Modeling of Carbon Mortar Color Expression Using Artificial Neural Network

Concrete coloring is a well-established practice in construction and architecture. Pigments and white Portland cement are commonly used to achieve decorative effects while maintaining structural integrity. However, predicting the final color outcome based on ingredient ratios remains a challenge. Prior research has shown that color evaluation in concrete can be measured using tristimulus values in the CIELAB space. Yet, no prior work had resolved how to effectively model these color outcomes using machine learning techniques. This gap motivated the exploration of artificial neural networks as a predictive tool. Existing methods rely on trial-and-error approaches, which are time-consuming and inefficient. The need for a reliable model to predict color parameters based on mix design is evident. This paper addresses that need by introducing a novel application of ANNs in the field of concrete color modeling. The study contributes to both materials science and machine learning applications in construction.

The aim of this study was to develop a predictive model for black mortar color outcomes using artificial neural networks. The specific problem addressed is the difficulty in accurately predicting color parameters based on pigment and carbon black ratios in mortar mixtures. The motivation stems from the inefficiency of traditional trial-and-error methods in concrete coloring. By using ANNs, the researchers sought to create a more efficient and accurate approach to color prediction. The study focused on black-colored mortars, which are commonly used in architectural applications. The goal was to determine whether ANNs could reliably model color parameters using input variables such as pigment, carbon black, and white Portland cement. The study also aimed to evaluate the effectiveness of ANNs in capturing complex relationships between mix proportions and color outcomes. This approach could reduce the need for extensive laboratory testing in concrete color formulation.

The study utilized a dataset from laboratory experiments involving nine different mortar mixtures. Each mixture varied in the ratios of pigment, carbon black, and white Portland cement. The experimental setup included measuring color parameters at ten locations on each mortar sample. The data collected included tristimulus values L*, a*, and b* in the CIELAB color space. These values represent luminosity, redness/greenness, and yellowness/blueness, respectively. The data were formatted into three input parameters for the ANN model: pigment, carbon black, and white Portland cement. The output parameter was the color evaluation of the black mortar. The ANN model was constructed using these input-output relationships. The model was trained and tested using the collected data to evaluate its predictive accuracy. The methodology focused on using machine learning to model a complex physical process.

The results demonstrated that the ANN model could effectively predict the color parameters of black mortars based on ingredient ratios. The model was trained using data from nine different mixtures, each with ten surface measurements. The predicted color values showed strong alignment with the measured tristimulus values. The L* values indicated the luminosity of the mortars, ranging from 0 to 100. The a* and b* values captured the redness/greenness and yellowness/blueness, respectively. The model achieved high accuracy in predicting these parameters. The study found that the ANN approach provided a reliable alternative to traditional methods. The model's performance was evaluated using standard metrics for neural network validation. The results suggest that ANNs can capture the complex interactions between pigment, carbon black, and cement in determining color outcomes. These findings support the use of machine learning in concrete color modeling.

The authors concluded that artificial neural networks can serve as a viable alternative for predicting color parameters in black mortars. The study showed that the model could accurately predict L*, a*, and b* values based on input variables. The results suggest that ANNs can capture the non-linear relationships between mix proportions and color outcomes. The authors propose that this approach could reduce the need for extensive laboratory testing in concrete color formulation. The study supports the use of machine learning in construction material modeling. The findings align with the authors' hypothesis that ANNs can be used to predict color outcomes in mortars. The authors did not claim that this model is the only solution but emphasized its potential as a predictive tool. The study's implications are specific to the application of ANNs in concrete color modeling.