Distilling middle-age cement hydration kinetics from observed data using phased hybrid evolution

Abstract

The hydration of cement is of great importance to the formation of the microstructure and development of strength. However, the complex nature of cement hydration results in that the existing manually derived model from some assumptions that are made to simplify the problem and make it mathematically and computationally tractable is not satisfactory in comparison with experimental results. In this paper, the middle-age hydration kinetics is distilled from observed data reversely using phased hybrid evolution method. The task that distils the hydration kinetics is divided into two phases and combines different algorithms. Furthermore, some strategies are also adopted for enhancement. To solve the problem of high time complexity, the searching process is accelerated by graphics processing units in parallel. The middle-age cement hydration kinetics model is distilled successfully from observed data. Studies show that the simulation result is close to the same with experimental results according to the distilled kinetics.