Prediction of tension state of bolt based on LPSO-GRNN
author:LIANG Wei, CHEN Zhixiong, OU YANG Zhongjie, et al.
Check PDF documents
Prediction of tension state of bolt based on LPSOGRNN
LIANG Wei1, 2,3, CHEN Zhixiong1, 4, OU YANG Zhongjie1, GONG Shengwei5,
ZHONG Jianhua1, 4, ZHONG Shuncong1, 4, LIAO Huazhong3
(1.College of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, China; 2.Fujian Provincial Key Laboratory
of Force Measurement and Testing (Fujian Institute of Metrology), Fuzhou 350100, China; 3.Xiamen Industrial Technology Research
Institute, Xiamen 361001, China; 4.Fujian Provincial Key Laboratory of Terahertz Functional Devices and Intelligent Sensing, Fuzhou
350108, China; 5.School of Advanced Technology, Xi'an Jiaotong-Liverpool University, Suzhou 215123, China)
Abstract: Aiming at solving the matter that the internal bolts of the axleload dynamic truck scale were loose and falling off due to the frequent loading and unloading cycle impact of heavy goods vehicles under the service state, a bolt tightening state prediction model based on the Lévy flight improved particle swarm algorithm optimization generalized regression neural network model (LPSO-GRNN model) combined with vibration signal feature extraction was proposed. And combining with the feature extraction of vibration signal, the model was applied to the state prediction of truck scale bolts. Firstly, the waveform index, peak index, pulse index, steepness index and other signal characteristics of the output vibration signal of the bolt under different tightening states were extracted, and they were jointly used as the input feature vector of the model. Then, Lévy flight was used to improve the optimization ability of the particle swarm optimization algorithm, and the global optimization ability of the algorithm was improved by generating random step sizes to avoid falling into jumping out of the local optimal value, and the smooth factor of the generalized regression neural network was optimized by the improved algorithm to obtain the global optimal smooth factor. Finally, by designing experiments, GRNN, PSO-GRNN and LPSO-GRNN were used to predict the bolt tightening state and compared it with the actual situation. The experimental comparison results show that the bolt tightening state prediction model established based on LPSO-GRNN has an accuracy of up to 95%, effectively improving the accuracy of bolt tightening prediction in weighing systems. This model addresses the problem of particle swarm optimization algorithm easily getting trapped in local optimal convergence. In conclusion, the proposed model provides an effective solution to the problem of bolt loosening and falling in dynamic axle weighing systems.
Key words: axleload dynamic truck scale; LPSOGRNN prediction model; bolt tension; vibration signal feature extraction; generalized regression neural network (GRNN); particle swarm optimization (PSO); Lévy flight