ROOT CAUSE TRACEABILITY ENGINE

Walk the failure
backward through
the digital thread
to its origin.

A bearing fractures at 14,200 hours. Traditional root cause analysis takes weeks of manual investigation. Trace walks the product knowledge graph backward from field symptom to design origin in seconds.

REVERSE GRAPH TRAVERSAL — FIELD EVENT FE-2024-3891
FIELD
Bearing housing fracture — Unit SN-48720, 14,200 hours
Fatigue crack at fillet radius · Nordic Pulp AB, Sundsvall · Field service report
AS-BUILT
As-built: PN-4420-A Rev C, lot 2024-0891
Ti-6Al-4V, tensile 924 MPa (spec: 895 min) · Supplier: Nordic Titanium AS · No deviations
MFG
5-axis CNC: fillet radius 1.48mm (nominal 1.50mm, tol 1.40–1.60)
Within spec · Ra 0.8µm · No NCR · Cpk 1.34
CHANGE
ECO-2023-0892: Fillet radius reduced 2.0mm → 1.5mm (weight optimization)
Approved 2023-04-12 · Impact analysis: no qualification impact · FEA re-run assumed steady-state
SIM
FEA: SIM-STR-2023-047 — steady-state, max stress 412 MPa, FoS 2.2
No cyclic fatigue analysis performed · Load case: steady-state only · Simulation gap identified
ROOT
Root cause: Simulation assumed steady-state — field telemetry shows 40% higher cyclic amplitude
Origin: ECO-2023-0892 · Gap: no fatigue analysis at reduced fillet · Corrective: ECR-2024-0447
THE INVESTIGATION GAP

Traditional root cause analysis is an archaeological expedition through disconnected systems. Trace makes it a graph query.

Every week spent investigating a field failure is a week the root cause remains unaddressed in the active fleet.

Weeks
Average time for manual root cause investigation across PLM, ERP, QMS, and field service systems
INDUSTRY AVERAGE
Seconds
Time for Trace to walk the product knowledge graph backward from field failure to design origin
AXIOM ARCHITECTURE
4
Causal domains investigated simultaneously: design, manufacturing, material, and operating conditions
MULTI-CAUSAL ANALYSIS
40%
Typical gap between assumed and actual cyclic loading when IoT telemetry is compared to simulation assumptions
ECHO DEPLOYMENT DATA

The most valuable question in product engineering is not “what failed?” — it is “why did the design allow this failure?” A bearing fracture is a symptom. The root cause might be a fillet radius reduced during weight optimization. Or a simulation that assumed steady-state loads when the field experiences cyclic loading. Or a material batch whose properties drifted. Trace finds the answer by walking backward through every layer of the digital thread.

Trace performs a reverse traversal of the Axiom product knowledge graph. Starting from a field failure event, it walks backward through the as-built configuration (Lattice), manufacturing process records (Forge ERP), engineering change history (Cascade), simulation validation results (Nexus), requirement specifications (Meridian), and the original design decisions — identifying every decision point that contributed to the failure condition. The traversal is not an investigation. It is a computation.

WHY TRACE

Five capabilities that transform root cause analysis from weeks of investigation to seconds of graph traversal.

Reverse Digital Thread Traversal
Starting from a field event, Trace walks backward through every layer: failure → as-built → manufacturing → engineering changes → simulation assumptions → design decisions → requirements. Every step is a graph edge, not a manual investigation.
Six-layer reverse traversal from symptom to design origin in seconds
As-Built vs. As-Designed Comparison
Compares the actual configuration of the failed unit against the intended design configuration. Identifies manufacturing substitutions, process deviations, or rework that may have contributed to the failure.
Exact configuration delta between what was designed and what was built
Multi-Causal Decomposition
Field failures are rarely single-cause. Trace identifies every contributing factor across four domains: design decisions, manufacturing process variations, material property deviations, and actual operating conditions that exceeded design assumptions.
Four-domain causal analysis: design, manufacturing, material, operating conditions
Simulation Assumption Audit
Cross-references simulation boundary conditions against actual field telemetry from IoT sensors. When the field sees loads the simulation never modeled, Trace identifies the gap as a potential root cause contributor.
Simulation assumptions validated against field reality via IoT telemetry
Affected Fleet Identification
Once the root cause is identified, Trace queries the entire fleet population to find all units with the same configuration, same manufacturing lot, same supplier batch, or same design revision — quantifying the affected population for proactive corrective action.
Fleet exposure computed in seconds: how many units are at risk
ROOT CAUSE INTELLIGENCE ENGINES

Eight engines. Every failure traced to its origin.

