Forecasting Turbulent Jets Using Univariate Time Series Models: A Study on , , ARMA, and

This chapter concentrates on implementing machine learning techniques to forecast univariate time series data. It explores fundamental forecasting models such as AR , MA , ARMA, and ARIMA , examining their mathematical foundations, fitting precision, correlations, residual analysis, and performance evaluations. The study employed experimental data derived from analyzing the behavior of a vertical buoyant water jet within a cylindrical tank equipped with both inlet and outlet features. Temperature measurements were taken at nine vertical points using K-type thermocouples. Applying these models to the experimental data for forecasting univariate temperature time series revealed the AR and ARIMA models as preferred options. Detailed analysis highlighted instances where the AR model excelled, notably in case 1 with all sensors except sensor 8, while the ARIMA model outperformed in case 2 (excluding sensors 2, 4, 5, 7, and 8), case 3 (excluding sensor 1), case 4 (excluding sensors 1, 5, 6, and 7), and case 5 (excluding sensors 1 and 4). These findings emphasize the efficacy of these models in comprehending and predicting detailed patterns within temperature time series, providing valuable insights for future forecasting endeavors.