I need to determine the "health" of car batteries in the field.

My objective is to predict when a car battery is likely to fail from data collected while the vehicle is in use. I am targeting 3 states - good Battery, Battery needs charging, Bad battery needs replacing.

In preparation, I will be testing batteries with various states of charge in a garage environment measuring voltage and current using automotive battery testers, current clamps and Oscilloscope and checking if the engine starts.

In the field I have the capability to measure voltage at 100Hz at every cranking event for a period of 5 seconds. In addition I can measure engine temperature, ambient temperature (but not battery Temp), Trip details - Duration, time of trip, engine size, time for engine to start.

I am unable to measure current remotely in the field.

The intention is to then "Prove" that the status of the battery condition in the field matches that of the measure in the garage.

Does my proposed experiment seem logical?

Are there any other factors that i should consider? I can then determine if I am capable of measuring them.

Any other input on the subject of developing the algorithm would also be welcome

  • \$\begingroup\$ (assuming la batteries here) Things like simple voltage or acidity tests will tell you a good estimate of the percentage of a battery, but the capacity will only be available if you draw some of it and calculate/estimate on how much you have drawn. \$\endgroup\$ – PlasmaHH Dec 17 '15 at 13:20
  • \$\begingroup\$ How large is your sample of batteries going to be? \$\endgroup\$ – pjc50 Dec 17 '15 at 13:28
  • \$\begingroup\$ Have you tried googling for a method? \$\endgroup\$ – Andy aka Dec 17 '15 at 14:10
  • \$\begingroup\$ Kepp in mind that you have a significant temperature dependency. The cranking will require more current in cold wheather while at the same time the battery delivers less. \$\endgroup\$ – Turbo J Dec 17 '15 at 15:43
  • \$\begingroup\$ Sample size is 10K vehicles when in the field for a 1 year period, therefore many thousands of cranking events will be recorded. \$\endgroup\$ – Gareth Adams Dec 17 '15 at 16:27

The algorithm is simple enough. Just correlate the temperature of the battery to the source impedance at that temperature. A battery's internal impedance increase as either the anode or cathode material is worn down by repeated chemical reactions. You will have to measure the batteries at the "same charge" which is to say that a LA battery with little charge, but is still a good battery will have a higher impedance than a fully charged bad battery. Either test them when they're all freshly charged or when they're all completely spent.

You can find the impedance of a battery by testing it's open circuit voltage, Vo, and testing the voltage across a known resistor, Vl and R. You plug all that information into this equation: internal Resistance = [R*(Vo-Vl)]/Vl

This isn't a very practical method unfortunately since you'd have to disconnect the battery to run this test. However, it's the only one I know. Internal Impedance is really the only tell-tale sign of a bad battery.

| improve this answer | |
  • \$\begingroup\$ What about testing the AH capacity? \$\endgroup\$ – Autistic Dec 22 '15 at 10:48
  • \$\begingroup\$ @autistic, car batteries rarely do anything more than short shallow discharges during engine cranking, so very little opportunity to measure AH capacity. Measuring internal resistance (how it changes over time, and as measured during recharge &/or cranking) is a key indicator of cell health, but even that can vary significantly from one manufacturer/model of battery to another as to what threshold indicates imminent failure. i'm not sure there's a reliable way to do what the OP is wanting across a range of not-pre-characterised batteries. \$\endgroup\$ – Techydude Dec 22 '15 at 22:21
  • \$\begingroup\$ @Techydude, You are correct. However the model can make up for that if you include manufacturer with that information. \$\endgroup\$ – Dave Dec 22 '15 at 22:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.