Prediction of Damage to Electronic Devices Based On Power Consumption Characters Using Artifician Neural Network

Abstract

This study was conducted to determine the ability of artificial neural network in predicting the damage of electronic devices based on the character of electric power consumption. each electronic device has standard input and power consumption when the condition of the device is good. This test is performed using artificial neural network backpropagation with sigmod activation function.  To conduct training on the system, we perform the recording of power consumption of electronic devices that are considered in good condition. Data is recorded as a volt ampere line in a given time unit. From the results of testing and implementation of artificial neural network is able to recognize the good pattern of device consumption by using 300 or more hidden network. Artificial neural network backpropagation successfully detects abnormal power consumption with a success rate of over eighty percent.