CAJ | 학술논문

将混凝土坝的应力监测数据视为一组含有不同频率信号的时间序列，针对信号中的低频信号和高频信号，采用提升小波的多尺度分解法提取混凝土坝应力的时效分量、水压和温度分量以及噪声分量，采用计算信噪比极大值的方法提取周期性分量中的水压、温度分量和噪声分量，得到最优的去噪效果。使用RBF神经网络对去噪信号进行建模，预测结果表明，该模型能够很好地反映混凝土坝应力变化的趋势和规律，可应用于混凝土坝安全监测中。
장혼응토패적응력감측수거시위일조함유불동빈솔신호적시간서렬，침대신호중적저빈신호화고빈신호，채용제승소파적다척도분해법제취혼응토패응력적시효분량、수압화온도분량이급조성분량，채용계산신조비겁대치적방법제취주기성분량중적수압、온도분량화조성분량，득도최우적거조효과。사용RBF신경망락대거조신호진행건모，예측결과표명，해모형능구흔호지반영혼응토패응력변화적추세화규률，가응용우혼응토패안전감측중。
The observation data of concrete dam stress is regarded as a set of time series which contains different frequency signals. Aiming at the signal with high frequency or low frequency, a new method of lifting wavelet multi-scale decomposition is proposed to extract the time dependent component, the water pressure and temperature component as well as the noise component from the observation date of dam stress, and then the method of calculating maximum SNR is used to extract the components of water pressure and temperature and noise component from cyclical component, which can get the most optimal denoising effect. The RBF neural network is finally used to model the denoising signals for stress prediction. The prediction results show that the model can well reflect the trends and patterns of stress changes of concrete dam.