首页|New Machine Learning Study Findings Have Been Reported by Investigators at University of Oklahoma (Modeling of Necking Area Reduction of Carbon Steel In Hydrogen Environment Using Machine Learning Approach)
New Machine Learning Study Findings Have Been Reported by Investigators at University of Oklahoma (Modeling of Necking Area Reduction of Carbon Steel In Hydrogen Environment Using Machine Learning Approach)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Elsevier
Investigators publish new report on Machine Learning. According to news reporting from Norman, Oklahoma, by NewsRx journalists, research stated, “Low carbon and low alloy steel pipes, prevalent in natural gas transmission systems due to their affordability, weldability, and strength, are confronted with significant challenges such as hydrogen embrittlement (HE) when transitioning towards hydrogen energy systems. This necessitates innovative predictive strategies to overcome these hurdles.” Funders for this research include Pipeline and Hazardous Materials Safety Administration (PHMSA) of the U.S. Department of Transportation (DOT), University of Oklahoma. The news correspondents obtained a quote from the research from the University of Oklahoma, “Most existing research on HE has focused on a limited range of low carbon and low alloy steels under specific experimental conditions and has been constrained by the limitations of experimental facilities. To expand the research scope, our study incorporated seven machine learning (ML) techniques: Random Forest (RF), Decision Tree (DT), Extreme Gradient Boosting (XGBoost), Gradient Boosting (GB), Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost), and Artificial Neural Networking (ANN). The aim is to predict the HE in terms of the degradation of mechanical properties, specifically the reduction in area. Drawing from tensile test data obtained from 47 distinct low carbon and low alloy steels under pressurized hydrogen gas conditions, we constructed and evaluated a range of ML models, with the aim of identifying the most efficacious one for our study. Our results indicated that the CatBoost ML model offered the best prediction of the reduction in the area of these steels in a hydrogen environment. The CatBoost model provided a low Mean Absolute Error (MAE) of 7.32, Mean Square Error (MSE) of 83.78, Root Mean Square Error (RMSE) of 9.15, and a coefficient of determination (R2) value of 77.62% for the training data and 72.50% for the testing data. Furthermore, the CatBoost model identified hydrogen gas pressure and the steel’s ultimate tensile strength as the most influential parameters, contributing 47.4% and 19.2% respectively to the prediction of HE.”
NormanOklahomaUnited StatesNorth and Central AmericaCyborgsElementsEmerging TechnologiesGasesHydrogenInorganic ChemicalsMachine LearningUniversity of Oklahoma
NETL
NSTL
Elsevier
