Research On Performance Prediction Of Basalt Fiber Concrete Based On Machine Learning

With the continuous advancement of composite material theory,compounded materials have become an effective way to solve the inherent defects of materials.Basalt fiber,as a new green inorganic fiber material,has attracted much attention for its excellent mechanical properties,excellent chemical stability,thermal expansion coefficient similar to that of concrete and high-cost performance,and has become a research hotspot in the field of fiber-reinforced concrete.Although composite materials have more excellent properties,at the same time their compositions are becoming more and more complex,and the traditional performance prediction formulas are facing serious challenges.In recent years,with the rapid development of artificial intelligence technology,machine learning technology has begun to be widely used in the field of civil engineering.With its powerful learning and regression capabilities,machine learning technology has become a new favorite in the field of performance prediction.To reduce the engineering cost and the workload of proportional testing,this thesis introduces machine learning algorithms used to learn the relationship between concrete composition and performance to provide new methods and approaches for concrete performance prediction and design.This thesis mainly focuses on the mechanical properties of basalt fiber concrete for experimental research,and then collects a large amount of research literature about the performance data of basalt fiber concrete and combines with machine learning algorithm to carry out the research of basalt fiber concrete performance prediction and analysis,the main research content is as follows:(1)The effects of basalt fiber length(12 mm,18 mm),fiber volume fraction(0.04%,0.06%,0.08%)and mixed length admixture method on the mechanical properties of concrete were studied by uniaxial compressive test and four-point bending test,and the damage patterns of specimens were observed experimentally to obtain the key strength and strain of concrete with different fiber length and volume rate The results show that the strength and strain of concrete with different fiber lengths and volume ratios can be obtained.The results showed that the compressive strength of the specimens was the highest when the dose was 0.06% and the fiber lengths were 12 mm and 18 mm mixed in a 1:1 ratio;the flexural strength of the specimens was the highest when the dose was 0.04%and the fiber lengths were 12 mm and 18 mm mixed in a 1:1 ratio;in terms of mechanical properties,the mixture of two different lengths of basalt fibers was better than that of single-length fiber concrete.In terms of mechanical properties,the mixture of two different lengths of basalt fibers was better than that of single length fiber concrete.(2)The performance of the hyperparameter tuned prediction models using the grid search method and the genetic algorithm is compared.The results show that the performance of the prediction model after hyperparameter tuning by the genetic algorithm is better than that of the model tuned by the grid search method.The genetic algorithm can search the hyperparameter space better and find better hyperparameter combinations,thus improving the model performance.(3)A dataset of basalt fiber concrete components and strengths was collected based on published literature,and various machine learning algorithms were applied to develop concrete compressive and flexural strength prediction models.The results show that the stacked integrated model has the best predicted compressive strength performance;the extreme gradient lifting model has the best predicted flexural strength performance,and the characteristic importance analysis shows that the water-cement ratio and sand-rock ratio are the most important factors affecting the compressive strength of basalt fiber concrete,and the cement set ratio and water-cement ratio have the greatest effect on the tensile strength of basalt fiber concrete.(4)Based on the dataset,multiple machine learning algorithms were used to construct a prediction model for the dynamic mechanical strength growth factor of basalt fiber concrete.The results show that the stacked integrated model performs the best and can accurately predict the dynamic mechanical strength growth factor of basalt fiber concrete.Through feature importance analysis,strain rate and water-cement ratio were identified as the main factors affecting the dynamic mechanical properties of basalt fiber concrete.(5)Based on the data set,a model for predicting the relative dynamic elastic modulus of basalt fiber concrete was developed using various machine learning algorithms.It was shown that the stacked integrated model performed best in predicting the relative dynamic elastic modulus.The number of freeze-thaw cycles and compressive strength were determined to be the most important factors influencing the durability of basalt fiber concrete in freeze-thaw cycles through feature importance analysis.