Adaptive Gradient-Based Luenberger Observer Implemented for Electric Drive with Elastic Joint

摘要

In this paper work of the Luenberger observer applied for electric drive with complex mechanical part is analyzed. Comparing to classical solution, additional adaptation of gain matrix was introduced. Starting point for gradient-based on-line parameter recalculation is determined using metaheuristic algorithm - Grey Wolf Optimizer. Two state variables, the most often used in control structures applied for two-mass system, are estimated: load speed and shaft torque. Mentioned methods lead to precise calculations of signals and improvement of results after time constants changes. Moreover, initial phase related to adjustment of observer parameters is shortened. Model of the adaptive observer was firstly prepared, then simulations were realized. Final stage of described project, presents experimental verification, whole algorithm was implemented in processor od dSPACE 1103 board and experimental tests were done (using two DC motors).