Name & Scope

[arthurrge]

Archived discussion here.

Purpose of Computation-AI Subcommittee

Advances in computational (digital / virtualized) methods are transforming the state of the art in science and engineering applied to aeronautics and astronautics - resulting in novel practices with the potential for paradigm shift.

1. AI/ML methods and data-derived models (including surrogates, generative, inverse, etc.)

2. HPC/advanced computing (see CFD 2030 bullet below)

3. Storage/processing abundance delivering scalable/affordable infrastructure for greatly-expanded problem scope

Synergies

The AIAA CFD 2030 (e.g., Fact Sheet) Vision Elements cover a lot of the target ground already

5. physics-based, predictive modeling (which can/should extend to co-simulation and coupled physics)

6. quantification of errors and uncertainties arising from physical modeling, mesh and discretization, etc.

7. automation in all steps of the analysis process: geometry creation, meshing, large databases of simulation results, extraction and understanding of simulation information, etc.

8. impact/opportunity from emerging HPC architecture advances, including: memory hierarchies (latencies/bandwidths), programming paradigms and runtime environments, etc.

but are not extended beyond CFD, leaving application gaps such as for materials modeling, part lifing, etc.

The AIAA DEIC community can leverage and expand upon adjacent efforts and synergies, such as:

• AIAA/NASA CFD 2030,
• AIAA DEIC DECO,
• NAFEMS (ASSESS Initiative, HPC, Multiphysics)
• NASA/ICME 2040
• SERC AI4SE & SE4AI
• Dept. of Energy Exascale Computing Project (legacy / go-forward)
• etc.

toward valuable guidance in shared efforts to advance the state of the art in methods and practices for this technical community.

Name Evolution? (vs. Computation+AI)

• @aramdraper 's proposal CoDADE (Computational Design and Analysis for Digital Engineering)
• @NigeTaylo's proposed ModelOps - concern over "model" being vague, noting difficulties in groups like DSM (digital system model) struggling with this
• "Digital" can be commonplace, where I find "Computational" speaks to applying capable
• I have some concern over excluding "AI" which has become a rallying and prominent pillar. NASA's digital transformation strategic framework notes 6 "technology foundations".
  • Intelligent Automation (read: sophisticated productivity agents)
  • Model-based Anything
  • AI / ML
  • (as well as) Zero-Trust Architecture, Internet of Things, and Extended Reality
• I understand the concern with "Design and Analysis" being narrow. One struggle has been expanding the scope of employing these tools to be inclusive of manufacturing and sustainment (which, honestly, is where all the data are!) Even if the intent is to pull those data back into prescriptive recommendations, the technology is employed prominently downstream from Design. I do not see a compelling need to reduce the scope of engineering to design, and technically in many cases these methods will employ synthesis - the opposite of analysis.
• Alternately, simply citing something like: "Novel" or "Breakthrough" Methods catches the gist of the scope.

(Looking to active participants to take note / contribute in the new thread - @aramdraper @Papaghena @coleyc @NigeTaylo @smarnold59)

[arthurrge]

Looking at methods and practices...

Computational Design and Analysis Framework Development

• Include: multiphysics (coupled, co-simulated), multiscale, multifidelity, multidisciplinary, transdisciplinary, multimodal, hybrid AI+Mod/Sim, maturity assessment, UQ, V&V readiness/facility, etc.
• Exclude: VVUQ study/formalism

Impacts of Computational Design and Analysis on Enterprise Data and Software Architecture

• Include: modes and states of data (e.g., unstructured text, structured data, boundary conditions, geometry, tolerances, visual) – fusion of synthetic + measured data, consistency across domains (aero vs other, civil vs military)
• Exclude: software verification / quality assurance, enterprise architecture implications (ERP, PLM, MES, MRO, etc.) – enterprise thread, data and software governance

System Co-Design

• Include: cyber-physical tradeoffs based on mod/sim, software, and data requirements (e.g., impact of HPC, different types of models), barriers of scale / extensibility – including generative and inverse methods, (exploration vs mapping) of trade-off spaces, life cycle management across full life cycle – including traceability, reproducibility, explainability, provenance, intellectual debt
• Exclude: infrastructure and governance (cloud vs datacenter)

Impacts of Socio-Technical Interactions on Computational Design and Analysis Strategy

• Include: use of semantic technologies for supporting human-machine interactions (conversational speech to text, LLMs, etc.), utilitarian-driven requirements generation, technical/modeling literacy of decision-makers and impact on mod/sim requirements generation
• Exclude: economic analysis of the impact of automation