A CFD-driven surrogate modeling framework for rapid erosion wear prediction in elbow pipes using a COA-optimized CNN–LSTM network

Yi Li

doi:10.1016/j.advengsoft.2026.104205

A CFD-driven surrogate modeling framework for rapid erosion wear prediction in elbow pipes using a COA-optimized CNN–LSTM network

2026-05-11

0

PRE

AI

Y

Yi Li

X

Xiaoyong Li

Z

Zhaohui He

J

Jincheng Chen

Z

Zhe Lin *

DOI：10.1016/j.advengsoft.2026.104205

Original

Original request for help

Save

Abstract

En 中文

• A COA-driven CNN-LSTM surrogate framework is proposed for elbow erosion prediction. • COA serves as a unified global optimizer for both temporal and spatial branches. • A cascaded pathway links operating conditions, particle dynamics, and wall erosion. • The model is trained and validated using 625 experimentally validated CFD cases. • The proposed model achieves R² ≈ 0.972 and reduces RMSE and MAE by about 60 %.

Keywords:

elbow erosion

CNN-LSTM network

COA optimization

surrogate modeling

CFD validation

A CFD-driven surrogate modeling framework for rapid erosion wear prediction in elbow pipes using a COA-optimized CNN–LSTM network

Abstract

Journal

Organization

Cited Papers

Citing Papers