Engineer reviewing structural calculations on a laptop with AI assistance
BIM & AI By Muhammad Haseeb  |  Structural Engineering  |  14 min read

Using ChatGPT & AI for Structural Design Checks: A Practical Engineer's Guide

Structural engineers are using AI tools daily — not to replace judgment, but to cut the grunt work out of preliminary design, code lookups, and calculation scripting. ChatGPT, Claude, and a handful of specialized platforms can handle beam sizing estimates, load combination tables, and even Python scripts for iterative checks. The catch: they hallucinate code clauses with complete confidence. This guide covers what actually works, what doesn't, and exactly how to prompt AI tools to get useful structural outputs.

📋 Table of Contents

▼
  1. What AI can actually do in structural engineering
  2. AI tools compared: ChatGPT vs Claude vs SkyCiv AI
  3. 10 practical use cases with real prompt examples
    1. Steel beam design check (AS4100 / AISC)
    2. Load combination generation
    3. Column slenderness and buckling check
    4. Concrete slab preliminary sizing
    5. Bolted connection checks
    6. Python scripting for iterative design
    7. Seismic base shear calculation
    8. Calculation report drafting
    9. Code clause navigation
    10. Peer review support
  4. Prompting strategies that produce reliable outputs
  5. What AI gets wrong — the critical limits
  6. Professional responsibility and liability
  7. Integrating AI into your daily engineering workflow
  8. The next generation: AI-integrated structural platforms
  9. Frequently asked questions
  10. References and further reading

What AI can actually do in structural engineering

The question most engineers ask first is whether AI tools understand structural mechanics at a level that's actually useful. The honest answer is: yes, for a narrow band of tasks, and no, for anything that requires site judgment or current code knowledge.

Large language models like GPT-4o and Claude 3.5 were trained on enormous volumes of engineering text — textbooks, research papers, design guides, forum discussions. They have absorbed beam theory, load path concepts, code formatting patterns, and calculation methodologies. What they lack is the ability to verify their outputs against live code documents, or to sense the construction context that changes how a calculation should be framed.

Research Finding
A 2024 survey of 400+ structural engineers by ASCE found that 61% were already using AI tools in some capacity — mostly for writing, code lookups, and scripting. Only 8% used AI to generate primary design calculations without independent verification.

Here's a useful way to split what AI does well versus where it falls over:

✅

Preliminary member sizing

Quick section estimates based on span, load, and target utilisation — good enough to start a model.

✅

Load combination tables

Generate complete load combo tables per AS/NZS 1170.1, ASCE 7, or Eurocode with correct factors.

✅

Code clause lookup

Find relevant clauses in AS4100, ACI 318-19, AISC 360-22, BS EN 1993 — though always verify.

✅

Python / Excel scripting

Write calculation scripts that automate repetitive section checks, interpolation, or design iteration.

✅

Report and note drafting

Draft calculation preambles, basis of design sections, and technical specifications from your notes.

✅

Formula derivation and checking

Verify algebraic steps in manual calculations and spot unit errors before they propagate.

❌

FEA and complex analysis

Cannot run finite element models, handle geometric non-linearity, or perform staged construction analysis.

❌

Site-specific judgment

No knowledge of soil conditions, existing structure constraints, or construction method implications.

❌

Current code amendments

Training data has a cutoff. Amendments to AS 4100-2020 or ACI 318-19 issued after that date are invisible to the model.

❌

EOR legal sign-off

AI cannot accept professional liability. A registered engineer must certify every calculation that goes on a stamped drawing.

AI tools compared: ChatGPT vs Claude vs SkyCiv AI

Not all AI tools perform equally for structural work. The table below reflects actual engineering use — not marketing claims. "Structural accuracy" refers to whether the tool's outputs match hand-checked results without prompting corrections.

Tool Best For Structural Accuracy Code Knowledge Scripting Free Tier
ChatGPT GPT-4o
OpenAI		 Code lookups, load combos, report drafting Moderate AS, AISC, ACI, EC Excellent Yes (limited)
Claude Sonnet
Anthropic		 Long calculation documents, complex reasoning chains Moderate AS, AISC, ACI, EC Excellent Yes (limited)
SkyCiv AI
SkyCiv		 Structural analysis with verified calculation engine High AS, AISC, ACI, EC Built-in Free trial
GitHub Copilot
Microsoft		 Python / Dynamo scripting for structural tasks N/A (code only) Limited Excellent Yes
Google Gemini Pro
Google		 Multi-modal tasks (reading PDF drawings) Moderate Limited Good Yes
ClearCalcs AI
ClearCalcs		 Residential/light commercial design checks High AS, NZS, AISC N/A Free trial
Pro Tip
SkyCiv and ClearCalcs are worth the subscription for anything going on stamped drawings. Use ChatGPT or Claude for the thinking work — code lookups, preliminary sizing decisions, scripting — and a verified calculation platform for the certified output.

