M.Sc. Tezi Görüntüleme

      In this study, performances of classical and fuzzy time series approaches, which are applied to time series data obtained from a product line, are compared. As a result, an approach to avoid possible anomalies of product line and to provide preventive maintenance capabilities to Industry 4.0 is tried to be achieved. Since The AdaptiveNetwork-Based Fuzzy Inference System (ANFIS) was used to estimate chaotic time series, it was particularly applied on time series data obtained from production line using sensors. The first degree Sugeno fuzzy inference system method has been used and suitable membership functions have been determined for the dataset. The Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) were calculated to assess estimation performance. ARIMA models, which is the one of the classical time series approaches, were determined by Box-Jenkins method and same performance metrics are calculated. The RMSE and MAPE values obtained by using ANFIS method and ARIMA models were compared. Finally, it was observed that the RMSE and MAPE obtained by first degree Sugeno fuzzy inference system method gave the best results on Industrial 4.0 data. In addition, the fuzzy inference system method produced the closest estimation values to actual values.