神经网络及回归型支持向量融合健康评估模型

Health Assessment Model Based on the Fusion of Neural Network and Support Vector Regression

健康评估是IVHM研究中的关键技术之一,基于关键特征参数监测无法解决复杂组件或系统的健康评估建模问题,而基于解析法的健康评估方法对构造特征参数的数学模型要求极高,工程价值不大。本文提出了一种基于仿真的健康评估建模新方法,该方法通过组件或系统在各种健康状态条件下仿真,生成样本数据,利用BP神经网络和支持向量机的非线性映射特性,以测量信息为基础分别构造了两种健康评估模型,考虑到单一模型缺陷,再将神经网络和支持向量机训练模型进行决策融合处理,提出了一种新的健康评估模型,并以石英挠性加速度计为例进行了建模研究与验证。结果表明:测量信息完备情况下,两种单一模型均能满足健康状态评估要求;测量信息不充分时,通过对两种模型进行决策融合处理,也可取得较好的健康状态评估效果。

Health assessment is one of the key techniques in IVHM research. Key characteristic pa- rameters monitoring can not solve the health assessment problem of complex components or system, and health assessment based on analytic method with high demand for mathematical model has little engineer- ing value. This paper puts forward a new health assessment modeling method based on simulation. This method generates the sample data by assembly or system simulation in variety of health condition. Using the nonlinear mapping properties of BP neural network and support vector machine, two health evaluation model based on measurement information are constructed. Considering the single model defect, a health assessment model is presented based on the decision fusion of neural network and support vector machine training model, and the research and validation of modeling as an example by quartz accelerometer. The results show that for the complete measurement information, two kinds of single model can meet the re- quirements of health status assessment, and for inadequate measurement information, by decision fusion of the two models, there will be also gotten a good assessment of health effects.