基于随机森林算法的虚拟仿真实验室仪器故障预警方法  

Fault early warning method of virtual simulation laboratory instrument based on random forest algorithm

作　　者：李梅琴[1] LI Meiqin(Training and Experiment Management Center,Minxi Vocational and Technical College,Longyan 364021,China)

机构地区：[1]闽西职业技术学院实训实验管理中心,福建龙岩364021

出　　处：《山东理工大学学报（自然科学版）》2023年第6期63-68,共6页Journal of Shandong University of Technology:Natural Science Edition

摘　　要：针对仪器故障频率的预估能力较低,影响预警时效的问题,基于随机森林算法,设计虚拟仿真实验室仪器故障预警方法。以集成神经网络对应仪器故障特征,对较弱置信度的信号作加强处理;应用随机森林算法,构建仪器故障预警模型,以信息熵和信息增益定位不同类型故障信号,以递归形式组建分类树,确定故障特征与设备之间关系,发出故障预警提示,完成虚拟仿真实验室仪器故障预警。测试结果表明,在余弦信号分别为246.12、425.87和648.03 Hz时,对故障信号的产生频率进行预估,估计误差可以控制在0.0025 Hz以下;当对故障信号加入噪声后,频率最大预估误差只有0.0801 Hz,能够对故障信号进行准确判断,及时做出故障预警。Aiming at the problem that the prediction ability of instrument failure frequency is low,which affects the effectiveness of early warning,an instrument failure early warning method for virtual simulation laboratory is designed based on random forest algorithm.The integrated neural network is used to enhance the signals with weak confidence.An instrument fault early warning model is established using the random forest algorithm.Different types of fault signals are located by information entropy and information gain.A classification tree is established in the form of recursion to determine the relationship between fault characteristics and equipment,issue a fault early warning prompt,and complete the instrument fault early warning in the virtual simulation laboratory.The test results show that when the cosine signal is 246.12,425.87 and 648.03 Hz,respectively,the estimation error of the frequency of the fault signal can be controlled below 0.0025 Hz.When the noise is added to the fault signal,the maximum estimation error of the frequency is only 0.0801 Hz,which can accurately judge the fault signal and make fault early warning in time.

关 键 词：虚拟仿真实验室 随机森林算法 实验室仪器 故障预警 信息增益 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]