|
Title
|
Predicting discharge coefficient of weir–orifice in closed conduit using a neuro-fuzzy model improved by multi-phase PSOGSA
|
|
Type
|
JournalPaper
|
|
Keywords
|
Combined weir-orifices; dimensionless experimental discharge; modeling; particle swarm optimization; gravity search algorithm; neuro-fuzzy system
|
|
Abstract
|
This study investigates the viability of a strong algorithm (PSOGSA) merging particle swarm optimization (PSO) and gravity search algorithm (GSA) in tuning adaptive neuro fuzzy system (ANFIS) parameters for modeling dimensionless experimental discharge of combined weir-orifices. The results are compared with the standard ANFIS and two hybrid models ANFIS tuned with PSO and GSA. The models are assessed by applying several dimensionless input parameters, consisting h/D (the ratio of upstream water depth to channel diameter), W/D (the ratio of orifice opening height to channel diameter), H/D (the ratio of plate height to channel diameter) and using comparison indices such as root mean square error and mean absolute error. The outcomes reveal that the new ANFIS-PSOGSA method provides superior accuracy in modeling dimensionless experimental discharge over the ANFIS-PSO, ANFIS-GSA and standard ANFIS method. Among the input parameters, the h/D was found to be the most effective input on modeling dimensionless experimental discharge while involving the H/D parameter deteriorated the models’ performances. The relative root mean square error differences between ANFIS-PSOGSA and ANFIS are found as 50% and 68.29% for pipe A and B, respectively. By implementing the ANFIS-PSOGSA, the accuracy of ANFIS-PSO and ANFIS-GSA is also improved in modeling dimensionless experimental discharge by 45.71% and 29.63% in pipe A and by 63.89% and 45.83% in pipe B with respect to root mean square error.
|
|
Researchers
|
Seyedeh Aniseh Ghazaii (Not In First Six Researchers), Maziar Osmani (Not In First Six Researchers), Hong-Liang Dai (Fifth Researcher), Mohammad reza nikpour (Fourth Researcher), Ozgur Kisi (Third Researcher), Payam Khosravinia (Second Researcher), Rana Muhammad Adnan (First Researcher)
|