Predictive maintanence fundamental vibration based on artificial neural network at machinery

Key words: Neural Network, Back-Propagation, Predictive Maintanence, Vibration Analyze Since some approaches are approved in order to get the max yield benefit and production wastes for machinery. One of this approache is the maintanence based on monitoring situation or called as early notice dynamic maintanence or as predictive maintanence. With this respect, machinery conditions are determined by periodically and continuous measurements. Previously the breakdown period depends on manufacturing is determined by measurment and control. Considering to the controls, machines are maintanence at proper times. The reason and improvement of the breakdown is trained by the analysis obtained from data. Thus prevent the unexpected breakdown stopsa re provided. In this study, training set was composed using ISO-2372 which is the evaluation standards of medium measure machines and was applied for training of Artificial Neural Networks (ANN). Networks consist of three layers which are an input layer, a hidden layer and an output layer. First data matrix which size is 720x9 was composed the table of ISO-2372 for the network training. Nevertheless a test matrix which is 200x5 size was composed from the training set. Training was tested by changing different types of hidden layers which were consist of 5,10,15,25,50 and 75 neurons. Test results were compared with each others and the best performance was found. Back-Propagation was used in training. Suitability was determined with comparing result valiues and real table values. Finaly result graphics were obtained using output values. xiv