289 / 2023-02-11 19:34:06

ChatGPT in Structural Engineering Research

ChatGPT,Structural Engineering,Research,Structural Design,FRP

ChatGPT is a conversational language model developed by OpenAI, based on the transformer architecture. It was trained on a massive dataset of diverse conversations to generate human-like text on a wide range of topics. It has created significant impacts on various fields. This presentation aims to discuss the impact of ChatGP on research work in structural engineering and summarize its advantages and disadvantages. It is found that ChatGPT can well solve structural engineering problems in design. Meanwhile, ChatGPT has powerful information retrieval capabilities, which can be applied to discover research gaps. However, it also has some disadvantages that should be considered. Due to a lack of creativity to a dependency on data quality, these disadvantages can impact the quality and reliability of the work produced.



In this work, ChatGPT was tested for its ability to solve structural engineering problems by some questions. For instance, ChatGPT was asked about how to design reinforced concrete beams. It can provide accurate and detailed design methods. On this basis, ChatGPT can solve the design examples given in ACI code. Even if some data in the question is changed, ChatGPT can still calculate the result. This means that in addition to a certain information retrieval ability, ChatGPT also has a certain degree of self-learning ability, which is beyond the traditional search engine, such as Google and Baidu. ChatGPT can be used by engineers to generate quick design sketches and prototypes for their products, reducing the time required for design iteration.



Besides, ChatGPT was evaluated for its ability to discover and solve research gaps in this study. In the case of the area of fiber reinforced polymer (FRP) structures, it can summarize its research gaps and divide them into three areas: material science, structural behavior, and durability and sustainability. However, when queried regarding methods to resolve these difficulties, ChatGPT can provide only generalized and inapplicable information, and is unable to furnish specific and efficacious strategies to address them. In addition, when presented with inquiries such as assessing the contributions of a particular academic or identifying the preeminent scholars in a specific discipline, ChatGPT frequently delivers unsatisfactory responses and may even be unable to supply a reply.



The cause of these outcomes can be attributed to the absence of creativity in ChatGPT and its reliability on the caliber of the data it was trained on. This implies that it is unable to generate innovative concepts or produce truly novel content. ChatGPT’s responses are derived from the training data and are limited to the information contained therein. If the training data is prejudiced, obsolete, or of subpar quality, the responses generated by ChatGPT will embody these biases. This can lead to erroneous or unsuitable responses. Thus, ChatGPT can be used to retrieve existing information and assist engineers and researchers to solve problems with fixed answers or solutions. However, it is currently incapable of substituting for researchers in executing innovative work, and it lacks the cognitive capability to assess an individual or entity by taking into account a multitude of factors in a comprehensive manner. Therefore, there is a long way to be developed for effective uses of ChatGPT in structural engineering research.