机理与数据融合的螺栓连接松脱预测

Prediction of bolt connection loosening based on mechanism and data fusion

针对螺栓松脱状态下夹紧力变化难以精确预测的问题,提出一种机理引导数据的螺栓连接松脱特性预测方法。首先结合螺栓受载时的力学状态建立松脱过程机理模型,通过参数试验法对机理模型中各特征进行敏感度分析,提出松脱敏感度评价指标以获取松脱过程的关键特征;进一步考虑螺栓松脱的非线性特征及不确定性规律,提出一种基于高斯过程回归的螺栓松脱特性预测模型,并对该模型进行了验证。结果表明:与传统回归模型相比,该模型不但可获取预紧力平均值的变化情况,而且可同步描述概率意义上的预紧力变化置信区间,为螺栓松脱特性准确预测提供了保证;螺栓松脱试验及预测数据具有良好的一致性,证明了该模型的合理性。

It is difficult to accurately predict the change of clamping force in bolt loosening state.Aiming at the this problem,based on the data-driven method guided by bolt loosening mechanism,a prediction method for bolt loosening characteristics was proposed.The mechanism model of the loosening process was established in combination with the mechanical state of bolt.The sensitivity analysis of each feature in the mechanism model was carried out through the parameter test method,and the evaluation index was proposed to obtain the crucial features of the loosening process.Furthermore,considering the nonlinear and uncertain characteristics of bolt loosening,a prediction model of bolt loosening characteristics based on Gaussian Process Regression(GPR)was proposed and verified.The results showed that compared with the traditional regression model,the proposed model could not only obtain the change of the mean value of preload but also describe the confidence interval of preload change in the sense of probability synchronously,which provided a guarantee for the accurate prediction of bolt loosening characteristics;the model proved to be reasonable by the excellent consistency of bolt loosening test data and prediction data.