Structural dynamic properties such as natural frequency depend not only on damage but also environmental condition (i.e., Temperature). Without removing the variation of environmental condition in the damage detection, false-positive or negative damage diagnosis may occur so that structural health monitoring becomes unreliable. One of methods used to solve this problem is to construct regression model based on structural responses with the environmental factors. However, it is difficult to determine where and which environmental variables to measure. One alternative is to remove the variability due to environmental variation without measurement of environmental variables. However, the performance of this method is depending on how to define the reference data set. Generally, there is no prior information on reference condition (i.e., healthy condition) during data mining. Reference condition is determined based on subjective perspective with human-intervention. To overcome the drawback of current methods, this paper investigates adaptive PCA technique for the monitoring of structural damage detection under environmental change. This method is not required to determine the reference condition and measure the environmental variables. Proposed method is tested on numerically simulated data for a range of noise in measurement under environmental variation.

The selection of appropriate ground motions and reasonable modification are becoming increasingly critical in reliable prediction on seismic performance of structures. A widely used amplitude scaling approach is not sufficient for robust structural evaluation considering a site specific seismic hazard because only one spectral value is matched to the design spectrum typically at the structural fundamental period. Hence alternative approaches for ground motion selection and modifications have been suggested. However, there is no means to evaluate such methodologies yet. In this study, it is focused to describe the main questions resided in the amplitude scaling approach and to propose a regression model for structural damage as point of comparison. Spectrum compatible approach whose resulting spectrum matches the design spectrum at the entire range of the structural period is considered as alternative to be compared to the amplitude scaling approach. The design spectrum is generated according to ASCE7-05.

This report analyzed the types of damage occurring in small scale safety vulnerable buildings inspected in 2012, by both structural types and elapsed time. The results of the analysis show that more than 70% of damage on finishing or usability by non-structural elements is recoverable for its intended performance. Facilities over 20 years from compleetion have more than a 30% structure damage ratio, than can be dangerous for structural safety, and requires periodic check-ups and repair/reinforcement work to recover its performance.

한 구조물이 손상을 입으면 그 구조물의 동적응답(고유진동수, 가속도, 변형률)이 변하게 된다. 이와 같이 변하는 동적응답을 응답신호로 계측하고 이들 데이터를 신경망에 적용하여 구조물의 손상을 평가하는 방법이 신경망손상평가법이다. 현재까지 정형화된 특정한 경우의 연구가 주로 이루어져 있지만 일반적인 신경망손상평가법의 특성에 관한 연구나 실용 가능성과 장단점에 관한 충분한 연구가 부족하다. 따라서 본 연구는 신경망에 다양한 동적응답을 적용하는데 있어 신경망손상평가법의 일반적인 특성과 적용의 문제점을 연구하였다. 신경망손상평가법은 일정한 가진력을 손상이 있는 구조물에 가하고 그로부터 얻은 응답신호를 이용하여 신경망을 학습을 시킨 후, 임의의 손상이 있는 구조물에 동일한 가진력을 가하여 얻은 응답신호를 이용하여 손상의 위치와 정도를 찾는 것이 현재까지의 연구였다. 그러나 일반적으로 구조물에 작용하는 가진력은 일정하지 않다. 따라서 동일한 가진력에 의해 학습된 신경망에 가진력의 변화가 있는 경우에도 손상을 파악하는지 평가하였다. 모든 응답신호는 모형실험을 통하여 획득하였다.