Abstract: The fusion of mechanism and data is crucial to accurate and efficient assessment of the dam life cycle state and reservoir regulation. This paper discusses the major problems in dam construction and the development of fusion models, and suggests three types of structure for a mechanism-data-driven model－series, parallel and hybrid－along with a brief description of its basic characteristics and applicability. Then, the application and applicability are demonstrated in detail through an example of temperature field analysis for a concrete arch dam. Results show that this fusion model is more accurate and stronger in analysis, prediction and generalization of dam construction and operation, and it is applicable to complicated dynamical-evolving data. Compared with the traditional model, all the three model structures are advantageous. The mechanism-data-driven model provides a new perspective and paradigm for solving the problems of parameter inversion, monitoring and analysis, and strategy optimization in the stages of dam construction, operation and maintenance.

Key words: dam, mechanism-data, mechanism-data-driven model, basic structure, applicability