Development of an Optimization Algorithm for Designing Low-Carbon Concrete Materials Standardization with Blockchain Technology and Ensemble Machine Learning Methods

Abstract

This study presents SmartMix Web3, a framework combining ensemble machine learning and blockchain technology to optimize low-carbon concrete design. It addresses two key challenges: (1) the limitations of conventional models in predicting concrete performance, and (2) ensuring data reliability and overcoming collaboration issues in AI-driven sustainable construction. Validated with 61 real-world experiments in Cameroon and 752 mix designs, the framework shows major improvements in predictive accuracy and decentralized trust. To address the first research question, a stacked ensemble model comprising Extreme Gradient Boosting (XGBoost)–Random Forest and a Convolutional Neural Network (CNN) was developed, achieving a 22% reduction in Root Mean Square Error (RMSE) for compressive strength prediction and embodied carbon estimation compared to traditional methods. The 29% reduction in Mean Absolute Error (MAE) results confirms the superiority of Extreme Learning Machine (EML) in low-carbon concrete performance prediction. For the second research question, SmartMix Web3 employs blockchain to ensure tamper-proof traceability and promote collaboration. Deployed on Ethereum, it automates verification of tokenized Environmental Product Declarations via smart contracts, reducing disputes and preserving data integrity. Federated learning supports decentralized training across nine batching plants, with Secure Hash Algorithm (SHA)-256 checks ensuring privacy. Field implementation in Cameroon yielded annual cost savings of FCFA 24.3 million and a 99.87 kgCO₂/m³ reduction per mix design. By uniting EML precision with blockchain transparency, SmartMix Web3 offers practical and scalable benefits for sustainable construction in developing economies.

1. Introduction

Researchers in civil engineering are exploring eco-friendly construction materials, solid waste reutilization, and CO₂-based concrete curing. Life-cycle assessment and emission factor calculations are commonly used to evaluate environmental impacts [1,2]. Globally, architects follow standard mix design guidelines (JGJ 55-2011) [3], determining component proportions through iterative calculations and trial batches. Despite refinements, these traditional methods remain resource inefficient. With advances in computation, machine learning is increasingly applied to enhance prediction accuracy in engineering [4], proving effective for modeling nonlinear structural and material properties [5,6]. Nonetheless, algorithmic limitations persist, highlighting the need for careful algorithm selection.

Recently, various AI techniques have been applied to assess concrete performance [7]. AI’s nonlinear processing and ability to analyze multidimensional variables enable effective evaluation of complex interactions in concrete. A hybrid model proposed for stock price forecasting demonstrates the potential of integrating multiple AI methods to improve predictive accuracy in dynamic systems [8,9]. However, challenges remain, such as developing robust, high-quality datasets and selecting optimal algorithms for performance prediction. Overcoming these issues can enhance model reliability. Furthermore, analyzing concrete’s sustainable performance can optimize material ratios, cut costs, and support sustainable construction practices [10,11].

To improve viscosity and workability, admixtures are added to traditional concrete while keeping or reducing the water–cement (w/c) ratio. This enhances flow and deformation without sacrificing rheology or long-term durability [12,13,14,15]. Life-cycle studies show that admixtures like viscosity-modifying agents, plasticizers, and microencapsulated phase-change materials (MPCM) reduce cement consumption without affecting hydro mechanical performance [16,17,18,19,20], thus lowering carbon emissions in line with United Nations Sustainable Development Goals (UN SDGs) and the 27th Conference of the Parties (COP27) goals [21,22,23,24,25,26]. COP27 also highlights cleaner concrete production to advance SDG 9, supporting SDGs 3 and 11 on health and sustainable urbanization [26].

Self-compacting concrete (SCC), requiring less than 350 kg/m³ of cement versus 350–450 kg/m³ in conventional concrete, is a key solution. SCC typically incorporates supplementary cementation materials (SCMs) like silica fume, fly ash, rice husk ash, metakaolin, kaolin, GGBS, and quarry dust from waste sources [27,28]. These SCMs contribute to strength development through pozzolanic reactions, forming additional calcium silicate hydrate (C-S-H) that densifies the microstructure and improves durability. SCC also improves particle packing and stability, allowing self-flow without increasing the w/c ratio. Additionally, replacing limestone filler with Polyethylene Terephthalate (PET) waste powder enhances porous concrete’s stability and mechanical properties while reducing pollution [29].

Low-carbon concrete adheres to standardized quality control protocols and regulatory compliance, with automated inspection systems enhancing accuracy and objectivity [30]. However, construction contract management remains challenged by transparency, compliance, and operational inefficiencies [31], often hindering stakeholder collaboration. Blockchain offers a promising solution by streamlining the planning, design, execution, and maintenance phases of construction projects [32].

Fly ash-based geopolymer concrete (GPC) is increasingly used, yet lacks an empirical model to accurately predict compressive strength across diverse mix parameters. Key influencing factors include curing age, plasticizer dosage, NaOH concentration, and sodium silicate (Na₂SiO₃) content [33], leading to variability in strength predictions. The rapid advancement of soft computing has enabled the development of efficient predictive models using machine learning techniques.

Traditional methods, based on physical testing and finite element simulations, often fall short in modeling multi-component materials like concrete. This study highlights recent advancements in machine learning software and their application in enhancing predictive accuracy within the construction industry.

To overcome the limitations of traditional concrete proportioning methods, researchers have integrated machine learning techniques into mix design prediction models [34]. One study employed the Extreme Learning Machine (ELM) method to predict the strength of rubber concrete, demonstrating superior accuracy and generalization over conventional models. Further research applied advanced optimization methods to design lightweight aggregate concrete, evaluating the performance of five algorithms: Support Vector Machines (SVMs), Artificial Neural Networks (ANNs), Decision Trees (DTs), Gaussian Process Regression (GPR), and XGBoost. Among these, GPR yielded the most accurate results [35].

Neural networks have been employed to predict the 28-day compressive strength of self-compacting concrete with admixtures [36]. Other studies have compared machine learning models such as Random Forest, SVM, and neural networks for residual strength prediction, identifying the most effective approaches [37]. A regularized machine learning model was also developed to accurately forecast the compressive strength of high-performance concrete [38], with results showing that algorithm optimization can significantly improve performance [39,40]. Genetic algorithms, known for their powerful search capabilities, have been used to enhance neural network efficiency [41].

In recent developments, researchers have utilized convolutional neural networks (CNNs) and XGBoost for deep learning-based fracture detection in low-carbon concrete, achieving high accuracy [42,43]. Moreover, a comparative study of deep learning models, CNNs [44,45], Long Short-Term Memory networks (LSTM) [46], Gated Recurrent Units (GRUs) [47], and Bidirectional LSTM (BiLSTM) [48] for predicting high-performance concrete strength revealed that the GRU model provided the most accurate predictions.

While AI-powered design tools are gaining traction in construction materials research, critical barriers such as data reliability, transparency, and limited stakeholder collaboration continue to hinder their widespread adoption. Notably, existing research does not sufficiently explore how blockchain can be integrated with AI models to ensure traceability, data integrity, and trust within low-carbon concrete supply chains [49]. The intersection of AI and blockchain offers a promising yet underexplored avenue for advancing material design and construction practices [50].

This study addresses that gap by evaluating the advantages of ensemble machine learning techniques over traditional regression and simulation models in predicting the performance of low-carbon concrete. Furthermore, it introduces the SmartMix Web3 framework, which combines AI and blockchain to enhance data reliability, traceability, and multi-stakeholder collaboration in AI-driven sustainable concrete design.

The research questions for this research are as follows:

(1). How can ensemble machine learning methods enhance the accuracy of predicting the compressive strength and sustainability performance of low-carbon concrete compared to traditional regression and simulation models?

(2). What are the key challenges in adopting AI-driven low-carbon concrete d sign, and how can blockchain-enhanced traceability address issues like data reliability and stakeholder collaboration in the construction industry?

To tackle the complex task of estimating compressive strength in low-carbon concrete with specific proportions, researchers are employing various artificial intelligence techniques, including machine learning, deep learning, and genetic algorithms or their elite strategy variants. However, AI applications in this field tend to be one-sided, with most studies focusing either on performance prediction or on single/multi-objective optimization, without fully integrating both. The following abbreviations are listed in Table A1.

The abundance of prediction algorithms also complicates their selection and improvement. In optimization, performance and cost control are typically prioritized, while environmental impact is often overlooked. This research introduces an innovative optimization algorithm that combines blockchain technology with ensemble machine learning methods to enhance both prediction and standardization of compressive strength in low-carbon concrete.

