MW Battery Analyticsby Bayesian Cybersecurity
Request assessment

MW Battery Analytics powered by the MW Battery Engine

Understand what your battery data is really telling you.

We read the BMS/CAN data you already record, find early ageing signals, check they are physically real, and tell you when to act. For BESS operators, battery makers and fleets.

Request a free assessmentSee the evidence

Pack view

Reference pack example
Chemistry
LFP
Cells
16 in series
Log
20.6 h
BMS state of health1.000BMS reading
MW manifold health0.976Bayesian posterior
Cell 1 dQ/dV shift−13 mVMeasured
  • Lithium-inventory loss suspect
  • Limits discharge
  • Next to the T5 hot spot
  • No signal (0 V)
  • Nominal

Cell 1LLI suspect

dQ/dV peak 13 mV below the pack median (threshold −5 mV).

Earliest sign of lithium-inventory loss. The BMS still reports this cell healthy. Put it on a watch list and verify.

Select a cell. States are findings from one 20.6-hour BMS log of a 16-cell LFP pack; they are not a live system.

Your BMS reports

1.000

state of health, on the reference pack

Why one number isn’t enough

A healthy-looking number can hide an ageing cell.

Your BMS stays the baseline and the safety layer. But one number can’t tell you:

  • What is actually ageing
  • Why it is ageing
  • Which cell is different
  • How certain the signal is
  • Whether the model is guessing
  • What to do, and when
What it does

Battery data in. Evidence, diagnosis and a decision out.

Every assessment answers three questions, in order — each finding with an action and a time horizon.

  1. What is happening?

    Which mechanism is moving, in which cell.

    Reference pack: cell 1 shows early lithium-inventory loss.

  2. Is it real?

    Physics checks tell battery behaviour from a sensor fault.

    Reference pack: cell 16’s 0 V is a wiring fault, not a dead cell.

  3. When should I act?

    Each finding carries an action and a time horizon.

    Reference pack: inspect T5 in days; verify cell 1 in weeks.

  • Uses the data you already have

    Works on the standard BMS/CAN log. No new sensors, no lab test, no downtime.

  • One day of data is enough to start

    The reference analysis used a 20.6-hour log at one sample per minute: 1,313 rows.

  • Auditable

    Every measured number can be recomputed from the raw CSV. Model estimates are kept separate, so they never pass as measurements.

  • Hard to fool

    CAN has no authentication, but physics is hard to forge: a spoofed state of charge must still match the current that flowed, and faked temperatures must still follow load heating.

What it answers that a BMS can’tThe full decision guide
Reference pack example: model forecast, not a guarantee

When the BMS still says healthy.

Cell 1’s dQ/dV peak sat 13 mV below the pack median while the BMS read 1.000 — the earliest sign of lithium-inventory loss, about 1,600 forecast cycles before end of life.

dQ/dV curves: the pack median peak sits at 3.376 V; cell 1's peak sits 13 mV lower at 3.363 V. Peak width (FWHM) is 30 mV.

Schematic. Peak positions and width use the reference pack’s values; the curve shapes are illustrative, not the cell’s measured curve.

Why it shows up first

LFP’s flat voltage plateau barely moves as a cell ages, so the BMS sees nothing. The dQ/dV peak moves early, and the engine reads it from ordinary CAN data.

BMS state of health
1.000
Pack manifold health
0.976
Cell 1 peak shift
−13 mV
Flag threshold
−5 mV
See all eight findings and the full case
How it works

The MW Battery Engine, in six steps

Physics does the heavy lifting; Bayesian inference says how sure we are.

  1. BMS / CAN data

    The log you already record: cell voltages, pack current, temperatures, BMS state of charge and health.

    Your data16S LFP, 1,313 rows at 1 sample/min

  2. Measured signals

    Resistance, relaxation and dQ/dV peaks fitted from the raw log. Anyone with the CSV can recompute them.

