by Felix Jordan

Manufacturing battery cells for electric cars is challenging. That’s because production yield is typically low, the final quality validation requires a long test period, and the overall process is costly. This article explains how predictive quality analytics increased production yield by 16% for battery cells.

Elisa Smart Factory team conducted the project at eLab, which is the electromobility research center at the Aachen University, Germany.

What’s the quality challenge in battery cell manufacturing?

The manufacturing of Lithium-Ion battery cells for electric cars is, in principle, a straightforward process.

First, anode and cathode electrodes are produced in several different sub-processes from a mixture of different raw materials. Then, they are packaged into battery cells, filled with electrolyte, closed and sealed. Finally, the battery cells go through the end-of-line testing.

In reality, the process is much more complicated. One of the most challenging part of the process is the time it takes to verify the final quality of the produced battery cells. In fact, it can take up to three weeks to complete the end-of-line testing. And, it’s only after this point the manufacturer can determine if the final product is good enough for the battery pack production.

In case of non-sufficient quality, the manufacturer has to scrap the battery cell as hazardous waste. This is due to lack of cost-neutral recycling options. As a result, scarce, non-renewable and expensive raw materials such as lithium, cobalt, nickel sulfate, copper, aluminium, and graphite end up as waste.

The global average first-time-yield (FTY) for battery cells is as high as 15 percent. For this reason, battery cell manufacturing is costly and slow.

How to increase manufacturing quality of battery cells?

eLab (Electric Mobility Lab) at Aachen University, is working on the future of battery production. Also for them, quality issues as the primary challenge in the manufacturing process. In fact, it prohibits cost-efficient manufacturing and thus slows down the adoption of environmentally-friendly electric cars. On the quest for developing a more efficient manufacturing process, eLab partnered with Elisa Smart Factory data scientist team.

Predictive quality analytics

eLab Elisa Smart Factory 3D Digital Twin

Why did we use predictive quality analytics?

Predictive quality analytics makes it possible to improve manufacturing quality, among other benefits. It’s the process of extracting useful insights from a manufacturing process by applying statistical algorithms and machine learning to determine patterns that can predict problems in advance. Thus, it was an appropriate tool to solve eLab’s quality challenge.

Predictive quality analytics process description

Elisa’s data scientist team followed the cross-industry standard process for data mining, the six-step CRISP-DM process. It is the most widely-used analytics model among data mining experts and below is the process outline:

  1. Business understanding – The first step is to form a clear business understanding of battery cell production and to set the right goals. To do this, the Elisa team assessed the overall situation, whereby they defined the quality drivers in the process, identified the data points affecting the battery cell quality and determined the parameters, which best describe battery cell quality.
  2. Data understanding – The second step is data understanding, as in the CRIP-DM process. It involved a gap-analysis of what kind of data is available vs. what data is required. In this case, the team covered the data gap by installing a quality camera on the manufacturing line.
  3. Data preparation – The third step is to clean and integrate the data into the same format. Also, this included cross-checking the timestamps to avoid data collection losses.
  4. Modeling – The fourth step is modeling, which was the most challenging phase for the data scientists. This is because there are thousands of ways to analyze data. Consequently, the data scientists had to try out different algorithms to see the various outcomes they provide when applied to the data collected from the battery cell manufacturing line.
  5. Evaluation – The fifth step, evaluation of results is critical. It involved investigating if the results are valid, and whether they enabled predicting the quality of the battery cells.
  6. Deployment – In the final deployment phase, the team defined the optimal parameters for setting up the production equipment and machines for optimal production quality. This included the right viscosity parameters, among many others.

What were the predictive quality analytics benefits?

Below are the benefits of the Elisa’s Predictive Quality Analytics project:

  • The scrap rate at the eLab battery cell production line has dropped by 16 percent, as the team can now predict the quality of the battery cells early in the process.
  • Increased overall efficiency, as the battery team can now identify the affected battery cells early on and remove them aside from the line.
  • Less production waste as the team is able to recycle the raw materials that of the battery cells that have been moved from the line early in the process.

If you are interested in further examples of how we’ve turned data into actionable insights with predictive analytics, download the pharmaceutical manufacturing process case study.

Share This