Blockchain integration ensures data transparency, security, and traceability across the production lifecycle. Ensemble machine learning—using XGBoost, Random Forest, and 1D-CNN—boosts prediction accuracy and reliability by aggregating multiple models to minimize errors. This synergy not only optimizes material performance but also supports carbon reduction in construction, promoting sustainable practices.

This study is arranged as follows: Section 1 presents the literature review on low-carbon concrete optimization; Section 2 details the proposed model and its parameters; Section 3 outlines methodologies and hybrid algorithm implementation; Section 4 discusses results; and Section 5 concludes the work.

2. Model

2.1. Detailed Description of the Proposed Model

One of the primary hurdles in developing concrete, especially when it comes to fly ash-based geopolymer concrete, is the optimization of the mix design. This process entails determining the ideal combination of materials to achieve desired characteristics, such as high compressive strength, while also minimizing carbon emissions [51,52]. With increasing environmental concerns, the focus of concrete research has shifted from solely enhancing mechanical performance to also reducing ecological impact [53]. Traditionally, optimizing mix proportions has relied on a trial-and-error approach, where numerous experimental batches are tested with varying variables. However, as the number of variables grows, the required number of trials increases significantly, making the process both time consuming and resource intensive. Additionally, this method often yields satisfactory results but rarely finds the most efficient or optimal solution.

To overcome the limitations of experimental techniques, the design of concrete mixtures can be improved by employing computational models that link mixture proportions to properties such as compressive strength.

Upon completing this study, it was clearly established that integrating machine learning and blockchain technology into construction projects enhances safety, improves productivity, and offers several additional benefits. This research addresses two distinct questions, each supported by different datasets.

The first question of how ensemble machine learning methods can improve the accuracy of predicting compressive strength and sustainability performance of low-carbon concrete compared to traditional models is explored using a comprehensive experimental dataset of 752 samples. This dataset includes measured concrete components such as Blast Furnace Slag, Fly Ash, Superplasticizer, Water, NaOH, Coarse Aggregate, Cement, Na₂SiO₃, Fine Aggregate, and Age. These features serve as inputs to train and validate the machine learning models, with compressive strength as the target variable.

The second question, examining the challenges in adopting AI-driven low-carbon concrete design and how blockchain-enhanced traceability can improve data reliability and stakeholder collaboration, is investigated using a smaller, focused dataset of 61 samples collected from Cameroon. This dataset supports an in-depth exploration of practical implementation challenges within the construction industry.

2.2. Machine Learning Component: Random Forest Regressor (RFR)

The RFR is an ensemble machine learning method that constructs multiple decision trees during training and outputs the mean prediction of the individual trees, effectively handling non-linear relationships and reducing overfitting. Feature selection involves identifying the most effective set of input variables by removing those that are irrelevant or only weakly correlated with the target outcome. This process results in an optimal subset drawn from all available input features. Redundant variables can negatively impact prediction performance by introducing noise and decreasing accuracy. Implementing feature selection helps to minimize data noise, enhance modeling efficiency, and boost predictive accuracy. Traditionally, this process has relied heavily on evaluating the importance of each variable [54].

A higher mean square value indicates a stronger impact of the feature variable on the target outcome. In the random forest algorithm, one of the most refined methods for feature selection involves evaluating permutation importance. This technique determines the significance of each feature by measuring the change in the model’s prediction accuracy before and after randomly shuffling the values of that feature. If the feature is strongly related to the target variable, randomizing it will significantly reduce the model’s accuracy, indicating a high level of importance due to its strong correlation with the output. The implementation of Random Forest is achieved through bagging decision trees by employing random split selection [55].

2.3. Model Analysis

The improved prediction accuracy in this study stems from the integration of XGBoost, Random Forest, and 1D CNN within a stacked ensemble. XGBoost, contributing about 45%, effectively captures complex non-linear relationships; Random Forest (25%) adds robustness and generalization across noisy datasets; and 1D CNN (30%) detects latent patterns in sequential data often missed by tree-based models. The stacked architecture leverages a meta-learner to combine these models’ outputs, enhancing overall accuracy. This approach resulted in a 22% reduction in RMSE for compressive strength and a 29% reduction in MAE for embodied carbon, demonstrating the strength of model complementarity in optimizing low-carbon concrete mix designs.

2.4. Blockchain Technology Component

The lack of digitization has caused many challenges in the construction industry. Nearly 90% of mega projects exceed their original budgets. Several advancements aim to address these issues. Cost estimation software provides accurate budget projections using historical data. Building Information Modeling (BIM) enhances construction processes by simulating scenarios, detecting problems early, and improving scheduling accuracy [56].

Quality control software ensures standardized procedures and regulatory compliance, while automated inspection systems enhance accuracy. However, challenges remain, particularly in contract administration, where transparency, compliance, and inefficiencies create obstacles [57].

These issues hinder stakeholder collaboration in construction projects. Blockchain technology offers a potential solution by transforming project planning, design, execution, and maintenance, improving efficiency and trust. Advancements in information technology, especially blockchain, are improving quality traceability and enhancing quality management in the construction of buildings. Blockchain enables innovation through its key features, including decentralization, distributed storage, traceability, transparency, and security. It has already transformed industries such as supply chains, the construction industry, and environmental management. By improving visibility and traceability, blockchain helps track product processes with greater transparency [58].

Sustainable construction is more complex than traditional ones, requiring a secure and transparent system to trace different issues during construction. Blockchain provides a tamper-resistant infrastructure that records and tracks data using IoT sensors. This allows seamless monitoring of materials, machines, and operators from production to construction sites. As a reliable digital record-keeping system, blockchain eliminates the limitations of traditional quality traceability models, such as centralization and data tampering. Its features enhance accountability and strengthen quality management in the construction industry [59].

Blockchain is a form of Distributed Ledger Technology (DLT) that securely records encrypted data in connected “blocks” across a network. It continuously expands as new blocks are added.

Blockchain can enhance supply chain management, project bidding, contract handling, and permit management. By providing a secure and transparent record-keeping system, it improves efficiency and accountability in construction processes. Despite its benefits, challenges exist, including the need for standardized data formats, stakeholder collaboration, and training. Scalability is also a concern, especially for large projects. Several studies have explored blockchain’s role in construction. Researchers have examined its applications, including smart contracts for dispute resolution and lifecycle tracking in buildings.

2.5. Hybrid Model Workflow

The workflow begins with collecting concrete composition data, which is recorded on the blockchain for security and verification. These data are then used by the XGBoost, Random Forest, and 1D-CNN to predict the compressive strength of low-carbon concrete. The predicted output is stored back in the blockchain, ensuring transparency and facilitating easy validation and auditing. A feedback loop is established, allowing new data from actual compressive strength tests to be added to the blockchain, further training and refining the ML model to improve its accuracy over time.

The proposed model integrates Machine Learning (ML), specifically XGBoost, Random Forest, and 1D-CNN, with Blockchain technology to predict and enhance the compressive strength of low-carbon concrete. This hybrid approach leverages the strengths of both technologies to ensure accurate predictions and secure, transparent data handling.

2.6. Data Feature and Selection

In Table A2 the dataset includes 752 concrete mixes analyzed for 11 parameters. Cement averages 257 kg/m³ (range: 0–611), with the supplementary materials showing high variability (slag: 29 kg/m³, fly ash: 163 kg/m³)—50% use no slag and 25% no fly ash, indicating sustainability potential. Aggregates (coarse: 672 kg/m³, fine: 805 kg/m³) form the bulk of mixes. Water averages 153 kg/m³ with minimal superplasticizer use (3 kg/m³). Half the mixes omit alkali activators, suggesting specialized use. Most samples (50%) were tested at standard 28-day curing, though ages varied (0.2–364 days). Compressive strength averages 38 MPa (range: 1.8–444 MPa), with 75% of mixes achieving 22–44 MPa. Correlation analysis reveals positive cement–water (0.49) and negative cement–slag (−0.43) relationships.

The correlation matrix plot in Figure 1 illustrates the relationships between various input features and the compressive strength of low-carbon concrete materials, using a color gradient to represent correlation coefficients ranging from −1 to 1. Diagonal cells, showing a correlation of 1.00, indicate a perfect correlation of each feature with itself. Correlation coefficients near 1 signify a strong positive relationship. Some key observations are Cement’s moderate positive correlation with Water (0.49) and a weak positive correlation with Compressive Strength (0.11). Blast Furnace Slag shows a moderate negative correlation with Cement (−0.43) and a weak negative correlation with Compressive Strength (−0.22). Fly Ash has a weak negative correlation with Compressive Strength (−0.15), suggesting a slight reduction in strength with increased proportion. Age demonstrates a relatively higher positive correlation with Compressive Strength (0.22), indicating that increased curing time enhances concrete strength. NaOH and Na₂SiO₃ exhibit weak correlations with most features and Compressive Strength, implying minimal linear impact on strength. The final column highlights the correlation of each input feature with Compressive Strength, with Age showing the highest positive correlation, underscoring its crucial role in strength development. Consequently, this correlation matrix offers valuable insights into the influence of different material components and curing time on the compressive strength of low-carbon concrete, aiding in optimizing material composition for improved performance.

