| dc.description.abstract | Any industrial machinery in its working state produces vibrations that can be read out using vibration sensors, or also known as accelerometers. This vibration cycle can occur thousands of times per second, which will cause a large volume of data to be collected, characterizing time séries of data. Thus, during any operation of the equipment, and based on some research already carried out in the area of Industry 4.0 and Internet of Things, it is possible to identify patterns of cyclic vibrations that characterize the operation of the equipment being performed. In other related research, some fault prediction models are proposed, whether using neural networks or some other artificial intelligence algorithm, but the proposed model cannot always be applied universally, leaving some very specific equipment outside of these implementations. And even if the model is applied, for a neural network to be effective, it needs to be trained with a reference database. Thus, the proposed model suggests the creation of operating templates for industrial equipment of the wirebonders type, using as a basis only the equipment’s vibration data. With the creation of these templates through the cataloging, quantification and sequencing of vibration signatures, it was possible to generate an index to measure the quality of the operations of the evaluated equipment, in this unsupervised system model, but which demonstrated great potential to become a system supervised, either to predict failures or predict normalities, depending on the type of operation template cataloged. In the case of the applied experiment, it was decided to catalog any operation and compare it or other operations that were executing the same processes as the cataloged operation, with the purpose of generating a scalar quality index, based not only on the similarity of vibration signatures, but also in their sequencing and quantification. Thus, when comparing the cataloged template with the other two operations, two scalar quality indexes were obtained: the first compared operation obtained a quality index of 87.25% and the second obtained an index of 95.36%. Thus, this work achieved the proposed objective, having as its main scientific contribution the ability to generate quality indexes per operation in machines of the wirebonders type, opening other possibilities of research that may be related to other types of equipment. | pt_BR |
