Element - Automatically Detecting Anomalies in Battery Test Results

Georghios Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Dataiku DSS Adv Designer, Registered Posts: 15 ✭✭✭


Dr. Georghios Joseph
Jamshid Barry
Ofelia Martinez
Rek Chong
Sofia Bernola

Country: United Kingdom

Organization: Element

The Element Materials Technology Group is one of the world’s leading global providers of testing, inspection, and certification services for a diverse range of products, materials, and technologies in advanced industrial supply chains where failure in use is not an option. Headquartered in London, UK, Element’s c.9,000 scientists, engineers, and technologists work across a global network of over 270+ locations, support customers from early R&D, through complex regulatory approvals, and into production ensuring their products are safe and sustainable, and achieve market access.

Awards Categories:

  • Best Acceleration Use Case
  • Best MLOps Use Case
  • Best Approach for Building Trust in AI

Business Challenge:

One of our laboratory teams manually screens battery test results, sometimes after the data has been made available to the customer. Irregularities when conducting charge/discharge tests, such as testing equipment (cycler), unsecured connections, or misbehaving battery units due to internal faults, are common (25%~30%).

The customer is granted access to the test results through an online platform. Here, they have access to analytical capabilities where anomalies in battery test results can be screened for visually. The lab is subsequently informed where the customer typically requests re-testing of the unit(s) free of charge. Re-testing costs personnel time and uses revenue-generating cycler channels.

The aim of the project is to detect anomalies automatically in battery test results near real-time and inform personnel to halt the tests early. This can be achieved using statistical testing and machine learning.


Business Solution:

Dataiku is a one-stop platform for applied data science and analytics methods and techniques. It helps with development and delivery/deployment speeds. Currently, five individuals are part of the team: Director of Data Science, Head of Data Analytics, Lead Data Scientist, Data Engineer, and Data Analyst(s).

To develop our solution, we used data preparation recipes, groupby, stack, filter/sample, python recipes, clustering for anomaly detection, statsmodels, and insights/dashboard for delivery.


Day-to-day Change:

The solution is an intermediator service that screens battery test results in seconds instead of hours. The statistical test runs on streaming data detecting signal irregularities. As soon as an issue is raised, a member of staff investigates, remediates the situation, and restarts the test if the battery unit is non-faulty. Otherwise, the unit is placed safely, waiting for its return to the customer.

Business Area: Internal Operations

Use Case Stage: In Progress

Value Generated:

The benefits of this solution include:

  1. Screening batteries in seconds, automatically.
  2. Reducing customer attrition, with less waiting for the customer.
  3. Reducing battery test result screening by up to 95%. This permits us to a) claim back 90% of technician time and b) reduce labour costs by up to 90% per year.
  4. Halting tests early. This is a) safer for our fellow staff and b) has increased testing throughput by up to 25%.

The MVP can be developed and tested in three months, with six months to break even.

Value Brought by Dataiku:

Benefits brought by Dataiku when developing this solution include:

  • Development speed.
  • Collaboration with team members.
  • Handy visualisations.
  • All tech we used was under one roof.
  • Team upskilling through Dataiku Academy courses.

Value Type:

  • Improve customer/employee satisfaction
  • Increase revenue
  • Reduce cost
  • Reduce risk
  • Save time
  • Increase trust

Value Range: Thousands of $

Setup Info
      Help me…