Forthe purpose of identifying situation of winding clamping force, this paperpresents a method of transformer winding pressed state detection based on Cross Recurrence Plot (CRP) and Recurrence Quantification Analysis (RQA) by analyzing the change of transformer surface vibration. Firstly, it makes use of phase space reconstruction to reconstruct phase curves of multivariate and univariate vibration signals based on recurrence characteristics. And then, CRP and RQA are used to analyze phase curves qualitatively and quantitatively. Consequently, winding pressed state can be detected successfully. The analysis results of vibration signals show that the change of diagonal structure of CRP can qualitatively reflect the change of winding pressed state. Meanwhile, the RQA metrics of multivariate and univariate vibration signals can detect winding pressed state quantitatively from the overall and local perspective. The results provide a theoretical basis for monitoring winding looseness fault from the viewpoint of nonlinear dynamics. This work is supported by National Natural Science Foundation of China (No. 51577050).