基于K近邻算法的钢筋混凝土柱地震破坏模式判别方法

钢筋混凝土（RC）柱在地震力的作用下会发生不同的破坏模式，不同的破坏模式会有不同的损伤特点。所以，有必要针对不同地震破坏模式提出有效的判别方法。首先基于SMOTE算法使数据样本达到均衡，其次根据ENN算法，筛选了判别弯曲破坏和非弯曲破坏、弯剪破坏和剪切破坏的最佳参数；再次通过TomekLinks算法合理剔除噪音样本重构均衡数据，最后基于kNN算法建立了两阶段kNN模型，达到了准确判别RC柱地震破坏模式的目的，并通过与传统kNN模型、传统经验方法进行对比分析，验证了模型的优异性。研究结果表明：该方法通过选取筛选最佳参数，在提高判别准确率的同时简化了传统机器学习判别模型；本模型提出的两阶段kNN模型对三种破坏模式的判别准确率均可达90%以上，比传统kNN模型高10%左右，比传统经验判别方法高20%左右。

Seismic Failure Mode Identification Method of Reinforced Concrete Columns Based on kNN Algorithm

Reinforced concrete (RC) columns have different failure modes under seismic forces, and different failure modes have different damage characteristics. Therefore, it is necessary to put forward effective classification methods for different earthquake failure modes. First, make a balance of the data samples based on SMOTE. Secondly, select the best parameters for classsify flexure failure and non-flexure failure and flexur-shear failure and shearing failure based on ENN . Finally, a two-stage kNN model was established based on the kNN algorithm to accurately classsify the RC column earthquake failure mode. The model''s excellent performance was verified by comparing with the traditional kNN model and the traditional empirical method. The results show that this method can improve the classsify accuracy and simplify the traditional machine learning discriminant model by selecting the optimal parameters. The classsify accuracy of the two-stage kNN model proposed in this model can reach more than 90% for the three failure modes, which is about 10% higher than the traditional kNN model and about 20% higher than the traditional empirical discrimination method.