基于孪生网络的仿真模型智能排序评估方法  

Intelligent ranking evaluation method of simulation models based on siamese network

作　　者：杨帆[1] 马萍[1] 李伟[1] 杨明[1] YANG Fan;MA Ping;LI Wei;YANG Ming(Control and Simulation Center,Harbin Institute of Technology,Harbin 150080,China)

机构地区：[1]哈尔滨工业大学控制与仿真中心,黑龙江哈尔滨150080

出　　处：《系统工程与电子技术》2023年第7期2060-2068,共9页Systems Engineering and Electronics

基　　金：国家重点研发计划(2018YFB1701600)资助课题。

摘　　要：针对动态输出多样本条件下的多备选仿真模型择优问题,提出了一种基于孪生卷积神经网络(siamese convolutional neural network,SCNN)的仿真模型智能排序评估方法。首先,将仿真数据和参考数据的一致性度量问题转化为二者特征一致性度量问题。其次,在分析评估数据特点和对比试验结果的基础上,确定采用SCNN实现评估数据的特征提取。接下来,给出基于SCNN的仿真模型排序评估方法,包括网络结构初步设计、网络参数训练调优和仿真模型排序评估三部分。最后,通过实例应用,验证了该方法在评估数据特征提取和仿真模型排序评估方面的有效性。Aiming at ranking and selection of simulation models which dynamic outputs and multiple samples,an intelligent ranking evaluation method of simulation models based on siamese convolutional neural network(SCNN)is proposed.Firstly,the consistency measurement problem of simulation data and reference data is transformed into a feature consistency measurement problem.Secondly,based on the analysis of the characteristics of the evaluation data and the comparative test results,it is determined to use SCNN to achieve feature extraction of the evaluation data.Next,a ranking evaluation method for simulation models based on SCNN is presented,which includes three parts:preliminary design of network structure,training and optimization of network parameters,and ranking evaluation of simulation models.Finally,an example application verifies the effectiveness of the proposed method in evaluation data feature extraction and simulation models ranking and selection.

关 键 词：仿真模型排序评估 孪生卷积神经网络 多样本数据 特征提取 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]