From reverse graph traversal through simulation assumption auditing to automated root cause report generation — Trace operates eight engines that transform weeks of investigation into seconds of computation.

01
Reverse Graph Traversal Engine
Six-layer backward walk · Field → As-Built → MFG → Change → Simulation → Design Origin
Cascade’s Impact Graph Analysis walks the product knowledge graph forward: given a proposed change, what will it affect? Trace inverts this traversal. Given a field failure, it walks backward: what design decision, manufacturing process, material selection, or engineering change produced the failed configuration? The reverse traversal follows typed relationships in the inverse direction: from a field event to its as-built serial record, from the serial record to the manufacturing work order, from the work order to the BOM revision, from the BOM revision to the engineering change that introduced it, from the engineering change to the simulation that validated it, and from the simulation to the requirement it was intended to satisfy. At every node in the traversal, Trace annotates the trace path with the specific data at that layer: the actual measured dimension, the actual material batch, the actual approval date, the actual simulation load case.
Inverse relationship traversal: Every typed relationship in the Axiom knowledge graph can be traversed in both directions. The “is_manufactured_from” edge from a component to a work order becomes “was_manufactured_as” when traversed backward. Trace’s query language supports directional traversal natively
Annotated trace path: The traversal output is not a list of nodes — it is a fully annotated path showing the specific data values at each layer. The engineer sees the actual measured fillet radius (1.48mm), not just “manufacturing process”. This transforms the traversal output into an investigation report
Multi-path exploration: When multiple causal paths exist (e.g., both a design change AND a material batch deviation), Trace identifies all paths and ranks them by probability of causation based on statistical correlation with similar failure patterns in the knowledge base
Temporal consistency: The traversal resolves every node at the point in time the failed unit was manufactured — not at the current revision. If the BOM has been revised six times since the failed unit was built, Trace shows the BOM revision that was effective when the unit was assembled
6
Layers traversed (field to design)
<5s
Full reverse traversal time
Multi
Path exploration for multi-causal failures
Time
Resolved at point of manufacture
02
As-Built vs. As-Designed Comparator
Serialized configuration delta · Substitution detection · Deviation impact assessment · Shop floor variance analysis
What was designed is not always what was built. Manufacturing substitutions, approved deviations, and shop-floor rework create configuration deltas between the as-designed BOM and the as-built serial record. A component substituted due to stock-out. A dimension reworked after initial machining. A material lot with properties near the specification boundary. Any of these deltas might contribute to a field failure, and most root cause investigations spend days discovering them. Trace computes the as-built vs. as-designed delta automatically by comparing the failed unit’s serialized build record (from Lattice’s as-built traceability) against the BOM revision that was effective at the time of manufacture (from Lattice’s temporal BOM resolution).
Component-level delta: For every component in the failed unit, Trace compares the as-built part number + revision against the as-designed BOM entry. Substitutions, alternate components from the AVL, and non-standard revisions are flagged
Process deviation correlation: Links approved manufacturing deviations (from Forge ERP NCR records) to the specific serial numbers they affected. If the failed unit was built under an approved deviation, Trace surfaces the deviation as a potential contributor
Material property variance: Compares the material test report values for the failed unit’s components against the population statistics. Properties near specification boundaries (>2σ from population mean) are flagged as potential contributors
Serial
Unit-level as-built resolution
Delta
Component-by-component comparison
NCR
Deviation linkage to affected SNs
Material property boundary flagging
03
Multi-Causal Decomposition
Four-domain analysis · Weighted contribution scoring · Ishikawa auto-generation · Contributing factor ranking
Field failures almost never have a single root cause. A fatigue fracture might result from a combination of a reduced fillet radius (design), a surface finish near the upper limit (manufacturing), a material batch with lower-than-typical fatigue life (material), and operating loads 40% higher than assumed (operating conditions). Trace decomposes every failure into contributions from four causal domains and assigns weighted contribution scores based on deviation magnitude, statistical correlation with failure population, and historical failure pattern similarity.
Design contribution: Evaluates whether the design geometry, tolerance allocation, material selection, or safety factor at the failure location was adequate for the actual operating conditions. Cross-references simulation results (Nexus) against field telemetry (Echo IoT)
