Aiming at the problem of performance dispersion of domestic turboshaft engines in batch production， a model-building method integrating manufacturing geometric parameters， component performance， and overall engine performance is proposed. Firstly， the manufacturing geometric parameters and factory performance parameters of important components from 200 new engines are obtained， and the key manufacturing geometric parameters of important engine components are screened by the Spearman correlation coefficient method. Secondly， based on the factory performance parameters of new engines and component-level models， the characteristic diagrams of important engine components are modified using component characteristic correction factors. Based on the particle swarm optimized support vector regression （PSO-SVR） method， the corresponding relationship between the key manufacturing geometric parameters of important engine components and component characteristic correction factors is established. Finally， a model of key manufacturing geometric parameters of important engine components-component performance-overall engine performance is established. The verification results show that the prediction error of the model for the overall engine performance is less than 5% at the 25% rated power， and less than 3% at 50%， 75%， 95%， 100%， and 105% rated powers. The model can simulate the influence of key manufacturing geometric parameters on the dispersion of engine performance and enable the prediction of overall engine performance before engine testing.