Abstract: To address diesel engine intake performance optimization, an intake port was optimized using response surface methodology （RSM） and high-density sample set approximation. First of all, six key structural parameters were designed via orthogonal testing. Global sensitivity analysis was employed to identify dominant factors significantly influencing flow coefficient and swirl ratio. Subsequently, a small sample set was generated using Latin hypercube sampling （LHS）, and regression prediction models for flow coefficient and swirl ratio were constructed. It was found that the flow coefficient model achieved the prediction accuracy of 99.69%, while the swirl ratio model exhibited relatively lower accuracy, which is insufficient for stringent design requirements. To enhance the accuracy, a high-density sample set approximation method was introduced. A large sample set containing 100 000 samples was established through high-precision flow coefficient regression fitting, and the main parameter range corresponding to flow coefficients more than 0.89 was selected. Based on this, a high-density sample set with a design deviation of less than 10% was used to approximate the calculation of the swirl ratio. Ultimately, design schemes exhibiting excellent flow coefficient and swirl ratio performance were obtained.