Map engineering tools to authoritative data and lifecycle responsibilities.
Specify metadata, provenance, version, configuration, and schema rules.
Explain what STEP / ISO 10303 supports and what data exchange still cannot guarantee.
Use JSON, CSV, and SQL as teaching tools for a queryable, AI-ready evidence architecture.
Evidence standard
Complete all four lessons, reproduce the worked checks, run the lab, and correct the weekly quiz. Treat AI output as candidate evidence until independently verified.
CAD, CAE, PLM, requirements, and test-data landscapes
Why this lesson matters
Engineering information fragments because each tool optimizes a local job, organization, and data model. Integration begins with responsibility, not software procurement.
Learning objectives
Define and distinguish PDM/PLM and Repository.
Apply the lesson method to the worked cad, cae, plm, requirements, and test-data landscapes case.
Evaluate evidence, uncertainty, and AI-assisted output before making a claim.
Readiness check
Before continuing, explain what decision this topic supports and name one upstream source that must be controlled.
Check your response
A sound answer names a specific engineering decision, its configuration, and a controlled requirement, model, dataset, interface, or standard that constrains the work.
Core idea
Map tools by the engineering objects they author, consume, transform, approve, and archive. Then assign authority and interfaces for requirements, geometry, materials, analyses, manufacturing, test, measurement, and decisions.
Key concepts
PDM/PLM
Systems that control product definitions, configurations, changes, and lifecycle records.
Repository
A managed location for artifacts, not automatically an authoritative or semantically integrated system.
System of record
The governed source for a defined information class and process.
Data lineage
The recorded path from source through transformations to derived data and decisions.
Step-by-step explanation
Inventory engineering objects and decisions before listing tools.
For each object, name author, approver, consumers, update frequency, and retention need.
Assign systems of record by scope and lifecycle state.
Map transformations and manual handoffs, including unit and identifier changes.
Prioritize integration risks where high-consequence decisions depend on repeated re-entry or ambiguous ownership.
Worked example
A company stores requirements in spreadsheets, CAD in PDM, analyses on personal drives, test data on a lab server, and decisions in meeting minutes.
1
Identify controlled objects in each location and distinguish storage from authority.
2
Trace one release decision across requirement, CAD revision, analysis run, test dataset, and review minute.
3
Mark manual ID translation, copied inputs, missing owner, inaccessible raw data, and ambiguous status.
4
Prioritize the release chain before attempting a company-wide data lake.
Result. The first architecture target is a narrow, high-value evidence chain with explicit ownership and stable links, not migration of every file.
Independent check. Every critical object has one defined authority for its scope and every transformation has an owner and validation rule.
Common misconceptions
Misconception
Correction
PLM is automatically the single source of truth
Authority depends on governed scope and actual process. PLM may link to specialist sources rather than contain every detail.
A tool output closes the question
A result remains a candidate until its inputs, method, configuration, uncertainty, and relevance have been checked.
Diagnostic questions
What should be integrated first?
Information flows with high decision consequence, frequent change, weak ownership, or costly manual reconciliation.
What would make this work reproducible?
Controlled inputs, method or code, versions, assumptions, outputs, and a stated interpretation tied to the decision.
Practice ladder
Basic
Build a tool-to-object responsibility matrix for a five-tool project.
Intermediate
Trace one requirement through CAD, analysis, test, and decision systems.
Advanced
Prioritize three integration investments using consequence, re-entry rate, detectability, and change frequency.
AI-assisted engineering task
Ask AI to classify a tool inventory by authored objects and candidate authorities. Do not allow it to infer current governance from product names.
How to prove the AI output yourself
Interview actual data owners.
Walk one real change end to end.
Compare documented workflow with files and approvals actually used.
Retrieval and spaced review
Answer closed-notes today, then again after 1, 3, 7, and 30 days.
Define PDM/PLM.
Systems that control product definitions, configurations, changes, and lifecycle records.
What role does Repository play here?
