Chapter at a Glance

Connected engineering information lets us follow dependencies and question what a change affects.

Learn / Chapter 03

Digital Threads and Knowledge Graphs

A proposal now passes. Which earlier conclusions can we still use?

Learning objective: trace a conclusion to its definitions and execution context, then explain which checks a change requires.

Prerequisite: Engineering Information Models.

A conclusion needs a context

The monitoring team still needs useful coverage of an industrial area. Two proposals appear to resolve the reference configuration’s failure: relax the coverage criterion or assume a larger sensor footprint. A reviewer asks whether the earlier analysis supports either proposal. Looking only at PASS, or at an unchanged requirement ID, cannot answer that question.

We need to reconstruct how the conclusion was reached. Which criterion was evaluated? Which configuration supplied the inputs? Which model interpreted them, under which assumptions? Which source revision and execution produced the record? These questions connect operational intent, engineering definition and execution evidence without treating them as interchangeable.

In this repository, a digital thread is that inspectable set of connections. A knowledge graph describes entities joined by relationships with explicit meanings. Here a small table is enough to inspect those meanings. The useful engineering capability is following and questioning the connections; a particular graph tool does not establish their correctness.

Follow backward before reasoning forward

Begin with an actual execution record when reviewing retained evidence. Read its exact evaluated definitions, calculation inputs, criterion and method. Check the model version, source commit, working-tree status, timestamp and execution context. A record from a dirty local checkout cannot be reproduced from its commit alone; its captured definitions matter too. A CI label does not remove the need to inspect what ran.

Next follow the requirement’s motivating need. The source_need relationship explains why the requirement exists; it is not a numerical input to the coverage formula. Follow the verification bindings to the scenario, executable model and registered test. The scenario’s model reference tells us how its input fields are interpreted. Each step answers a different question.

The table below projects validated fields from the canonical definitions bundled with this site. Field paths are relative to their source record. It describes definition relationships, not a saved execution. No retained execution record is loaded on this page. Use the reference evidence instructions or proposal analysis instructions to inspect a real record.

Definition connections from this site build
SourceFieldTargetMeaning
REQ-COV-001source_needNEED-001Motivating operational need
REQ-COV-001verification.scenario_idSCN-001Reference inputs for verification
REQ-COV-001verification.model_idMODEL-COV-001Model used to estimate the measure
REQ-COV-001verification.test_idTEST-COV-001Registered reference test
SCN-001model_idMODEL-COV-001Model interpreting these inputs
TEST-COV-001requirement_idREQ-COV-001Criterion evaluated by the reference test
TEST-COV-001scenario_idSCN-001Reference scenario evaluated by the test
TEST-COV-001model_idMODEL-COV-001Executable model used by the test

Connectivity is a starting point

A stable identifier lets us discuss a record across revisions. It does not prove that its contents stayed the same. Both proposal trials keep the existing identifiers and relationships while changing a definition. The same table can therefore accompany different numerical evaluations. To decide whether evidence applies, compare the evaluated definition and execution context, not merely the connected IDs.

Forward tracing identifies candidates for review. It does not imply that every connected calculation must rerun. A threshold feeds the evaluator; the radius feeds the coverage calculation, whose output feeds the evaluator. The motivating need supports interpretation and justification. Treating all arrows as identical would obscure these distinctions and could miss a necessary engineering review.

For a threshold-only change, geometry remains unchanged, but the criterion needs fresh evaluation. For a radius change, calculate coverage again and then evaluate it. This workspace deliberately computes both baseline and proposal through the same model on every valid trial; it does not implement selective reuse or caching. The distinction concerns why each check is needed, not an optimization promised by the interface.

Experiment: explain what must be checked

  1. Reset. Follow the requirement’s source_need and verification bindings in the table. Predict which connections will change in each trial.
  2. Try the threshold example. Inspect the exact differences and validated proposed definition. Explain why the geometry can stay the same.
  3. Reset, then try the radius example. Explain why fresh geometric calculation is needed even though the relationship table is unchanged.
  4. Break the need reference. Observe the diagnostic and cleared result, then restore it. Does a resolvable need justify the threshold?
  5. Suppose a wall blocks part of the area. Without editing a number, identify which assumption and conclusion need review.

A definition change, not an approval

Temporary, unapproved browser proposal. This workspace changes a copy of the definitions. It does not edit YAML, execute TEST-COV-001, retain evidence or approve a requirement. Reset restores the baseline; leaving or reloading discards edits.

