AI holds the key to limiting the risk of EV battery fires

Vehicle fires, irrespective of the powertrain, have been an ongoing issue. However, the focal point of recent discussions has centred on incidents related to electrified vehicles. Consequently, vehicle manufacturers, along with the broader automotive sector, confront the task of restoring the confidence of consumers, some of whom may have been adversely influenced by the latest media coverage.

"Indeed, EV battery fires are exceptionally uncommon, but even a single incident is excessive," states Dr. Umut Genc, CEO at Eatron Technologies. "As a sector, it is imperative that we ensure the incidence of severe battery failures is eliminated and remains so. Our sophisticated, integrated, and secure automotive-grade battery management software has illustrated that AI is instrumental in realising this objective."

The underlying reasons for battery failure can be intricate, often stemming from a myriad of causes. One prevalent cause – lithium plating – arises when metallic lithium accumulates around the anode. This is predominantly observed during rapid charging at reduced temperatures, and these accumulations degrade battery performance over time. If not addressed, this could result in the formation of dendrites, filamentous structures capable of perforating the separator between the anode and the cathode, leading to an internal short circuit. This can subsequently prompt a swift self-discharge, potentially triggering thermal runaway, a relentless chain reaction challenging to control.

Identifying lithium plating without physically accessing and inspecting the battery cell, which is largely unfeasible once fitted in a vehicle, remains a significant research challenge. Although various methodologies have been devised over time, each possesses its own constraints, especially in differentiating lithium plating from other degradation processes.

Nevertheless, through the deployment of artificial intelligence, Eatron Technologies has demonstrated the capability to not only identify lithium plating more efficiently but also to foresee its potential occurrence.

"By employing a method known as feature extraction, we convert the primary health data sourced from the battery into a structure that simplifies anomaly detection. When integrated with our exclusive AI framework that precisely characterises battery behaviour, our AI diagnostic tools can anticipate cell malfunctions before they materialise, boasting an accuracy of up to 90% with no false positives," elaborates Dr. Umut Genc.

Anticipating a malfunction prior to its occurrence provides an opportunity to address it in a more effective and timely manner. This could entail modifying battery management to limit further damage in the immediate future or, in the long term, organising a maintenance appointment at a time suitable for the vehicle owner.

"Importantly, regardless of the chosen response by the manufacturer, the malfunction has been averted, ensuring that the incidents we have witnessed in recent times will soon become a thing of the past."