基于AdaBoost集成学习的强震动观测抗干扰技术研究

Research on Anti-jamming Technology of Strong Seismograph Based on Machine Learning

为提高强震仪的抗干扰能力,基于分类、决策的机器学习中的AdaBoost集成学习方法,设计一种强震动数据抗干扰算法,以解决基于决策树的强震动数据抗干扰算法存在的易过拟合、分类准确度不高等问题。从天然地震动与人工干扰下的强震动数据中提取出若干个特征(波形对称度、卓越频率、最大增长速度等),形成一一对应的训练样本特征集与事件属性集;初始化权重分布,持续利用AdaBoost技术更新样本权重分布,以增加较难分辨样本的权重值,然后将若干个弱分类器训练为一个强分类器,达到提高强震仪抗干扰准确度的目的。此方法分类准确度较高,具有较强的环境适应性,对于推动强震观测仪器智能化实现、促进土木工程结构防震减灾技术发展具有一定现实意义。

In order to improve the anti-jamming ability of the seismograph under the excitation of external disturbance environment,an anti-jamming algorithm for strong vibration data is designed based on AdaBoost to improve the over-fitting and low classification accuracy of decision tree method.Several features(waveform symmetry,predominant frequency,maximum growth rate,etc.)are extracted from the strong motion data,and gained a corresponding training sample feature set and event attribute set;the weight distribution is initialized,and the weight distribution is continuously updated by using AdaBoost technology to increase the weight of difficult samples.The weights of samples are distinguished,and then several weak classifiers are trained as a strong classifier to improve the anti-jamming accuracy of strong seismograph.This algorithm has high classification accuracy and strong environmental adaptability.It is of practical significance to promote the intelligent realization of strong-motion earthquake observation instruments and the development of earthquake prevention and disaster reduction technology for civil engineering structures.