    MeasuredCell 1 dQ/dV peak −13 mV

  3. Physics checks

    Ten integrity laws test whether the log is physically consistent: Ohm, Joule heating, charge counting and more.

    Physics-checked1 of 10 laws fired: thermal (MP8)

  4. Bayesian model

    Places each cell on a learned map of physically consistent states, with its own uncertainty.

    Bayesian posteriorManifold health 0.976, MW distance 1.7σ

  5. Forecast + uncertainty

    Many simulated futures through the map give a range for remaining life, not a single guess.

    Bayesian posterior1,620 cycles, 90% window 1,503–1,700

  6. Decision

    Each signal becomes an action and a time horizon your team can plan against.

    Your actionCell 1: verify and budget a spare, in weeks

Inside the engine, and the three kinds of number
How might it evolve?

Remaining life is a range, not a promise.

The distribution is more useful than the mean.

Remaining useful life distribution for the reference pack: mean 1,620 cycles, 90% credible window 1,503 to 1,700 cycles, probability of end of life before 1,000 cycles is 0%, OEM rating 2,000 cycles.

End-of-life posterior for the reference pack. Mean, window, P(EOL < 1,000) and the OEM line are the pack’s values; the curve shape is illustrative.
Estimate
1,620 cycles mean, about 5 years at 0.88/day. Bayesian posterior
Uncertainty
90% window 1,503–1,700 cycles. P(end of life before 1,000) = 0%.
Assumptions
Duty as logged. Leans on the 2,000-cycle OEM label (+926), outside the training range.
Decision
Planning grade — for budgets and warranty odds. Never a guarantee.
Why you can trust it

Knowing when not to trust a forecast is part of the analytics.

We quote every number with its uncertainty, and say openly where the limits are.

  • One reference pack, one 20.6-hour log so far. No lab capacity test yet.
  • The life forecast leans on the OEM cycle rating, partly outside the training range.
  • The credible window covers model uncertainty only.
  • Reference work to date is on LFP.
  • Not a safety system. It never overrides a BMS warning.
All documented limits and methodology

Who is behind it

Decades of applied Bayesian statistics meet clean-energy and battery leadership. We work with battery startups founded by globally leading researchers.

About the team
  • Rahul ModakCo-founder & CEO, Technical Lead
  • Dr. Rahul WalawalkarCo-founder & Executive ChairmanPhD, Carnegie Mellon University
  • Rajas JoshiCo-founder & COO
An extra layer

Physics is hard to forge.

CAN has no authentication, but a spoofed reading still has to obey physics. The same checks that catch a dead sense wire also expose manipulated data.

Questions

Before you send us a log

Does it replace our BMS?

No. The BMS stays your baseline and safety layer. MW reads the same data and adds the context behind the number: which mechanism, which cell, how sure, and what to do.

What data do you need?

An ordinary BMS/CAN log — cell voltages, current, temperatures, SoC and health. One day is enough to start. No lab test required.

Is the life number guaranteed?

No. It is a planning estimate given as a range, with its assumptions shown. More logs narrow it.

Is it a safety system?

No. It cannot detect thermal runaway, shorts, venting or fire, and never overrides a BMS warning. It is decision support for maintenance and planning teams.

Which chemistries?

Reference work to date is on LFP. Tell us yours and we will say plainly what we can and cannot support.

What does the free assessment include?

We analyse one log and send a report: findings with evidence, confidence, an action and horizon for each, and the limits of one log.

Get a free MW Battery Assessment.

Send one BMS/CAN log. Get back what is ageing, how sure we are, and what to do next.

  1. Share a logAny BMS/CAN export with cell voltages, current and temperatures.
  2. We run the engineMeasured signals, physics checks, Bayesian model, life forecast.
  3. You get the reportFindings, confidence, actions and time horizons — with the limits.

Prefer email? rahul.modak@bayesiananalytics.in

About the battery system

We use your details only to reply. Logs are shared separately, after we agree how.