ECSS System Engineering — Design Methods and Models (space-systems/ecss/e10-design-methods)
Use when the task is to show that the design methods, tools and models
producing a design output under ECSS-E-ST-10C clause 5.4.1.3 were
selected as appropriate to that output and shown adequate before the
output was relied on.
Domain quick reference
- Methods split into two families whose credibility rests on different
evidence. A model-based method (analytical derivation, numerical
simulation) is credible through a validation record. A rule-based
method (empirical correlation, heritage design rule) is credible
through a documented precedent whose original conditions bound the
present case. A method type outside both families is an input error.
- A model's validation status is three-valued, not two: validated
inside the domain the record covers, validated but applied outside
that domain, or unvalidated. The middle state is the one that gets
lost -- a real validation record applied to a case it never covered.
- Each model-based method contributes a confirmation weight: its base
confidence scaled by its validation status. Out-of-domain validation
contributes partial weight rather than none, because the record
still says something; an unvalidated model contributes nothing at
all, whatever base confidence it was assigned.
- The confirmation the output needs comes from the method selection
plan, not from the methods. A threshold that has never been captured
is itself a finding as soon as the methods contribute any weight --
there is no default bar to clear.
- A rule-based method whose precedent does not bound the case is
permitted, but only against a recorded rationale for relying on it.
Absence of that rationale is the finding, not the use of the rule.
- Compliance is conjunctive: the model-confirmation and rule-based
traceability lists must both be empty. A well-confirmed model does
not license an unjustified design rule beside it.
Workflow
- Sort each method behind the design output into the model-based or
rule-based family; reject an unrecognized method type.
- For each model-based method, determine its validation status from
its validation record and whether the present case falls inside the
domain that record covers.
- Weight each model's base confidence by its validation status and
sum the contributions for the output.
- Compare the total confirmation weight against the threshold
recorded in the method selection plan; record a finding for a
shortfall, and for a threshold never captured.
- For the rule-based methods, determine whether any precedent fails
to bound the case, and if so confirm a rationale is on record.
- The design output complies only when both finding lists are empty.
Pitfalls
- Collapsing validation to a yes/no flag, which files an out-of-domain
model with the fully validated ones and lets a record earned in one
regime carry a design case in another.
- Letting a high base confidence stand in for validation. Base
confidence is scaled by the validation status, so an unvalidated
model contributes nothing however confident its author.
- Assuming a default confirmation threshold when the method selection
plan never recorded one. The missing threshold is the finding; an
invented bar makes any method set appear adequate.
- Rejecting a rule-based method because its precedent does not bound
the case. Clause 5.4.1.3 permits the reliance -- what it requires is
the recorded rationale, so the check looks for the rationale.
- Offsetting an unjustified design rule against strong model
confirmation. The two checks are independent and both must pass.
