一种水泥基材料的关系补全方法、系统、设备及介质
By employing self-adversarial loss and self-interference loss methods, a set of triples for cement-based materials is constructed to train the encoder and decoder. This solves the problems of information loss and sample imbalance in the research and development of cement-based materials, and improves the accuracy and generalization ability of relation prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF JINAN
- Filing Date
- 2026-05-13
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional cement-based material research and development relies on trial and error and experience tables, resulting in long experimental cycles, high costs, and difficulty in effectively modeling the complex interactions between formulations and processes. Knowledge graphs in cement-based materials suffer from information loss and sample imbalance, leading to poor relationship prediction capabilities.
By employing self-adversarial loss and self-interference loss, an encoder and decoder are trained by constructing a set of triples for cement-based materials. Semi-supervised training and self-interference decoder are used to enhance relation embedding features, prevent overfitting, and improve relation prediction accuracy.
It effectively reduces information loss, improves the accuracy and generalization ability of predicting relationships in cement-based materials, and can better characterize the relationship between material composition, process and performance.
Smart Images
Figure CN122177321B_ABST