We are exploring how to predict battery capacity Q [Ah] with a high accuracy of over 95% in a few minutes.  Now that we made a breakthrough. 

 There are two techniques for diagnosing battery degradation, one using AC the other using DC.

 A typical example of the use of AC  is the EIS (Electrochemical Impedance Spectroscopy) test method.

• We have also tried the EIS test method but have not been able to obtain preferable results. Yes, it is effective,  but does not meet our target prediction accuracy, and similar performance was seen in the actual cases in the market. Due to difficulties such as AC-specific stray impedance and the fact that the battery parameters change over time after application of electricity, we decided to explore new methods other than the electrochemical approach. It was a long time ago.
  

Our product  uses DC

• Our co-founder Dr. Takagi was involved in the degradation of in-vehicle batteries while working for Toyota Motor Corporation, and moved to Toyota Systems Corporation, where he became the inventor of this mathematical and statistical approach and patented it (applied  in 2022, previous method). The mathematical-statistical approach is based on knowledge of electrical engineering and mathematics.
  

•  However, this technology assumes the highest grade of data quality, measured by expensive Source Measurement Units(SMUs) with low noise. The more sophisticated the data, the more data points, the better the prediction results. To achieve a 95% level of prediction accuracy, super expensive investments are required, e.g. in a constant temperature environment. In short, the previous method is vulnerable to noise and data errors.
  

•  In practice, noise, missing data, errors, etc., arising from the measurement environment and hardware cannot be completely eliminated. Of course, the better the quality of the data, the better. We wanted to achieve a 95% level in the reality of SMUs and measurement environments used by normal remanufacturers. Here comes as a new method – patent to be applied in 2024.
  

 In real business, prediction accuracy has the greatest impact on profitability

	EIS testing method *	Previous method *2 Invented by Takagi in 2022. (Patents owned by Toyota Systems Corp.)	*3 Invented by Takagi & Narui in 2024
Expected prediction accuracy	65%±	85%±	90%＋
Required time duration.	At least 5 to 15 min.	3 min.	2 min.
Required resolution of data.	1.00[mΩ] or less.	0.01[mV]・0.01[mA] or less. & 10[mSec] or less.	12[mV]・12[mA] or less. & 100[mSec] or less.
Influence of the DUT’s SOC status.	Affected by.	Not affected.	Not affected.
Acquired data sample.			(Not disclosed)
Systems	On-premises.	SaaS Online only.	SaaS Both Online and Offline.
PC specitication	Enabele to run  ML such as Neural Network.	Windows PC Microsoft Edge	Chromebook Google Chrome
Applicable solutions	Module measurement. Pack measurement.	Module measurement.	Module measurement. Pack measurement On-board diagnosis

*1 There are many variations of the EIS method, the above is just one example.
*2 This previous method boasts an accuracy that overwhelms the EIS method, but has the weakness of being vulnerable to data noise. If 95% accuracy were to be achieved, it would require world-class measuring equipment and fixtures in a constant-temperature facility, which is not realistic.
*3 Our invention is patent application in preparation.