In order to analyze the tensile stage of the sample by magnetic memory detection technology, a variety of information entropy is introduced to process and analyze the magnetic memory signal. The information entropy of the magnetic signal under different tensile loads was calculated, and used as a criterion to judge the state of the tensile sample. An information fusion strategy coupled with Dempster synthesis rule was proposed to improve the accuracy and reliability of the state recognition. Based on the Dempster synthesis rule and D-S evidence theory, a state identification algorithm was developed by using Matlab software, which can easily obtain the deformation state of different regions on the sample surface. The experimental results show that the data analysis method can quickly and accurately judge the tensile stage of the sample.