A managed location for artifacts, not automatically an authoritative or semantically integrated system.
What must a reviewer be able to reconstruct?
Every critical object has one defined authority for its scope and every transformation has an owner and validation rule.
End-of-lesson summary
Map tools by the engineering objects they author, consume, transform, approve, and archive. Then assign authority and interfaces for requirements, geometry, materials, analyses, manufacturing, test, measurement, and decisions.
Student notes
For each tool, write what it owns, what it copies, what it transforms, and which decision would fail if it were wrong.
Use the current project environment, including shared drives and email, instead of an ideal enterprise architecture.
6.2
Metadata, versioning, provenance, and configuration management
Why this lesson matters
A valid file can still be applied to the wrong product, revision, operating condition, or decision.
Learning objectives
Define and distinguish Metadata and Version.
Apply the lesson method to the worked metadata, versioning, provenance, and configuration management case.
Evaluate evidence, uncertainty, and AI-assisted output before making a claim.
Readiness check
Before continuing, explain what decision this topic supports and name one upstream source that must be controlled.
Check your response
A sound answer names a specific engineering decision, its configuration, and a controlled requirement, model, dataset, interface, or standard that constrains the work.
Core idea
Metadata makes identity and context queryable. Versioning records artifact evolution; configuration management identifies compatible sets; provenance records origin and transformation. These controls overlap but are not interchangeable.
Key concepts
Metadata
Structured data describing identity, meaning, ownership, status, context, and relationships.
Version
A distinguishable state in an artifact's history.
Baseline
An approved configuration used as a reference for controlled change.
Configuration item
An element placed under configuration control because its identity and changes matter.
Step-by-step explanation
Define stable identity separately from version.
Record semantic fields such as units, reference frame, coordinate system, sample rate, and quantity definition.
Create baselines that name compatible requirement, design, model, test, and software versions.
Record derivation and transformation lineage.
Use change control to review impact, approve disposition, and preserve history.
Worked example
Simulation run R44 used CAD-C, material card MAT-6, solver 2025.2, mesh script v3, and load case LC-07B. CAD-D exists, but R44 remains displayed as the latest stress result.
1
Do not call R44 current merely because its timestamp is newest.
2
Compare the release baseline with R44's configuration tuple.
3
Mark R44 valid for baseline C and suspect for baseline D until geometry impact is assessed.
4
Preserve the old result for decision history and create a new run rather than relabeling R44.
Result. Version and configuration semantics prevent a correct historical result from becoming incorrect current evidence.
Independent check. A query for baseline D cannot silently return R44 as applicable without an explicit reviewed equivalence.
Common misconceptions
Misconception
Correction
Latest timestamp means applicable
Applicability depends on configuration and validity conditions, not recency alone.
A tool output closes the question
A result remains a candidate until its inputs, method, configuration, uncertainty, and relevance have been checked.
Diagnostic questions
Why preserve an obsolete result?
It explains earlier decisions, supports audit, and may remain valid for the historical baseline.
What would make this work reproducible?
Controlled inputs, method or code, versions, assumptions, outputs, and a stated interpretation tied to the decision.
Practice ladder
Basic
Distinguish file version, product revision, baseline, and run identifier.
Intermediate
Create a configuration tuple for an FEA run and a physical test.
Advanced
Design equivalence rules for reusing a test after a minor drawing change.
AI-assisted engineering task
Ask AI to extract candidate configuration metadata from run logs, returning unknown for missing values.
How to prove the AI output yourself
Compare extracted fields with native tool metadata.
Check controlled baseline membership.
Reproduce the run or validate a checksum for critical inputs.
Retrieval and spaced review
Answer closed-notes today, then again after 1, 3, 7, and 30 days.
Define Metadata.
Structured data describing identity, meaning, ownership, status, context, and relationships.
What role does Version play here?
A distinguishable state in an artifact's history.
What must a reviewer be able to reconstruct?
A query for baseline D cannot silently return R44 as applicable without an explicit reviewed equivalence.