Canonical baseline from this site build

Requirement REQ-COV-001: 50%. Scenario SCN-001: radius 35, center (50, 50), area 100 × 100 model units.

Under reference scenario SCN-001 and the V0 model assumptions, the monitoring configuration shall cover at least 50 percent of the defined area geometrically.

FAIL — 38.52% estimated coverage. Margin: -11.48 percentage points.

This is a build-time calculation, not a displayed CI verification artifact. The baseline stays unchanged while you edit.

Temporary proposal controls

Threshold: 0–100%. Radius: 0–200 model units. Choosing a different change restores that field’s baseline value.

This edit applies only to the temporary copy. Its source_need must resolve to the existing need.

Enable JavaScript to validate and evaluate a temporary proposal.

Exact field differences

The requirement statement and basis change with the threshold so the proposed definition stays consistent. These differences describe the current trial, not a committed revision.

Canonical value → temporary proposed value
FieldBaselineProposal
Inspect the validated proposed definition
No proposal evaluated yet.

Both calculations use MODEL-COV-001 v0.1.0: a 100 × 100 cell-center area estimate. No physical sensor calibration, terrain, occlusion, directional effects, time, uncertainty, faults or communications are modeled. A PASS establishes neither operational suitability nor an optimal design.

Compare your impact reasoning

The threshold trial changes the requirement criterion, statement and unapproved basis. It needs evaluation against that exact definition; it has not improved coverage. The radius trial changes the scenario and needs calculation plus evaluation. Neither changes the eight relationship roles. A broken reference blocks validation and analysis. A wall challenges the ideal unobstructed footprint assumption: this model cannot calculate the blocked region, so engineering review needs a suitable model or other evidence. Restoring a reference or obtaining PASS supplies neither that evidence nor requirement approval.

Review assumptions as well as numbers

The ideal sensor covers a circle without terrain, occlusion or directional effects. If an obstruction challenges that interpretation, repeating the same inputs can repeat the same number while losing relevance to the monitoring decision. There is no justified correction factor in this model. Do not manufacture one to preserve a convenient conclusion.

Likewise, satisfying the geometric criterion does not choose an optimal design. Cost, installation, reliability and actual sensing performance are absent. The thread helps expose where additional justification belongs. It cannot supply missing operational knowledge or turn an unapproved proposal into an accepted system definition.

Keep the execution attached to its definitions

For a repository trial, follow the isolated proposal exercise from an identified checkout. Install with npm ci, run npm run check, generate reference evidence with npm run evidence, then analyze the separate fixtures with npm run proposals. Preserve the canonical baseline and its reference assertions.

Inspect the generated proposal record’s complete baseline and proposed definitions, differences and calculation. Its metadata records the actual revision and working-tree state. It explicitly does not claim execution of TEST-COV-001 against the proposal. The registered reference test continues independently against the canonical configuration. Browser trials remain temporary, unapproved copies with no retained execution artifact.

CI artifacts are revision-bound, expire after 30 days and may require sign-in. Reproduce at the recorded revision instead of substituting today’s sources for an older execution. A fresh result needs its own context even when its number matches an earlier one. That discipline makes a thread useful for review rather than merely persuasive to look at.

Check your reasoning

Answer before revealing each explanation.

1. All eight links are unchanged. Does earlier evidence apply?

Not necessarily. IDs and relationship roles can stay fixed while definitions change. Inspect exact evaluated definitions and execution context.

2. Which check differs between a threshold change and a radius change?

A threshold change requires evaluation of a changed criterion with unchanged geometry. A radius change requires fresh coverage calculation and evaluation. The workspace computes both paths rather than caching geometry.

3. Validation succeeds, but an obstruction is discovered. What now?

Review the unobstructed-footprint assumption and the conclusion’s applicability. Structural validity cannot justify a physical assumption. This model supplies no obstruction calculation.

4. Does a path from need to test prove the requirement is verified?

No. It shows registered relationships. Verification requires inspecting an actual execution, its definitions, method and outcome; suitability still requires engineering judgment.

Next: make review repeatable

We can now follow a conclusion backward and explain a proposed change forward. Future automation should preserve those meanings, explicit limits and review responsibilities. Chapter 4: Executable Engineering Models opens the numerical calculation behind these checks. Connecting more records alone does not establish which engineering decisions are justified.

← Chapter 2 · Learning path · Inspect engineering sources →