研究発表

免震床の制御においては，加速度だけでなく，相対変位を低減させることも重要である．著者らの研究グループは，建物と設備の協調制御について検討を行っているが，建物に設置された加速度センサーの測定値を共有するためには通信機器が必要となりコストがかかる問題がある．そこで，建物の状態量を共有しないローカルな測定環境で免震床上の設備の応答を低減することを目的とし，局所センサーにより測定した設備機器の状態量を用いて，免震床の相対速度および相対変位から計算されるエネルギー総和を評価関数とした多項式分散制御の有効性について数値解析と震動台実験により検討を行った．その結果，多項式分散制御は，LQG理論を用いた線形分散制御よりも免震床の絶対加速度および相対変位を低減することができ，建物内の全センサーを用いたLQG理論による線形集中制御と同等程度の応答低減性能を有する可能性があることが明らかとなった．

地震動をより適切に評価するため卓越方向を考慮した水平二方向を決定し，統計的グリーン関数法による模擬地震動の作成法を提案した．2011年東北地方太平洋沖地震の余震記録を用い，加速度の二乗平均値を計算し卓越方向を探索し，差の小さい断層走向方向と断層直交方向の組み合わせを使用する二方向として採用した．またBooreの包絡関数のパラメータである地震動継続時間と振幅のピークまでの時間について，包・川瀬(2007)と同様の方法に従い地震波の区間最大値に包絡線をフィッティングさせ最適な値を得て，二段階回帰の方法により震源距離とマグニチュードで表した回帰式を得た．

To employ neural networks in structural health monitoring, a large volume of training data is required. The data can be obtained by a generative learning method based on high-precision simulation of seismic responses. We investigate appropriate machine learning parameters such as the numbers of layer nodes in a deep neural network and the number of epochs for training. Response records measured in shaking table tests of a steel frame structure are used as training, validation, and test data. Our results show that response data of only the translational component can achieve high accuracy of damage pattern classification. Regarding the length of response data, clear difference is not observed in two durations of 1 s and 2 s. We propose a method to detect unlearned damage pattern data using the Mahalanobis distance.

When an earthquake occurs, the people in and/or contents of buildings may suffer damage due to the sliding or overturning of furniture and equipment or the falling of objects even if the structure of the building experiences no damage. In this study, using numerical analyses, we investigate a model of effective equipment caster wheel control to reduce damage caused by the moving and overturning of equipment during earthquakes and propose a control method to switch the lock and free conditions of the caster wheel. In the numerical analyses, a simulator library of three-dimensional rigid body dynamics is used and two-direction earthquake ground motion is input into a model of a 10-story steel frame building. The horizontal displacement and the falling down of the equipment model are expressed depending on the caster lock condition.

A reliable index is proposed to assess damage severity of a reinforced concrete (RC) beam based on an experiment. The index is fractal dimension of time-history of the fractal dimension of response waveform to impact excitation, which is calculated by the box counting method. The proposed index can evaluate damage level that requires repair with good accuracy. The finite element dynamic analysis shows the proposed index has a potential to evaluate damage level of an RC beam with a different boundary condition from the experiment.