The Application of Neuroshell 2 in Predicting the Compressive Strength and Workability of Concrete

The paper deals with a problem of predicting the compressive strength and concrete workability by means of the software package NeuroShell 2. Eleven input parameters which describe properties of the concrete components, and two output parameters which specify the concrete properties are defined. On the basis of sixty examples, the appropriate architecture is selected and the respective net parameters are adjusted. The ˝well-trained˝neural net enables establishing of the dependance of two properties (compressive strength and concrete workability) described by five output parameters, as well as eleven input parameters which specify the concrete components'properties. thus, the compressive strength can be estimated immediately instead of the estimates done by means of the experiments after concrete hardening (28 days of age).