Underground power cables carry no visible warning signs before they fail. Insulation degrades silently over years — through partial discharge activity, thermal cycling, and moisture ingress — until a sudden fault blacks out entire grid zones. Excavating and repairing a single underground cable fault costs utilities $340,000–$1.2 million and takes 8 to 22 hours. AI-based cable monitoring changes this equation entirely: by continuously analyzing partial discharge patterns, thermal signatures, and load current data, AI detects insulation degradation 6–14 weeks before a destructive fault occurs. Utilities deploying AI cable health monitoring report 68% fewer unplanned cable outages, 54% reduction in emergency excavation events, and 41% lower cable maintenance costs annually. Connect your cable network to OxMaint — sign up free today.
Grid Infrastructure · AI Cable Health · Predictive Fault Detection
AI-Based Underground Cable Monitoring & Fault Prediction
Detect partial discharge buildup, insulation degradation, and cable aging before they become grid-disrupting faults. Real-time AI monitoring for XLPE and paper-insulated HV/MV cables.
6–14 Week Fault Warning
Partial Discharge AI
XLPE & PILC Cables
HV/MV Grid Ready
$780K
Average cost of a single underground cable fault — excavation, repair, lost revenue
68%
Reduction in unplanned cable outages with continuous AI health monitoring
60%
Of all cable failures originate at joints — the primary AI monitoring focus point
14 wks
Maximum advance warning window AI provides before XLPE insulation failure
Cable Degradation Science
How Underground Cables Fail — And When AI Intervenes
Cable insulation does not fail suddenly. It degrades through a predictable multi-stage process. AI monitoring detects the early stages — when intervention is cheap and scheduled — rather than the terminal stage when a destructive fault is inevitable.
Stage 1
Micro-void Formation
Years 1–8
Thermal cycling and mechanical stress create microscopic voids in XLPE insulation. No visible degradation. Conventional monitoring: silent.
AI detects: Subtle PD pulse inception at <5 pC
→
Stage 2
Electrical Tree Growth
Years 8–15
Partial discharge activity begins eroding insulation channels. PD magnitude increases. Thermal hotspots appear at joints. Threshold alarms still silent.
AI detects: PD pattern shift + thermal gradient anomaly
→
Stage 3
Insulation Breakdown
Weeks before fault
Rapid PD escalation. Tree channels approach conductor. Dielectric strength collapsing. SCADA threshold alarms begin — but intervention window is closing.
AI flags: Critical — schedule excavation within 2–4 weeks
→
Stage 4
Destructive Fault
No warning
Full insulation puncture. Phase-to-ground fault. Grid outage. Emergency excavation. $340K–$1.2M repair. 8–22 hours of supply interruption.
Without AI: First indication is total failure
Monitoring Signals
What AI Reads From Your Cables — Signal by Signal
AI cable health monitoring is not a single sensor — it is a multi-signal fusion model that correlates partial discharge, thermal, and electrical data into a unified cable health index for every segment and joint in your network.
Partial Discharge
Primary Signal
PD pulse rate, magnitude (pC), and pattern classification (corona, surface, internal) are the earliest and most reliable indicators of insulation degradation in XLPE cables.
Detection lead time before fault
Up to 14 weeks
Distributed Temperature
Key Signal
Fiber-optic DTS measures temperature along the full cable length at 1-meter spatial resolution. Hotspots at joints, duct blockages, and overloaded segments are detected automatically.
Spatial resolution for fault location
±1 meter accuracy
Load Current Signature
Key Signal
Current harmonic analysis detects sheath faults, asymmetric loading, and metallic screen corrosion. No additional sensors required — uses existing current transformer installations.
Sheath fault detection accuracy
86% detection rate
Dielectric Loss (tan δ)
Primary Signal
Increasing dielectric loss factor directly maps to moisture ingress and insulation aging in paper-insulated cables. AI trends tan δ over time to predict end-of-life with 91% accuracy.
Insulation aging prediction accuracy
91% accuracy
Fault Classification
Fault Types AI Identifies — and What It Recommends
AI does not just raise an alarm — it tells you what type of fault is developing, where it is located on the cable route, and what specific maintenance action is required.
Fault Type
Primary Signal
AI Lead Time
Recommended Action
Severity
XLPE Internal Void
PD pulse pattern — internal discharge
6–14 weeks
Targeted cable section replacement
Critical
Joint Overheating
DTS thermal hotspot at joint location
3–6 weeks
Joint inspection, resin replacement
Critical
Metallic Sheath Corrosion
Screen current asymmetry + PD elevation
8–18 weeks
Sheath integrity test, cathodic protection
High
Moisture Ingress (PILC)
Dielectric loss trend — tan δ elevation
10–20 weeks
Thermal drying, end-cap resealing
High
Duct Thermal Blockage
DTS ampacity derating — thermal resistance
4–8 weeks
Duct ventilation, derating implementation
Medium
Electrical Tree Propagation
PD pattern — tree-type discharge signature
2–6 weeks
Emergency cable replacement scheduling
Critical
OxMaint Cable Health Platform
Turn Cable Fault Predictions Into Scheduled Maintenance
OxMaint links your cable AI monitoring outputs to automatic work order generation — with fault type, GPS location, severity tier, and recommended action pre-filled and routed to the right team.
