Evaluation On Remaining Life Of Retired Sucker Rod Based On Magnetic Memory Technique

At present, the sucker rod pumping take a great proportion of many oil production methods, and the oil well will be forced to shut down once a sucker rod is broken off. Therefore, sucker rods were replaced after the deadline, but most of them still could be repaired except for a few severe distorted sucker rods. Today, reproducibility of most sucker rods were detected by visual inspection because of the lack of detection means, which lead to low repair rate and problems that many blanks of sucker rods which could be repaired by the remanufacturing. So, remanufacturing blanks of sucker rods need to be screened. The remaining life is an important indicator which used to estimate the remanufacturing value. Therefore, a method used to evaluate the remaining life of retired sucker robs, had an important significance to guide the screening of the remanufacturing blanks, and had an important practical engineering value will be presented in this paper.In this paper, the finite element software Comsol was adopted to simulate the theory of magnetic memory testing. Then, experiments were made to research on the variation regularities of magnetic memory signals. Next, a model of support vector machine was used to evaluate the remaining life of retired sucker robs based on the characteristic values extracted in experimental data. At last, the model of support vector machine was optimized to improve the accuracy of the evaluation. To be specific, the main contents are explored as follows.Based on the model of force-magnetic coupling, the finite element software Comsol was adopted to simulate the detecting mechanism of magnetic memory testing. Through analysis of magneto-statics and magneto-mechanical, leakage field distribute under the effect of geomagnetic field and under the common effect of force and geomagnetic field was simulated respectively. The mechanical-magneto effect was proved through comparison.Through tension-tension fatigue experiments of notched specimens of sucker rod, and magnetic memory detector was used to test magnetic memory signals on specimens surface, variation law of magnetic memory signals with different number of cycles was researched. The characteristic values of tangential amplitude, normal peak-peak value and maximum value of composite magnetic field gradient were extracted. The results show that there is a corresponding relation between the characteristic values and fatigue remaining life.On the basis, characteristic values indicating fatigue damage were extracted further: values of tangential amplitude, normal peak-peak value and maximum value of composite magnetic field gradient, tangential average wave energy, normal average wave energy, coefficient of tangential gradient limit state and variation rate of normal regional signal strength. A model of support vector machine was established, the characteristic values were imputed and the uniform percent service life of sucker rods were outputted in the model. The results show that there is a ideal distribution trend between predicted values and measured values. Besides, the influence of parameters in support vector machine to prediction accuracy was discussed.On the purpose of improving the prediction accuracy, the technique of genetic algorithm, partical swarm optimization and grid search technique were applied to optimize the core parameters in support vector machine and the penalty factor. The prediction accuracy was improved significantly after optimization.