Inverse dynamic control via simulation of feedback control by artificial neural networks for a crane system

Cranes are systems highly used in industrial applications to transport heavy loads. The nonlinear behavior, and the crane own dynamic produce vibrations during the motion. In order to improve the system performance and specifically to ensure stability and a control with minimum vibrations, this work proposes a new strategy based on IDCS (Inverse Dynamic Control Simulation) using machine learning with Artificial Neural Networks ANN that gives the possibility to learn the inverse dynamic model of the crane and apply the feedback information based on the inverse dynamic control simulation. IDCS allows a suitable signal control avoiding noise and need for extra sensors in the feedback loop. The ANN creates the inverse dynamic model of the crane. The training architecture is developed to learn the inverse dynamic model of the crane in different operational points, and is used as feedforward control with the feedback dynamic in the IDCS scheme. Simulations and experiments are conducted with an industrial crane, and results show that the proposed method decreases vibration and position error. The proposed ANN-IDCS showed suitable performance compared to other controllers such as analytical inverse model control (AIC), Dual Matching Control (DMM) and shaped reference in feedforward (FRC).