Identification of a MIMO Twin Rotor System Using an Artificial Neural Network Trained by PSO

Twin Rotor Mimo System (TRMS) is a dynamic system with multiple inputs and multiple nonlinear outputs that simulates the action of a helicopter, in this type of system there is a complexity at the time of describing the operation through a transfer function with conventional methods, due to the development of mathematics. For the identification of this type of systems there are alternative methods such as artificial intelligence, specifically Artificial Neural Networks (ANN). The nonlinear autoregressive network with exogenous inputs (NARX) allows modeling nonlinear dynamic systems because it takes prior values of inputs and outputs in different layers. The weights of this network were improved by particle swarm optimization (PSO), as these were considered as particles to find their best position within the search space. For this identification, a data set relating the input to the output of the TRMS at a given time was used through the MATLAB software with its Neural Network Time Series app library and it was obtained as a result that the output signal of the equipment was similar to the estimated output signal of the neural network, optimizing the computational cost and the training time. The algorithm that was developed has the versatility to identify the response of linear systems.