改进ANFIS对静压箱热误差建模研究  

Thermal Error Modeling of Plenum by Improved ANFIS

作　　者：钱雨鲲 李岩舟[1] 杨正昊 秦承斌 王佳宁 吴媚 QIAN Yukun;LI Yanzhou;YANG Zhenghao;QIN Chengbin;WANG Jianing;WU Mei(Mechanical Engineering College,Guangxi University,Nanning 530004,China;Guilin Grice Technology Co.,Ltd.,Guilin 541004,Guangxi,China;Guangxi Zhuang Autonomous Region Institute of Metrology&Test,Nanning 530004,China)

机构地区：[1]广西大学机械工程学院,南宁530004 [2]桂林格莱斯科技有限公司,广西桂林541004 [3]广西壮族自治区计量检测研究院,南宁530004

出　　处：《机械科学与技术》2024年第10期1778-1785,共8页Mechanical Science and Technology for Aerospace Engineering

基　　金：广西科技重大专项桂科(AA18242011)。

摘　　要：为了减小静压箱排气孔温度不均匀对薄膜拉伸加工时的影响,通过建立热误差模型,来分析静压箱在不同输入参数下排气孔的温场情况。采用SOM-GRA相结合的综合算法得出最优测温点,以保证输入模型的数据具有代表性,将测温点数量由20降至3。利用ANFIS模型建立静压箱的热误差模型,并通过RF算法优化ANFIS中隶属度函数数量参数,将实验验证过的数值模拟数据作为输入的训练数据。预测结果表明较原ANFIS模型、BP模型和RBF模型MAE值分别下降了22.43%、59.97%和49.87%,该优化预测模型具有更高的精度。To reduce the influence of the uneven temperature of the vent hole of the static plenum on the film stretching process,a thermal error model was established to analyze the temperature field of the vent hole of the static plenum under different input parameters.The optimal temperature measurement points are obtained by a comprehensive algorithm combining SOM-GRA(Self-organizing map-grey relational analysis)to ensure that the data input into the model is representative,and the number of temperature measurement points is reduced from 20 to 3.The thermal error model of the static plenum was established with the ANFIS(Adaptive-network-based fuzzy inference system)model,and the number parameters of the membership function in ANFIS were optimized by the RF(Random forest)algorithm,and the numerical simulation data verified by the experiment was used as the input training data.The prediction results show that compared with the original ANFIS model,the BP(Back propagation)model and the RBF(Radial basis function)model,the MAE(Mean absolute error)values have decreased by 22.43%,59.97%and 49.87%,respectively,and the optimized prediction model has higher accuracy.

关 键 词：热误差模型 自组织映射网络 灰色关联分析 随机森林 自适应神经模糊推理系统 

分 类 号：TK124[动力工程及工程热物理—工程热物理]