합성곱 신경망을 이용한 다층 구조물의 층강성 예측

다국어 초록 (Multilingual Abstract)

The method for predicting the story stiffness of building structures using convolutional neural network is proposed, and it is verified using a five-story structure example. A random number generator is used to determine the stiffness value of each story, and a total of 1000 models are obtained by repeating this independently. Linear time history analysis is performed on the generated model to collect data for training and testing. The acceleration history response of the top is wavelet-transformed and used as an input image, and the stiffness values of each story used for the corresponding modeling are set as the output value. As a result of applying the example, it is found that the proposed method predicts the behavior and dynamic characteristics of structures similarly, although the degree of error is different for each variable. To reduce this error, a method of applying a genetic algorithm to the predicted value is presented, and the improvement effect of this is confirmed.

목차 (Table of Contents)

• Abstract
• 1. 서론
• 1.1 연구배경
• 1.2 연구범위 및 방법
• 2. 이론
• 2.1 합성곱 신경망
• 2.2 웨이블릿 변환
• 2.3 유전자 알고리즘
• 3. 합성곱 신경망을 이용한 층강성 예측 방법
• 4. 예제 검증
• 4.1 개요
• 4.2 결과 분석
• 5. 결론
• REFERENCES