3. Research Methodology

To address the first research question on improving prediction accuracy, we adopted a stacked ensemble model combining XGBoost, Random Forest, and a 1D CNN. This approach overcomes the limitations of traditional regression methods, such as weak generalization and poor handling of complex or noisy data [60,61]. XGBoost ensures high accuracy through gradient boosting, Random Forest adds robustness against overfitting, and 1D CNN captures latent patterns in structured data.

As shown below, we have the flow chart for system architecture diagram with machine learning and web application in Figure 2, which represents a system that begins with machine learning training, where 752 research mix designs are processed through an ensemble model (XGBoost, Random Forest, and 1D-CNN) to predict compressive strength and sustainability metrics. The trained model is serialized (.pkl), encrypted with AES-256, and hashed using SHA-256 for integrity verification.

Engineers then access the React/Next.js frontend, authenticating via Meta Mask wallet (Decentralized ID), which triggers smart contract verification on the Ethereum blockchain. Once authenticated, users upload encrypted models to IPFS storage, with the Content ID (CID) permanently recorded on-chain through the MixDesignRegistry smart contract. For predictions, the system verifies user permissions on-chain before processing requests through the Django backend. Model downloads require private key authentication, with all transactions immutably logged.

For predictions, users are validated on-chain using smart contract logic before the Django backend service processes the model inputs. The system ensures end-to-end privacy and access control by verifying user permissions through the blockchain and requiring private key-based authentication for secure downloads. All major interactions, including model uploads, predictions, and downloads, are permanently logged on the blockchain, maintaining a decentralized and auditable trail of events.

Additionally, a Key Management System (KMS), potentially involving a Hardware Security Module (HSM), supports secure handling of encryption keys and access policies, while a blockchain gateway facilitates the interaction between the web application and the Ethereum network. This architecture integrates machine learning, web technologies, and blockchain in a seamless and secure workflow optimized for engineering design predictions and digital asset integrity.

3.1. Machine Learning

The SmartMix Web3 framework integrates ensemble machine learning (EML) with blockchain technology to advance low-carbon concrete design by accurately predicting compressive strength and minimizing embodied carbon. A hybrid model—combining XGBoost, Random Forest, and a 1D Convolutional Neural Network (1D-CNN)—is trained on 752 historical mix designs to identify influential parameters such as the water–binder ratio and curing temperature. The workflow begins with blockchain-secured data acquisition, where mix parameters are tokenized to ensure traceability and integrity. A federated learning-inspired preprocessing stage is incorporated to preserve data privacy across distributed sources, supported by SHA-256 hashing to verify data integrity without exposing raw inputs. The ensemble model leverages XGBoost for feature importance, Random Forest for robustness against over fitting, and a 1D-CNN for capturing time-dependent curing patterns. This integrated approach addresses both research questions, enhancing prediction accuracy (RQ1) and improving data reliability (RQ2). Case studies from Cameroon demonstrate how decentralized, privacy-aware AI can support sustainable concrete practices while enabling transparent, auditable collaboration among stakeholders.

Building on this foundation, the SmartMix Web3 framework introduces a secure and transparent model deployment and usage protocol powered by blockchain and decentralized identity (DID) mechanisms. As illustrated in Figure 3, the workflow ensures that every model transaction is authenticated, encrypted, and verifiable.

The process begins when a user connects their MetaMask wallet to the web-based frontend. This triggers a request for authentication to the backend, which subsequently interacts with the blockchain to issue a signed authentication token. This token represents a validated digital identity and is used to ensure that only authorized users can submit or access predictive models.

Once authenticated, the user uploads the serialized machine learning model (.pkl file) along with the token. The backend verifies the token’s integrity and confirms the user’s permission by checking the blockchain registry. Upon successful verification, the .pkl file is stored in encrypted form, ensuring both confidentiality and integrity.

For model usage or result retrieval, such as downloading predictions or model files, the user initiates a download request. To unlock access, the user submits a private key, which is securely processed by the backend to decrypt the .pkl file. A download link is then generated and returned to the user through the frontend interface.

This secure pipeline facilitates a seamless interaction between users and the system while upholding the highest standards of confidentiality, authenticity, and transparency. Decentralized authentication is achieved through MetaMask, leveraging Decentralized Identifiers (DIDs) to ensure that only verified users can initiate model transactions. Once authorized, all model files are handled with robust encryption protocols, including AES-256, supported by a secure key management infrastructure. Every upload and download operation is recorded on the blockchain via smart contracts, providing a tamper-proof record of activity. This approach guarantees an immutable audit trail that reinforces trust, enables regulatory compliance, and supports collaboration among diverse stakeholders without compromising data ownership or security.

3.2. Blockchain Technology

Blockchain has increasingly been applied in the construction industry to address long-standing challenges related to transparency, traceability, and trust among stakeholders. Prior studies have demonstrated how blockchain can securely log transactions, verify the provenance of construction materials, and ensure compliance with environmental standards. For instance, Ref. [62] developed a conceptual framework in which blockchain enhances trust through immutable data sharing across project participants. Ref. [63] Conducted a systematic review highlighting blockchain’s application in supply chain transparency, where material origins, certifications, and deliveries are recorded in a tamper-resistant ledger. Similarly, Ref. [64] showed how blockchain facilitates transparent construction management by securely storing records related to contracts, material flows, and inspections. Ref. [65] Emphasized blockchain’s potential to support sustainability claims by validating environmental product data through decentralized smart contracts.

Building on these foundations, Figure 3 represents the system leveraging the Ethereum blockchain to ensure secure, tamper-proof operations through smart contract-mediated authentication, where users connect via Meta Mask to generate cryptographically signed tokens that verify their identity before uploading or accessing machine learning models. Once authenticated, all model transactions (uploads/downloads) are logged on-chain via IPFS hashes, enabling immutable traceability and dispute resolution faster than traditional methods. Private key encryption ensures only authorized users decrypt stored models, while tokenized Environmental Product Declarations (EPDs) automate carbon credit verification, creating a transparent, auditable framework for low-carbon concrete optimization.

4. Results and Discussion

To evaluate concrete strength development and ultimate capacity, this study employed an innovative health-monitoring approach using Reusable Smart Bolt (RSB) piezoelectric sensors for concrete condition assessment [66]. Building upon this foundation, we developed the SmartMix Web3 Application Interface (Figure 4A), an advanced blockchain-integrated concrete strength prediction system that combines machine learning (ML) with decentralized Web3 infrastructure to improve transparency and prediction accuracy. The other system enhances traditional non-destructive testing (NDT) methods, including ultrasonic pulse velocity and rebound number measurements for the compressive strength of concrete [63]. This research proceeds by processing key concrete mix parameters (cement content, fly ash percentage, superplasticizer dosage, etc.) to generate precise strength predictions (achieving 45 MPa compressive strength in validation tests). The platform features a secure framework where users can upload ML models (as .pkt files) with all prediction results permanently recorded on the blockchain through a custom SepolicETH token, ensuring tamper-proof performance verification. For enhanced security, SmartMix Web3 incorporates multi-wallet authentication (compatible with MetaMask and Coinbase) and private key-controlled data access, maintaining strict version control for all ML models with blockchain-certified accuracy metrics. This integration of experimental NDT methodologies with decentralized blockchain technology represents a significant advancement in concrete testing and quality assurance protocols.

The architecture illustrated in Figure 4B represents the operational backbone of the SmartMix Web3 framework, detailing the secure and decentralized flow of model management from authentication to encrypted storage and controlled access. The system begins with the user initiating a connection through MetaMask or Coinbase Wallet, thereby authenticating their decentralized identity (DID) via Web3 protocols. Once authenticated, the frontend sends an authorization request to the backend, which generates a signed token validated through interaction with the blockchain. This signed token not only confirms the user’s identity but also governs access rights across subsequent stages. Following successful authentication, the user proceeds to upload the trained machine learning model (in. pkl format) to the platform. This model, having already undergone rigorous validation and encryption, is submitted along with the signed token to the backend. Here, the token is verified and the model is securely stored in an encrypted form, using AES-256 encryption standards. This entire transaction is immutably recorded on the blockchain via a custom SepolicETH token, ensuring the integrity and traceability of every uploaded file.

