Reduce high reject rates with AI algorithms

Lithium-ion batteries require expensive raw materials and undergo a complex production process with high scrap rates. One solution is digitalization - the implementation of digital technologies to collect and process the large amounts of data generated in modern production. They are used to gain insights that bring benefits to the production process.

Meeting rising demand

One can only estimate how dramatically the lithium-ion battery sector will grow over the next ten years. Rising demand from the electric vehicle sector has already contributed to the global market growing from around USD 40 billion in 2018 to more than USD 60 billion in 2022. The growth trend is expected to accelerate. According to forecasts, the market will be worth more than a quarter of a trillion dollars by the end of the decade.

However, the sheer scale of this expansion brings with it a number of challenges. Not only will competition increase sharply, but the raw materials used to manufacture lithium-ion batteries will become even scarcer and more expensive as demand increases. Lithium, cobalt and nickel must be mined and processed before they can be used for production. The IEA forecasts that the total demand for minerals required for the production of electric vehicles will increase 30-fold between 2020 and 2040. In order to meet the demand for electric vehicles, production must be made as efficient as possible. One bottle-neck in production is the implementation of effective quality control and the reduction of high reject rates.

Very high reject rate

Around ten percent of finished products do not meet the minimum requirements and end up as scrap. In many cases, this rate can even rise to 30 percent. This is one of the reasons why the battery usually accounts for up to 60 percent of the total cost of an electric vehicle. It is also unacceptable for sustainability reasons. However, the production of lithium-ion batteries is complex; the reject rates are not due to the negligence or irresponsible behavior of the manufacturers, but rather the manufacturing process and the implementation of conventional quality control measures are demanding.

Influence of the coating thickness on the energy capacity

First, strips or foils are made from metal - copper for the anode and aluminum for the cathode. The size of these strips varies depending on the design and specifications of the batteries produced; they are usually several hundred meters long, often over a kilometer at larger production sites. The width ranges from a few centimeters to more than a meter. To convert them into batteries, the strips are coated with a thin layer of active material slurries. Lithium cobalt oxide, lithium iron phosphate or other lithium metal oxides are used for the cathode, while the anode is coated with graphite or silicon-based materials. Regardless of the materials used, the coating thickness has a major impact on the energy capacity and ion transport efficiency of the finished battery. Thicker coatings can store more active material, which can increase the overall energy storage capacity of the battery. However, coatings that are too thick can impede the movement of lithium ions between the electrodes, which reduces the charging and discharging speed of the battery.

Consistent coating thicknesses over a large area

This balance between thickness for increased energy storage and the need for efficient ion transport is critical. It has a direct impact on battery performance, especially in EV applications that require rapid energy delivery. If a coating is too thick or too thin, large sections of a tape may be unsuitable for use in a finished product. Although the ideal thickness varies depending on the manufacturer's requirements, cathode coatings are usually in the range of 100 to 200 μm, while anode coatings are slightly thinner at 70 to 120 μm; equivalent to the thickness of a human hair. It is considered a challenge to obtain a consistently thin film over a large area. Many factors can influence the thickness of the end product - from the temperature of the film to the humidity in the plant. In order to capture this complex, constantly changing data, the implementation of suitable production technologies and sensors can be helpful.

AI-driven solution Melsoft Mailab

The principle of predictive maintenance can also improve battery production. It is possible to determine which parameters need to be checked in order to obtain high-quality cells. This increases the value of functional batteries and reduces the amount of materials that need to be scrapped. Mitsubishi Electric has already applied this knowledge to real production lines for lithium-ion batteries. The team focused on the aspect of uneven layer thickness and collected data from 127 different parameters to determine which ones might be related. With the help of Mitsubishi Electric's AI-driven solution Melsoft Mailab, it was found that four factors correlate strongly with variations in layer thickness: Voltage, coating pressure, excess temperature and distance from the coating opening. The team developed a diagnostic rule to detect the thickness. The experts then combined it with industrial automation technology, such as voltage regulators, to carefully monitor and change the parameters. This ensures that as much of the strip as possible meets the quality standards.