Research On Parameter Tuning-free Strategy For Permanent Magnet AC Servo Systems

Since 2000,domestic servo drive technology has developed rapidly,starting with no capacity and developing to maturity,and fostering the emergence of many excellent Chinese enterprises.From the perspective of technical level and market share,local brands have occupied a place in the low-end market,but there is still a certain gap between Chinese companies and the top foreign manufacturers in the performance and intelligence level of servo drives.The high-end servo market is still occupied by European,American,and Japanese products.Therefore,the research and development of high-performance permanent magnet synchronous motor(PMSM)alternating current(AC)servo systems is an essential part of achieving the goal of “Made in China 2025” and breaking the trade and technical barriers of high-precision machining machine tools from abroad.The automatic adjustment of servo control parameters,which can avoid tedious manual debugging and optimize the performance of these systems in different applications,is the most feature and most essential function of high-end intelligent servo system.Nowadays,conventional servo parameter self-tuning technologies still have some residual problems,such as strong trajectory dependence,poor robustness,and incomplete functionality.Thus,this paper proposes a series of modification methods for the traditional modular functions of servo drives,including mechanical resonance damping and parameter optimization,and makes a systematically in-depth study of the intelligent parameter tuning-free strategy.First,this paper analyzes the mechanism of the velocity measurement error of photoelectric encoders.Through the deduction strategy of fractional pulse accumulation,the noise frequency domain model was established,which can make up for the deficiency of the traditional time-domain model and can prove that the velocity measurement noise does not have the characteristics of Gaussian white noise in most situations.According to the frequency-domain features of the noise for classic M-method and MT-method,two improved strategies were proposed,which are independent of mechanical and electrical models and parameters.The proposed M-method adaptive finite impulse response filter method and improved single-phase MT-method speed measurement strategy with low-speed expansion can effectively suppress the quantization error of encoder and pulse quadruple noise.At the same time,the noise model of encoders was applied to the inertia identification algorithm,and the boundary value of the inertia identification error,which was caused by speed difference noise,was deduced.Then,the relationship between speed measurement period and identification accuracy was established.A variable-period speed differential strategy was proposed based on this criterion,which significantly reduced the identification error under low acceleration conditions.In addition,considering that the traditional integration method is only suitable for speed commands with fixed periods,a new type of speed zero-crossing criterion was proposed based on the characteristics of the position control mode.This variable-period inertia update structure broke the shackles of signal orthogonality,established an adaptive iterative updating system for the identification results,expanded the application potential of the algorithm,and improved th e integration-method-based inertia identification to a quasi-online level.In terms of mechanical resonance suppression,this paper analyzed the resonance deviation phenomenon and the failure mechanism of traditional online adaptive notch filters and established a more comprehensive nonlinear mechanical resonance model in the discrete domain.By analyzing the influence of current control delay,discrete control,and controller saturation nonlinearity,the conclusion that the resonance frequency is not equal to the oscillation frequency was obtained.Next,in order to improve the resonance suppression performance of the classical bi-quad filter,a series of improvement strategies were proposed in the aspects of parameter identification,discretization distortion correction,transient switching shock,etc.Based on the characteristic that the resonance and anti-resonance frequencies appear in pairs in the elastic transmission system,a twins-point search strategy was designed,which can accurately obtain complete resonance information and the parameters of bi-quad filters after one-turn search;Then,inspired by the discretization process of notch filters,a parameter mapping Tustin method for bi-quad filter discretization was proposed.Compared with the traditi onal pre-warped Tustin and zero-pole matching method,the expected characteristics of bi-quad filters can be reproduced more accurately in the discrete domain;Furthermore,by separating the transfer function,the equivalent time constant of the bi-quad filter was obtained.Based on this index,a bi-quad filter buffer strategy was designed.Owing to this warm-up process,the transient torque impact caused by filter switching under load conditions is effectively suppressed.Moreover,in order to expand the application range of the traditional rule-based auto-tuning methods and to avoid the dependence on specific trajectories such as square waves and triangle waves,this paper applied the adaptive infinite impulse response filter,which is used in the online identification of resonance frequency in the conventional sense,to system state judgment.By extracting the covariance coefficient of the filter,real-time judgment of system stability is achieved from the frequency domain.This new index is compatible with most of the working trajectories,which can greatly improve the adaptability of the rule-based tuning algorithm.Based on optimized inertia identification,mechanical resonance suppression,and the parameter searching rule,a trajectory planning module with controllable displacement,speed and acceleration was designed.Then,by integrating modular algorithms and planning signal flows,an intelligent parameter tuning-free strategy was proposed.Compared with the traditional self-tuning algorithm,the tuning-free strategy covers a broader range of parameter adjustments and can achieve a more compatible parameter auto-tuning without manual intervention.Furthermore,the tuning-free method was also applied to the position-current dual-loop control structure based on the position loop proportional-integral-lead(PI-lead)controller,and a better position follow-up performance was achieved.Since the direct current(DC)gain of the phase leading correction part is not a unit,the paper proposed improved dual-limitation and module inversion structures,which could completely solve the problem of failures of the conditional integral and tracking back calculation methods under the traditional positive sequence PI-lead architecture.Analysis and experimental results showed that the inversed structure can improve the Anti-Windup(AW)performance of the controller and greatly enhance the system stability.Additionally,the traditional root mean square error(RMSE)index is modified based on the high-order follow-up characteristics of the dual-loop position control structure.The incremental time-domain criterion of the fast root mean square error(FRMSE)was proposed,which can shorten the oscillation time of the system and reduce the sharp mechanical noise during the process of parameter search.Finally,the experimental comparison between the three-closed-loop and dual-closed-loop position control structures was given in this paper,and the theoretical innovation and engineering application value of all the algorithms proposed in this paper were verified.