基于uml的系统鲁棒性建模

Modeling of system robustness based on uml

系统野点会对系统数据鲁棒性检测精度产生影响,为了提高系统数据检测精度,基于uml建模语言,利用PLS结构提出了一种新型系统鲁棒性检测建模方法。通过构造出的径向基神经网络结构以及输入和输出端,对系统的输出数据进行预测。利用隐马尔克夫检测模型,对系统数据输出数据的预测值和系统真实值进行数据对比分析,达到检测系统野点检测的目的。根据RBFN所及具备的无限趋近输入端数据的特性,将RBFN的输出点替换系统数据野点。再根据剔除野点后的系统数据,通过设置PLS系统结构,实现基于uml的系统鲁棒性建模。实验数据表明,在系统数据野点为高密度状态下,设计的去野点鲁棒性建模方法比传统鲁棒性建模方法监测精度可以提高29%,当数据处于低密度状态下,监测精度会提高17%。可以明显提高系统鲁棒性建模精度。

In order to improve the accuracy of system modeling in the case of system data having wild points,a new method of system robustness modeling is proposed based on the uml modeling language using PLS structure. Firstly,a radial basis neural network with input and output structure is constructed to predict the output data of the system. The hidden markov detection model is used to compare and analyze the output predictive value and the real value of the system data,so as to achieve the purpose of detecting the system wild point. The RBFN itself has the property of approximating the system output,so the system field point is replaced by the output point of RBFN. Finally,the dynamic PLS structure is introduced to realize the uml-based system robustness modeling. Experimental data show that in the condition of system data outliers for high density,design modeling method to outliers robustness robustness than the traditional modeling methods monitoring accuracy can be increased by 29%,when the data is in a state of low density,the monitoring precision will be increased by 17%. The system robustness modeling accuracy can be improved obviously.