Building upon this foundation of secure storage, the system also facilitates model retrieval and prediction result access through a similarly robust process. To retrieve prediction results from Figure 4B’s SmartMix Web3 System Architecture or re-access uploaded models, users must request a secure download. This triggers a multi-stage verification process, beginning with the submission of a private key. Once the backend confirms key authenticity, the encrypted .pkl file is decrypted, and a secure download link is generated for the user. Notably, all of these operations’ upload, storage, access, and download are transparently tracked on the blockchain, reinforcing auditability and eliminating the risk of tampering or unauthorized model use.

The incorporation of multi-wallet authentication and private key-controlled data access underscores the SmartMix Web3 Application Interface’s emphasis on secure, user-centric design. Combined with the system’s compatibility with decentralized storage (e.g., IPFS or AWS S3) and smart contracts for version control, this framework enables trustworthy, traceable, and reproducible predictions. When integrated with experimental Non-Destructive Testing (NDT) techniques, including ultrasonic pulse velocity and rebound hammer methods, the SmartMix Web3 Application Interface system represents a next-generation tool for real-time concrete quality monitoring and verification in both laboratory and field settings.

The integration of blockchain and ML in the SmartMix Web3 Application Interface enhances trust by immutably recording model performance via SepolicETH, preventing manipulation. Secure wallet authentication and encrypted downloads protect data integrity, while versioned model tracking supports iterative improvements. Future applications could include smart contract-based rewards for high-accuracy models, IoT integration for real-time data, and expansion into other construction materials. This system sets a precedent for decentralized, AI-driven engineering solutions by merging predictive analytics with Web3 security.

The SmartMix Web3 Application Interface delivers transformative business benefits for construction companies by integrating blockchain and AI to optimize low-carbon concrete design, as demonstrated in 61 Cameroonian projects achieving FCFA 24.3 million (≈USD 40,514) annual savings through material efficiency in Table A3. The system’s Ethereum-based architecture ensures tamper-proof tracking of mix designs via SepolicETH tokens (0.05 SepolicETH benchmarks), reducing disputes while improving prediction accuracy by 22% RMSE—directly cutting trial batch costs and accelerating projects. Its federated learning framework enables secure collaboration across batching plants (SHA-256 verified) while maintaining data privacy, reducing emissions, and enhancing ESG compliance. Version-controlled ML models with cryptographic authentication

(MetaMask/Coinbase) [67] simultaneously lower costs and carbon footprints. Beyond immediate operational gains, SmartMix Web3 System Architecture creates future revenue potential through smart contract rewards for high-accuracy models and positions adopters as innovators in sustainable construction tech, offering both financial advantages (24.3 million FCFA/year) and market differentiation in the Web3-enabled construction era.

Figure 5 is a histogram titled “Error Distribution,” which illustrates the distribution of prediction errors for a model predicting compressive strength. This plot provides insights into the spread and nature of the errors made by the model. The horizontal axis (X-axis) represents the prediction error, calculated as the difference between the predicted and actual values. The vertical axis (Y-axis) shows the frequency or count of occurrences for each error value. The histogram bars indicate the frequency of errors within specific ranges, while the overlaid Kernel Density Estimate (KDE) curve provides a continuous estimate of the error distribution.

The histogram reveals a peak near zero, indicating that most prediction errors are small, suggesting that the model’s predictions are accurate. The distribution appears approximately symmetric around zero, implying that the model does not consistently over-predict or under-predict. However, there are some errors on both negative and positive sides, showing that the model occasionally underestimates and overestimates the compressive strength. The distribution’s tails extend further on the positive side, indicating a few instances where the model significantly overestimated the actual values.

The distribution’s symmetry around the zero mark suggests that there is no consistent bias in the model’s predictions; it neither systematically over-predicts nor under-predicts the compressive strength. This balance is crucial for maintaining the reliability of predictions across different scenarios.

Despite the general accuracy, the presence of errors on both the negative and positive sides reveals that the model occasionally misjudges the compressive strength, either underestimating or overestimating it. This variability is a common challenge in predictive modeling.

Notably, the tails of the distribution extend further on the positive side. This asymmetry indicates that while underestimations are relatively controlled, there are a few instances where the model significantly overestimates the actual values. These outliers suggest that while the model performs well overall, there is potential for refinement to address these rare but impactful overestimations, enhancing the model’s precision and reliability.

While the proposed blockchain-integrated framework demonstrates significant promise for enhancing transparency, security, and environmental accountability in low-carbon concrete optimization, several limitations should be acknowledged. First, the current implementation has been specifically tailored for geopolymer and fly-ash-based concrete formulations. The applicability of the method to other concrete types, such as high-performance concrete, fiber-reinforced composites, or alternative cementitious materials, may require adaptation of the machine learning models and validation data to capture different material behaviors and environmental impacts.

Second, the current blockchain implementation utilizes the Ethereum network and MetaMask for user authentication and transaction logging. While this provides a secure and decentralized infrastructure, it introduces scalability and accessibility challenges, particularly in environments with high transaction volumes or limited internet connectivity, such as remote construction zones. Notably, transaction fees (gas costs) and network latency may impair real-time responsiveness and increase operational costs.

To address these concerns, future versions of the SmartMix Web3 framework will explore integration with Layer 2 scaling solutions (e.g., Polygon, Optimism) that significantly reduce gas costs and enhance throughput while maintaining compatibility with Ethereum [68]. Additionally, the adoption of lightweight, permissioned blockchain platforms like Hyperledger Fabric or Quorum may offer better performance in constrained environments by allowing offline consensus, faster transaction processing, and greater control over infrastructure deployment [69].

Third, while the tokenization of Environmental Product Declarations (EPDs) effectively automates carbon credit verification, it currently presumes harmonized regulatory standards and consistent carbon accounting methodologies. In practice, regional variation in certification protocols and environmental regulations could limit interoperability, requiring local customization and smart contract adaptation.

Finally, although the use of private key encryption safeguards model data and transaction integrity, it places the responsibility of key management on the end-user, which could introduce vulnerabilities in the event of key loss or compromise. To mitigate this, future work will explore decentralized identity management (DID) and secure key recovery mechanisms.

Despite these limitations, the proposed system presents a flexible and extensible architecture that can be tailored to various construction materials, contexts, and regulatory environments. Future research will focus on increasing dataset diversity, improving blockchain scalability and decentralization strategies, and adapting the platform for multi-material lifecycle assessment.

Table A4 presents data from 61 real-world concrete mix experiments conducted in Cameroon using traditional construction methods. These experiments were designed to reflect typical on-site practices in developing countries, where concrete is often prepared manually with minimal optimization. The mix ratios explored included common proportions such as 1:2:4, 1:3:3, and 1:1.5:3, with water-to-cement (w/c) ratios varying between 0.5 and 0.80. Each experiment detailed the quantities of materials used, water, cement, river sand, quarry sand, and two grades of gravel (5/15 mm and 15/25 mm), alongside the compressive strength achieved after 28 days, the total material weight per mix, and the total cost in FCFA.

A critical analysis of the data reveals that the total amount of cement used across the 61 experiments was approximately 21,350 kg. This figure was calculated by summing the individual cement usage in each mix, most of which used either 350 kg or 400 kg per batch. Given that the current market price of cement in Cameroon averages FCFA 100 per kilogram, the total cost of cement alone during these traditional experiments amounts to FCFA 2,135,000. This significant expense represents not only a financial burden but also an environmental concern.

Cement is one of the largest contributors to carbon emissions in the construction industry. For every kilogram of cement produced, approximately 0.9 kg of CO₂ is emitted into the atmosphere. Based on this standard, the total CO₂ emissions generated from cement use in these 61 experiments amount to roughly 19,215 kg (or 19.2 metric tons) of carbon dioxide. These emissions are particularly concerning, considering that the mixes were not optimized for performance, leading to material waste and inconsistent strength results. For instance, despite using the same mix ratio and material quantities, the compressive strength varied significantly across mixes, ranging from as high as 25.8 MPa to as low as 19.4 MPa. This performance inconsistency highlights the inefficiency and unpredictability of traditional methods.

In contrast, the SmartMix Web3 framework addresses these inefficiencies through the integration of ensemble machine learning and blockchain technology. Unlike the 61 traditional mixes, a dataset of 752 optimized mix designs was used to train and validate SmartMix Web3, resulting in a predictive model that reduces RMSE in compressive strength prediction by 22% and MAE in embodied carbon estimation by 29% compared to conventional regression and simulation methods. The system’s deployment ensures tamper-proof data logging through blockchain and encourages decentralized collaboration among batching plants using federated learning and SHA-256 encryption.

