Concrete Quality Prediction using Machine Learning

AbstractThe mechanical properties must be forecasted accurately to meet the high standards of design mostly seen in the construction field. Concrete is particularly problematic for such predictions because its composition is intricate and different aspects affect its outcome in rather idiosyncratic manners such as curing conditions or mix elements. It is for such reasons that conventional techniques of estimator models, statistical and empirical using linearity and nonlinearity regression tools fail to perceive these complex inter relational characteristics. Because of this limitation, they have long and time-consuming experiments in order to get the results. In order to overcome these problems the focus is gradually changing to using the Machine Learning (ML) models with the purposely aimed increase in the predictive accuracy and efficiency. Some of the models being considered for the purpose includeLasso, Ridge, XGBOOST, Decision Trees, and the evolutionary algorithms. This work provides a critical evaluation and analysis of several proposed ML models for predicting concrete mechanical strength with emphasis made on discovered research gaps andpotential future research directions. In the end it seeks to contribute towards the development and use of operational actual prediction techniques that are normal in industries.