Manufacturing contribution: Assesses whether manufacturing process variations (within tolerance but biased toward limits) contributed to the failure. Compares the failed unit’s process data against the population to identify process-driven risk factors
Material contribution: Evaluates material batch properties (mechanical, chemical, metallurgical) from the failed unit against the specification range and population distribution. Identifies batches with properties that, while compliant, are statistically associated with higher failure rates
Operating condition contribution: Compares actual operating envelope (from IoT telemetry or customer-reported conditions) against the design qualification envelope. Identifies exceedances in temperature, load, speed, or environmental exposure that exceed design assumptions
4
Causal domains decomposed
Score
Weighted contribution per domain
Auto
Ishikawa diagram generation
Rank
Contributing factors by probability
04
Simulation Assumption Audit
Boundary condition vs. field reality · Load case gap detection · FEA-to-telemetry overlay · Missing analysis identification
The most insidious root cause is a simulation that was correct for the wrong assumptions. A steady-state FEA that shows a factor of safety of 2.2 is meaningless if the field experiences cyclic loading the simulation never modeled. A thermal analysis that assumes ambient temperature is misleading if the product operates in a 65°C enclosure. Trace audits every simulation linked to the failure location by extracting the boundary conditions and load cases from the Nexus simulation record and comparing them against actual field telemetry data from Echo’s IoT platform. Load magnitudes, cycle counts, temperature ranges, and vibration spectra from the field are overlaid against simulation assumptions — and gaps are flagged as potential root cause contributors.
Load case coverage analysis: Compares the set of load cases analyzed in simulation against the actual loading conditions observed in the field. Identifies load scenarios (thermal cycling, vibration harmonics, pressure transients) that occur in the field but were not represented in any simulation
Magnitude comparison: For each load case that was simulated, compares the assumed magnitude against the actual field measurement. A simulation assuming 500N cyclic load when the field measures 700N is a quantified root cause contributor
Missing analysis detection: Identifies analysis types that should have been performed but were not. For a fatigue failure: was a fatigue analysis performed? For a corrosion failure: was an environmental exposure analysis performed? Gaps in analysis coverage are explicit root cause paths
IoT
Field telemetry overlay on sim assumptions
Gap
Missing load case identification
40%
Typical sim-to-field gap discovered
Nexus
Native simulation record integration
05
Manufacturing Process Deviation Analysis
SPC data correlation · Tool wear trending · Process parameter drift · NCR-to-failure linkage
The manufacturing process is a noise source that the design must tolerate — but when process variation conspires with design margin to produce failures, finding the connection requires data that spans both domains. Trace connects field failure data from Echo with manufacturing process data from Forge ERP to identify process-driven failure contributors. For the failed unit’s serial number, Trace retrieves the complete manufacturing history: CNC parameters, in-process measurements, tool identifiers, operator IDs, machine identifiers, and environmental conditions. Statistical comparison against the broader population reveals whether the failed unit was manufactured under conditions that differ from the norm.
SPC data retrieval: For every critical dimension on the failed component, Trace retrieves the in-process measurement from the manufacturing SPC system and plots it against the population distribution. Measurements near specification limits are flagged as process-driven risk contributors
Tool wear correlation: Links the failed component to the specific cutting tool that machined it. If that tool was near end-of-life (high cumulative cut length), Trace flags tool wear as a potential contributor to surface finish degradation or dimensional drift at the failure location
Environmental condition linkage: Retrieves the ambient temperature, humidity, and coolant temperature during the manufacturing shift that produced the failed component. Extreme environmental conditions during manufacturing can affect material properties and residual stress
SPC
In-process measurement retrieval per SN
Tool
Wear state at time of manufacture
Env
Shop floor conditions during production
NCR
Deviation linkage to field outcomes
06
Material & Supplier Batch Correlation
Heat lot traceability · Material cert cross-reference · Batch-to-fleet mapping · Supplier trend detection
Material properties vary batch to batch. Within specification does not mean identical. A titanium heat lot with tensile strength at 924 MPa (spec minimum: 895) may behave differently from one at 980 MPa. Trace links the failed unit’s component to its specific material batch (heat lot, melt lot, batch number) through the supplier traceability chain, retrieves the material test report, and compares the batch properties against the population distribution of all batches received from that supplier. When a specific batch shows properties associated with higher failure rates across the fleet, Trace identifies the batch-level root cause and maps every unit in the fleet that received components from the same batch.
