Research on performance prediction of cement mortar based on dynamic machine learning
Supplementary cementitious materials have become indispensable components of cement-based materials.The variation of technical properties caused by their origins and formation conditions has brought challenges to the efficiency and accuracy of material de-sign.In this study,the influence of raw material and composition characteristics such as binder dosage,pozzolanic activity,particle size and packing,water requirement ratio,specific surface area,loss on ignition,density on cement mortar was comprehensively considered,and performance prediction was conducted based on dynamic machine learning.The parameterized expression of material composition and its correlation weightings with material performances were explored,and the prediction model based on the technical properties of cementi-tious materials was proposed.The results show that the numerical model based on dynamic machine learning can accurately and efficiently predict the fresh and mechanical properties of cement mortar.Through the correlation analysis of technical properties of cementitious mate-rials,the combination of key characteristic parameters that affect the properties of cement mortar is determined,and the generalization abil-ity of the model is improved.The prediction accuracy of the performance prediction model based on key characteristic parameters is 83%in fluidity,while the prediction accuracy of rheology and compressive strength is no less than 96%.This method is expected to reduce or elim-inate the influence of raw material quality fluctuation on material design and trial efficiency and provide reference for intelligent composi-tion design of building materials.
万方数据
