Lamb波高斯混合模型螺栓松动损伤检测

Bolt Looseness Damage Detection using Lamb Wave Gaussian Mixture Model

螺栓连接广泛应用于多种领域,及时发现螺栓松动的位置是结构健康监测的重要课题之一。利用粘贴在铝板上的压电阵列采集Lamb波信号,提取特征参数集建立高斯混合模型。通过采集监测区域内螺栓连接结构的各种松动工况的数据建立完备的基准数据库,更新实时数据建立动态高斯混合模型,基于高斯混合模型之间概率密度分布之间的相似度最大准则,判断监测区域的各个螺栓松动情况。实验结果表明,螺栓松紧状态一致的测试样本与训练样本之间的高斯混合模型概率分布相似度值达到0.99以上,明显高于工况不匹配的相似度,该方法可有效判断监测区域每个螺栓的松紧状态。

Bolted⁃joints are widely used in many fields.The discovery of the position of bolt looseness in time is one of the most important topics of structural health monitoring.The Gaussian mixture model is established by using the feature parameter sets,which are extracted from the Lamb wave signal collected by the piezoelectric array attached to an aluminum plate.The complete reference database is established by collecting the data of various looseness working conditions in the bolted⁃joint structure of the monitoring area.The real⁃time data is updated to establish the dynamic Gaussian mixture model,and the looseness of each bolt in the monitoring area is judged based on the maximum similarity criterion between the probability density distributions of the Gaussian mixture model.The experimental results show that the probability distribution similarity of the Gaussian mixture model between the test sample and the training sample in a consistent bolt tightness state is above 0.99,which is obviously higher than the probability distribution similarity in the inconsistent working condition.This method can effectively judge the state of each bolt in the monitoring area.