Full Graph Expansion Plan
Purpose: describe how the current 36-node, 41-edge expanded concept graph
(concept_graph_expanded.json) is upgraded into a fuller Atlas graph that
includes every Markdown node currently authored under 02_concepts/,
04_material_systems/, and 05_experiments/.
1. Current graph status
Source: 07_visualization/graph_validation_report.md.
- Nodes in expanded graph: 36
- Edges in expanded graph: 41
- Markdown nodes not yet in the graph: 53 (see
graph_validation_report.mdsection "Content nodes not yet represented in expanded graph").
The expanded graph is a curated high-level backbone that captures the Chapter 1–10 argument of Victor C. Li's ECC book. It intentionally omits application-specific nodes, test-specific nodes, and durability/sustainability subnodes so that the top-level narrative stays readable.
2. Design goals for the fuller graph
- Every Markdown node under
02_concepts/,04_material_systems/, and05_experiments/should be reachable as a graph node. - The book-anchored backbone from
concept_graph_expanded.jsonmust be preserved unchanged. New nodes and edges are additive. - New edges must correspond to a documented relation in
07_visualization/relationship_rules.mdor to a cross-link already implied by07_visualization/cross_link_map.md. - Edges that are not yet supported by book text should carry
status = candidateand be validated againstchapter_key_evidencefiles or explicit book text before promotion tostatus = confirmed.
3. Node addition strategy
Split the 53 omitted nodes into four ingestion tiers so the fuller graph can be built in reviewable slices.
Tier 1 — micromechanics and mechanics subnodes
Nodes:
micromechanics_based_design, strain_hardening_criteria,
fiber_bridging_law, flaw_design, distributed_microcracking,
tension_stiffening, long_term_strain_capacity,
direct_tensile_test, crack_width_distribution,
compression_behavior, flexural_test, shear_behavior,
fatigue_behavior, creep_behavior.
Rationale: these nodes densify the Chapter 2 and Chapter 4 core of the backbone. They already exist as Markdown and are heavily referenced from concept files.
Tier 2 — durability and sustainability subnodes
Nodes:
permeability, sorptivity, freeze_thaw, abrasion_wear,
cover_spalling, restrained_shrinkage, reflective_cracking,
life_cycle_analysis, bridge_deck_lca, alternative_binder,
industrial_waste_streams, green_fiber.
Rationale: these connect to Chapter 7 and Chapter 8 evidence and refine the durability and sustainability side of the backbone.
Tier 3 — multifunctional and smart infrastructure subnodes
Nodes:
self_healing_mechanisms, self_healing_robustness,
piezoresistivity, electrical_impedance_tomography,
phase_change_materials, photocatalytic_function,
functional_additive_effects, structural_recovery.
Rationale: these refine the Chapter 10 multifunctional cluster and connect into resilience and smart infrastructure.
Tier 4 — material systems and application-specific nodes
Nodes:
R_ECC subgraph, coupling_beams, link_slab, composite_bridge_deck,
tunnel_lining, water_tunnel_lining, patch_repair, dam_repair,
irrigation_channel_repair, earth_retaining_wall_repair,
external_insulation_wall, rigid_frame_railway_bridge,
steel_concrete_interaction_zone, pavement_overlay,
self_consolidating_ecc, sprayable_ecc, extruded_ecc,
low_fiber_egc.
Rationale: these are book-level Chapter 9 application demonstrations and material variants. They plug into the applications sub-tree and to their governing performance domain (resilience, durability, or sustainability).
4. Edge derivation rules
For each new node, candidate edges are proposed using this precedence:
- If the node file already contains a "Linked nodes" or "연결 개념" section,
emit one edge per link with the semantically closest relation from
relationship_rules.md. - If the node belongs to a cluster in
cross_link_map.md, emit edges to the cluster's governing concept (for exampledurability,resilience,sustainability,applications,multifunctional_ecc). - If the node is an experiment, emit
measures/observesedges to the mechanical property or behavior it evaluates. - If the node is a material system, emit an
is_aedge to the closest parent material class (R_ECC,green_ecc,multifunctional_ecc, orecc) anddemonstratesedges to the application it targets.
All Tier 1–4 edges are emitted with status = candidate in
full_graph_candidate_edges.csv unless the exact (source, target, relation)
triple already exists in concept_graph_expanded.json, in which case
status = confirmed_expanded.
5. Deliverables
full_graph_candidate_edges.csv(this checkpoint) — candidate edge list.- Follow-up checkpoints:
full_graph.json— merged node+edge JSON that supersedesconcept_graph_expanded.jsonwithout deleting it.full_graph.md— human-readable rendering, grouped by tier.full_graph_validation_report.md— validation of degree, orphans, and unresolved link targets.
6. Validation checkpoints
Before promoting status = candidate edges to status = confirmed:
- Verify the source Markdown file exists and contains the referenced link.
- Verify the target node has a Markdown file or is an approved graph-only backbone node.
- Verify that the relation label is one of the labels enumerated in
relationship_rules.md. - If the relation encodes a numeric or experimental claim, ensure the
supporting evidence is captured in a chapter evidence file or is marked
needs verificationin the concept file.
7. Out-of-scope for this checkpoint
- Merging paper cards (
03_papers/*) into the graph. Papers will be linked via a separatecites/supportsedge layer in a later checkpoint. - Lab-position nodes under
06_lab_position/. These are deferred perPROJECT_RULES.md. - Teaching module nodes under
08_teaching/. These will be added when the teaching layer stabilizes.