Behavior contract (gate 3)
The method-family, validation-status, precedent-bounding, confirmation-
weight, model-confirmation and rule-based traceability logic is
exercised by the gate 3 contract test:
scripts/test_e10_design_methods.py against
scripts/e10_design_methods_logic.py (stdlib unittest, offline). Run:
python3 scripts/test_e10_design_methods.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and
paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
1---2name: e10-design-methods3description: "Use when select and validate the design methods, tools and models behind a design output under ECSS-E-ST-10C clause 5.4.1.3: separate model-based methods from rule-based ones, establish each model's validation status against the domain of applicability of the present design case, weight the confirmation each model contributes and compare the total against the threshold in the method selection plan, and confirm a recorded rationale exists wherever a design rule is applied outside its documented precedent. Trigger: ecss, e-st-10-system-scope, design-methods, model-validation, domain-of-applicability, design-rules, heritage-precedent, confirmation-weight."4license: Apache-2.05compliance: STANDARDS-REF6standards:7 - id: ecss8 reference-only: true9gated: false10domain: space-systems11pack: space-systems12compatibility: "agentskills.io SKILL.md; any SKILL.md host (Claude Code, Hermes, OpenClaw)"13metadata:14 source: https://skillmd.com/skills/ashfordeou/e10-design-methods15 domain: space-systems16 subdomain: ecss17 tags: [ecss, e-st-10-system-scope, design-methods, model-validation, domain-of-applicability, design-rules, confirmation-weight]18 version: 0.1.019 author: Aero Agent Skills20---2122# ECSS System Engineering — Design Methods and Models (space-systems/ecss/e10-design-methods)2324Use when the task is to show that the design methods, tools and models25producing a design output under ECSS-E-ST-10C clause 5.4.1.3 were26selected as appropriate to that output and shown adequate before the27output was relied on.2829## Domain quick reference3031- Methods split into two families whose credibility rests on different32 evidence. A model-based method (analytical derivation, numerical33 simulation) is credible through a validation record. A rule-based34 method (empirical correlation, heritage design rule) is credible35 through a documented precedent whose original conditions bound the36 present case. A method type outside both families is an input error.37- A model's validation status is three-valued, not two: validated38 inside the domain the record covers, validated but applied outside39 that domain, or unvalidated. The middle state is the one that gets40 lost -- a real validation record applied to a case it never covered.41- Each model-based method contributes a confirmation weight: its base42 confidence scaled by its validation status. Out-of-domain validation43 contributes partial weight rather than none, because the record44 still says something; an unvalidated model contributes nothing at45 all, whatever base confidence it was assigned.46- The confirmation the output needs comes from the method selection47 plan, not from the methods. A threshold that has never been captured48 is itself a finding as soon as the methods contribute any weight --49 there is no default bar to clear.50- A rule-based method whose precedent does not bound the case is51 permitted, but only against a recorded rationale for relying on it.52 Absence of that rationale is the finding, not the use of the rule.53- Compliance is conjunctive: the model-confirmation and rule-based54 traceability lists must both be empty. A well-confirmed model does55 not license an unjustified design rule beside it.5657## Workflow58591. Sort each method behind the design output into the model-based or60 rule-based family; reject an unrecognized method type.612. For each model-based method, determine its validation status from62 its validation record and whether the present case falls inside the63 domain that record covers.643. Weight each model's base confidence by its validation status and65 sum the contributions for the output.664. Compare the total confirmation weight against the threshold67 recorded in the method selection plan; record a finding for a68 shortfall, and for a threshold never captured.695. For the rule-based methods, determine whether any precedent fails70 to bound the case, and if so confirm a rationale is on record.716. The design output complies only when both finding lists are empty.7273## Pitfalls7475- Collapsing validation to a yes/no flag, which files an out-of-domain76 model with the fully validated ones and lets a record earned in one77 regime carry a design case in another.78- Letting a high base confidence stand in for validation. Base79 confidence is scaled by the validation status, so an unvalidated80 model contributes nothing however confident its author.81- Assuming a default confirmation threshold when the method selection82 plan never recorded one. The missing threshold is the finding; an83 invented bar makes any method set appear adequate.84- Rejecting a rule-based method because its precedent does not bound85 the case. Clause 5.4.1.3 permits the reliance -- what it requires is86 the recorded rationale, so the check looks for the rationale.87- Offsetting an unjustified design rule against strong model88 confirmation. The two checks are independent and both must pass.8990## Behavior contract (gate 3)9192The method-family, validation-status, precedent-bounding, confirmation-93weight, model-confirmation and rule-based traceability logic is94exercised by the gate 3 contract test:95scripts/test_e10_design_methods.py against96scripts/e10_design_methods_logic.py (stdlib unittest, offline). Run:97python3 scripts/test_e10_design_methods.py9899## Compliance100101- ECSS standards are freely downloadable (ESA); cite the source and102 paraphrase per standards-map.yaml.103- compliance: STANDARDS-REF, gated: false.