Keywords

1 Motivation

The advancing climate change necessitates a shift from the use of conventional energy sources to renewable ones, which is accompanied by the indispensability of efficient storage of electrical energy, leading to a rapid increase in demand for battery cells [1, 2]. In the context of battery cell production, sustainability is proving to be a key challenge. Especially the use of digital twins can offer a solution for this challenge [3]. One specific approach that can lead to reducing the resource-intensive initial period for identifying machine parameters is the use of virtual experiments. This is to use less of the materials in production that may themselves have to be sourced from non-sustainable sources. Thus, the input/output ratio can be improved and a more sustainable battery cell production can be achieved. The concept of virtual experiments in battery cell production based on the digital twin is presented in this paper and the concept is evaluated in the context of electrode production.

2 State of the Art

2.1 Battery Cell Production

A battery cell consists of a cathode, an anode, collectors, a separator, the electrolyte, and a housing [4]. For many applications that rely on mobile energy storage, the lithium-ion batteries have emerged as the best solution in recent years. Production of the cells is divided into three main segments: electrode production, cell assembly, and activation or formation [5]. In electrode production – on which the explanations in this contribution focus – the production for anode and cathode follows the same process steps. The electrode foils are coated with a paste of active materials – the slurry – dried, calendered, slitted into narrower electrode strips, and finally vacuum dried [4, 6].

2.2 Sustainability

Sustainability is broadly defined as “the ability to meet the needs of the present without compromising the ability of future generation to meet their own needs” [7]. Today ecosystems are exceeding the capacity for self-regulation from man-made disruptions. Climate change, overexploitation of land and natural resources, and loss of biodiversity are outcomes of these disruptions in ecosystems [8]. Therefore the use of resources must be adequately planned and emissions to air, water, and soil have to be measured and minimized. The ISO normed method of life cycle assessment (LCA) captures all material, energy, and emission flows that occur during the life cycle of a product system. From this ecological status-quo analysis, different scenarios for optimization can be investigated to enhance sustainability performance. Sustainability always bases on the three pillars of ecology, economy, and society [9]. This implies that sustainable and economic targets are not – as often perceived – contrary concepts, but are integrated in each other, therefore creating a strong interdependence. Nonetheless each pillar of sustainability can be targeted individually, but consequences for the other pillars have to be taken into account.

2.3 Digitalization for a Sustainable Battery Cell Production

Battery cell production is very energy-intensive and requires critical raw materials [10]. Digitalization can help to enable sustainable battery cell production in all pillars: economic, environmental, and social. The basis for this is created by the digital twin, which is a digital representation of physical objects. It includes the properties, states, and behavior of the object via data, models, and information [11]. The data generated during production can be aggregated in the respective digital twins, stored, and thus used for a variety of services – e. g. for simulations, or as input data for an LCA.

Predictive maintenance or adaptive production can have a direct impact on economic key figures in the form of cost reductions. For this, the linking and analysis of data from the machine and product twin is essential, as this is the only way to make reliable predictions based on the production data.

Social sustainability aspects can also be addressed by digitalized battery production. For example, the creation of transparency regarding the entire value chain of battery production is an essential factor in avoiding social injustice in the future. This transparency can be enabled by the digital twin of the product, which contains data on the type of production or the origin of the materials. It serves also as a basis for automized LCA or the Battery Passport which is targeted by the European Union [12].

Furthermore, to enable the goal of ecologically sustainable production, data-based simulation use cases can be realized that increase the input/output ratios in cell manufacturing in the long term through optimized use of resources and energy.

Simulation is intended to mimic the behavior of a physical scenario, whose effects are applied to a physical system [13]. In the field of battery cell production, there is related work addressing concepts for simulating impacts of the entire process chain on the battery performance such as [13]. Moreover, they state that various interdependencies in the complex battery cell production process are still unknown, and thus stressing the desire for more transparency [13]. Schönemann et al. [14] focus on a holistic framework for coupling multiple submodels of a battery cell production system, enabling transparency on process chain level. Thomitzek et al. [15] propose a multi-paradigm simulation for the entire process chain, focusing on energy demand. However, besides these holistic approaches on battery process chain level, no platform has been developed so far using a combination of simulation models that allows for optimizing machine parameters while considering the peculiarities of the specific process step, its intermediate product quality, and its interdependency to the final battery performance. To narrow this gap, we propose to establish a user-friendly virtual experiments platform for realistic experiments at the process step level. It would enable process experts to plan experiments more efficiently and support the production of high-quality intermediate products for the subsequent process step. The use of virtual experiments can help to reduce the resource-intensive setting of machine parameters for production [16].

