Context
Our client is a major car manufacturer, famous for the quality of its vehicles. He has recently unveiled a new range of electric vehicles. In a market in constant evolution, reliability has become a key factor for users. The client therefore wanted to explore the feasibility of establishing a monitoring system to ensure maximum vehicle availability and reduce downtime caused by unexpected technical failures. To anticipate these failures before they lead to visible or costly consequences is a crucial challenge for ensuring customer satisfaction.
Needs
The client’s objective is clear:
“To implement a data analysis system capable of detecting and even anticipating—malfunctions in the batteries of this new range of electric vehicles during the testing phase.”
The client is looking for a solution for its R&D team that will allow him to analyze Energy Storage System (ESS) behavior and and analyze road test results.
For that, he needed an autonomous, easy-to-deploy tool that can quickly to transform large volumes of sensor data into actionable insights for diagnostics and system reliability improvement. An additional challenge is that, since the batteries are entirely new, there is no historical failure data available.
Soluce
Data from around ten vehicles was provided for analysis using DiagFit. Reference models were generated for these trucks based on data collected from test benches. The data scientists had flagged the moments they considered the most “abnormal,” which served as a baseline for anomaly detection.
This allowed the client to compare the results produced by DiagFit with their own analyses, helping to validate the detections and refine them further.
Results
The project was carried out using complex data, recorded by numerous sensors embedded in the batteries. Thanks to DiagFit, the data was prepared, cleaned, and various fault detection models were tested.
The observed results were as follows:
– 6 anomalies identified by the R&D team were easily detected by DiagFit.
– DiagFit also detected an additional anomaly that had not initially been identified by the client’s teams, but was later confirmed upon review.
Being able to detect failures that are not easily spotted by internal teams is a real advantage for the client, who now plans to continue exploring new datasets and use cases. Want to try DiagFit?