10 practical use cases with real prompt examples

The following use cases come from actual engineering workflows. Each includes a prompt template you can adapt directly.

1. Steel beam design check (AS4100 / AISC 360)

AI handles preliminary bending and shear checks well. The key is giving it every input up front — section properties, span, supports, loads. If you leave anything out, it assumes values and doesn't always tell you.

ChatGPT prompt # Steel beam check — AS4100-2020 I need a bending moment capacity check for a steel beam per AS4100-2020. Section: 310UB46.2 Span: 6.5 m, simply supported Restraints: Full lateral restraint at both supports, no intermediate restraint Loading: UDL dead load = 12 kN/m, live load = 8 kN/m (unfactored) Load combo: 1.2G + 1.5Q per AS/NZS 1170.1-2002 Steps required: 1. Calculate design bending moment (M*) 2. Determine section moment capacity (φMs) and member moment capacity (φMb) 3. Perform shear check 4. Check deflection at serviceability (L/250 for total load, L/500 for live) 5. Summarise compliance — PASS or FAIL for each check 6. Reference specific AS4100 clause numbers for each step Use fy = 320 MPa, E = 200 GPa. Show full working.

That prompt structure — section, span, restraints, loading, combo, specific outputs required — produces a step-by-step calculation that's far more useful than asking "check this beam." The clause references let you verify the working against the actual standard.

M* ≤ φMb = φ · αm · αs · Ms ≤ φMs
AS4100-2020 Cl. 5.6 — Member moment capacity for unrestrained beams. φ = 0.9, αm = moment modification factor, αs = slenderness reduction factor
Verification Required
AI typically handles αm correctly for simple load cases. It struggles with αm for non-uniform moment diagrams unless you provide the moment values at L/4, L/2, and 3L/4 points explicitly. Always check the αm value it produces.

2. Load combination generation

Generating a full load combination table is tedious work that AI does quickly and accurately — provided you specify the code and relevant load types.

ChatGPT prompt # Load combinations per AS/NZS 1170.1-2002 Generate a complete ultimate limit state (ULS) and serviceability limit state (SLS) load combination table per AS/NZS 1170.1-2002. Load types present on this structure: - G = permanent/dead load - Q = imposed/live load (office occupancy, category B) - W = wind load (ULS and SLS wind pressures) - E = earthquake load Format as a table with columns: Combination ID | Load factors | Usage Include the short-term and long-term SLS combinations separately. Note which combinations govern for: (a) members in compression, (b) uplift checks, (c) foundation bearing pressure.
Combination Expression (ULS) Typical Governing Case
1 1.35G Self-weight dominated members
2 1.2G + 1.5Q Gravity load governed beams & columns
3 1.2G + ψc·Q + Wu Wind uplift + live load concurrent
4 0.9G + Wu Net uplift — holding-down bolts, slab uplift
5 G + ψE·Q + Eu Seismic zones — lateral + gravity concurrent
SLS-ST G + ψs·Q Short-term deflection, crack width
SLS-LT G + ψl·Q Long-term creep, permanent deflection

3. Column slenderness and buckling check

AI is reasonably reliable for Euler buckling, effective length factors, and slenderness limit checks. Where it fails: when the effective length factor depends on joint stiffness that requires a sway/non-sway classification the AI cannot assess without the full frame geometry.

Prompt Check a steel column for member capacity per AISC 360-22 Chapter E. Section: W10x49 Length: 4.2 m (13.8 ft) Boundary: Pinned-pinned about strong axis (Kx = 1.0) Fixed-pinned about weak axis (Ky = 0.7) Axial load: P* = 850 kN (factored ULS) Fy: 345 MPa (A992 steel) Calculate: 1. Effective slenderness ratios (KL/r) about both axes 2. Critical buckling stress Fcr per AISC 360-22 Eq. E3-2 or E3-3 3. Design compressive strength φcPn (φc = 0.9) 4. Utilisation ratio P*/φcPn 5. State PASS or FAIL

4. Concrete slab preliminary sizing

For preliminary span-to-depth ratios and bar spacing estimates per ACI 318-19 or AS 3600-2018, AI is useful. It should not be used to generate final bar schedules without running a proper reinforced concrete design tool.

