Optimization of detection conditions is defined as maximizing the amount of heat generated in crack area, in order to perform better in the Eddy Current Pulsed Thermography(ECPT). Aiming at standardizing the method of optimization in ECPT, and a single metal plate specimen with a specific crack was taken as the investigated subject. Response signal increased with the excitation time and excitation intensity, and it had a tendency to enhance first and then weaken with the increase of lift-off distance analyzed by results of simulation and experiment. A multivariate nonlinear regression model was proposed to estimate response signal under specific detection conditions, and the quantitative relation between response signal and different detection conditions was determined. Finally, the Particle Swarm Optimization(PSO) algorithm was introduced to optimize the detection conditions, and the distribution of response signal and Probability of Detection(POD) with different detection conditions were drawn. The research results provide theoretical guidance for optimization of detection conditions in ECPT.