Paraconsistent Artificial Neural Network For Structuring Statistical Process Control In Electrical Engineering

In this study, we present an algorithmic structure based on paraconsistent annotated logic (PAL) that can simulate the calculi of average values present in a dataset and detect the variations of the average using only PAL concepts. We call the structure as paraconsistent artificial neural network for extraction of moving average (PANnet(MovAVG)). As an example of its application, we use PANnet(MovAVG) to assist in the analysis of a final product quality index related to electrical engineering. To obtain the final result, we applied PANnet(MovAVG) to simulate the statistical behavior of the Statistical Process Control (SPC) by comparing values obtained with a ranking that establishes quality index standards based on electrical power distribution. First, tests were conducted using data with random values to verify the behavior of PANnet(MovAVG) and to set the optimum number of algorithms to form an optimized computational structure. Then, we used a database with actual electric voltage values generated by an electrical power system of an electrical power utility grid in Brazil. In the various tests, PANnet(MovAVG) appropriately detected changes and identified variations of electric voltage in 220-V transmission lines. The results show that PANnet(MovAVG) can be used to construct an efficient architecture for determining and monitoring quality scores with applications in various areas of engineering, especially for detecting quality index in an electricity distribution network.