Prompt — AS 3600-2018 Provide a preliminary design for a one-way reinforced concrete slab. Span: 4.8 m (simply supported) Superimposed dead load: 1.5 kPa Live load: 3.0 kPa (office) Concrete: f'c = 32 MPa Steel: fsy = 500 MPa Cover: 20 mm (interior exposure, Euroclass B1 equivalent) 1. Select slab thickness from AS 3600-2018 Cl. 9.3.4 span-to-depth ratios 2. Calculate design moment M* for 1 m width strip 3. Estimate main bar size and spacing (N-bars) 4. Check minimum steel per Cl. 9.1.1 5. Check crack control per Cl. 9.4.1 Show all working. State clause references.
Important Limitation
AI does not know your slab's deflection history, construction sequence, or adjacent span continuity unless you tell it. Moment redistribution and two-way action effects require a proper structural model, not an AI text response.

5. Bolted connection checks

Connection design is one area where AI delivers useful preliminary checks — bolt group centroids, eccentricity effects, and bearing capacity — but struggles with weld throat geometry and multiplanar connections.

Prompt — AISC 360-22 Check a bolted shear tab connection per AISC 360-22. Shear tab: PL 3/8 × 3 × 9, A36 steel Bolts: (3) ¾" A325-N bolts in standard holes, single shear Beam: W18x35 A992, coped 2" depth × 4" length Shear force: Vu = 40 kips (factored) Check: 1. Bolt shear capacity (φRn per J3.6) 2. Bolt bearing on plate (J3.10) 3. Block shear on plate (J4.3) 4. Gross shear yielding of plate (J4.2) 5. Net shear fracture of plate (J4.2) 6. Flexural yielding of plate (due to eccentricity) Summarise pass/fail for each limit state.

6. Python scripting for iterative design

This is arguably the most productive use of AI for engineers. Writing Python scripts to iterate through section sizes, plot interaction diagrams, or automate load takeoffs takes hours manually. AI gets a working script in minutes.

Python scripting prompt Write a Python script that: 1. Takes a user-defined UDL (dead load G, live load Q) and beam span 2. Applies AS/NZS 1170.1 ULS combination: 1.2G + 1.5Q 3. Iterates through a list of common UB sections (310UB32, 310UB46, 360UB45, 360UB57, 410UB54) 4. For each section, checks: - Moment capacity φMs = φ · fy · Zx (compact section, full restraint) - Shear capacity φVv = φ · 0.6 · fy · Aw 5. Prints a table showing section, M* / φMs ratio, V* / φVv ratio, and PASS/FAIL 6. Highlights the lightest section that passes both checks Use Python with pandas and tabulate libraries for output formatting. Section properties should be hardcoded from the OneSteel catalogue.

How structural engineers are integrating AI into daily design workflows — tools, prompts, and verification strategies.

7. Seismic base shear calculation

Equivalent static force method calculations are well within AI capability. Dynamic response spectrum analysis is not — that requires a structural model, not a text conversation.

Prompt — ASCE 7-22 Equivalent Static Force Calculate seismic base shear using ASCE 7-22 Equivalent Lateral Force procedure. Building data: - Seismic Design Category: D - Occupancy Category: II (Ie = 1.0) - Site Class: D - SS = 1.50g, S1 = 0.60g (from USGS) - Building weight W = 2,800 kips - Structural system: Special moment frame (R = 8, Cd = 5.5, Ω0 = 3) - Building height: 5 storeys, hn = 65 ft Steps: 1. Calculate SMS, SM1, SDS, SD1 (Cl. 11.4) 2. Calculate approximate fundamental period Ta (Cl. 12.8.2.1, Ct and x for moment frame) 3. Calculate Cs (Cl. 12.8.1.1) — check upper and lower bounds 4. Calculate V = Cs × W 5. Distribute V over building height using Cvx (Cl. 12.8.3) 6. Show full working with equation references

8. Calculation report drafting

Engineers spend a disproportionate amount of time writing — calculation preambles, basis of design narratives, letter of recommendations. AI handles these quickly, and the output usually only needs light editing.

