Comparison of DSP and FPGA realization of neural speed estimator for 2-mass system

In this paper two hardware applications of neural state estimator for the two-mass drive system are presented. One of them is realized in a Digital Signal Processor (DSP), and the second consists in parallel implementation of neural networks in a reconfigurable Field-Programmable Gate Array (FPGA) placed in CompactRIO controller. Described solutions of neural network implementations are used for the estimation of the load-side speed of an electrical drive with elastic joint. Several design and programming problems are presented. Developed neural estimators are tested in the laboratory drive. Application of the neural network with FPGA gives better results, especially on drive system transients. Obtained results show high accuracy of the load speed estimation with the presented neural estimator. High precision of calculation is also obtained in the presence of the load time constant changes.