<p>Technological development in today’s industry leads to much faster cyclic processes and reduction in material usage. Modern industry implies that the mechanical energy required for performing mechanical work in the technological process is provided from a controlled multi motor drive with different types of mechanical coupling of individual drives. With rigid coupling, a continuously increasing production speed might cause the inclusion of torsional characteristics of material within frequency range of the controller. Such a system can be modeled as two mass resonant mechanical system, with the motor presenting the first or driving mass, and the tool presenting the load mass. This paper studies model predictive control algorithm (MPC) as a proven approach for torsional oscillation suppression. Mechanical parameters estimations are established using neural network (NN) to assist MPC algorithm, while Luenberger state observer estimates torsional torque and load speed for required speed control. MPC algorithm and NN are implemented in software in the loop (SIL) experiment to demonstrate a proof of concept for this type of control. MPC is executed as a highly optimized real-time software using industrial PLC and NN is designed using MATLAB and executed using external server. Together with NN, digital twin (DT) algorithm, possessing ability to estimate, check, predict, test and adjust mechanical system parameters is developed from scratch on the external server. The paper presents a unique solution for industry application and proves the proposed algorithm’s stability and quality of speed control for a wide range of mechanical parameters variations under the external influences.</p>