08Apr 2019

Let’s say you love to ride your bicycle and that you want to measure speed without using fancy computers and GPS. What would you do?

High-school physics to the rescue! All we need is the circumference of the wheel then count the number of rotations the wheel makes in a certain amount of time that we can measure with our stopwatch. The speed, v, is calculated as the number of rotations, N, multiplied by the circumference, L — that’s the total distance traveled by the wheel — divided by the measured time, T. Put in one simple equation:

We replace the circumference with the radius, R, because it is easier to measure the radius of the wheel:

The speed equation becomes:

Therefore, measure the wheel radius, then count the number of rotations, and clock them with a watch…et voila, you can measure speed.

You quickly realize that you have an approximation of speed because it fails to take into account other factors that can introduce errors, for example temperature. On a hot day, the wheel expands a little making the radius longer.  Then you realize that the thickness of the rubber tire is not exact — it varies across manufacturers. With aging, the rubber gets worn out. It becomes thinner and consequently the radius is a little smaller. You may think these are small effects but if you are racing, they can make the difference between winning and losing.

So what is the relevance of a bicycle wheel to a battery?

Scientists understand the electrochemistry inside a battery.  They represent this science with many complex equations — like Fick’s law, Tafel’s equation, and several other mathematical forms. Yet, these equations remain insufficient to describe batteries in real life. 

Much like the wheel, there are significant variations in manufacturing across batteries from the same manufacturer or from different manufacturers. Temperature dependence, aging, presence of defects…etc. are significant additional considerations that impact the performance and safety of the battery.

Capturing these “real-life” considerations is what makes a model of the battery useful.  By “model” I mean a sufficiently accurate representation of the battery that one can use to make meaningful conclusions. For example, a good model can be used to predict the end of life of the battery. It can be used to identify counterfeit batteries or find defective batteries before they become a fire hazard.

Developing the model entails collecting data — millions of measurements — to capture manufacturing variations, temperature dependence, defects…etc. It takes a long time to collect statistically meaningful data across different types of batteries, from different manufacturers and across a board range of operating conditions.

The battery model is not static — it must improve over time or it becomes obsolete. One must keep updating it so it learns and adapts to newer battery materials, newer battery designs and manufacturing processes. This learning process can be in the test lab, or it can be in the field — in other words, intelligent algorithms can learn from batteries deployed in smartphones or other devices already in the hands of users.

Possessing intelligent algorithms and useful battery models is a powerful combination to make key predictions about the battery’s health and safety…that can make the difference between a safe battery and a fire.