Structural design is one of the important elements in the construction of a building in order to produce a safe, strong and economical building. Especially in the design of upper structures (in the form of floors, beams, columns, walls and roofs). In beam structures, design and feasibility testing regarding dimensions and the use of reinforcement are safe and economical. Based on the allowable normal stress, it can be seen whether the normal stress is feasible or not if a certain input value is entered, so that information on the feasibility of this quantity can be applied directly to the building. Therefore, it is necessary to have a simulation process regarding the feasibility of stress so that the building to be created can meet the criteria for being a safe, strong and economical building. Overall the analysis process uses Machine Learning methods. By using three types of data, namely algorithm training data, validation data and testing data with a comparison percentage of training and testing data of 80%:20%. The training process produces very high accuracy. This can be seen in the R-Squared value of 0.99 while the RMSE is 1.30e-06. Meanwhile, in the data testing process, high accuracy was also obtained, for R-Squared of 1.0 and RMSE 3.87e-16. After the training and testing process has obtained very good results, we proceed to the beam feasibility testing process by entering certain input values, to produce the normal stress magnitude (Mpa) along with the confidence level of each normal stress prediction result.