Prompt Write a "Basis of Design" section for a structural engineering calculation package. Project: 3-storey office building, steel frame with concrete slab Location: Brisbane, QLD, Australia Standards: AS/NZS 1170 (Loading), AS 4100 (Steel), AS 3600 (Concrete), AS 4600 (Cold-formed) Geotechnical: Footing design based on geotechnical report dated [DATE], allowable bearing capacity 150 kPa Cover: - Applicable codes and standards (list with edition years) - Material specifications (steel grades, concrete strength) - Design life and exposure classification - Loading assumptions (occupancy category, wind region, seismic zone) - Scope of calculations and exclusions Keep it formal, third-person, suitable for a stamped calculation package.

9. Code clause navigation

Scanning a 400-page design standard to find the clause about minimum eccentricity in column design wastes real time. AI locates these clauses in seconds — but you should read the actual clause before relying on what it says.

Prompt In AS 3600-2018, what clause covers: (a) minimum eccentricity for columns (b) requirements for closed ties in compression members (c) lap splice length for N-bars in tension zones For each, give the clause number, a one-sentence summary of the requirement, and the relevant formula or table reference.

10. Peer review support

Paste a set of hand calculations into ChatGPT or Claude and ask it to check the logic and unit consistency. It catches things that slip past tired eyes — wrong sign conventions, incorrect moment factor, skipped steps. It's not a substitute for a second engineer, but it's a useful first pass.

Prompt Review the following structural calculation for logical errors, unit consistency, and missing checks. [Paste your calculation here] Check for: 1. Correct load combination applied 2. Unit consistency throughout 3. Missing code-required checks (e.g., shear, deflection, stability) 4. Incorrect formula application 5. Sign convention errors List any issues found, with the line or step where the error occurs.

Prompting strategies that produce reliable outputs

The quality of what AI gives you is determined almost entirely by how you ask. Structural engineering prompts have specific requirements that general-purpose AI users don't think about.

1

Specify the exact code edition

Say "AS4100-2020" not "Australian steel code." Say "ACI 318-19" not "ACI code." AI has training data for multiple code editions and will default to whichever it has the most data on — which may be an older version.

2

Give all inputs in one message

Section designation, span, support conditions, all applied loads (factored and unfactored), material properties, and exposure class. If you omit something, the model assumes a value and may not flag it.

3

Ask for step-by-step working with clause references

Add "Show all working and reference specific clause numbers." This forces the AI to commit to a calculation path and cite clauses you can check. Without this instruction, you get an answer with no audit trail.

4

Request a PASS/FAIL summary

Tell it to conclude with a compliance summary for each limit state. This prevents vague outputs where you have to interpret the result yourself.

5

Verify clause numbers independently

Never trust an AI-quoted clause number without opening the actual standard. Hallucinated clause numbers are the most common and dangerous error in AI-assisted structural work.

6

Ask it to flag what it cannot check

Include "State any inputs you had to assume and any checks you cannot perform without additional information." This surfaces gaps rather than letting the AI paper over them.

Efficiency Tip
Save your standard prompt templates as text files. A beam check prompt, a column check prompt, a load combo prompt — each tailored to your jurisdiction's code. You'll cut the time to get a useful AI response from 5 minutes to 30 seconds.

How AI tools rank for common structural tasks

Approximate capability rating (1 = limited, 10 = highly reliable) based on engineering community testing and peer-reviewed assessments:

Report writing
9.2 / 10
Load combinations
8.5 / 10
Python scripting
8.8 / 10
Code clause lookup
7.0 / 10
Preliminary sizing
7.5 / 10
Connection design
5.8 / 10
Seismic ESF method
6.5 / 10
Detailed slab design
4.2 / 10
FEA / dynamic analysis
0.8 / 10

What AI gets wrong — the critical limits

Engineers who have integrated AI into their workflows consistently report the same failure modes. Knowing these before you start saves time and prevents errors reaching construction.

❌ Hallucinated clause numbers

AI invents plausible-sounding clause references (e.g., "AS4100-2020 Cl. 6.3.4") that either don't exist or contain different content. This is the most dangerous failure mode.

❌ Incorrect effective length factors

Ke for columns depends on sway/non-sway frame classification and joint stiffness ratios that AI cannot assess without the full structure geometry.

❌ Wrong load factors for edge cases

Unusual combinations — prestressed concrete, machinery vibration, crane loads — often use reduction factors or allowable stress methods that AI confuses or mixes between codes.

❌ Section property errors

AI sometimes uses incorrect Zx, Ix, or J values from memory rather than catalogue values. Always cross-check section properties from the actual steel section tables (e.g., OneSteel, AISC Manual).

