GP algorithm of correlation dimension computation is ameliorated which overcomes the shortage of traditional one. Improved process of GP algorithm takes the influence of temporal correlative pairs of points on correlation dimension into account and promotes the computational efficiency prominently. Iterative SVD method is applied to remove the influence of noise on the result of correlation dimension. The faults of steam flow turbulence and oil film disturbance which occur in 600MW Steam Turbine Generator are analyzed and whose correlation dimensions are computed. More distinct quantitative index than FFT is gained to distinguish two faults and its of little importance to apply correlation dimension to study the influence of various factors on steam flow turbulence fault for nonexistence of convergent floor in correlation integral curve, which presents a new way to learn the operational function of large capacity steam turbine generator and carry out comprehensive condition monitoring.