International Journal of Applied Science and Technology

ISSN 2221-0997 (Print), 2221-1004 (Online) 10.30845/ijast

Speed Forecast of DC Motor Using Artificial Neural Network
Adepoju G. A., Aborisade, D.O; Eluwole O. T.

Artificial Neural Network (ANN) has achieved a lot of attention as well as gained enormous popularity over the last two decades owing to its vast applications both in industry and academia. This paper explicitly and effectively examines the application of ANN to speed forecast of dc motors. The angular speed of a dc motor can be determined by prediction with given voltage as input parameter. By using motor parameter values and assigning state variables as armature current, angular speed, and rotor displacement; the dc motor transfer function relating angular speed to voltage was obtained. With the input (voltage) and speed as the target. A simulation process was carried out on the transfer function using one of the most commonly used and versatile technical computing facilities (MATLAB & SIMULINK, version 7.0) to generate various input/target pairs for the ANN training and testing. Levenberg-Marquardt standard back-propagation algorithm with normalized preprocessed data was used to predict and show the pattern of correlation of input (voltage) in relation to the output (speed). The ANN developed comprised 4 layers: an input layer, 2 hidden layers, and an output layer. The input layer has 20 neurons; the first hidden layer has 20 neurons, the second hidden layer has 18 neurons, while the output layer has a single neuron. The ANN was able to develop a good mapping between the input and the target. The results obtained indicated 100% correlation and an absolute mean error (AME) of 0.38% during testing with unfamiliar input/target data pairs. The former shows a high reliability of the network to predict the output while the latter represents on an average, a very high degree of accuracy in the prediction.

Full Text: PDF