❌ Missing limit states

Unless you explicitly list what to check, AI may skip limit states — particularly local buckling, web crippling, or second-order effects in slender members.

❌ Overconfident failure mode

AI presents wrong answers with exactly the same confident tone as correct ones. There is no hesitation or uncertainty flag. You cannot tell from the tone whether a result is reliable.

✅ Known good: load combination tables

When you specify the code and load types, AI generates correct ULS/SLS combinations accurately and consistently. This is one of its most reliable structural outputs.

✅ Known good: formula application

Given explicit formula inputs, AI applies equations correctly. It's the clause identification and assumption-setting where errors creep in, not the arithmetic.

Real-world testing of ChatGPT on structural engineering problems — what passes, what fails, and what you should never trust without verification.

Professional responsibility and liability

The legal position is clear in every jurisdiction with a licensed engineering profession: the engineer of record is responsible for all calculations on a stamped drawing, regardless of what tool generated them.

The Australian Engineers Australia Code of Ethics, ASCE Code of Ethics, and UK Engineering Council standards all require that engineers apply independent professional judgment to every output they certify. AI is a tool in the same category as structural software — you are responsible for verifying its outputs, not trusting them by default.

Scenario Appropriate AI Use Liability Position
Preliminary beam sizing for architectural coordination Use AI estimate, note "preliminary only" Acceptable
Load combination table for calculation package Use AI output, verify against code before including Acceptable with verification
Python script for section selection (run by engineer) AI writes script, engineer reviews logic and runs it Acceptable
Design calculation for stamped drawing, unchecked AI output Not acceptable Professional liability risk
AI-generated code clause reference, unchecked Not acceptable Risk of non-compliance

Some firms are now drafting internal AI use policies that require engineers to document which parts of a calculation were AI-assisted and what verification steps were taken. This is sound practice regardless of whether your jurisdiction mandates it yet.

Integrating AI into your daily engineering workflow

The engineers who get the most out of AI tools treat them like a knowledgeable assistant, not an oracle. The assistant drafts, the engineer decides. Here's how that looks day to day:

Morning: project intake. A new project comes in — a 4-storey retail building in seismic zone 2B. You ask ChatGPT to generate a preliminary loading summary: gravity loads by floor, wind load parameters for the city, seismic hazard level. This takes 3 minutes instead of 20. You review and adjust.

Midday: calculation scripting. You need to check 12 different beam spans with varying UDLs. You prompt GitHub Copilot in VS Code to write a Python script that loops through the spans, calculates M* and V*, and checks them against tabulated φMs and φVv values for a chosen section. You review the script logic, run it, and have your results in a table within 15 minutes.

Late afternoon: report writing. You have hand calculations done. You paste your notes into Claude and ask for a formal calculation preamble. You edit the output for accuracy, add your registered engineer details, and have a professional-looking document in 20 minutes instead of an hour.

Time Savings
Engineers in firms that have adopted AI-assisted workflows report saving 15–25% of documentation and preliminary design time — equivalent to roughly one hour per eight-hour day. The savings are concentrated in report writing, load takeoffs, and section selection iteration.
🏗️

Structural Engineering Services

Need structural design checks, calculation packages, or peer review for your project? Muhammad Haseeb (MIEAust) provides structural engineering services for residential, commercial, and industrial projects. Available for international remote consulting.

View Portfolio →    LinkedIn

The next generation: AI-integrated structural platforms

General-purpose AI tools like ChatGPT are a starting point. The more significant shift is happening inside dedicated structural software, where AI connects to verified calculation engines rather than generating freeform text.

SkyCiv. SkyCiv's AI assistant can interpret natural language descriptions of structures, set up analysis models, and run verified AISC/AS/Eurocode checks — with the AI acting as an interface layer, not the calculation engine itself. This avoids the hallucination problem because the AI translates your intent into inputs for a certified solver.

Autodesk Forma and Speckle. Autodesk's Forma platform is adding AI-driven structural concept generation, where load path suggestions and structural system options come from ML models trained on thousands of completed building designs. Speckle provides the data layer that lets AI tools access structural model data across platforms.

ETABS and SAP2000 integrations. CSi is embedding AI features for automated model checking and report generation — catching common modelling errors before they reach the analysis stage.

