Data-driven optimization of calcination parameters for low-grade clays in LC³ production: Integrating experimental characterization, explainable machine learning, and multi-criteria decision analysis
Construction and Building Materials, cilt.539, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 539
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.conbuildmat.2026.147419
- Dergi Adı: Construction and Building Materials
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
- Anahtar Kelimeler: Calcination optimization, Explainable machine learning, Leave-one-condition-out validation, Limestone calcined clay cement (LC3), Low-grade clay, TOPSIS
- Bilecik Şeyh Edebali Üniversitesi Adresli: Evet
Özet
The depletion of conventional supplementary cementitious materials (SCMs), such as fly ash and ground granulated blast furnace slag, necessitates alternative binders for sustainable cement production. Limestone Calcined Clay Cement (LC3) is a promising option, yet optimizing calcination for heterogeneous low-grade clays remains difficult given mineralogical variability. This study proposes an integrated experimental–computational framework combining pozzolanic activity tests, replicate-aware explainable machine learning (ML), and TOPSIS-based decision analysis to evaluate calcination conditions for four low-grade clays (Vez, Lim, WG, and WO) from Bilecik, Türkiye. Clays were calcined at 600–900 °C for 30–120 min, and performance was assessed using the Strength Activity Index (SAI, ASTM C311) and Frattini tests (TS EN 196–5). To avoid overestimating performance on replicated data, models used leave-one-condition-out validation. The temperature branch predicted compressive strength (condition-level R² = 0.792; MAE = 2.38 MPa), and the time branch predicted activation gain (R² = 0.822; MAE = 4.44 SAI points). Ablation-based XAI showed that processing level and curing age controlled the temperature branch, whereas clay type dominated the duration-dependent activation response. TOPSIS identified Lim clay calcined at 800 °C for 30 min as the best-performing compromise among the screened alternatives, with a closeness coefficient of 0.928. The ranking remained stable across SAI weights of 0.30–0.70 and under a combined-burden formulation. SEM qualitatively supported these trends, with more developed hydrate coverage in high-performing mixtures. Overall, the workflow provides a transparent, experimentally bounded strategy for identifying performance-compatible, process-efficient calcination regimes for heterogeneous low-grade clays.