Bolts are widely used in engineering structures and mechanical systems,as the weak part of the whole system,once the bolts are loose,it may cause disastrous consequences,so it is necessary to monitor the tightening state of bolts in time.As the internal microstructure of bolts is different in different tightening states,the acoustic signals generated by knocking bolts with different tightening forces are also different to some degree.Bolts with seven different tightening torques were struck.The spectral diagram of the acoustic signal at 0.26 s after percussion was compared with the spectral diagram of the first and the second order IMF components after empirical mode decomposition.In the experimental results,the rule that the main frequency value of the original knocking signal directly taking the FAST Fourier transform changes with the tightening torque is not obvious.After the original signal is processed by the empirical mode decomposition,the frequency values of the peak points in the spectrum graph of the first and the second mode functions of the knocking signal increase with the increase of the tightening torque.The study shows that the connection state of the bolt can be identified effectively by analyzing the peak frequency of the first and the second order IMF components through empirical mode decomposition.