Online Parameter Autotuning And Optimization For Servomechanism

Equivalent moment of inertia should be considered while tuning the speedregulator of a servomechanism. The proportional and integral gains decide thestability and dynamic response of the overall system. Online self-tuningtechniques for servo controller generally include inertia identification and PItuning, neither of which is expected to dominate the servo commands duringworking conditions. The thesis gives the study results on completely onlineparameter self-tuning based on a3.3kW permanent magnet synchronous motor(PMSM) servo platform.Firstly, the paper presents two different approaches to identify equivalentinertia online in the simulation. As some inertia-identification algorithmsincluding covariance matrices require long word length and which is beyond thescope of embedded processor, the accumulated truncation error will aggravatenegative impact of noise on the identification results. Thus, a new approach ininertia identification based on forgetting factor recursive square root (FFRSR)will be proposed as a third one. The experimental results show the proposedmethod do works. The estimation accuracy does not vary with different workingconditions. Besides the thesis also proposes some effective improvements so thatit can be adaptable to occasional change in load torque.Secondly, a set of PI tuning formulae is produced in Chapter4, whichcontains crossover frequency, phase margin and the equivalent moment of inertia.Many simulation results illustrate the feasibility of controlling phase margin andbandwidth. The paper then does research on the definite relationship betweenmaximum bandwidth limit and moment of inertia. In addition, the genericcalculation formula of maximum bandwidth limit is well obtained on the premiseof keeping phase margin of60°.Finally, this thesis gives the corresponding derivation process of PI tuningformulae for a normalized digital control system. Then the FFRSR inertiaidentification algorithm, the generic calculation formula of maximum bandwidthlimit and the derived PI tuning formulae for a normalized digital control systemare combined into a module that realizes the completely online parameterself-tuning free of human intervention. Adequate experimental results show thatthe tuning process is completely stable and well done. The dynamic response ofservomechanism is well improved.