Diagnosis of Physical Faults in Profibus-Dp Networks through Artificial Neuronal Networks

The present research exposes a PROFIBUS-DP fault diagnosis system based on Artificial Neural Networks, the model was trained with the keras package, specialized in the creation and training of deep learning models. Measurements of six possible scenarios were used, five of them correspond to common faults and the last one is the waveform when the network is working properly, these were captured with an oscilloscope for model training. The system provides a probable diagnosis based on data obtained with the oscilloscope connected to the PROFIBUS-DP network. The waveform indicates which type of failure it corresponds to, these are saved in a file that is sent to an interface that contains the previously trained classification model. The program gives a percentage of what type of failure may be occurring. The model was tested with variants in the test network, obtaining favorable results achieving an accuracy of 0.7225.