Batch-to-fleet mapping: Given a suspect material batch, Trace identifies every serial number in the fleet that contains a component from that batch. This affected-population computation is critical for proactive recall, inspection, or maintenance scheduling decisions
Multi-property correlation: Does not evaluate material properties in isolation. Correlates tensile strength, yield strength, elongation, hardness, and chemistry simultaneously. A batch with individually-compliant properties that form an unusual combination (high strength + low elongation) may indicate a microstructural anomaly
Supplier quality trending: Aggregates batch-level data across all deliveries from a supplier over time. Identifies long-term drift in material properties that precedes failures — connecting to Foresight’s supplier quality drift detection for proactive intervention
Batch
Heat/melt lot traceability to serial number
Fleet
Affected population from suspect batch
Multi
Property correlation (not single-value)
Trend
Supplier quality drift linkage
07
Failure Pattern Knowledge Base
Historical failure library · Pattern matching · Similarity scoring · Cross-product learning
Every root cause investigation Trace performs adds to a growing library of failure patterns. When a new failure arrives, Trace searches the knowledge base for historically similar failures based on failure mode, component type, material, manufacturing process, design geometry, and operating conditions. If a similar failure has been investigated before, the prior root cause analysis is surfaced immediately — potentially resolving the new investigation in minutes rather than weeks. The knowledge base operates across product families and programs: a fatigue failure pattern learned from an industrial pump bearing applies to a compressor bearing in a different product line that shares the same geometry and material class.
Similarity scoring: New failures are compared against the knowledge base using multi-dimensional similarity: failure mode (fracture, wear, corrosion), component geometry class, material family, manufacturing process type, and operating condition profile. High-similarity matches surface the prior root cause as a starting hypothesis
Cross-product learning: Root cause patterns are not confined to the product that experienced the failure. A thermal fatigue pattern discovered in a gas turbine blade may apply to a heat exchanger tube in a different product line. The knowledge base indexes by failure mechanism, not by product identity
Continuous enrichment: Every completed root cause investigation automatically enriches the knowledge base. The failure mode, contributing factors, weighted contributions, corrective action, and verification results are indexed for future retrieval. The knowledge base improves with every failure analyzed
Library
Growing historical failure pattern index
Match
Similarity scoring for instant hypotheses
Cross
Product learning (mechanism-indexed)
Auto
Knowledge base enrichment per investigation
08
Automated Root Cause Report Generation
Regulatory-ready 8D/A3 reports · CAPA linkage · Corrective ECR auto-generation · Fleet action recommendations
The investigation is only valuable if it produces action. Trace auto-generates a complete root cause investigation report in industry-standard formats (8D, A3, 5-Why) from the traversal data, causal decomposition, simulation audit, and manufacturing analysis. The report includes the annotated trace path, contributing factor ranking, affected fleet population, recommended corrective actions, and links to the automatically-generated draft ECR in Cascade. For regulated industries, the report satisfies FDA MDR, FAA SDR, and IATF 16949 CAPA requirements — because every data element in the report is traceable to its source in the governed product knowledge graph.
8D report auto-generation: Populates all eight disciplines from governed data: D1 (team — from RACI), D2 (problem description — from field event), D3 (containment — from fleet exposure analysis), D4 (root cause — from causal decomposition), D5 (corrective actions — from ECR linkage), D6 (verification), D7 (prevention), D8 (closure)
CAPA-to-ECR closed loop: The corrective action identified in the root cause analysis automatically generates a draft ECR in Cascade with the root cause evidence attached. The ECR traces back to the field event, through the investigation, to the corrective design change — closing the loop from field failure to design improvement
Fleet action recommendation: Based on the affected population analysis, Trace recommends fleet-level actions: inspection campaign (for latent failures), proactive replacement (for imminent failures), monitoring enhancement (for slow-developing conditions), or no action (for unique manufacturing anomalies with no population risk)
8D
Auto-generated from traversal data
CAPA
To ECR closed loop via Cascade
Fleet
Action recommendations (inspect/replace/monitor)
MDR
FDA/FAA/IATF regulatory report compliance
DEPLOYMENT EVIDENCE