End-of-lesson summary
Metadata makes identity and context queryable. Versioning records artifact evolution; configuration management identifies compatible sets; provenance records origin and transformation. These controls overlap but are not interchangeable.
Student notes
Write the complete configuration tuple at the top of every analysis and test note.
Use several meanings of version in one example. Students should feel why a single version column is insufficient.
6.3
STEP / ISO 10303 and product-data interoperability
Why this lesson matters
Geometry that opens successfully in another tool may still lose product and manufacturing information, assembly structure, units, validation properties, or design intent.
Learning objectives
Define and distinguish STEP and Application protocol.
Apply the lesson method to the worked step / iso 10303 and product-data interoperability case.
Evaluate evidence, uncertainty, and AI-assisted output before making a claim.
Readiness check
Before continuing, explain what decision this topic supports and name one upstream source that must be controlled.
Check your response
A sound answer names a specific engineering decision, its configuration, and a controlled requirement, model, dataset, interface, or standard that constrains the work.
Core idea
STEP, formally ISO 10303, provides standardized product-model data representations and exchange mechanisms used across CAD, CAE, PDM, manufacturing, and inspection. Conformance to a format improves interoperability but does not guarantee complete or correct transfer for every intended use.
Key concepts
STEP
The ISO 10303 family for computer-interpretable representation and exchange of product data.
Application protocol
A scoped information model for a class of product-data use, such as AP242.
PMI
Product and manufacturing information such as dimensions, tolerances, datums, and annotations.
Validation property
A reference value, such as area, volume, or centroid, used to check translation fidelity.
Step-by-step explanation
Define the downstream use before selecting exchange content.
Specify the required STEP application protocol, edition, geometry, assembly, PMI, and metadata.
Export with units and validation properties.
Import into the target and perform semantic and geometric conformance checks.
Record translator versions, warnings, deviations, and acceptance disposition.
Worked example
A bracket STEP file imports with the expected shape. The receiving system reports volume 178.6 cm³, while the source CAD reports 181.0 cm³.
Check unit conversion, suppressed features, geometry healing, and translator warnings.
3
Compare mass properties, face count, PMI association, assembly identity, and critical dimensions against the intended downstream use.
4
Reject or conditionally accept based on predeclared tolerances and consequence, not visual similarity.
Result. A viewable solid is not sufficient evidence of a faithful product-data exchange. The 1.33% volume gap triggers investigation under the translation acceptance plan.
Independent check. Source and target agree on required validation properties and semantic PMI within controlled criteria.
Common misconceptions
Misconception
Correction
Successful import proves interoperability
The file may open while losing semantics, identity, precision, PMI association, or validation properties.
A tool output closes the question
A result remains a candidate until its inputs, method, configuration, uncertainty, and relevance have been checked.
Diagnostic questions
What should determine exchange checks?
The downstream engineering use and consequence of missing or changed information.
What would make this work reproducible?
Controlled inputs, method or code, versions, assumptions, outputs, and a stated interpretation tied to the decision.
Practice ladder
Basic
Explain why STL and STEP are not interchangeable for controlled product definition.
Intermediate
Design a translation acceptance checklist for FEA preprocessing.
Advanced
Decide which PMI and validation properties must survive a supplier-to-inspection workflow.
AI-assisted engineering task
Ask AI to summarize translator logs and group warnings by geometry, PMI, units, assembly, and metadata. It may not waive warnings.
How to prove the AI output yourself
Compare validation properties numerically.
Inspect critical features and PMI association.
Run downstream reference tests and record translator versions.
Retrieval and spaced review
Answer closed-notes today, then again after 1, 3, 7, and 30 days.
Define STEP.
The ISO 10303 family for computer-interpretable representation and exchange of product data.
What role does Application protocol play here?
A scoped information model for a class of product-data use, such as AP242.
What must a reviewer be able to reconstruct?
Source and target agree on required validation properties and semantic PMI within controlled criteria.
