Prediction of Healing Performance of Autogenous Healing Concrete Using Machine Learning

Abstract

Cracks typically develop in concrete due to shrinkage, loading actions, and weather conditions; and may occur anytime in its life span. Autogenous healing concrete is a type of self-healing concrete that can automatically heal cracks based on physical or chemical reactions in concrete matrix. It is imperative to investigate the healing performance that autogenous healing concrete possesses, to assess the extent of the cracking and to predict the extent of healing. In the research of self-healing concrete, testing the healing performance of concrete in a laboratory is costly, and a mass of instances may be needed to explore reliable concrete design. This study is thus the world’s first to establish six types of machine learning algorithms, which are capable of predicting the healing performance (HP) of self-healing concrete. These algorithms involve an artificial neural network (ANN), a k-nearest neighbours (kNN), a gradient boosting regression (GBR), a decision tree regression (DTR), a support vector regression (SVR) and a random forest (RF). Parameters of these algorithms are tuned utilising grid search algorithm (GSA) and genetic algorithm (GA). The prediction performance indicated by coefficient of determination (R2) and root mean square error (RMSE) measures of these algorithms are evaluated on the basis of 1417 data sets from the open literature. The results show that GSA-GBR performs higher prediction performance (R2GSA-GBR = 0.958) and stronger robustness (RMSEGSA-GBR = 0.202) than the other five types of algorithms employed to predict the healing performance of autogenous healing concrete. Therefore, reliable prediction accuracy of the healing performance and efficient assistance on the design of autogenous healing concrete can be achieved.

An Overview of AI-Assisted Design-on-Simulation Technology for Reliability Life Prediction of Advanced Packaging

“…Machine learning involves the use of AI theory combined with big data to guide computers for training and learning; eventually, a simple AI model with input and output relationships will be developed to help researchers make design decisions [ 21 , 22 , 23 , 24 ]. Machine learning [ 25 , 26 , 27 , 28 ] can be applied for regression or classification models using either supervised or unsupervised learning. In this review, because the input datasets are labeled, the learning algorithm for predicting the reliability life is considered supervised.…”

Advancements in bacteria based self-healing concrete and the promise of modelling

“…The output of the model was one single neuron representing the final crack width. Their work was followed up by more sophisticated ML approaches [172] based on a similarly large experimental data sets from literature (1417 data points), when six algorithms were compared for their ability to predict the autogenous healing performance of concrete. A similar example was very recently reported by Chen et al [173], with the difference that they have generated their own experimental data.…”

Interpretable Machine Learning for Prediction of Post-Fire Self-Healing of Concrete

“…Good precision was achieved, with R 2 greater than 0.85; however, only four input variables were employed: initial crack width, fly ash, silica fume, and hydrated lime powder [ 34 ]. Six different ML algorithms, i.e., ANN, k-nearest neighbors, decision tree regression, Support Vector Regression, and two ensemble models (gradient boosting regression and Random Forest) were trained on an extensive database of more than 1400 records to predict autogenous self-healing of concrete with very high accuracy [ 35 ]. Sixteen predictors were included: type and dosage of healing material; fiber diameter, length, and tensile strength; the initial cracking data and initial cracking width; the time for healing; the healing condition (environmental exposure); the amount and type of cement; the amount of superplasticizer; fine aggregates; fly ash; slag; and the water–binder ratio [ 35 ].…”