Introduction
Against the backdrop of the global sanitaryware industry's accelerated transition toward digitalization and decarbonization in 2026, raw material R&D is undergoing an unprecedented paradigm shift. Traditional ceramic formulation development has heavily relied on the experience and intuition of veteran engineers. Today, the intervention of digital technologies is pushing ceramic formulation R&D to brand-new heights.
1. Materials Informatics: Breaking the "Black Box" of Formulation R&D
Materials Informatics (MI) is an interdisciplinary field that combines computer science and data science with traditional mineralogy and inorganic chemistry. Its ultimate goal is to establish a digital correlation network of "Composition-Processing-Structure-Property."
A Paradigm Shift in R&D Efficiency: According to public data from materials digitalization laboratories, introducing MI processes can reduce physical experiments for new materials by up to 70%.
High-Throughput Screening: When faced with millions of permutations of natural mineral proportions and particle size distributions, AI can replace humans in initial screenings, drastically reducing time consumption.
Faster Time-to-Market: It compresses the development and validation cycle of traditional ceramic formulations from months or even years down to just weeks, allowing new products to seize the initiative in market competition.
2. Explainable AI: Precisely Predicting Sintering and Forming Behaviors
Ceramic manufacturing exhibits highly non-linear characteristics. Minor fluctuations in raw material chemical compositions and particle size distributions are magnified exponentially after high-temperature sintering above 1200°C. The core of digital R&D lies in using machine learning to predict these complex changes.
Machine learning models do not guess blindly; they optimize complex objective functions at the micro level. Through this rigorous machine learning, traditional empirical ceramic rules are successfully mapped onto the digital world.
Eliminating the AI Black Box via SHAP Analysis: According to the latest literature on ScienceDirect, utilizing SHAP (SHapley Additive exPlanations) models intuitively reveals specific causal relationships. In certain clay formulations, the weight of silica ($SiO_2$) content on water absorption is around 30%. Meanwhile, alumina ($Al_2O_3$) and firing temperature are the keys to determining bending strength (accounting for about 60%). This digital-based causal inference provides precise navigation for fine-tuning raw material formulations.
3. Digital Twins of Raw Materials and the Adaptive Loop of Casting Processes
The digitalization of raw material R&D cannot just stay in the laboratory. The true synergy lies in bridging the "static formulation data" on the raw material side with the "dynamic process data" on the production line.
Digital Twins of Raw Materials: We do not just use machine learning to predict material properties in the lab; we seamlessly integrate these predictive models into the control systems of Sunlets' high-pressure casting equipment.
Real-Time Process Adaptation: When a batch of slip experiences a minor 2% fluctuation in particle size distribution or mineral composition, the AI algorithm instantly calculates its impact on the green body's demolding strength. The system then automatically fine-tunes the physical pressure curve of the equipment (e.g., dynamically correcting from 1.2 MPa to 1.25 MPa).
The Ultimate Closed Loop: This ability to let the process self-adapt to material fluctuations allows us to "hedge" against raw material batch defects in real-time right on the production floor. This is the pinnacle of digital R&D.