| When the servo system drives the load through the connecting shaft,once the frequency of the input signal reaches or exceeds the bandwidth of the servo system,mechanical resonance will occur.After analyzing the vector control structure of servo system,the dynamic equation of flexible load system is established,the fundamental cause of mechanical resonance is pointed out.The influence of stiffness coefficient,motor inertia and load inertia on resonance characteristics is analyzed.The analysis of the effect of mechanical resonance on the system performance shows the importance of the research on resonance suppression.After analyzing the defects of common notch filters,an off-line improved notch filter is proposed.According to the resonant model structure,the calculation methods of frequency,depth and width of notch center are deduced through the inherent parameters of servo system,such as inertia,damping coefficient and stiffness coefficient,and the parameters are set.The off-line tuned improved notch filter is verified by simulation.It can restrain mechanical resonance better.Considering that the off-line tuning method depends on the intrinsic parameters of the system,an adaptive notch filter based on BP neural network is proposed.The momentum term is added to the traditional BP neural network to prevent the algorithm from falling into the local minimum.The slow convergence speed of the algorithm is solved by using adaptive learning rate,the activation function of the algorithm is improved,and the threshold is added to the output layer to improve the stability of the system.Through simulation experiments,this method can quickly adjust the parameters of notch filter to achieve the optimal notch effect when the frequency changes.The hardware and software platform of the servo system is established to offline tune the notch filter parameters and suppress the resonance.The experimental results verify the effectiveness of the improved notch filter and its offline tuning method. |
PDF Full Text Request