Three investigations. Root cause in seconds, not weeks.

INDUSTRIAL EQUIPMENT · ROTATING MACHINERY
Bearing housing fatigue fracture traced to weight optimization ECO in 47 seconds
14,200 operating hours · Ti-6Al-4V housing · 6-layer reverse traversal
A bearing housing fractured at 14,200 hours in a pulp mill application. Traditional investigation would have taken 3-4 weeks across PLM, ERP, and simulation systems. Trace walked the product knowledge graph backward in 47 seconds: field failure → as-built record (Rev C, lot 2024-0891) → manufacturing data (fillet radius 1.48mm, within spec) → engineering change ECO-2023-0892 (fillet reduced from 2.0mm to 1.5mm for weight savings) → simulation SIM-STR-2023-047 (steady-state FEA only, no fatigue analysis) → root cause: cyclic loading amplitude 40% higher than steady-state assumption. The simulation assumed steady-state loads; the field experienced high-amplitude cyclic loading at the reduced fillet. Corrective ECR filed within 2 hours of failure report.
47s
Full root cause traversal time
6
Layers traversed to design origin
2 hr
Failure report to corrective ECR
AEROSPACE · ENGINE COMPONENTS · ITAR CONTROLLED
Material batch correlation identifies 340 at-risk units across 12 engine programs from single titanium heat lot
Fleet-wide batch mapping · 4 supplier batches analyzed · Proactive inspection campaign
A turbine blade cracked during a routine borescope inspection. Trace’s reverse traversal identified the failed blade’s material batch (heat lot TN-2023-4407 from a titanium supplier) and found that the batch exhibited unusually low elongation (8.2% vs. 12.4% population mean, though above the 6% specification minimum). Trace’s batch-to-fleet mapping identified 340 blades across 12 engine programs manufactured from the same heat lot. A targeted borescope inspection campaign on the 340 units found 7 additional blades with incipient cracking — enabling proactive replacement before in-flight failure. The supplier corrective action traced the low elongation to a furnace calibration drift in their heat treatment process.
340
At-risk units identified from batch
7
Incipient cracks found proactively
Zero
In-flight events (all caught pre-failure)
MEDICAL DEVICES · CLASS III IMPLANTABLE
Simulation assumption audit reveals 60% underestimation of cyclic fatigue in implantable cardiac lead
FDA MDR filed · 1,200 affected units · Proactive monitoring protocol established
A Class III cardiac lead exhibited premature conductor fatigue at 18 months (design life: 10 years). Trace’s simulation assumption audit compared the fatigue analysis boundary conditions against patient activity telemetry from the device’s embedded accelerometer. The original simulation assumed 70 bending cycles per minute at the lead anchor point. Patient telemetry data from 240 patients showed actual bending rates averaging 112 cycles per minute — a 60% underestimation. Trace identified 1,200 implanted units with the same lead design and generated an FDA MDR (Medical Device Report) with complete traceability from patient telemetry through simulation assumptions to the design specification. A proactive monitoring protocol was established for all affected patients, and a corrective ECR was filed to redesign the conductor geometry for the actual loading environment.
60%
Cyclic fatigue underestimation found
1,200
Affected units identified for monitoring
Auto
FDA MDR generated from trace data

“Forty-seven seconds. From the field service report to the design decision that caused the failure. The fillet radius reduction for weight savings. The simulation that only modeled steady-state loads. The actual cyclic amplitude from IoT telemetry that was forty percent higher than assumed. All connected. All traceable. What used to take my team three to four weeks of digging through five different systems, Trace computed in less than a minute.”

Director of Product Reliability
INDUSTRIAL ROTATING EQUIPMENT · 50,000 INSTALLED UNITS

“We found seven blades with incipient cracking across twelve engine programs. All from the same titanium heat lot. All with elongation values that were technically within specification but statistically anomalous. Without Trace’s batch-to-fleet mapping, we would have discovered those cracks one at a time during routine overhauls — or worse, as in-flight events. The proactive inspection campaign cost a fraction of what a single uncontained failure investigation would have.”

VP of Fleet Safety & Airworthiness
AEROSPACE ENGINE OEM · 12 ENGINE PROGRAMS · ITAR CONTROLLED

Stop investigating
failures manually.
Start tracing them
computationally.

Submit a field event. Watch Trace walk the product knowledge graph backward through every layer — from symptom to design origin — in seconds.

Or contact the Trace engineering team at trace@brindwell.com