The content of silicon and iron in primary aluminum were detected by self-built LIBS device. The micro-morphology analysis of the primary aluminum sample was carried out before the experiment, It was found the distribution of silicon element is relatively uniform in the primary aluminum except little silicon has agglomerated in partial areas, the iron elements mostly appeared in agglomerated form, and there was no obvious distribution rule. The effects of laser energy on plasma spectrum characteristics were investigated in the paper. It was found that the signal-to-noise ratio of silicon and iron analytical lines increased firstly and then decreased with the increasing of laser energy, when the laser energy reached 160mJ, the signal-to-noise ratio was maximized, so the laser energy 160 mJ is the more reasonable experimental condition. The calibration model was established based on the CC method using two standard samples (pure aluminum standard samples and self-selected standard samples) under the above reasonable experimental conditions. The results showed that the calibration curve established by the self-selected sample was not ideal compared with the calibration curve established by standard samples, and there were large errors in the results, The fit goodness of the iron element calibration model is only 0.821 3, and the relative standard deviation is also large. When the pure aluminum standard samples were used, under the condition of fixed sample, the fit goodness of the calibration curves for silicon and iron elements was 0.961 1 and 0.974 1, respectively, and the relative standard deviations were 8.85% and 9.43%, respectively. Errors expressed by error bars increased with the increasing of silicon and iron contents in the pure aluminum standard samples. Under the condition of rotating sample pool, the fit goodness of the calibration curves for silicon and iron was 0.978 5 and 0.988, respectively, and the relative standard deviations were 3.78% and 3.4%, respectively. The calibration results showed that the fit goodness was significantly improved and the relative standard deviation was also reduced compared with fixed sample pool condition. The calibration model was significantly better than the model established by the self-selected samples. The 25 samples were detected using two different calibration models by LIBS, the relative errors for the results obtained from different models were compared, and the content of pure aluminum samples has a larger concentration gradient and a wider distribution, so the models obtained from pure aluminum samples have relatively poor adaptability to low-iron primary aluminum samples, while the calibration model established by the self-selected samples is not ideal, but the measurement adaptability for low-iron primary aluminum samples is relatively good. The plasma generated by laser-induced primary aluminum was diagnosed. The plasma temperature was calculated to be approximately 34 100.14 K from the Boltzmann diagram of several magnesium ion lines, The plasma electron density was estimated to be 1.69×1017 cm-3 by the Stark broadening of a line of magnesium, which confirms that the assumption that the plasma obtained from laser induced raw aluminium is in a local thermodynamic equilibrium state is valid.