Durability prognostication of ferroconcrete structures on the basis of neural indistinct networks

Abstract

The paper considers possible application of modern prognostication techniques as an element of a quality control system. Applied mathematical tools are the artificial indistinct neural networks with the inverse distribution of a TSK type architecture error. The analysis is made of the factors influencing the ferroconcrete durability. The selected input characteristics are: the sand fineness module, the number of of a lamellar and needle-shaped grains in crushed stone, cement volume weight, of a cement stone strength. The output parameter is the arithmetical mean value of the destroying force by the results of three experiments. The MS Access database was formed on the basis of the laboratory logbooks of the production input control. Two groups of tuples are formed: for training of indistinct neural network and for adequacy tests of the trained network. Mathematical model showed the efficiency of testing. The average error value was 9.6 kg/cm2 or 2%.

Prediction of Mechanical Properties of Highly Functional Lightweight Fiber-Reinforced Concrete Based on Deep Neural Network and Ensemble Regression Trees Methods

“…As for scientific deficits, some pioneering studies were carried out earlier in Russia and the CIS, in particular, reference [ 46 ], the authors of which used a fuzzy neural network as a model for predicting the strength of reinforced concrete products. The mathematical model showed its effectiveness in testing.…”

“…The mathematical model showed its effectiveness in testing. The average error was 0.96 MPa or 2% [ 46 ].…”