The societal impact of adopting the SmartMix Web3 system is profound. First, by reducing unnecessary cement use through precision prediction, the method directly contributes to cost savings, demonstrated by annual savings of FCFA 24.3 million in material selection. Second, the environmental impact is substantially lowered, with a reduction of 99.87 kg of CO₂ per cubic meter of concrete. If implemented on a national scale, this could translate into tens of thousands of metric tons of CO₂ avoided annually. Furthermore, SmartMix Web3 promotes data transparency, stakeholder collaboration, and policy compliance by enabling smart contract–driven validation of environmental declarations.

In conclusion, Appendix A illustrates the real costs, both financial and environmental, of traditional concrete practices in Cameroon. The total cement usage of 21.35 tons and the resulting 19.2 tons of CO₂ emissions from just 61 tests underline the urgent need for change. SmartMix Web3 emerges as a practical, scalable solution, combining technological innovation with tangible societal benefits. By embracing this method, construction sectors in developing countries can move toward sustainable development, reduce carbon footprints, and improve economic efficiency.

5. Conclusions

SmartMix Web3 integrates blockchain and AI to optimize low-carbon concrete, combining accurate machine learning predictions with secure, transparent Web3 infrastructure. The system enables engineers to submit material inputs and receive verified strength predictions with cryptographic proof, while tracking model accuracy via blockchain tokens. Features like private key authentication, multi-wallet support, and immutable version control ensure trust and prevent data manipulation. Validated by 61 Cameroonian experiments, SmartMix Web3 reduces carbon emissions and prediction costs, improving accuracy by 22% RMSE. This scalable framework advances sustainable construction by merging materials science with decentralized technology. Future research will concentrate on enhancing the global adaptability of the proposed framework by facilitating cross-chain blockchain interoperability, expanding and diversifying material databases to support broader applicability, and incorporating advanced artificial intelligence methodologies to further improve predictive accuracy and promote sustainability in concrete design.

Author Contributions

Conceptualization, Z.E.N. and Y.-H.L.; Formal analysis, Z.E.N. and Y.-H.L.; Investigation, Z.E.N.; Methodology, Z.E.N. and Y.-H.L.; Project administration, Y.-H.L.; Software, Z.E.N.; Supervision, Y.-H.L.; Writing—original draft, Z.E.N.; Writing—review & editing, and Y.-H.L. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

Data will be made available upon request.

Conflicts of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures

Figure 1 Correlation plot for compressive strength of low carbon concrete material.

Figure 2 System architecture diagram with machine learning and web application.

Figure 3 Blockchain integration request authentication.

Figure 4 (A) SmartMix Web3 Application Interface. (B) SmartMix Web3 System Architecture.

Figure 5 Error distribution in predicted compressive strength.

Appendix A

Table A1

Concrete blockchain abbreviations.

Abbreviation	Full Term	Description	Measurement Unit
w/c	Water–Cement Ratio	Ratio of water to cement in concrete mix	Decimal (0.5–0.8)
CPJ	Cement Grade	Portland cement classification (CPJ 35 or CPJ 42.5 used in study)	-
W	Water	Total water content in mix	kg/m3
C	Cement	Total cement content	kg/m3
R Sand	River Sand	Fine aggregate sourced from rivers	kg/m3
Q Sand	Quarry Sand	Fine aggregate sourced from quarries	kg/m3
G 5/15	Gravel 5–15 mm	Coarse aggregate (5–15 mm particle size)	kg/m3
G 15/25	Gravel 15–25 mm	Coarse aggregate (15–25 mm particle size)	kg/m3
EPD	Environmental Product Declaration	Tokenized sustainability credential for carbon tracking
DID	Decentralized Identifier	Blockchain-based user authentication standard
IPFS	Interplanetary File System	Distributed storage for model files
HSM	Hardware Security Module	Secure hardware for private key management

Table A2

Statistics analysis of proposed algorithm.

Parameter	Cement	Blast	Fly Ash	Water	SuperPlasticizer	Coarse Aggregate	Fine Aggregate	Na2SiO3	NAOH	Age	Compressive Strength
Count	752	752	752	752	752	752	752	752	752	752	752
Mean	257.3103	29.38705	162.7484	153.3245	3.140528	671.9908	804.9908	6.074035	9.745806	47.22367	37.91569
STD	164.9837	103.7698	183.3163	71.36635	5.310532	512.863	390.8386	23.28694	60.72815	70.27353	39.86671
Min	0	0	0	0	0	0	0	0	0	0.2	1.76
25%	120	0	0	132	0	0	606	0	0	7	21.9175
50%	293.6	0	120	170	0.46	932	735	0	0	28	32.595
75%	380	0	240	194	3.3325	1099.498	868	0	0	56	44.125
Max	611	900	1245	242.5	23.6	1370	1603	194	684.8	364	444

Table A3

Carbon reductions.

Ratio	N0 of Bags	Total Cost (FCFA)	Cost in USD	CO2 Emission (8%)
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:1.5:3	1.50	5000.00	8.33	0.67
1:1.5:3	1.50	5000.00	8.33	0.67
1:1.5:3	1.30	5000.00	8.33	0.67
1:1.5:3	1.50	5000.00	8.33	0.67
1:1.5:3	1.50	5000.00	8.33	0.67
1:1.5:3	1.50	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:2:4	2.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
1:3:3	3.00	5000.00	8.33	0.67
				40.67

Table A4

Annual concrete strength predictions and cost savings analysis.

