Pack3401 health materially worse than Pack3001 across DRI, range, and warranty — despite being younger on average. 44 batteries warranty-claim eligible.
Service Waste
10.4%
Service dispatches are false alarms. One BMS firmware patch eliminates the category.
Detection Gap
0 alerts
BMS fired zero cell imbalance alerts. Platform is sole detection layer for 20.6% of faults.
Priority Queue — Action Required
Service Groups
Fleet Attribution — Degradation Drivers
RUL Distribution
Preliminary estimate. Direction is reliable; exact timing is not.
// INTELLIGENCE APPLIED
Fleet batteries (326)
⌕
← SHOW ALL BATTERIES
Select a battery →
How is it now?
—
Where is it going?
—
What do I do?
—
Select a battery from the list.
Deepdive
DEKF · SOC Correction
BMS vs physics model
SOH Three Methods
Capacity fade comparison
Physics Relations
SOH · EFC · Range · Spread
Degradation Curves
Pack3401 vs Pack3001
Signal Plotter
Configurable VWF explorer
Analytics
Fleet signal explorer
Model Catalogue
Status · MAPE · Gates
Intelligence Gap ›
The blind zone · control plane
Intelligence scenario
DEKF SOC Correction — BMS vs Physics Model
Dual Extended Kalman Filter corrects firmware-level BMS bias. GE segment overreads by 4.44% on average. Every range estimate in the fleet is affected.
DEKF SOC
—
Physics estimate
BMS SOC
—
Firmware reading
BMS Bias
—
Overread correction
DoD Corrected
—
vs observed
LFP SOC MAPE 25.1% — physics constraint from flat OCV plateau (3.2–3.4V across 20–80% SoC). Not a software limitation. DEKF correction removes systematic firmware bias only.
Fleet DEKF Coverage
Covered
327
of 361 batteries
Coverage
90.6%
Active DEKF correction
GE Avg Bias
−4.44%
BMS overread
SG Avg Bias
−1.0%
BMS overread
SOH — Three Independent Methods
True capacity can only be estimated — never directly measured in the field. Three methods triangulate the truth. Conservative method takes the minimum. BMS is permanently excluded for LFP.
Conservative
—
Primary · use this
KPS Proxy
—
Range efficiency
Coulomb
—
Ah counting
BMS
—
Excluded for LFP
SOH vs Cycle Count (EFC) — Capacity Fade Curve
Expected LFP fade: ~20% over 1,200 EFC. Pack3401 shows accelerated fade independent of cycle count — design gap signature.
COLOR BY
Range vs SOH — The Operational Consequence of Degradation
Range is the operator's metric. SOH is the chemistry metric. Their relationship is the bridge between what the battery can do and what it delivers.
Cell Spread vs DRI — Pack Homogeneity Signal
Higher cell spread → lower DRI. Fleet median spread 247mV. Critical threshold 500mV. BMS alert threshold set to 800mV — effectively disabled.
SOH vs Age — Fleet Aging Profile
Pack3401 shows steeper age-SOH slope. Same age, different health — design gap separates the two populations clearly.
Pack3401 vs Pack3001 — Design Gap
Most important finding in the fleet. Pack3401 shows materially worse health than Pack3001 across DRI, range, and warranty exposure — despite being younger on average. Operating conditions are near-identical across both packs. This is a design signature, not a usage pattern.
Pack3001 avg DRI
53
178 batteries
Pack3401 avg DRI
45
63 batteries
Warranty exposure
67% vs 43%
Pack3401 vs Pack3001
Claim eligible
44
Pack3401 batteries
Commissioning-Stage Cell-Spread Defect — Detectable at Week 4
6 batteries showed commissioning cell spread >300mV. Not from use — from the factory. Platform flags this 12–16 months before warranty claims accumulate.
Defect batteries
6
Flagged at week 4
Threshold
300mV
Commissioning spread
Detection lead
12–16mo
Before claims
Fleet median spread
247mV
All batteries
Configurable Signal Plotter
Any weekly feature, plotted over time for the selected battery.
SIGNALCOMPARE
Weekly aggregation. Coverage varies by signal and battery. Some signals require minimum EFC or full-charge cycles.
Fleet Analytics — Signal Explorer
Plot any two metrics across the fleet. Color by pack, city, or tier. Click a point to open its passport.
X AXISY AXISCOLOR BYPACKCITY
Model Catalogue — Production Status
All models with accuracy metrics, gates, and current status. Physics-first pipeline: physics validates before ML predicts.
// enerlytik · The intelligence layer
The control plane for every battery on Earth.
Nearly every battery on Earth is operating blind. Across grid storage, EVs, drones, and robotics — batteries degrade in unforeseen ways, fail unpredictably, and deliver only a fraction of their true potential. The industry's answer has been more batteries, heavier packs, greater redundancy. Exponentially increasing costs. Hardware compensating for what software could never see.
"Our AI is not watching the battery. It is running it. The answer was never more batteries — it is smarter ones."
The Blind Zone — What BMS Sees vs What the Platform Sees
BMS — One Dimension
Fires cell imbalance alerts 0 times across 192M IoT rows and 65 weeks
Reports SOH 97–100% on batteries with true SOH of 53%
Overreads SOC by 4.44% — every range estimate is wrong
Cannot distinguish pack design failure from operator behaviour
Generates 10.4% false-alarm service dispatches — ₹400K/year waste
Sees one battery. Cannot learn from 361.
enerlytik — Five Dimensions
Detects cell imbalance in 20.6% of service returns — 4–8 weeks ahead of BMS
Conservative SOH from 3 independent methods — physics-validated, not firmware guesswork
DEKF corrects SOC bias at fleet level — every range claim is corrected
Attribution engine separates design fault from operator fault — evidence-grade
Firmware false alarm identification — one threshold patch saves ₹400K/year
Every new battery makes every existing battery's model smarter. Network effect.
The Numbers — What Staying Blind Costs
0
BMS cell alerts fired
Across 192M IoT rows · 65 weeks · 361 batteries. Platform detected what firmware missed in 100% of cases.
44
Pack3401 warranty-eligible
Pack3401 health materially worse than Pack3001 (DRI 45 vs 53, warranty exposure 67% vs 43%) under similar conditions. Invisible without cross-fleet intelligence.
10.4%
Service dispatches are false alarms
78% share BMS firmware v2.1.3. One threshold patch eliminates the category. No hardware change needed.
4.4km
Range overstatement per trip
BMS SoC reads 5.5% high. Every driver navigates with an optimistic gauge. DEKF corrects this in software.
12–16mo
Earlier defect detection
Commissioning-stage cell-spread defects flagged at week 4. Standard warranty claims arrive 12–16 months later.
4 of 10
Urgent replacements were surprises
Not on the operator's known-bad list. Platform-discovered. RUL 0–8 weeks. Operator would have had stranded vehicles.
The Control Plane — Monitoring → Optimization → Control
//
OEM Intelligence · LFP Fleet