Jochen Linke,et al. Synergistic effects of ELMs and steady state H and H/He irradiation on tungsten[J]. Fusion Engineering and Design,2015,98-99:2020-2024.
[5] 林小峰,孔伟凯.基于ELM的水泥立磨生料细度ADP控制[J].系统仿真学报,2016,28(11):2764-2770.