Optimization of ANFIS network using firefly algorithm for simulating discharge coefficient of side orifices

Abstract

Side orifices are installed on side walls of the main channel to regulate and measure the flow. Generally, discharge coefficient is the most important hydraulic parameter of side orifices. In this study, using ANFIS and firefly algorithm, discharge coefficient of rectangular and circular side orifices was simulated. Firstly, effective parameters were defined then six ANFIS and ANFIS-FA models were introduced. Monte Carlo simulations were utilized to survey the ability of numerical models, and the k-fold cross-validation approach was used to validate numerical results. Then, a sensitivity analysis was employed to introduce the superior model. The superior model predicted discharge coefficient with reasonable accuracy. This model estimated discharge coefficient in terms of all input variables. For example, for ANFIS model, R and SI were computed 0.832 and 0.029, respectively. In addition, for the hybrid model, RMSE, MARE and SI were calculated as 0.017, 0.017 and 0.027, respectively. Additionally, the ratio of width of the main channel to the side orifice diameter (B/D) was identified as the most effective parameter.

Study for Predicting Land Surface Temperature (LST) Using Landsat Data: A Comparison of Four Algorithms

“…e statistical evaluation of the model error was conducted, and the values of R 2 , RMSE, and MAE were found to be 0.91, 1.42°C, and 0.90°C, respectively. Furthermore, Akaike's information criterion (AIC) [49] of the models was calculated and found to be −11.72 and 4.23 for ANFIS and WNN models, respectively. e results for the training stage show that the ANFIS model is the optimal model for predicting the LST values, while the WNN is the worst model.…”

Timer Based Optimized Clustering for Lifetime Enhancement in WSN

“…Researchers have provided a method for optimizing ANFIS with the use of the firefly algorithm in [10]. The method shows improvement in processing in terms of accuracy and optimization of neurons which reduces the complexity as well.…”

“…Optimization of the ANFIS model is feasible by optimizing the number of neurons in the ANFIS network [10]. The main objective behind using ANFIS along with firefly is to optimize the neural network in ANFIS similar to the method in [4].…”

“…The rate of energy consumption normally can be estimated by considering Eq. (10). By considering Es, The timer value can be estimated by Eq.…”

“…The timer value depends on load handling capacity and energy consumption per second. The energy consumption per second is estimated by, Es = Eold -Enew/time_span (10) Where, Eold is residual energy for the last round of CH selection, Enew is residual energy for the current round of CH selection and time_span Eq. ( 11) is the time within two CH selection attempts.…”