Publication

Abstract : Strength, workability and durability of concrete are the major aspects that are considered in the construction field. In order to improve the fresh stage and hardened stage properties of concrete, chemical admixtures are added to concrete. The flow behaviour of cement paste will help in ensuring production of concrete with good workability. In this work experiment is carried out on the super plasticized Portland Pozzolana Cement (PPC) mix made using superplasticizer from four different family. Marsh cone flow time for PPC mix containing different percentage superplasticizer is determined experimentally and utilized for modelling. The Artificial Intelligence era is intervening to all engineering domain. In concrete technology the research orientation mostly directing towards the strength prediction, mixture design and crack pattern analysis. In this research multilayer neural network approach with backward propagation is used for predicting the flow behaviour of cement paste. These methods will enhance the model prediction accuracy than the simple regression analysis. Python environment with tensor flow approach is used here for developing the algorithm and for compiling Google co-lab were utilized. Multiple hidden layers used for developing the model. From the experimental studies 269 data were adopted. Cement content, water, superplasticizer amount (taken as the seven separate entities based on the family) were considered as input variables and flow time as out-put variable. The model predicted marsh cone time values reliably and it is statistically analyzed with root mean squared error value 5.3 %, mean absolute percentage error as 5.03, linear correlation coefficient as 0.99.