Computational Engineering / Digital fabrication, cloud and connected workflows
Computation somewhere else
Cloud solvers, monitoring, connected models — and the five risks that come with all of them.
Lesson review: expert verification remains open
This lesson has received a full-context editorial review and a check of its technical explanations. Selected arithmetic has been checked separately. This is not independent engineering sign-off.
Checks needed for project use
- A computational engineering reviewer must check algorithms, convergence claims, statistical assumptions and numerical examples beyond the cases explicitly recorded.
- Qualified legal/information-governance review remains required for contractual confidentiality, personal-data processing and allocation of duties.
- Current commercial quantum capabilities are not verified; no practical speed-up claim made.
- Fabrication tolerance analysis requires actual geometry, datums and installation allowances.
These are educational models and exercises. Check their assumptions before using the results. The full technical content and every diagram have not yet been independently verified.
Concrete and steel calculations retain their stated first-generation Eurocode teaching basis. Use the standards and amendments specified for your project; do not mix generations without an appropriate design basis.
Read the lesson register and outstanding review notesWhy this matters
Analysis increasingly runs somewhere other than the desk it was set up from, and project data increasingly lives somewhere other than the office that produced it. That brings real benefits: capacity on demand, collaboration across locations, and a single current version of the model.
It also relocates a set of questions that engineers used to answer implicitly by having the file on their own machine. The questions did not go away; they went somewhere less visible.
By the end of this lesson you should be able to
- Name the five risks in a connected workflow
- State the consent rule for sending project data to an external service
- Distinguish a stable principle from a dated capability claim
- Explain why version control matters more when the model is shared
- Say what a digital twin needs to be more than a visualisation
Connected workflows
Establish data location, access controls, recovery arrangements, version history and responsibilities before relying on an external service. These involve technical, contractual and organisational decisions. An approved service may already provide suitable controls; inspect its actual terms and configuration rather than assuming either protection or exposure.
Project information
Use authorised services for client material. Check contractual confidentiality, practice policy, permissions and applicable data-protection obligations before transfer. Personal-data processing needs an appropriate lawful basis; explicit consent is not the only possible basis, and client permission does not automatically settle every obligation. Retention, deletion and training use depend on the service and agreement. Seek the relevant information-governance or legal advice where unsettled.
Reproducibility
Record the exact input-model identifier, software and solver version, relevant settings and outputs. Preserve a recoverable snapshot. A date or the phrase current model may not identify the analysed state.
Digital twins
A digital twin links a representation of a physical asset with information about that asset. The useful connection may support monitoring, prediction or decisions; definitions do not universally require automatic recalibration of an analytical model. If parameters are updated from measurements, define the update method and validate predictions on suitable independent evidence.
Currency
Separate dated platform capabilities, standards and legal requirements from mathematical definitions and project policies. The course register is a record of evidence and unresolved checks, not proof that a claim labelled stable is correct. Read its current entries rather than relying on fixed counts in lesson prose.
Try it
Workflow risk exercise
One link in a connected workflow at a time. Identify the risks that apply, then see which they are and who owns them.
Architectural geometry → parametric model
Grid, levels and envelope imported to drive the structural generator.
Architectural geometry → parametric model
Grid, levels and envelope imported to drive the structural generator.
Which risks apply here?
Select the risks you think apply, then check. There is usually more than one, and the owner is rarely the person who set the link up.
This course’s currency register
16 time-sensitive claims are tracked. 8 are stable principles that do not date. 7 require verification against a current authoritative source, and are marked as such wherever they appear. 1 is reported as a 2022 position rather than as current practice.
- Stable principle
- 8
- Verified
- 0
- Requires verification
- 7
- As at 2022 — not checked since
- 1
- Superseded
- 0
Areas tracked
- Analysis methods: 1 claim
- Second-order effects: 1 claim
- Meshing: 1 claim
- AI and machine learning: 4 claims
- Cloud and connected working: 2 claims
- Quantum computing: 1 claim
- BIM and interoperability: 1 claim
- Design standards: 1 claim
- Embodied carbon: 2 claims
- Security and data protection: 1 claim
- Digital fabrication: 1 claim
What this shows: Moving computation elsewhere relocates the questions of location, access and liability — it does not remove them, and the analysis liability does not move at all.
Worked example
Five questions about one arrangement
Given
- A frame analysis is run on a vendor's cloud solver
- The model is authored in a shared environment used by three consultants and the contractor
- The completed structure will carry sensors feeding a monitoring platform
Find
The five risks, and who owns each
Practice
In a hypothetical register, 16 claims are classified as 8 stable principles, 7 requiring verification and 1 historical statement. None has a verified status. What percentage is marked verified?
Check yourself
An engineer pastes a client's structural calculations into a public AI assistant to have the arithmetic checked, then deletes the conversation. What is the position?
Practice
A workflow has five links, each with an owner. If two of the five have no owner, what percentage of the chain is unowned?
Practice
In the fictional eight-link exercise, three links are marked with confidentiality risk. How many links carry that flag?
Check yourself
For a predictive digital twin that updates parameters from measurements, what must be established?
Check yourself
A model that adjusts freely to match its measurements — what has gone wrong?
Summary
- Where is it, who can reach it, what if access is lost, what if it is wrong, who is liable
- None of the five is answered by the platform working well
- Transfer project information only through authorised arrangements with suitable contractual and information controls
- Record professional, organisational and service-provider responsibilities
- A shared current model makes single-version benefit and single-point-of-failure the same property
- A digital twin that would match any measurement has stopped being a prediction
- Consult the current evidence register for verified claims and unresolved checks