End-of-lesson summary
STEP, formally ISO 10303, provides standardized product-model data representations and exchange mechanisms used across CAD, CAE, PDM, manufacturing, and inspection. Conformance to a format improves interoperability but does not guarantee complete or correct transfer for every intended use.
Student notes
Never write 'imported successfully' without listing the content and properties actually checked.
If STEP tools are unavailable, use before-and-after metadata and validation-property tables. The lesson is evidence of exchange fidelity.
6.4
Schemas, JSON, CSV, SQL, and AI-ready engineering data
Why this lesson matters
AI systems amplify whatever structure, omissions, and semantic ambiguity they receive. Data architecture quality bounds AI-assisted engineering quality.
Learning objectives
Define and distinguish Schema and Primary key.
Apply the lesson method to the worked schemas, json, csv, sql, and ai-ready engineering data case.
Evaluate evidence, uncertainty, and AI-assisted output before making a claim.
Readiness check
Before continuing, explain what decision this topic supports and name one upstream source that must be controlled.
Check your response
A sound answer names a specific engineering decision, its configuration, and a controlled requirement, model, dataset, interface, or standard that constrains the work.
Core idea
A schema defines allowed entities, fields, types, units, relationships, and constraints. JSON is useful for nested interchange, CSV for flat tables, and SQL for constrained persistent relationships. Choose representations by data behavior, then validate at boundaries.
Key concepts
Schema
A machine-readable or documented contract for data structure and semantics.
Primary key
A stable value that uniquely identifies a database row or entity.
Foreign key
A constrained reference to another entity that enforces relationship integrity.
Data contract
Agreed structure, meaning, quality rules, ownership, and change policy for exchanged data.
Step-by-step explanation
Model entities and relationships from engineering questions.
Normalize persistent data enough to avoid inconsistent duplicates while preserving usable queries.
Version schemas and provide migrations for controlled changes.
Expose AI only to authorized, relevant fields and require outputs that preserve source identifiers.
Worked example
A CSV uses columns `value`, `units`, and `test`, but `value` sometimes means force, temperature, or pass/fail code. An AI summary compares rows numerically.
1
Stop the comparison because the schema does not identify quantity kind or data type.
2
Create fields for observation_id, quantity, value_numeric, value_text, unit, location, time, uncertainty, test_id, and configuration.
3
Use controlled quantity and unit vocabularies with type validation.
4
Preserve raw source and transformation lineage before generating summaries.
Result. AI readiness comes from explicit semantics and traceability, not from converting files to a format accepted by a model API.
Independent check. Invalid quantity-unit combinations and missing configuration fail validation before analysis or AI use.
Common misconceptions
Misconception
Correction
Structured data is automatically meaningful
Fields can be perfectly structured and semantically ambiguous, inconsistent, or disconnected from authority.
A tool output closes the question
A result remains a candidate until its inputs, method, configuration, uncertainty, and relevance have been checked.
Diagnostic questions
What makes data AI-ready?
Relevant controlled semantics, quality, provenance, permissions, configuration, and source-linked outputs.
What would make this work reproducible?
Controlled inputs, method or code, versions, assumptions, outputs, and a stated interpretation tied to the decision.
Practice ladder
Basic
Choose JSON, CSV, or SQL for five engineering data cases and justify each choice.
Intermediate
Design a schema for calibrated temperature observations.
Advanced
Plan a backward-compatible schema migration that adds uncertainty and location to existing test rows.
AI-assisted engineering task
Ask AI to propose a schema from sample records, then challenge it with nulls, mixed units, revisions, and one-to-many relationships.
How to prove the AI output yourself
Validate against real query needs.
Use database constraints and test cases.
Review semantics with domain owners.
Measure information loss during migration.
Retrieval and spaced review
Answer closed-notes today, then again after 1, 3, 7, and 30 days.
Define Schema.
A machine-readable or documented contract for data structure and semantics.
