加固式公共计算服务器健康管理系统设计

Design of health management system for reinforced public computing server

基于实时监测加固式公共计算服务器的健康状态,快速精确诊断舰载武器火控系统故障的目的,文中采用了VITA46(VersaModule Eurocar International Trade Association,VPX)总线架构,以龙芯3A3000为处理器控制芯片,设计了加固式公共计算服务器的健康管理系统。该系统采用隐马尔可夫模型(Hidden Markov Model,HMM)对加固式公共计算服务器进行状态监测和故障诊断,结合t-分布邻域嵌入(t-distributed Stochastic Neighbor Embedding,t-SNE)算法和线性判别分析(Linear Discriminant Analysis,LDA)算法对原始数据进行特征提取和降维处理,利用多智能体遗传算法(Multi-Agent Genetic Algorithm,MAGA)对HMM进行参数估计,通过计算(Kullback Leibler,KL)距离评估系统的健康状况,并根据HMM的最大似然概率得出故障状态类型。通过搭建仿真测试系统,实时监测加固式公共计算服务器的电压、电流、温度和工作状态等信息,得出上述方法有效可行,故障识别率提高了20%,为高效地维护系统设备的健康运行提供了便利。

Based on the purpose of real-time monitoring of the health status of the reinforced public computing server and rapid and accurate diagnosis of the fault of the shipboard weapon fire control system,this paper designs the health management system of the reinforced public computing server by adopting the VITA46(VPX)bus architecture and taking the loongson 3A3000 as the processor control chip.The system adopts the Hidden Markov Model(HMM)for state monitoring and fault diagnosis of the reinforced public computing server,and combines the t-distributed Stochastic neighbor embedding(tSNE)algorithm and Linear Discriminant Analysis(LDA)algorithm for feature extraction and dimensionreduction processing of the original data.Multi-Agent Genetic Algorithm(MAGA)was used to estimate HMM parameters,and the health status of the system was evaluated by Kullback Leibler(KL)distance,and the fault state type was obtained based on the maximum likelihood probability of HMM.By setting up a simulation test system to monitor the voltage,current,temperature and working status of the reinforced common computing server in real time,it is concluded that the above method is effective and feasible,and the fault identification rate is increased by 20%,which provides convenience for efficient maintenance of the healthy operation of the system equipment.