3 Conception of Virtual Experiments

3.1 Setting the Goal of Virtual Experiments

The idea of virtual experiments is to support the parameter optimization in terms of sustainability measures in a virtual space. In other words, virtual experiments are not simulations, but virtual representations of real experiments that can be manipulated in the digital world to a certain extent. Thus, virtual experiments represent a special application of simulation-based engineering of production processes that aim to decrease the time to develop a product with optimal parameters in a virtual space before producing it on the physical machine. [17] Therefore, among other sustainability-related measures, the use of input materials as well as energy consumption will be decreased using virtual experiments. Moreover, virtual experiments increase transparency by supporting to understand cause-effect relationships of machine input parameters and output parameters of the product. In this case, virtual experiments are applied for understanding the process by simulating various parameter sets of the depicted machine.

3.2 Virtual Experiments Platform Design

Figure 1 illustrates the big picture of the virtual experiments platform (VEP). The bottom layer consists of the physical assets, on which physical experiments are conducted. In production, these physical assets are the production machine and the product that is crafted on the respective machine. For both, the product as well as the machine, digital twins are used. Within the digital twin of the battery cell production, the data, models, and information of the relevant machine and product twins are aggregated. These aggregations are, for example, training data sets that link the machine parameters provided by the digital machine twin (e. g. of the coater) with the quality parameters that belong to the crafted product, e. g. the produced electrode foil. The quality parameters of the product are provided by the digital product twin. The data-based models and physical models that represent the characteristics of the particular process step are the basis of the VEP, which runs as a service of the digital twin of the battery cell production. The VEP allows for setting the input boundaries of machine parameters as well as parameters to be optimized (e. g. quality, energy, and material consumption). Based on the models, the parameters are shown within the VEP. The following step is to apply the optimized parameter set on the physical asset and validate the virtual experiment.

Fig. 1.
figure 1

© Studio Wiegel).

Big Picture of the Virtual Experiments Platform (photos

Full size image

3.3 Development of the Virtual Experiments Platform

In the following, an iterative development approach for the VEP is proposed. The general procedure bases on design thinking, which is useful for addressing complex issues by integrating the end user’s perspective [18, 19]. Design thinking has already been used in platform development to ensure that the needs of end users are considered [20].

Similar to the design thinking methodology, the development process for virtual experiments is executed in an iterative way so that a first version is available in early stages, which can then be improved. The first step of the development process of a VEP is to incorporate the perspective of the end user by identifying stakeholders and determining the need for virtual experiments similar to the Understand and Observe stages of design thinking [21]. Based on the understanding of the user needs, the platform’s scope and requirements can be defined in the next step. For this purpose, it is necessary to assess which experiments should be represented within the platform by identifying the critical cause-effect relationships for sustainability through expert interviews or quantitative analysis. Moreover, the complexity of the experiments needs to be defined by specifying whether the parameters within experiments should be optimized with respect to a single optimization objective or multiple objective variables in parallel. As sustainability in production is dependent on several factors in parallel, it can be defined as a multi-objective optimization problem.

To establish the data basis for a VEP, physical experiments have to be conducted in the second stage. The reason for this is two-fold. On the one hand, data-driven models such as machine learning algorithms need data for generating well-performing models. One the other hand, even for physical models a data basis is needed since expert knowledge is required for these models which can only be generated through physical experiments. Furthermore, physical models have to be validated on real data. For a VEP whose primary goal is to increase sustainability, there exist approaches that allow the data basis to be generated in a sustainable way by using as few input materials as possible while maximizing the exploration of the decision space [22, 23].

Once the data basis is ensured by conducting physical experiments, the frontend of the platform can be developed and connected with the models that are generated and validated on the data basis. The frontend should at least allow to set the input boundaries of the experiment and the target parameters as illustrated in Fig. 1. Based on the prediction of the models, the optimized target parameters and corresponding input parameters should be visualized in the platform.

