LIBS (laser-induced breakdown spectroscopy) was used to detect Fe element content in soybean oil quantitatively. In this experiment, a series of soybean oil samples with different concentrations of Fe were used; LIBS spectra were collected with a two-channel high precision spectrometer. According to the LIBS spectrum of samples, two characteristic wavelength of Fe (404.58 and 406.36 nm) were determined, and different simple regression methods (exponential regression, linear regression and quadratic regression) were used to establish the quantitative analysis models of Fe content using each characteristic spectral line. The results indicate that the average relative error of Fe Ⅰ 404.58 and Fe Ⅰ 406.36 in simple exponential regression, linear regression and quadratic regression models were 29.49%, 8.93%, 8.70% and 28.95%, 8.63%, 8.44%, respectively. The results of Fe Ⅰ 406.36 regression models is better than that of Fe Ⅰ 404.58, and the quadratic regression model is optimal among the three regression models. According to these results, LIBS technology has certain feasibility for detecting Fe in soybean oil; the quadratic linear regression model can improve the prediction accuracy of Fe element effectively.