S/N	CPJ	MIX RATIO	w/c	W (kg)	C (kg)	R SAND (kg)	Q SAND (kg)	G 5/15 (kg)	G 15/25 (kg)	FCJ = 28 (Mpa)	Materials Sum	Total Cost (FCFA)
1	42.5	1:2:4	0.5–0.55	1820	350	640	0.00	480	720	24.1	4034.1	403,410
2	42.5	1:2:4	0.56–0.6	1750	350	640	0.00	480	720	24	3964.0	396,400
3	42.5	1:2:4	0.61–0.65	1800	350	640	0.00	480	720	23.9	4013.9	401,390
4	42.5	1:2:4	0.66–0.70	1820	350	640	0.00	480	720	23.8	4033.8	403,380
5	42.5	1:2:4	0.71–0.75	1750	350	640	0.00	480	720	22.8	3962.8	396,280
6	42.5	1:2:4	0.76–0.80	1750	350	640	0.00	480	720	22.1	3962.1	396,210
7	42.5	1:2:4	0.5–0.55	1820	350	320	320.00	480	720	24.5	4034.5	403,450
8	42.5	1:2:4	0.56–0.6	1750	350	320	320.00	480	720	24.35	3964.35	396,435
9	42.5	1:2:4	0.61–0.65	1800	350	320	320.00	480	720	24.2	4014.2	401,420
10	42.5	1:2:4	0.66–0.70	1820	350	320	320.00	480	720	24	4034.0	403,400
11	42.5	1:2:4	0.71–0.75	1750	350	320	320.00	480	720	23.2	3963.2	396,320
12	42.5	1:2:4	0.76–0.80	1750	350	320	320.00	480	720	22.5	3962.5	396,250
13	42.5	1:2:4	0.5–0.55	1820	350	256	384.00	480	720	24.6	4034.6	403,460
14	42.5	1:2:4	0.56–0.6	1750	350	256	384.00	480	720	24.2	3964.2	396,420
15	42.5	1:2:4	0.61–0.65	1800	350	256	384.00	480	720	23.8	4013.8	401,380
16	42.5	1:2:4	0.66–0.70	1820	350	256	384.00	480	720	23.3	4033.3	403,330
17	42.5	1:2:4	0.71–0.75	1750	350	256	384.00	480	720	22.6	3962.6	396,260
18	42.5	1:2:4	0.76–0.80	1820	350	256	384.00	480	720	22.1	4032.1	403,210
19	42.5	1:3:3	0.5–0.55	1750	350	640	0.00	480	720	22.5	3962.5	396,250
20	42.5	1:3:3	0.56–0.6	1820	350	640	0.00	480	720	22.1	4032.1	403,210
21	42.5	1:3:3	0.61–0.65	1750	350	640	0.00	480	720	21.6	3961.6	396,160
22	42.5	1:3:3	0.66–0.70	1750	350	640	0.00	480	720	21.5	3961.5	396,150
23	42.5	1:3:3	0.71–0.75	1820	350	640	0.00	480	720	20.5	4030.5	403,050
24	42.5	1:3:3	0.76–0.80	1820	350	640	0.00	480	720	20	4030.0	403,000
25	42.5	1:3:3	0.5–0.55	1750	350	320	320.00	480	720	22.7	3962.7	396,270
26	42.5	1:3:3	0.56–0.6	1750	350	320	320.00	480	720	22.3	3962.3	396,230
27	42.5	1:3:3	0.61–0.65	1820	350	320	320.00	480	720	21.7	4031.7	403,170
28	42.5	1:3:3	0.66–0.70	1750	350	320	320.00	480	720	21.1	3961.1	396,110
29	42.5	1:3:3	0.71–0.75	1800	350	320	320.00	480	720	20.5	4010.5	401,050
30	42.5	1:3:3	0.76–0.80	1820	350	320	320.00	480	720	19.8	4029.8	402,980
31	42.5	1:3:3	0.5–0.55	1750	350	256	384.00	480	720	22.9	3962.9	396,290
32	42.5	1:3:3	0.56–0.6	1750	350	256	384.00	480	720	22.5	3962.5	396,250
33	42.5	1:3:3	0.61–0.65	1820	350	256	384.00	480	720	22	4032.0	403,200
34	42.5	1:3:3	0.66–0.70	1750	350	256	384.00	480	720	21.4	3961.4	396,140
35	42.5	1:3:3	0.71–0.75	1800	350	256	384.00	480	720	20.8	4010.8	401,080
36	42.5	1:3:3	0.76–0.80	1820	350	256	384.00	480	720	20.1	4030.1	403,010
37	42.5	1:1.5:3	0.5–0.55	1750	400	590	0.00	450	700	25.8	3915.8	391,580
38	42.5	1:1.5:3	0.56–0.60	1750	400	590	0.00	450	700	24.5	3914.5	391,450
39	42.5	1:1.5:3	0.61–0.65	1720	400	590	0.00	450	700	25	3885.0	388,500
40	42.5	1:1.5:3	0.66–0.70	1720	400	590	0.00	450	700	24.4	3884.4	388,440
41	42.5	1:1.5:3	0.71–0.75	1720	400	590	0.00	450	700	23.9	3883.9	388,390
42	42.5	1:1.5:3	0.76–0.80	1720	400	590	0.00	450	700	23.1	3883.1	388,310
43	35	1:2:4	0.5–0.55	1750	350	640	0.00	480	720	22.6	3962.6	396,260
44	35	1:2:4	0.56–0.60	1750	350	640	0.00	480	720	22.1	3962.1	396,210
45	35	1:2:4	0.61–0.65	1820	350	640	0.00	480	720	21.7	4031.7	403,170
46	35	1:2:4	0.66–0.70	1750	350	640	0.00	480	720	21.2	3961.2	396,120
47	35	1:2:4	0.71–0.75	1800	350	640	0.00	480	720	20.5	4010.5	401,050
48	35	1:2:4	0.76–0.80	1820	350	640	0.00	480	720	19.4	4029.4	402,940
49	35	1:3:3	0.5–0.55	1750	350	640	0.00	480	720	22.6	3962.6	396,260
50	35	1:3:3	0.56–0.60	1750	350	640	0.00	480	720	22.2	3962.2	396,220
51	35	1:3;3	0.61–0.65	1820	350	640	0.00	480	720	21.7	4031.7	403,170
52	35	1:3:3	0.66–0.70	1750	350	640	0.00	480	720	21.2	3961.2	396,120
53	35	1:3:3	0.71–0.75	1800	350	640	0.00	480	720	20.5	4010.5	401,050
54	35	1:3:3	0.76–0.80	1820	350	640	0.00	480	720	20	4030.0	403,000
											21,516,245	35,860.22

References

1. Jacobs, A.; Hitch, M. Experimental mineral carbonation: Approaches to accelerate CO₂ sequestration in mine waste materials. Int. J. Min. Reclam. Environ.; 2011; 25, pp. 321-331. [DOI: https://dx.doi.org/10.1080/17480930.2011.608883]

2. Ferdaus, M.; Dam, T.; Anavatti, S.; Das, S. Digital technologies for a net-zero energy future: A comprehensive review. Renew. Sustain. Energy Rev.; 2024; 202, 114681. [DOI: https://dx.doi.org/10.1016/j.rser.2024.114681]

3. Li, X.; Zhang, L.; Tang, Y.; Chen, Q.; Sun, W.; Fang, X.; Tao, Y.; Shang, B. Recent Advances in Machine Learning for Building Envelopes: From Prediction to Optimization. SRN Electron. J.; 2025; Available online: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5079321 (accessed on 20 July 2025).

4. Schmidt, J.; Marques, M.R.G.; Botti, S.; Marques, M.A.L. Recent advances and applications of machine learning in materials science. Comput. Mater. Sci.; 2019; 156, pp. 443-458.

5. Li, J.; Ni, W.; Wang, X.; Zhu, S.; Wei, X.; Jiang, F.; Zeng, H.; Hitch, M. Mechanical activation of medium basicity steel slag under dry condition for carbonation curing. J. Build. Eng.; 2022; 50, 104123. [DOI: https://dx.doi.org/10.1016/j.jobe.2022.104123]

6. Xi, C.; Cao, S.-J. Challenges and Future Development Paths of Low Carbon Building Design: A Review. Buildings; 2022; 12, 163. [DOI: https://dx.doi.org/10.3390/buildings12020163]

7. Chau, C.-K.; Leung, T.M.; Ng, W.Y. A Review on Life Cycle Assessment, Life Cycle Energy Assessment and Life Cycle Carbon Emissions Assessment on buildings. Appl. Energy; 2015; 143, pp. 395-413. [DOI: https://dx.doi.org/10.1016/j.apenergy.2015.01.023]

8. Mirmozaffari, M.; Yazdani, M.; Boskabadi, A.; Dolatsara, H.A.; Kabirifar, K.; Golilarz, N.A. A Novel Machine Learning Approach Combined with Optimization Models for Eco-efficiency Evaluation. Appl. Sci.; 2020; 10, 5210. [DOI: https://dx.doi.org/10.3390/app10155210]

9. Shamsabadi, E.A.; Roshan, N.; Hadigheh, S.A.; Nehdi, M.L.; Khodabakhshian, A.; Ghalehnovi, M. Machine learning-based compressive strength modelling of concrete incorporating waste marble powder. Constr. Build. Mater.; 2022; 324, 126592. [DOI: https://dx.doi.org/10.1016/j.conbuildmat.2022.126592]

10. Almansour, F.; Al-Bukhaiti, N. Predicting the strength of fiber reinforced polymer materials externally bonded to masonry using artificial intelligent techniques. Front. Struct. Civ. Eng.; 2025; 19, pp. 242-261. [DOI: https://dx.doi.org/10.1007/s11709-025-1136-0]

11. Wang, S.; Xia, P.; Chen, K.; Gong, F.; Wang, H.; Wang, Q.; Zhao, Y.; Jin, W. Prediction and optimization model of sustainable concrete properties using machine learning, deep learning and swarm intelligence: A review. J. Build. Eng.; 2023; 80, 108065. [DOI: https://dx.doi.org/10.1016/j.jobe.2023.108065]

12. Cao, G.; Li, Z. Numerical flow simulation of fresh concrete with viscous granular material model and smoothed particle hydrodynamics. Cem. Concr. Res.; 2017; 100, pp. 263-274. [DOI: https://dx.doi.org/10.1016/j.cemconres.2017.07.005]

13. Wang, H.; Xu, J.; Song, Y.; Hou, M.; Xu, Y. Development of low-cost engineered cementitious composites using Yellow River silt and unoiled PVA fiber. Constr. Build. Mater.; 2024; 425, 136063. [DOI: https://dx.doi.org/10.1016/j.conbuildmat.2024.136063]

14. Mortar, N.A.M.; Kamarudin, H.; Rafiza, R.A.; Meor, T.; Rosnita, M. Compressive Strength of Fly Ash Geopolymer Concrete by Varying Sodium Hydroxide Molarity and Aggregate to Binder Ratio. IOP Conference Series: Materials Science and Engineering, Proceedings of the 2nd Joint Conference on Green Engineering Technology & Applied Computing, Bangkok, Thailand, 4–5 February 2020; IOP Publishing: Bristol, UK, 2020; Volume 864.

