Cable-specific physics models on top of our 15-module platform engine. Every output traces to a citeable authoritative source — FERC and PUC defensible by design, not black-box scoring.
1
AHI CompositeACTIVATES WITH UTILITY DATA Asset Health Index
What it doesFuses cable age, insulation type, load history, environmental exposure, splice history, and historical failures into a single 1–5 score per cable segment.
MethodMulti-factor weighted fusion with confidence intervals — IEEE health-index methodology adapted for distribution cable.
Why utilities careThe single number an Asset Mgmt VP hands to Capex Planning to prioritize replacement programs. The meeting-table number.
2
RUL SurvivalDEMONSTRATED ON BENCHMARK Cox Proportional-Hazards
What it doesEstimates Remaining Useful Life with confidence intervals — not a single point, but a distribution.
MethodCox proportional-hazards survival model — IEEE-recognized standard cable-survival framework for two decades. Not invented; cited.
Data consumedCable attributes + failure history from utility OMS.
OutputPer-segment RUL distribution with confidence bands (e.g. "5–10 years with 80% confidence").
Why utilities careDefensible in DPU prudency reviews. Cited methodology, not vendor magic — CFOs can defend deferral decisions.
3
PD ForecasterACTIVATES WITH UTILITY DATA Partial Discharge Progression
What it doesIngests partial discharge test results (PDFs, scanned reports, or live monitor streams) and forecasts failure horizon based on PD pC/cycle trends.
MethodPD progression curve fitting + cross-cable benchmarking against IEEE data.
Data consumedUtility PD test history (we extract from PDFs / structured records) + any online PD monitor streams.
Why utilities carePD is the strongest insulation-degradation predictor known. Utility PD data is fragmented across decades of paper reports — we make it queryable.
4
Fault TriangulationACTIVATES WITH UTILITY DATA ~200 ft Fault Locator
What it doesLocalizes underground faults to ~200 ft using time-of-arrival physics across line sensors and AMI dark-sector clusters.
MethodTime-of-arrival multi-sensor triangulation + AMI topology narrowing — requires an installed sensor fleet.
Data consumedLine sensor data (if deployed) + AMI last-gasp data + GIS topology.
OutputFault location estimate within ~200 ft of true segment.
Why utilities careTranslates directly to patrol-time reduction (65% benchmark) and CAIDI improvement — fewer truck rolls, faster restoration.
5
Soil CorrosionLIVE IN THIS DEMO SSURGO → Sheath-Loss Kinetics
What it doesPulls soil pH, resistivity, chloride content, organic matter, and water-table depth from SSURGO at the cable-route geometry; computes lead-sheath corrosion rate in μm/yr.
MethodEstablished sheath-loss equations against SSURGO chemistry — geochemistry-based, not ML.
Data consumedSSURGO (we already have it nationally) + cable geometry from utility GIS.
Why utilities carePILC and lead-sheathed cables fail because of soil chemistry. Every utility has SSURGO access but very few are using it for cable risk modeling.
6
Flood HydrologyLIVE IN THIS DEMO FEMA NFHL Exposure
What it doesIdentifies cable segments in FEMA AE flood zones; quantifies percentage of each circuit in flood-exposed areas; combines with sea-level-rise projections.
Why utilities careCoastal exposure combined with sea-level-rise pressure makes flood-zone underground a growing reliability risk that has to be tracked.
7
Manhole RiskACTIVATES WITH UTILITY DATA Methane / CO IoT Classifier
What it doesClassifies risk of smoldering insulation / manhole explosions from combustible gas, temperature, and humidity — using IoT sensor data if deployed, otherwise inferred from cable type, age, and fault history.
MethodIoT classifier + Con Edison-validated thresholds from their NYC deployment.
Data consumedManhole IoT sensor streams (if deployed) + cable-condition data.
OutputPer-manhole risk index.
Why utilities careCon Ed has prevented manhole explosions in NYC using this class of sensor. Same risk exists in dense urban underground networks.
8
Capex DeferralACTIVATES WITH UTILITY DATA SAIDI per $ Optimization
What it doesGiven the prioritized cable risk list, computes which capex projects yield the most SAIDI improvement per dollar spent. Produces a ranked deferral schedule.
MethodMulti-objective optimization with SAIDI improvement, replacement cost, and risk-tolerance constraints.
OutputRanked deferral schedule with SAIDI/$ score per project.
Why utilities careThe dashboard CFOs and Capex Planning teams actually use to write next year's investment plan. Wins DPU prudency reviews.
9
AMI Dark-SectorACTIVATES WITH UTILITY DATA Last-Gasp Meter Clusters
What it doesUses AMI last-gasp signals and ping data to identify which cable segment likely failed during an outage event — narrows fault localization to ~200 ft using meter topology.
MethodTopology-aware clustering + temporal correlation across meter dropout events.
Data consumedUtility AMI data (already collected) + GIS topology.
Why utilities careUtilities already collect this data. Pure software value-add — zero new sensors needed.
Every output carries provenance JSON traceable to its source dataset (SSURGO, FEMA NFHL, NLDN, USGS, IEEE C57/C37). FERC- and PUC-defensible by design.
