Magnetic memory PE full-encapsulation sucker rod SVM stress prediction research
In order to detect the stress region of PE fully clad anti-corrosion layer pumping rods and solve the technical difficulties in the condition assessment of pumping rods.The magnetic memory signals of prefabricated point corrosion,face corrosion and groove corrosion of PE fully clad anticorrosion layer sucker rods were collected,and the relationship between the signal eigenvalues and the defect stress values,as well as the stress state was investigated by using support vector machines(SVMs)to establish a stress state prediction model,and a particle swarm optimisation(PSO)algorithm was introduced to optimise the parameters of the SVM model.Through the optimisation,the average error of predicting the stress state of the sucker rod using this model is only 3.638% .The results show that the support vector machine model established based on particle swarm optimization algorithm and seven kinds of magnetic memory signal feature quantities has high accuracy and is suitable for predicting the stress state of pumping rods with PE fully clad anti-corrosion layer.
PE fully coated anti-corrosion rodsupport vector machinesmetallic magnetic memorysignal characteristic quantityparticle swarm optimization