Experimental and Machine Learning Strategies with SHAP and PDP Analysis of Self-Healing Characteristics of Bio-concrete Using Bacillus Cereus and Steel Slag

Abstract Self-healing concrete is a cement-based substance that can mend itself from a hairline crack. This experiment examines the effectiveness of the gram-positive cell bacteria Bacillus cereus on the self-healing characteristics of steel-slag bio-concrete (SSBC). The study's ultimate goal is to discern the best proportions of bacteria to incorporate into bio-concrete by examining the fresh, hardened, and self-healing qualities, as well as machine learning (ML) approaches with SHAP and PDP analysis to forecast crack width. As a result, five distinct mix designs were developed by substituting steel slag for varying percentages of sand, e.g., 0%, 10%, 20%, 30%, and 40%. The behavior of steel-slag concrete (SSC) was evaluated through fresh, mechanical, and non-destructive tests (NDT) to determine the optimal percentage of steel slag. After finding the ideal percentage of steel slag, three different bacterial solutions of 12.5 × 105 CFU/ml, 25 × 105 CFU/ml, and 31.25 × 105 CFU/ml were mixed with calcium lactate as a feeder to make SSBC. The compressive strength of the mix, which consisted of 12.5 × 105 CFU/ml, 25 × 105 CFU/ml, and 31.25 × 105 CFU/ml B. cereus bacteria, along with a constant 30% steel slag, increased by 13.85%, 21.75%, and 31.57% at 28 days. In the NDT test, the assessment of the rebound hammer and ultrasonic pulse velocity (UPV) showed a similar tendency to the compressive strength test. Visual observation and scanning electron microscopy (SEM) were done to examine the crack-healing tendency of the SSBC. SEM images revealed that adding B. cereus improved the physical characteristics of the concrete by forming CaCO3 crystals, which heal the microcracks. The CatBoost model exhibits greater precision and reliability, even though the regression coefficient (R2) of the random forest (RF) and artificial neural networks (ANN) models is similarly outstanding.