Abstarct: the preheating rolling furnace is one of the equipments that consumes thermal energy, especially natural gas, in the steel industry. Evaluating the furnace performance and trying to improve its energy consumption can significantly affect the gas consumption of the steel chain. Therefore, the preheating rolling furnace of Khorasan Steel Complex Company has been selected as a case study so that by collecting information, a model can be created for this important equipment in the steel industry using artificial intelligence tools. Therefore, in this research, first by examining the furnace, the sensors required for the ingot input and output from the furnace have been installed. Then, complete data of the furnace has been recorded at specific time intervals. Then a neural network baxsed on the multilxayer perceptron network, which uses data regression for data modeling, was coded in MATLAB. Now, in order to train the neural network, it separates the obtained information into intervals of 5, 10 and 15 each time and calculates the average, standard deviation, skewness, elongation, minimum and maximum of each interval, Now, each of the information in the same intervals for the temperature of different zones, either one by one, or in multiples as input, and the information related to the temperature of the output ingot is entered into the neural network as an output, and the neural network We teach It should be noted that neural network coding uses 70 percent of the input data as network training and the rest for testing the trained network. Then it gives us a report of the performance of the data The closer the data is to zero, the better the training performance of the neural network is, and the best performance was achieved in the interval of 15 and only with the average. Now, according to the trained neural network, we analyze the relationship between the temperature of the zones and the temperature of the outlet billet, the temperature of the smoke with the temperature of the outlet billet, and the temperature of the smoke with the efficiency of the furnace, which causes the reduction of the changes in the temperature of the smoke and the temperature of the zones, reducing energy consumption. baxsed on the results obtained from neural network training, it can be concluded that by reducing the average temperature of each zone by 50 degrees, the output billet temperature does not change noticeably, but the average output smoke temperature decreases significantly and is equal to 549 degrees. According to efficiency calculations, this temperature reduction causes the efficiency to change from 54/63 percent to 61/21 percent, which causes an increase of 6/58 percent in furnace efficiency. The annual gas bill for the pre-rolling furnace is 80 billion Tomans. Now, considering the potential savings of 58.6 percent, approximately 5/264 billion Tomans can be saved per year. Increasing efficiency and reducing fuel consumption is one of the direct effects of this research, and considering the decrease in temperature in different areas of the furnace, another effect of this increase in efficiency is the increase in the life of the furnace refractories, which increases the furnace refractory replacement period and reduces downtime, due to which this life also increases the annual production rate.