Abstract: In order to overcome the interference of matrix effect and mineral effect on determination of 18 major and trace components in titanium ore samples, the samples were pressed into tablet at 1 500 kN for analysis of 10 trace components (phosphorus, gallium, chromium, vanadium, sulfur, rubidium, strontium, manganese, zinc and zirconium) by XRF, and the mixture flux (composed of lithium tetraborate and lithium metaborate at a mass ratio of 2:1) together with the sample at a mass ratio of 20:1 were mixed and melted to a fused bead at 1 050℃ for analysis of 8 major components (iron trioxide, titanium dioxide, silica oxide, aluminum trioxide, calcium oxide, magnesium oxide, potassium oxide and sodium oxide) by XRF. 24 calibration samples in a certain content gradient were prepared by using a mixture of six standard materials, thus to make the calibration curves. Spectral line overlap was corrected by empirical coefficient method. Matrix effect was corrected by theoretical α coefficient method and compton scattering internal standard method. The matrix absorption effect of iron was corrected by cobalt internal standard method. The results showed that after correction of comprehensive correction formula, the linear correlation coefficients of 18 components were all greater than 0.996 0. Three unknown samples were determined by this method, giving results consistent with alkali melting method. Detection limits (3s) of 10 trace and 8 major components were 1.2-21 mg·kg^-1 and 0.015%-0.66% respectively. 10 parallel samples of standard material were analyzed by the above method, giving results of RSDs (n=10) of trace components in the range of 0.21%-11% and of major components in the range of 0.22%-0.99%.