Object-oriented DSP implementation of neural state estimator for electrical drive with elastic coupling

The study presents results and procedure of object-oriented and test-driven implementation of neural-network-based state estimator. The presented algorithm has been developed for estimation of the state variables of the mechanical part of electric drive with elastic coupling. Estimated state variables – load speed and shaft stiffness torque – can be used in speed control process for reducing mechanical vibrations of working machine. The basic objective was to create a simple, extensible and readable program code, performing the task of state estimation of the considered system. The target platform is a DSP (Digital Signal Processor) from SHARC (Super Harvard architecture Single-Chip Computer) family, which allows for hardware acceleration of matrix operations. The IDE (Integrated Development Environment) available for the selected platform made it possible to write program in C++. The usage of UML (Unified Modelling Language) in the development of control software was discussed.