On-line Detection Technology And Residual Life Prediction Method Of Bolt Components Based On Metal Magnetic Memory

Bolt connection is widely used in aerospace,nuclear energy,civil engineering and other fields.Under the action of stress,temperature and other factors,the bolt connection is easy to fail.The common failure modes are bolt looseness,uneven bolt forces and fatigue failure of bolts,which can cause great loss of life and property.Therefore,experimental studies were conducted to explore the three types of common failure modes detected by metal magnetic memory(MMM)technology to prevent accidents caused by bolt failure.The bolt tensile tests were carried out to detect the bolt MMM signal under different stress.Twelve magnetic memory parameters,which characterize the stress,were extracted.The linear equations of the characteristic parameters and the bolt axial stress were obtained.The influence of bolt material,diameter and length on the linear equation is analyzed.At the same time,the torque load tests were carried out to analyze the influence of the flange and other factors on the bolt signal.Based on the 12 characteristic parameters that characterize the stress and analysis of the factors that influence the bolt signal,the bolt force calculation model is established,and the bolt force can be calculated according to this model to evaluate whether the bolt is loose.Bolt group tests were carried out to detect the bolt group signal.Based on the 12 characteristic parameters characterizing the stress,the distance between bolts in the bolt group and the center moment of bolts were defined.Based on the distance and the center moment,three methods to evaluate the force uniformity of the bolt group are proposed,which are respectively the force uniformity assessment of the bolt group based on the distance ratio,the stress uniformity assessment model of the bolt group based on the distance and the stress uniformity assessment model of the bolt group based on the center moment.These three methods can only assess the force uniformity of the bolt group itself,can not assess the force uniformity of other bolt groups of the same material and the force uniformity of bolt groups of other materials.The bolt fatigue tests were carried out to detect the bolt signal after different loading cycles and to analyze the change rule of MMM signal during the fatigue process.Based on 12 feature parameters,36 feature parameters were extracted.With 48 characteristic parameters as input and the residual fatigue life of the bolt as the output,a support vector machine(SVM)model was established to predict the residual fatigue life of the bolt.When the genetic algorithm is used to optimize the parameters,the mean square error of the model is the minimum and the prediction effect is the best.When the stress amplitude invariant stress ratio changes or the stress ratio invariant stress amplitude changes,the model prediction effect is still acceptable.