Development of automotive body which can detected structural damage and it's structural health monitoring

近年来，无线通信技术的发展能够为驾驶员提供有用的信息和主动安全。虽然许多研究人员试图通过车载电视摄像机或探测车系统来监测交通拥堵、天气状况、路面状况和汽车维护，但这些传感器信号在车辆运行时具有非平稳和瞬时特征。结构健康监测用于此目的。在结构健康监测中，内置在结构中的传感器的响应信号用于监测结构的状况。当使用应变、振动、声音和红外传感器对真实结构进行监测时，它们通常会受到环境和载荷变化的影响，导致现象与响应信号之间存在非线性关系。由于来自持续经历振动的类似车辆的结构的响应信号包含关于车辆的o…的信息更多的运行条件和环境，需要一种高精度的监测和分析车辆结构的技术，以准确地解释响应信号的变化。为了预测现象的发生，我们需要建立一个将响应信号与现象联系起来的模型。对于高度准确的预测，需要强大的泛化能力。一般的学习工具可以用于建模。这些技术采用的学习过程能够创建复杂的非线性模型，但它们需要适当地定义各种学习参数。即使是能够正确地对学习数据进行分类的模型也可能只产生局部解。为了得到具有很强泛化能力的模型，学习参数必须通过反复试算来确定，因此，本研究将统计学习和估计技术的支持向量机方法应用于结构健康诊断。由于基于统计学习理论建立具有很强泛化能力的模型的方法显然是一个优化问题，因此预计支持向量机将给出对现象的高精度预测。通过将支持向量机应用于依赖于各种因素的振动响应，并将其用于车辆结构健康诊断，对支持向量机进行了研究和评价。较少

In recent years, development of the wireless communication technology is able to provide useful information and active safety for the drivers. Though many researchers try to monitor a traffic congestion, weather condition, road surface condition and car maintenance by the on-board TV camera or the probe car system, sensor signals of these have non-stationary and transitory characteristics while the vehicle is in motion. Structural health monitoring is used for this purpose. In structural health monitoring, the response signals from sensors built into a structure are used to monitor the condition of the structure. When strain, vibration, sound and infrared sensors are used for monitoring real structures, they are generally affected by environmental and load changes often resulting in a nonlinear relationship between phenomena and the response signals. Since the response signals from a vehicle-like structure that continually experiences vibration contain information about the vehicle's o … More perating conditions and environment, a highly accurate technique for monitoring and analyzing the vehicle's structure is required to interpret changes in the response signal accurately.In order to predict the occurrence of phenomena we need to create a model that relates the response signals to the phenomena. Strong generalization capabilities are required for highly accurate predictions. General learning tools can be used for modeling. The learning processes employed by these techniques enable the creation of a complicated nonlinear model, but they require defining the various learning parameters appropriately. Even a model capable of classifying learning data correctly may yield only a local solution. In order to obtain a model that has great generalizability, the learning parameters must be determined by trial and error.In this study, therefore, the Support Vector Machines method that uses a statistical technique for learning and estimating is applied to structural health diagnosis. Since the method for creating a model of great generalizability based on a theory of statistical learning is clearly an optimization problem, SVM is expected to give highly accurate prediction of phenomena. We investigate and evaluate SVM by applying it to the vibration response, which is dependent on various factors, and using it to diagnose the structural health of a vehicle. Less