Why Earthflow Fuses Electrical + Environmental Data
A cable doesn't fail because of itself. It fails because of its environment. Underground risk only emerges when you fuse what the utility knows about its assets with what the environment reveals about the ground they're buried in.
Looking at an underground cable in isolation — its age, its insulation type, its load history — tells you one part of the story. But the cable doesn't fail because of itself. It fails because of its environment. Corrosive soil chemistry. Water infiltration in a flooded duct. A lightning surge upstream. A drought year that shifted ground moisture. A storm season that overloaded a parallel feeder. None of that is in your GIS or your OMS. It lives in soil chemistry, flood-zone maps, lightning records, elevation and terrain, soil-moisture sensing, wind-load profiles, and decades of storm history. Our thesis is simple: electrical data alone is not enough to understand underground risk. You have to fuse it with environmental data — at the cable-segment level — and run it through physics-grounded models.
⚡Electrical Data
GISCable geometry, age, insulation type, conductor, voltage class
AMI Dark-SectorLast-gasp clustering for outage-segment isolation
🌍What Earthflow Brings — the Environmental Data Layer
We've spent the last 18 months building this. You don't have to. The same authoritative graph powers Earthflow for Solar Development, Earthflow Underwriter (Reinsurance), and now Earthflow Underground — reusable across verticals, 100% spatial coverage for any North American utility.
Soil Data
What it providespH · resistivity · chloride · organic matter · water-table depth
CoverageNational · route-level
Flood-Hazard Zones
What it providesAE / X / VE zones · per-segment flood exposure
CoverageNational · periodic updates
Lightning Strikes
What it providesStrike density + magnitude per cable corridor
CoverageNational · 1989-present
Elevation & Terrain
What it providesElevation · slope · drainage
CoverageNational · 1-meter where available
Seismic Hazard
What it providesSeismic shear-velocity (Vs30) for ground-motion estimates
CoverageNational
Wind Profiles
What it providesWind speed · gust · load profiles for storm-season modeling
CoverageNational
Storm History
What it providesTornado · hail · wind events
CoverageNational · 70+ years
Vegetation & ET
What it providesVegetation index · evapotranspiration · surface temperature
CoverageNational · 250m / 500m
Soil Moisture
What it providesVolumetric soil-water content
CoverageNational · 10 km grid
Daily Climate
What it providesPrecipitation · temperature
CoverageNational · daily resolution
Industry Standards
What it providesCable-aging curves · transformer specs · reference benchmarks
CoverageIndustry literature (IEEE)
⚡What the Utility Provides — Critical · Nice-to-Have · Not Needed
For a v1 pilot, here's what we need from your side — and just as importantly, what we don't.
✓ Critical for v1
✓Cable network geometry — the spatial anchor for every analysis (utility GIS)
✓Cable attributes — age, type, conductor, voltage (where records exist)
✓Outage history — last 5–7 years from OMS / reliability database (the most valuable single dataset)
✓Feeder SCADA loading — load context per feeder, from your historian / PI
✓AMI meter data — topology + last-gasp clusters
~ Nice-to-have
~PD test history — we extract from PDFs / scanned reports; no need to digitize first
~Manhole inspection records — if catalogued
~Cable splice locations — where documented
~Manhole IoT sensor streams — if deployed; works fine without
~Field tan-delta test results — where available
✗ Not needed
✗Real-time SCADA streams — historical loading is enough
✗Customer billing data — out of scope
✗Operational control authority — this is decision-support, not control
✗Staff time on bulk data digitization — we do the unstructured extraction
🔧How We Get Started — Schema Auto-Mapper + Knowledge Graph
We are not going to spend six months mapping your GIS to our data model — that's the standard utility-analytics onboarding pain, and it's why most utility-vendor relationships start poorly. Three architectural choices make a 7-day live twin possible.
1
Schema Auto-Mapper
Detects field-name patterns automatically across the common utility GIS / OMS / SCADA / AMI schemas — ArcFM, Conduit Manager, generic geodatabases, and in-house variants. Produces a candidate mapping in hours, not months. We aim for 80% schema coverage in the first 48 hours; an analyst-led validation pass closes the remaining 20%.
2
Knowledge-Graph Architecture
No common data model required. We do not ask utilities to conform to one canonical schema. Earthflow runs on a semantic knowledge graph: your heterogeneous source systems plug in via the auto-mapper and are reconciled in graph space — not by force-fitting your fields into ours. This is the architectural decision that makes multi-utility onboarding tractable.
3
Adjustable Schema — Built for Bad Data
Underground data is not clean. We don't pretend otherwise. Every output carries a provenance JSON declaring which inputs contributed and which were imputed with industry defaults. Data gaps surface as explicit confidence intervals, not hidden behind a precise number. Your data doesn't have to be complete or correct for this to work — and we won't ask you to make it so.
7 days from data drop to live risk twin. That's the engineering choice that makes this possible.
70+ authoritative datasets at 100% spatial coverage for any North American utility — fused at the cable-segment level.