Skill and Scale Up: Machine Learning

From data flow to decision making: The process of Machine Learning

A distinction is made between various methods for optimizing production processes, including quality management and the area of “analysis and modelling”, which covers process optimization, error analysis and the use of statistical methods. This area also includes ML and AI. ML is a sub-area of AI that enables computers to learn from experience without applying explicit rules.  Essentially, the aim is to develop algorithms that recognize patterns in data to make predictions or decisions based on them. The aim is to enable computer systems to learn from data and improve their performance - without being explicitly programmed to do so.

An example: An AI is to be taught to distinguish a car from a lorry. Instead of giving the AI clear instructions on what a car or lorry looks like, the AI is shown many images of cars and lorries. As these images have objective similarities, such as the presence of certain colors, outlines or shapes, similar information is repeatedly recognized when processing the images. This results in specific patterns to which the system attributes the association “car” or “lorry” in this case. If the trained AI is now shown an image that it does not yet recognize, it compares the features it finds with known training data. If patterns from the new image match already known patterns, the algorithm recognizes this and decides whether it is a car or a truck.

In everyday life, we encounter ML not only using programs such as ChatGPT, but also in retail, for example, to understand purchasing behavior and make personalized recommendations. But also, in the production of battery cells. 

How battery cell production can benefit from Machine Learning

The digitalization of battery cell production offers enormous potential for optimization to meet the high-quality requirements and at the same time produce as cost-efficiently as possible. Data-driven applications, such as the use of ML, can thus make production more sustainable and optimize quality. For example, by predicting cell quality based on intermediate products or predicting device failures and maintenance requirements. ML is used in battery cell production in three clusters, among others: In the process, the machines, and systems and in the product, the battery cell. We take a closer look at the process below. 

Precise traceability and quality monitoring in the coating process

One exciting approach is the precise allocation of data to quality features and process parameters during the coating process. As explained in the blog post on the manufacturing process, the electrode paste is first applied to the carrier film in a wafer-thin layer before the electrode film passes through the drying channel.

The cells must first be clearly labelled on the electrode foil to ensure that they can be accurately traced. For this purpose, laser labelling with a data matrix code (DMC) is used and applied to the foil. This enables cell-specific data mapping.

The next step is quality control: possible coating defects, such as cracks or air bubbles, are not even visible to the human eye. A line scan camera is therefore equipped with a previously trained algorithm and can use the patterns it recognizes to check the film for defects and decide whether the film area should be processed further or excluded from further processing. The previously assigned codes make it possible to clearly localize the coating defects. This intelligent process control enables potential rejects to be recognized at an early stage, thereby contributing to the quality of the battery cell. 

Machine learning as the key to improving quality

The progressive integration of machine learning into the manufacturing process of battery cells marks a decisive turning point in production. Precise traceability and quality monitoring in the coating process are just one example of the intelligent use of data, algorithms, and AI. The use of ML in quality control, among other things, not only increases efficiency, but also raises the quality of battery cells to a new level. However, predicting device failures and maintenance requirements also helps to improve the production process. The silent revolution in the production halls is therefore not only becoming a reality, but also a driving force for more sustainable, cost-efficient, and high-quality battery production - a decisive step towards an intelligent future.