15. Sonebi, M.; Cevik, A.; Grünewald, S.; Walraven, J. Modelling the fresh properties of self-compacting concrete using support vector machine approach. Constr. Build. Mater.; 2016; 106, pp. 55-64. [DOI: https://dx.doi.org/10.1016/j.conbuildmat.2015.12.035]

16. Ahmad, S.; Umar, A.; Masood, A. An appraisal of the models for the prediction of compressive strength of thermally damaged self-compacting concrete. Mater. Today Proc.; 2022; 65, pp. 478-483. [DOI: https://dx.doi.org/10.1016/j.matpr.2022.03.022]

17. Onyelowe, K.C.; Ebid, A.M.; Riofrio, A.; Soleymani, A.; Baykara, H.; Kontoni, D.-P.N.; Mahdi, H.A.; Jahangir, H. Global warming potential-based life cycle assessment and optimization of the compressive strength of fly ash-silica fume concrete; environmental impact consideration. Front. Built Environ.; 2022; 8, 992552. [DOI: https://dx.doi.org/10.3389/fbuil.2022.992552]

18. Onyelowe, K.C.; Ebid, A.M.; Riofrio, A.; Baykara, H.; Soleymani, A.; Mahdi, H.A.; Jahangir, H.; Ibe, K. Multi-Objective Prediction of the Mechanical Properties and Environmental Impact Appraisals of Self-Healing Concrete for Sustainable Structures. Sustainability; 2022; 14, 9573. [DOI: https://dx.doi.org/10.3390/su14159573]

19. Onyelowe, K.C.; Kontoni, D.-P.N.; Ebid, A.M.; Dabbaghi, F.; Soleymani, A.; Jahangir, H.; Nehdi, M.L. Multi-Objective Optimization of Sustainable Concrete Containing Fly Ash Based on Environmental and Mechanical Considerations. Buildings; 2022; 12, 948. [DOI: https://dx.doi.org/10.3390/buildings12070948]

20. Onyelowe, K.C.; Ebid, A.M.; Baldovino, J.d.J.A.; E Onyia, M. Hydraulic conductivity predictive model of RHA-ameliorated laterite for solving landfill liner leachate, soil and water contamination and carbon emission problems. Int. J. Low-Carbon. Technol.; 2022; 17, pp. 1134-1144. [DOI: https://dx.doi.org/10.1093/ijlct/ctac077]

21. Onyelowe, K.C.; Ebid, A.M.; Onyia, M.E.; Amanamba, E.C. Estimating the swelling potential of non-carbon–based binder (NCBB)-treated clayey soil for sustainable green subgrade using AI (GP, ANN, and EPR) techniques. Int. J. Low-Carbon Technol.; 2022; 17, pp. 807-815. [DOI: https://dx.doi.org/10.1093/ijlct/ctac058]

22. Onyelowe, K.C.; Tome, S.; Ebid, A.M.; Usungedo, T.; Van, D.B.; Etim, R.K.; Onuoha, I.C.; Attah, I.C. Effect of desiccation on ashcrete (HSDA)-treated soft soil used as flexible pavement foundation: Zero carbon stabilizer approach. Int. J. Low-Carbon. Technol.; 2022; 17, pp. 563-570. [DOI: https://dx.doi.org/10.1093/ijlct/ctac042]

23. Onyelowe, K.C.; Gnananandarao, T.; Ebid, A.M.; Mahdi, H.A.; Ghadikolaee, M.R.; Al-Ajamee, M. Evaluating the Compressive Strength of Recycled Aggregate Concrete Using Novel Artificial Neural Network. Civ. Eng. J.; 2022; 8, pp. 1679-1693. [DOI: https://dx.doi.org/10.28991/CEJ-2022-08-08-011]

24. Alyhya, W.; Kulasegaram, S.; Karihaloo, B. Simulation of the flow of self-compacting concrete in the V-funnel by SPH. Cem. Concr. Res.; 2017; 100, pp. 47-59. [DOI: https://dx.doi.org/10.1016/j.cemconres.2017.05.021]

25. Atwoli, L.; Erhabor, G.E.; Gbakima, A.A.; Haileamlak, A.; Kayembe Ntumba, J.M.; Kigera, J.; Laybourn-Langton, L.; Mash, B.; Muhia, J.; Zielinski, C. . COP27 Climate Change Conference: Urgent action needed for Africa and the world: Wealthy nations must step up support for Africa and vulnerable countries in addressing past, present and future impacts of climate change. Int. J. Low-Carbon Technol.; 2022; 62, ezac492. [DOI: https://dx.doi.org/10.1111/nin.12532] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36268838]

26. Mahjoubi, S.; Barhemat, R.; Meng, W.; Bao, Y. AI-guided auto-discovery of low-carbon cost-effective ultra-high performance concrete (UHPC). Resour. Conserv. Recycl.; 2023; 189, 106741. [DOI: https://dx.doi.org/10.1016/j.resconrec.2022.106741]

27. Peng, X.; Tao, Y.; Wang, C.; Zhong, Z. Research on low-carbon supply chain decision-making under carbon cap-and-trade and subsidy policy. Int. J. Low-Carbon. Technol.; 2022; 17, pp. 1061-1074. [DOI: https://dx.doi.org/10.1093/ijlct/ctac086]

28. Peng, X.; Tao, Y.; Wang, C.; Zhong, Z. Research on low-carbon supply chain decision-making under different incentive models. Int. J. Low-Carbon. Technol.; 2022; 17, pp. 696-709. [DOI: https://dx.doi.org/10.1093/ijlct/ctac051]

29. Zhou, Y.; Wang, L.; Chen, Z.; Li, F. Influence of recycled plastic and fly ash fillers on the performance of porous asphalt mixtures. Constr. Build. Mater.; 2023; 369, 130473. [DOI: https://dx.doi.org/10.1016/j.conbuildmat.2023.130473]

30. DeRousseau, M.; Kasprzyk, J.; Srubar, W. Computational design optimization of concrete mixtures: A review. Cem. Concr. Res.; 2018; 109, pp. 42-53. [DOI: https://dx.doi.org/10.1016/j.cemconres.2018.04.007]

31. Young, B.A.; Hall, A.; Pilon, L.; Gupta, P.; Sant, G. Can the compressive strength of concrete be estimated from knowledge of the mixture proportions?: New insights from statistical analysis and machine learning methods. Cem. Concr. Res.; 2019; 115, pp. 379-388. [DOI: https://dx.doi.org/10.1016/j.cemconres.2018.09.006]

32. Provis, J. A concrete future. Commun. Eng.; 2022; 1, 32. [DOI: https://dx.doi.org/10.1038/s44172-022-00026-3]

33. Kocak, Y.; Nas, S. The effect of using fly ash on the strength and hydration characteristics of blended cements. Constr. Build. Mater.; 2014; 73, pp. 25-32. [DOI: https://dx.doi.org/10.1016/j.conbuildmat.2014.09.048]

34. Zhang, J.; Xu, J.; Liu, C.; Zheng, J. Prediction of Rubber Fiber Concrete Strength Using Extreme Learning Machine. Front. Mater.; 2021; 7, 582635. [DOI: https://dx.doi.org/10.3389/fmats.2020.582635]

35. Zou, Z.P. Prediction of elastic modulus of ordinary and high-strength concrete based on machine learning. Model. Simul.; 2024; 13, 2591. [DOI: https://dx.doi.org/10.12677/mos.2024.133236]

36. Hussain, F.; Khan, S.A.; Khushnood, R.A.; Hamza, A.; Rehman, F. Machine Learning-Based Predictive Modeling of Sustainable Lightweight Aggregate Concrete. Sustainability; 2022; 15, 641. [DOI: https://dx.doi.org/10.3390/su15010641]

37. Siddique, R.; Aggarwal, P.; Aggarwal, Y. Prediction of compressive strength of self-compacting concrete containing bottom ash using artificial neural networks. Adv. Eng. Softw.; 2011; 42, pp. 780-786. [DOI: https://dx.doi.org/10.1016/j.advengsoft.2011.05.016]

38. Zhang, W.; Lee, D.; Lee, J.; Lee, C. Residual strength of concrete subjected to fatigue based on machine learning technique. Struct. Concr.; 2022; 23, pp. 2274-2287. [DOI: https://dx.doi.org/10.1002/suco.202100082]

39. Al-Shamiri, A.K.; Yuan, T.-F.; Kim, J.H. Non-Tuned Machine Learning Approach for Predicting the Compressive Strength of High-Performance Concrete. Materials; 2020; 13, 1023. [DOI: https://dx.doi.org/10.3390/ma13051023]

40. Mahjoob, M.; Fazeli, S.S.; Milanlouei, S.; Tavassoli, L.S.; Mirmozaffari, M. A modified adaptive genetic algorithm for multi-product multi-period inventory routing problem. Sustain. Oper. Comput.; 2022; 3, pp. 1-9. [DOI: https://dx.doi.org/10.1016/j.susoc.2021.08.002]