Auto-Generated Work Order
AssetCircuit CB-07 — Joint J-14 @ km 4.2
Fault TypeJoint thermal overheating — DTS +18°C delta
PriorityCritical — 3-week intervention window
Location51.4892°N, 0.1234°W — Junction Box 14
ActionJoint inspection + resin compound replacement
Generated by OxMaint AI · 12 seconds after anomaly detection
Before vs After
How Utilities Operate Before and After AI Cable Monitoring
The operational difference between reactive cable management and AI-powered monitoring is measured in outage frequency, excavation costs, and crew deployment efficiency.
Without AI Monitoring
Cable fault discovery
After grid outage — reactive excavation
Average response time
8–22 hours emergency repair window
Repair cost per incident
$340K–$1.2M emergency excavation
Fault location accuracy
±50–200m requiring test trenches
Maintenance strategy
Calendar-based time overhauls — age only
With OxMaint AI Monitoring
Cable fault discovery
6–14 weeks before fault — scheduled intervention
Average response time
Planned maintenance window — zero emergency
Repair cost per incident
$28K–$95K planned section repair
Fault location accuracy
±1–3m via DTS + PD time-domain analysis
Maintenance strategy
Condition-triggered — intervene only when needed
Asset Coverage
Cable Types and Infrastructure AI Monitors
OxMaint AI covers the full range of underground cable infrastructure found in transmission and distribution networks — each with monitoring models tuned to its specific failure physics.
XLPE HV Transmission Cables
66kV – 500kV
PD pattern recognition
Thermal ampacity
Joint dielectric
AI warning lead time: Up to 14 weeks
XLPE MV Distribution Cables
11kV – 33kV
Insulation resistance trend
Sheath current
PD inception
AI warning lead time: 6–10 weeks
PILC Paper-Insulated Cables
Legacy networks
Dielectric loss tan δ
Moisture index
Thermal profiling
AI warning lead time: 10–20 weeks
Cable Joints & Terminations
All voltage levels
Joint temperature delta
PD at joint
Resin aging index
60% of all faults originate here: Highest priority
Submarine Power Cables
Inter-island / offshore
Mechanical stress index
Seawater ingress
Armour corrosion
AI warning lead time: 8–16 weeks
DC Link Cables (HVDC)
HVDC systems
Space charge accumulation
DC insulation aging
Polarity reversal stress
AI warning lead time: Up to 12 weeks
Common Questions
Frequently Asked Questions
How does AI detect partial discharge in underground cables remotely?
Partial discharge sensors — either permanently installed high-frequency current transformers (HFCTs) at cable terminations or distributed acoustic sensing along fiber-optic cables — capture PD pulse signals continuously. The AI model analyzes pulse shape, repetition rate, magnitude trend, and time-frequency patterns to classify the discharge type and track progression. Remotely mounted sensors require no excavation and can monitor a 10km cable circuit from two terminal points.
Can AI pinpoint the exact location of a developing fault on a 20km cable?
Yes, through two complementary methods. Distributed Temperature Sensing (DTS) measures temperature at 1-meter intervals along the full cable length, locating thermal anomalies to ±1–3m. For PD-based faults, time-domain reflectometry analysis of pulse travel time locates discharge sources to within ±5–10m on long cable circuits. Combined, these give excavation teams a precise dig location — eliminating exploratory trenching that can cost $80,000–$200,000 per event.
How does OxMaint use cable AI monitoring data to generate maintenance work orders?
When the AI monitoring system detects an anomaly crossing a configurable health score threshold, it pushes the event data — fault type, severity, GPS location, confidence score, and trend history — via API to OxMaint. OxMaint automatically generates a structured work order mapped to the correct cable asset record, pre-populated with fault classification, urgency tier, recommended intervention, and estimated excavation window. The maintenance team receives a fully actionable work order rather than a raw sensor alert.
See this in a live demo.
What is the typical deployment timeline for AI cable monitoring on a distribution network?
Sensor installation at cable terminations takes 2–4 days per circuit for HFCT-based PD monitoring — no excavation required. DTS requires fiber co-installation, typically done during planned maintenance windows. AI model baseline calibration requires 4–6 weeks of normal operation data to establish circuit-specific health signatures. Most utilities are receiving actionable cable health scores within 60 days of sensor installation, with the full predictive capability operational at 90 days.
Start Monitoring Before the Next Fault
Every Cable. Every Joint. Every Fault Warning — Weeks Ahead.
OxMaint connects your underground cable AI health monitoring to structured, prioritized, and location-precise maintenance work orders — so your grid team acts on intelligence, not emergencies.
