MMM Identifying Damage for Recycled Pumping Rod Base on SVM

Remanufacturing is an important way to develop recycling economy. However, the quality reliability testing is a bottleneck for the recycled product. Metal Magnetic Memory (MMM) technology is a valid method for early damage testing based on the magneto-mechanical effect. This paper aims at quantitative identifying damage for the recycled pumping rod（RPR）. The MMM rules of RPR damage are found through fatigue experiments. By using four-dimensional MMM parameter vectors, that is, peek-peek value，differential gradient，integral gradient and energy, are marked the damages. Furthermore，a multi-level damage classification model has been established base on optimal multi-classification algorithm of support vector machine (SVM) for RPR. The relationship is given between the damage degrees and MMM signals, which provides a theoretical basis and engineering tool of RPR reliability evaluation.