A comparative study of solo and hybrid data driven models for predicting bridge pier scour depth

Abstract

This paper comparatively investigates the capability of 10 solo and hybrid Data Driven Models (DDMs) in predicting the pier scour depth. These models include Support Vector Machine (SVM), Adaptive Neuro Fuzzy Inference System (ANFIS), Artificial Neural Network (ANN), Gene-Expression Programming (GEP), improved Group Method of Data Handling (GMDH1 and GMDH2) and two hybrid forms of GMDH in combination with Harmony Search (HS) and Shuffled Complex Evolutionary (SCE) algorithms (GMDH1-HS, GMDH1-SCE, GMDH2-HS, GMDH2-SCE), respectively. A large set of experimental data collected from the literature were used to evaluate the capability of applied models in prediction of bridge pier scour depth. The results of the developed DDMs were compared with two mathematical formulas of HEC-18 and HEC-I8- K4Mo. The performance of all utilized models was evaluated by statistical criteria of RMSE, MSRE, CE and R². The results indicated that ANFIS was the superior model in terms of all statistical criteria in both training (CE = 0.969, RMSE = 0.038, MSRE = 0.049 and R² = 0.971) and testing phases (CE = 0.986, RMSE = 0.036, MSRE = 0.011 and R² = 0.985). GMDH2-SCE, ANN, GMDH2-HS, SVM and GEP were placed in the next ranks, respectively. This study recommends using DDMs, as powerful tools, for accurate prediction of pier scour depth.

Keywords:

• Local scour
• Depth prediction
• Data Driven Models
• Sediment
• Data analysis

Disclosure statement

No potential conflict of interest was reported by the author(s).