AI in Structural Engineering: The Challenge of Final Drawings

AI in Structural Engineering: The Challenge of Producing Final Drawings A few thoughts as I try to build a complete engineering application from scratch. AI is helping us work faster in civil, structural, and bridge engineering. However, a large gap remains between a design concept and a reliable set of drawings for fabrication and construction. I see this process in four stages. 1. From sketches to 3D models AI can help us visualize ideas faster. A simple sketch and a short description can produce a 3D representation. We can use it to discuss options and refine the concept. AI now can do very very good this step. 2. From digital models to finite element models and FEA AI is very helpful for writing scripts, working with APIs, and automating model creation. The solver still performs the finite element analysis. Engineers still define the structural idealization, boundary conditions, loads, connections, and analysis assumptions. AI can speed up the process. AI now can do good this step :D Solvers built by vibe developers can certainly solve complex engineering problems. However, they still need to demonstrate that their solvers are reliable. 3. Code checks and design changes This stage offers many opportunities for automation. The equations, checking procedures, and acceptance criteria are relatively well defined. AI now can do good but Sometimes we need to change in our own way. There are approximate manual calculation methods - based on conservative assumptions - that are both easy to understand and sometimes better. :D And when a member fails a check, an experienced engineer still needs to decide how to modify the design. 4. From design results to shop drawings In my view, this remains one of the biggest challenges for AI in our industry. A finite element model usually lacks the information needed to produce shop drawings. It simplifies the structure for analysis. Fabrication and construction require many additional decisions. These include connection details, reinforcement layouts, splice locations, tolerances, transport segments, and erection sequences. A connection may have enough strength, yet leave no room for a tool to tighten the bolts. A reinforcement layout may provide the required steel area, yet be too congested for installation and concrete placement. A splice location may work well in the analysis, yet create major problems during construction. These decisions depend heavily on practical experience. For many developers who rely on “vibe coding,” this is the hardest barrier to overcome. You may be able to build good solutions for stages 1–3. Stage 4 remains much harder. So, keep it up, structural-viber-engineers :D