What role does Primary key play here?
A stable value that uniquely identifies a database row or entity.
What must a reviewer be able to reconstruct?
Invalid quantity-unit combinations and missing configuration fail validation before analysis or AI use.
End-of-lesson summary
A schema defines allowed entities, fields, types, units, relationships, and constraints. JSON is useful for nested interchange, CSV for flat tables, and SQL for constrained persistent relationships. Choose representations by data behavior, then validate at boundaries.
Student notes
Write three questions the data must answer before choosing tables or files.
Avoid presenting SQL normalization as the course goal. The goal is reliable engineering queries and controlled semantics.
LAB 6
Lab 6: Build a simple evidence dashboard report
Lab objective
Query structured evidence records and generate a static HTML report that shows coverage, status, configuration, and uncertainty flags.
Engineering context
A design review needs a reproducible status view without treating the dashboard as the authority.
Input data
Requirement and evidence dictionaries
Typed link records
A current baseline
Step-by-step task
Compute coverage
Flag stale or draft evidence
Generate an HTML table
Include source identifiers and a generation timestamp
Python code
from datetime import datetime, timezone
from html import escape
requirements = {"BRK-001": "mass <= 0.50 kg", "BRK-002": "limit load 2.0 kN"}
evidence = {
"E-MASS": {"status": "approved", "baseline": "C", "result": "0.46 kg"},
"E-STRESS": {"status": "reviewed", "baseline": "B", "result": "112 MPa"},
}
links = {"BRK-001": ["E-MASS"], "BRK-002": ["E-STRESS"]}
baseline = "C"
rows = []
for rid, text in requirements.items():
linked = links.get(rid, [])
flags = []
for eid in linked:
if evidence[eid]["baseline"] != baseline:
flags.append(f"{eid}: stale baseline")
rows.append((rid, text, ", ".join(linked) or "none", "; ".join(flags) or "none"))
body = "".join("<tr>" + "".join(f"<td>{escape(cell)}</td>" for cell in row) + "</tr>"
for row in rows)
stamp = datetime.now(timezone.utc).isoformat()
report = f"<h1>Evidence review</h1><p>Generated {stamp}</p><table>{body}</table>"
print(report)
Explanation of code
Step 1 compute coverage Step 2 flag stale or draft evidence Step 3 generate an HTML table Step 4 include source identifiers and a generation timestamp
Expected output
An HTML fragment with two requirement rows and a stale-baseline flag on E-STRESS.
Interpretation
The report is a generated view. Reviewers follow identifiers to authoritative evidence before accepting a claim.
Common errors
Hiding the baseline rule
Displaying pass without uncertainty or status
Letting the report become an editable system of record
Extension tasks
Write the report to a file
Add orphan evidence
Add risk-prioritized sorting and provenance links
Reflection questions
What is authoritative here?
Why include generation time?
What would make the dashboard misleading?
PROJECT
Mini-project 2: Product information architecture
Deliverable
A tool and authority map, schema, configuration tuple, STEP exchange acceptance plan, and generated evidence report for a small mechanical product.
Required checks
At least five systems or repositories, three controlled transformations, schema constraints, one translation validation property, and one deliberate failure test.
WEEK 6
Weekly quiz and concept check
Closed notes. Answer each item, then use the key to correct in a different color.
What should precede a tool inventory?
Distinguish version and configuration.
What does STEP address?
Why is visual geometry comparison insufficient?
What is a schema?
Why does poor data architecture weaken AI?
Answer key
1. The engineering objects, decisions, owners, and lifecycle responsibilities.
2. Version is an artifact state; configuration is a compatible set representing a product or analysis state.
3. Standardized computer-interpretable representation and exchange of product-model data.
4. It can miss semantic, PMI, assembly, unit, and validation-property differences.
5. A contract for data entities, fields, types, semantics, relationships, and constraints.
6. It supplies ambiguous, stale, untraceable, or unauthorized inputs and prevents evidence-linked verification.