The application of automatic acquisition of knowledge to mix design of concrete

In this paper, an automatic knowledge-acquisition system based on neural ne tworks is created to design concrete mix. This system consists of three mod els: the mix-design model, the slump-prediction model, and the strength-pre diction model; the first model is the core of the system with the other two models supporting the core. Each model is made up of a mix-design database , a knowledge base, a neural network-learning block, and a problem solution block. The automatic acquisition of knowledge is realized through the lear ning process of a neural network from sample mix designs. The knowledge bas e is the network itself. This system not only makes full use of the workabl e mix designs that are already in existence but also provides a quick means to predict the slump and 28-day compressive strength of ready made concret e. Examples and experimental work show that the application of the system t o concrete mix design is practicable. (C) 2000 Elsevier Science Ltd. All ri ghts reserved.