Document Type : Original Article

Authors

• Hadi Faghihmaleki ¹
• Mehdi Ghadiri ²

¹ Tonekabon. Mazandaran, Iran Mirkarimi St.

² Master's student in Civil Engineering, Fulty of Civil Engineering, Ayandegan Institute of Higher Education, Tonkaban, Iran

Abstract

Purpose: This study was conducted to predicte the resistance properties of concrete with the help of different types of neural networks. The data studied in this research was collected from the database of 127 mixing plans. The input data included the age of concrete in days, the amount of coarse grain, fine grain, cement, water, lubricant in kilograms per cubic centimeter, and the target data included compressive strength.

Methodology: The use of artificial intelligence as a modern method has a special place in engineering sciences. In this research, the data used were first normalized and then the desired data were trained using the Lorenberg Marquardt algorithm.

Findings: The evaluation criteria of artificial neural network models were obtained using evaluation and error and the results showed that the use of 10 hidden layers had the highest correlation coefficient and the lowest error. The structure of this network was multi-layered perceptron.lubricant in kilograms per cubic centimeter, and the target data included compressive strength.

Methodology: The use of artificial intelligence as a modern method has a special place in engineering sciences. In this research, the data used were first normalized and then the desired data were trained using the Lorenberg Marquardt algorithm.

Originality/Value: The results showed that for the constructed neural network, the value of correlation coefficient, mean root, error square and mean absolute error of the artificial neural network were 0.94 and 1.9, respectively.

Keywords

• Mechanical performance of concrete
• artificial intelligence
• artificial neural network
• Lunberg Marquardt algorithm

Main Subjects

• Urban Management