With respect to several drawbacks such as easily causing the loss of experts' risk assessment information，overlooking the failure mode and effect analysis（FMEA） team member's bounded rational behaviors and the relative importance among risk factors in traditional FMEA method. a probabilistic linguistic FMEA risk evaluation method based on regret theory and Complex Proportional Assessment（COPRAS） method was proposed. Firstly，the probabilistic linguistic term sets（PLTS） was used to characterize the ambiguity and uncertainty of experts risk assessment information；Secondly，the risk factor weights were determined by using comprehensive weighting method based on the probabilistic linguistic multiplicative Analytic Hierarchy Process（PL-MAHP） and the maximum deviation method；Thirdly，taking experts′ psychological behaviors into account，regret theory and COPRAS were adopted to prioritize failure modes. Finally，an empirical case concerning the risk evaluation of the grinding wheel frame was presented to demonstrate the practicality and effectiveness of the proposed method. Results derived from comparative and sensitivity analyses fully show that the risk priority ranking order determined by the proposed FMEA method is shown to be more rational and reliable by comparing with other existing methods.