41. Mirmozaffari, M.; Shadkam, E.; Khalili, S.M.; Kabirifar, K.; Yazdani, R.; Gashteroodkhani, T.A. A novel artificial intelligent approach: Comparison of machine learning tools and algorithms based on optimization DEA Malmquist productivity index for eco-efficiency evaluation. Int. J. Energy Sect. Manag.; 2021; 15, pp. 523-550. [DOI: https://dx.doi.org/10.1108/IJESM-02-2020-0003]

42. Jin, L.; Duan, J.; Fan, T.; Jiao, P.; Dong, T.; Wu, Q. Using GA—BP Coupling Algorithm to Predict the High-performance Concrete Mechanical Property. KSCE J. Civ. Eng.; 2023; 27, pp. 684-697. [DOI: https://dx.doi.org/10.1007/s12205-022-0912-9]

43. Arbaoui, A.; Ouahabi, A.; Jacques, S.; Hamiane, M. Concrete Cracks Detection and Monitoring Using Deep Learning-Based Multiresolution Analysis. Electronics; 2021; 10, 1772. [DOI: https://dx.doi.org/10.3390/electronics10151772]

44. Zhang, F.; Wen, B.; Niu, D.; Li, A.; Guo, B. Optimized Design of Low-Carbon Mix Ratio for Non-Dominated Sorting Genetic Algorithm II Concrete Based on Genetic Algorithm-Improved Back Propagation. Materials; 2024; 17, 4077. [DOI: https://dx.doi.org/10.3390/ma17164077]

45. Islam, N.; Kashem, A.; Das, P.; Ali, N.; Paul, S. Prediction of high-performance concrete compressive strength using deep learning techniques. Asian J. Civ. Eng.; 2024; 25, pp. 327-341. [DOI: https://dx.doi.org/10.1007/s42107-023-00778-z]

46. Farshadfar, Z.; Mucha, T.; Tanskanen, K. Leveraging Machine Learning for Advancing Circular Supply Chains: A Systematic Literature Review. Logistics; 2024; 8, 108. [DOI: https://dx.doi.org/10.3390/logistics8040108]

47. Tan, S.; Liu, G. A low carbon management model for regional energy economies based on blockchain technology. Heliyon; 2023; 9, e19966. [DOI: https://dx.doi.org/10.1016/j.heliyon.2023.e19966]

48. Jain, A.; Sharma, A.; Jately, V.; Azzopardi, B. Sustainable Energy Solutions with Artificial Intelligence, Blockchain Technology, and Internet of Things; CRC Press: Boca Raton, FL, USA, 2023.

49. Scott, D.J.; Broyd, T.; Ma, L. Exploratory literature review of blockchain in the construction industry. Autom. Constr.; 2021; 132, 103914. [DOI: https://dx.doi.org/10.1016/j.autcon.2021.103914]

50. Kiu, M.S.; Chia, F.C.; Wong, P.F. Exploring the potentials of blockchain application in construction industry: A systematic review. Int. J. Constr. Manag.; 2022; 22, pp. 2931-2940. [DOI: https://dx.doi.org/10.1080/15623599.2020.1833436]

51. Li, Z.; Yoon, J.; Zhang, R.; Rajabipour, F.; Srubar, W.V., III; Dabo, I.; Radlińska, A. Machine learning in concrete science: Applications, challenges, and best practices. NPJ Comput. Mater.; 2022; 8, 127. [DOI: https://dx.doi.org/10.1038/s41524-022-00810-x]

52. Kumar, M.; Samui, P.; Kumar, D.R.; Asteris, P.G. State-of-the-art XGBoost, RF and DNN based soft-computing models for PGPN piles. Géoméch. Geoengin.; 2024; 19, pp. 975-990. [DOI: https://dx.doi.org/10.1080/17486025.2024.2337702]

53. Asteris, P.G.; Lemonis, M.E.; Le, T.-T.; Tsavdaridis, K.D. Evaluation of the ultimate eccentric load of rectangular CFSTs using advanced neural network modeling. Eng. Struct.; 2021; 248, 113297. [DOI: https://dx.doi.org/10.1016/j.engstruct.2021.113297]

54. Omran, B.A.; Chen, Q.; Jin, R. Comparison of Data Mining Techniques for Predicting Compressive Strength of Environmentally Friendly Concrete. J. Comput. Civ. Eng.; 2016; 30, 04016029. [DOI: https://dx.doi.org/10.1061/(ASCE)CP.1943-5487.0000596]

55. Breiman, L. Random forests. Mach. Learn.; 2001; 45, pp. 5-32. [DOI: https://dx.doi.org/10.1023/A:1010933404324]

56. Redmond, A.; Hore, A.; Alshawi, M.; West, R. Exploring how information exchanges can be enhanced through Cloud BIM. Autom. Constr.; 2012; 24, pp. 175-183. [DOI: https://dx.doi.org/10.1016/j.autcon.2012.02.003]

57. Suichies, B. 2015. Why Blockchain Must Die in 2016. Available online: https://medium.com/@bsuichies/why-blockchain-must-die-in-2016-e992774c03b4 (accessed on 20 July 2025).

58. Scott, B. How Can Cryptocurrency and Blockchain Technology Play a Role in Building Social and Solidarity Finance; Working Paper No. 2016-1 UNRISD: Geneva, Switzerland, 2016.

59. Kakavand, H.; Kost De Sevres, N.; Chilton, B. The Blockchain Revolution: An Analysis of Regulation and Technology Related to Distributed Ledger Technologies. 2017. Available online: https://ssrn.com/abstract=2849251 (accessed on 20 July 2025).

60. Motlagh, S.A.T.; Naghizadehrokni, M. An Extended Multi-Model Regression Approach for Compressive Strength Prediction and Optimization of Concrete Mixtures. Appl. Soft Comput.; 2021; 112, 107777. [DOI: https://dx.doi.org/10.1016/j.conbuildmat.2022.126828]

61. Javed, M.F.; Ali, H.S.; Khan, A. Machine Learning–Based Prediction of Waste Foundry Sand Concrete Properties Using Hybrid SVR Optimization Techniques. Constr. Build. Mater.; 2024; 369, 130473.

62. Li, J.; Greenwood, D.; Kassem, M. Blockchain in the built environment: Analysing current applications and developing an emergent framework. Autom. Constr.; 2019; 102, pp. 288-307. [DOI: https://dx.doi.org/10.1016/j.autcon.2019.02.005]

63. Shojaei, A.; Zuo, J. Application of blockchain technology in the construction industry: A systematic review. Buildings; 2021; 11, 334.

64. Turk, Ž.; Klinc, R. Potentials of Blockchain Technology for Construction Management. Procedia Eng.; 2017; 196, pp. 638-645. [DOI: https://dx.doi.org/10.1016/j.proeng.2017.08.052]

65. Perera, S.; Nanayakkara, S.; Rodrigo, M.N.N.; Senaratne, S.; Weinand, R. Blockchain technology: Is it hype or real in the construction industry?. J. Ind. Inf. Integr.; 2020; 17, 100125. [DOI: https://dx.doi.org/10.1016/j.jii.2020.100125]

66. Khalid, S.W.; Al-Hadad, B.M.A. A comparison study of using polyethylene terephthalate and limestone fillers on porous asphalt mixture behaviours. Ain Shams Eng. J.; 2023; 15, 102426. [DOI: https://dx.doi.org/10.1016/j.asej.2023.102426]

67. Haq, M.; Khan, A.; Naqvi, T.; Yusuf, M.; Kamyab, H.; Chelliapan, S. Parametric analysis of RSB sensors for concrete strength monitoring using hybrid EMI and WP techniques: Numerical investigation. Ain Shams Eng. J.; 2023; 15, 102457. [DOI: https://dx.doi.org/10.1016/j.asej.2023.102457]

68. Al-Bassam, M.; Sonnino, A.; Bano, M.; Danezis, G. Blockchain Scalability and its Foundations in Distributed Systems. arXiv; 2020; arXiv: 2001.00826

69. Androulaki, E.; Barger, A.; Bortnikov, V.; Cachin, C.; Christidis, K.; De Caro, A.; Enyeart, D.; Ferris, C.; Laventman, G.; Manevich, Y. . Hyperledger Fabric: A Distributed Operating System for Permissioned Blockchains. arXiv; 2018; arXiv: arXiv:1801.10228[DOI: https://dx.doi.org/10.48550/arXiv.1801.10228]