Abstract: To investigate the influence of random variables, such as structural dimensions, material property parameters, and temperatures, on the equivalent stress of steam turbine rotor during material preparation, processing, installation and operation, taking the high pressure rotor in a 300 MW class steam turbine as the research object, the time and location of the maximum equivalent stress during cold start-up was obtained by calculating with finite element method. Meanwhile, combined with the randomness of variables, the relationship between the rotor maximum equivalent stress and the random variables was fitted with both of polynomial response surface method and support vector regression (SVR) response surface method, and analyses and comparisons were conducted on the fitting effects of above two methods. Additionally, the reliability of rotor equivalent stress and the sensitivity of various random variables during cold start-up process were investigated by Monte Carlo simulation method, while the sensitivity index order of six random variables including rotor radius, was determined. Results reveal that the fitting effect of SVR response surface method is better than that of polynomial response surface method, and the strength reliability indexes of above two response surface methods for the cold start-up curves are 96.99% and 94.16%, respectively. Among the six random variables, the elastic modulus of rotor material is identified as having the most significant impact on the equivalent stress of steam turbine rotor.