An experience based artificial neural network in the design optimization of steel frames

Abstract

The design of steel frames is an iterative process relying on the experience and decisions of the designer to achieve economical and safe designs. With recent advances in artificial intelligence, particularly, artificial neural networks, it became possible to train such networks to simulate an experienced designer. Thus, the aim of this contribution is to investigate the possibility of artificial neural networks gaining design experience and using such experience in predicting adequate and economical designs. To achieve this aim, an adaptive harmony search algorithm is used to obtain the optimum structural design of two-dimensional steel frames. Those designs are, in a sense, considered an experience, which are then used in training artificial neural networks. The trained networks are finally used in predicting the optimum solution of new problem variants. In total, 18684 samples based on 3114 two-dimensional frames were used to train multiple feed forward artificial neural networks, with a training, validation and testing ratios of 70%, 15% and 15%, respectively. The trained networks' performance was verified, and used in design predictions on interpolated and extrapolated cases. Considering the designs suggested by the artificial neural networks, 99% were adequate in the case of network verification. Furthermore, 97% and 93% of designs were adequate in the case of interpolation and extrapolation. Thus, artificial neural networks are able to learn from the design experience and provide good approximations for designs of variants even outside the training set. Such findings encourage the development of artificial intelligence assisted design systems that are capable of suggesting optimum or near-optimum designs for two-dimensional frames. Also, it could encourage further research for three-dimensional steel frames and more complex steel structure systems.

The Role of Artificial Intelligence in Civil Engineering Applications and Programs

“…1 These models can provide real-time monitoring and assist in proactive decision-making, enhancing construction site management and safety practices [5], [25]. 5.…”

“…By analyzing historical data, trends, and patterns, engineers can identify potential risks, anticipate challenges, and make proactive decisions to mitigate these risks. This approach improves project planning, resource allocation, and risk management strategies [25].…”

“…Data-driven decision-making assists engineers in ensuring compliance by analyzing and assessing design and construction parameters against the relevant regulations. This approach reduces the risk of non-compliance issues and legal complications [25]. 7.…”