Similar to the last phase of design thinking, the final process stage of VEP development entails the validation by users to assess whether the requirements are reached. If adjustments are necessary, the next iteration of the development cycle begins. The iterative process that has been described in this section is illustrated in Fig. 2.

Fig. 2.
figure 2

Virtual Experiments Platform Design Process.

Full size image

4 Application of Concept in Battery Cell Production

The developed concept of VEP was applied in battery cell production at a coating line of Fraunhofer Research Institution for Battery Cell Production FFB. Due to its modular structure, transferability to other production processes and environments is given.

4.1 Coating Line at FFB Workspace

In 2021, Fraunhofer started operating a coating line with clean room technology – the “FFB Workspace”. Its focus lies on innovative production procedures for slurry mixing, coating, and drying in the electrode production. The production machinery consists of a twin-screw extruder for continuous slurry mixing and a slot die coater with attached dryer for anode coating. As the research team experiments with continuous mixing techniques and different materials, the production recipes are frequently changed. Each change in the production recipe can lead to a different behavior of the machinery during the potentially material intensive and time-consuming ramp-up phase. The optimization of these ramp-up phases is one possible use case for which a VEP is developed via the methods described in Sect. 3.

4.2 Implementation of the Process at FFB Workspace

As described in Sect. 3.3 the design process for developing the VEP starts with identifying the users of the platform and their needs. For this purpose, interviews with experts for electrode production are conducted to define use cases for the slurry mixing and coating/drying process. Possible general use cases could be to find ramp-up recipes with reduced material throughput or to find faster ramp-up procedures for configuring the process parameters. The specific parameters to be optimized during slurry mixing or coating/drying will be chosen, building on preliminary research on cause-effect relationships in battery cell production [24]. Based on these requirements, the scope of the platform and the experiments can be defined. Real experiments on the extruder and/or coater are planned and conducted to create the needed data basis consisting of the process parameters and inline quality measurements. The data basis also includes additional offline quality measurements – e. g. the slurry viscosity – and input material qualities. The subsequent steps in the design process, which are the model creation, frontend development and the platform validation follow as described in Sect. 3.3. Table 1 lists the specific content underlying the VEP components for an example use case of the coating/drying process focusing on optimizing the product quality.

Table 1. VEP content for a use case of coating/drying process with focus on product quality
Full size table

4.3 Digital Infrastructure for the Realization of the VEP

The digital infrastructure to realize the VEP is based on the big picture presented in Fig. 1. The machinery is connected via an OPC UA interface [31]. The process parameters, inline quality parameters, and machine states are read out via OPC UA and stored in time series databases as part of its respective digital machine twin. Offline quality parameters and material data are stored in relational and document-oriented databases as part of the digital product twin.

The relevant data out of these databases is aggregated in data sets and transformed into suitable data formats in the combined digital twin of production. These data sets will then be used for validating the physical models or training the data-based models. The simulations with these models shall run as services on the combined digital twin where the communication with the VEP will be made possible via its API. The VEP can run outside of the digital twin, e. g. as a web service, depending on the needs of the users. The advantages and disadvantages of the VEP are summarized in Table 2.

Table 2. Advantages and Disadvantages of a Virtual Experiments Platform.
Full size table

5 Conclusion

In this paper, the need of sustainable battery cell production processes was presented. Digital twins were identified as enablers for a more sustainable battery cell production, as their data aggregation can be used as the basis for many services, ranging from visualization to advanced services such as the VEP introduced in this paper. The idea of virtual experiments is to accelerate parameter optimization in terms of sustainability measures and decrease the actual physical experiment effort. Thus, material consumption can be reduced. The user sets input boundaries and target parameters to be optimized and the platform returns the optimized parameter set which is then applied on the actual machine. For developing such a platform, an iterative six-step development approach based on design thinking was proposed. Similar to design thinking, an end user perspective is adopted. Finally, it was described how the virtual experiments platform can be applied in the battery cell production using the example of the electrode production in the “FFB Workspace”. Further research should lie on broadening the validation of the VEP conception to increase sustainability in the battery cell production.