Reinforcement learning for optimisation. Research from MIT, ETH Zürich, and the University of Melbourne is applying reinforcement learning to structural topology optimisation — finding minimum-material solutions to complex loading scenarios that human-guided optimisation routinely misses. These tools are not yet in production practice but are 2–4 years away from commercial release.

Platform AI Feature Status Link
SkyCiv Natural language structural model setup + code checks Live skyciv.com
Autodesk Forma AI structural concept generation, wind/solar analysis Live autodesk.com
ClearCalcs AI-assisted residential structural design checks Live clearcalcs.com
Speckle AI model review across BIM/structural platforms Beta speckle.systems
ETABS / SAP2000 AI model checking and automated report generation In development csiamerica.com

Frequently asked questions

Can ChatGPT perform structural calculations reliably? +
ChatGPT can perform basic structural calculations — beam bending, shear, deflection, and load combinations — but it can hallucinate code clause references. Every output must be independently verified against the actual design standard before use in construction documents. It's most reliable for load combination tables and scripting, least reliable for section property recall and complex multi-step design procedures.
Which AI tool is best for structural engineering? +
GPT-4o and Claude are the most capable general-purpose AI tools for code interpretation, calculation checking, and report writing. SkyCiv AI provides the most reliable structural outputs because it connects AI to a verified calculation engine rather than generating freeform text. For Python scripting, GitHub Copilot inside VS Code is the most efficient option.
Is it legal for a structural engineer to use ChatGPT in design? +
Yes, but the registered engineer of record remains fully responsible for all design outputs. AI is a tool, not a substitute for professional judgment. Most jurisdictions require a licensed engineer to review and certify any structural calculation before it appears on stamped drawings. Using AI does not reduce your professional liability — it adds a verification step you must document.
Can AI replace structural engineers? +
No. AI cannot assess site-specific conditions, exercise engineering judgment on ambiguous loading scenarios, or accept legal liability. It is most useful as a productivity tool for drafting, scripting, and preliminary checks — not as a replacement for professional engineering. The closest analogy is spreadsheet software: it changed how engineers work, but didn't replace the need for engineering judgment.
How do I verify AI-generated structural calculations? +
Cross-check every clause reference against the actual standard. Verify section properties against the manufacturer's published table (OneSteel, AISC Manual). Run a parallel hand calculation or use a certified software tool (ETABS, SkyCiv, ClearCalcs) for the same problem and compare results. Document any discrepancies before including the calculation in your package.
What is the biggest risk of using AI for structural design? +
Hallucinated clause numbers presented with full confidence. AI does not flag its own uncertainty. If it recalls a wrong clause number, it states it as fact. Engineers who don't check clause references before certifying a calculation may submit non-compliant documents without realising it. This is the most common and consequential failure mode in AI-assisted structural work.
Python for structural calculations
Python for Structural Engineering Calculations — A Practical Guide
 SkyCiv vs ETABS comparison
SkyCiv vs ETABS: Which Structural Analysis Software is Right for You?
 AS4100 steel beam design
AS4100-2020 Steel Beam Design: Step-by-Step Worked Example
ChatGPT Structural Engineering AI Design Checks AS4100 ACI 318 AISC 360 Structural AI Tools Python Structural GPT-4 Engineering SkyCiv AI Load Combinations

References and further reading

  1. Standards Australia. AS 4100-2020: Steel Structures. SAI Global. saiglobal.com
  2. Standards Australia. AS 3600-2018: Concrete Structures. SAI Global.
  3. Standards Australia / Standards New Zealand. AS/NZS 1170.1-2002: Structural Design Actions — Permanent, Imposed and Other Actions. SAI Global.
  4. American Concrete Institute. ACI 318-19: Building Code Requirements for Structural Concrete. concrete.org
  5. American Institute of Steel Construction. AISC 360-22: Specification for Structural Steel Buildings. aisc.org
  6. ASCE. (2022). ASCE 7-22: Minimum Design Loads and Associated Criteria for Buildings and Other Structures. asce.org
  7. Zheng, R., et al. (2023). "Evaluating large language models on structural engineering tasks." Journal of Structural Engineering. ascelibrary.org
  8. SkyCiv. (2024). "AI-assisted structural analysis: Current capabilities and limits." skyciv.com/docs
  9. Engineers Australia. Code of Ethics. engineersaustralia.org.au
  10. Autodesk. (2024). "Forma AI structural tools overview." autodesk.com
  11. Speckle Systems. (2024). "Connecting structural data across platforms." speckle.systems
More in BIM & AI