Font Size: a A A

Research On Detection Methods Of Looseness Properties Of Bolted Connections In Timber Structures Based On Deep Learning

Posted on:2023-05-08Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y B YuFull Text:PDF
GTID:1522306824991209Subject:Mechanical engineering
Abstract/Summary:
Bolted connections have the advantages of structural simplicity and disassembly convenience,which is widely used in heavy-duty timber engineering.Bolt loosening caused by the decrease of preload under cyclic loading is the key factor that results in the failure of the connection.Bolt preload and loosening rotation angle are important indexes for quantitative evaluation of bolted connections loosening performance.Therefore,instant fastening can be done by testing preload decrease and loose rotation angle change at early stage of bolted connections loosening to prevent the failure of timber structure connection,thus effectively preventing the safety problems caused by the overall loosening of timber engineering.This paper focuses on the test methods of bolted connections loosening performance in initial stage.The main work is as follows:Aiming at the theoretical calculation of preload decrease and rotation angle at the initial stage of bolted connections loosening,the loosening failure law of bolted connections under lateral load was studied based on Junker theory and experimental models.It is clear that preload and loosening rotation angle are important indexes to quantify the loosening performance of bolted connections.The influences of bolt-to-component and bolt-to-nut thread pair on bolted connections loosening were studied.Combined with the calculation methods of preload,the friction torque of bolt threads,the theoretical relationship between the friction torque of bolt head and end faces,and the relationship between loosening rotation angle and preload,the mathematical model of preload recession of bolted connection was deduced.The relation between preloads and rotation angles were theoretically calculated with bolts’ specifications and material parameters,which verified the feasibility of the looseness theory in timber structure.This research laid foundation for the subsequent test of preloads and looseness rotation angles.Aiming at the analysis method of bolt loosening performance of timber engineering under vibration load,a nondestructive testing way of wood mechanical properties combined with three methods was proposed.The elastic constants of wood were obtained by experiments,and the material modeling problem of bolted connection structure of timber engineering was solved.The finite element models of bolt and thread pair with thread angle were established.Then,the looseness calculation method under periodic transverse vibration was established by applying preload to bolt combined with contact analysis.The relationship between the downward trend of preload and the loosening rotation angle was calculated.According to the loosening analysis of four types of connection,the influencing factors of bolt loosening were analyzed.Through vibration test,the ultrasonic method was used to process the preload signal for verifying the loosening analysis.Aiming at the preload test of bolted connection in timber engineering,a preload test method was designed with deep learning technology.Combined with time reversal method,a ultrasonic Lamb wave testing system for bolted connection preload based on deep learning was developed.Three deep learning algorithms(LSTM,Wide Res Net50_2 and Dense Net121)and a time series signal classification algorithm SAX-VSM were used to process the test signals with different preloads,and then they were identified according to different classifications.The identification accuracy of the four algorithms under different preloads were obtained,and the optimal algorithm was found to improve the accuracy.Online preload detection of bolted connections in timber engineering were realized.Aiming at the problem of looseness rotation angle detection of bolted connections on timber engineering,the offline detection calculation method of looseness rotation angle was proposed.The offline detection model of looseness rotation angle of bolts was established with deep learning algorithm YOLO v3.Furthermore,according to the real-time coordinate change of spatial displacement of bolt in actual working conditions,the online detection model of looseness rotation angle of bolt was proposed with SSD algorithm.The looseness rotation angle of bolts was trained and tested,and the feasibility were verified.This method accurately measured the rotation angle of bolts’ looseness in timber engineering,and realized the online detection of bolted loosening properties,which improved the integrity of this study.The analysis and testing methods of bolt preload and rotation angle was studied in this paper.The two indexes were also quantitatively evaluated which provided theoretical foundation and application to the research of looseness properties of bolted connections on timber structure.This study can effectively prevent the safety problems of timber structure during employing.
Keywords/Search Tags:Deep learning, Bolted connection on timber structure, Looseness detection, Ultrasonic technology, Machine vision
Related items