The vision of the circular factory for the perpetual innovative product

1 Introduction

To decouple prosperity from resource consumption, society needs to rethink linear economic patterns. The scarcity of raw materials, coupled with the rapid growth of the world’s population, is a major challenge for society today [1], [2]. The “Earth Overshoot Day”, marking the date when humanity’s demand for natural resources exceeds the Earth’s regenerative capacity, has advanced from December 30th in 1970 to July 28th in 2022 [1]. This escalating global resource consumption implies that 1.75 earths would have been required in 2022 to meet humanity’s global resource demand [1]. Manufacturing companies, in particular, need to navigate the implementation of sustainable production patterns and rethink linear production patterns while adhering to the principles of economic behavior. However, a critical concern emerges when examining the global utilization of secondary materials. In 2018 only 9.1 % of the worldwide total material input could be met by secondary materials, highlighting a significant gap in resource efficiency [3]. This percentage further plummeted to 7.2 % in 2023, signifying a dramatic decrease attributed to the concurrent rise in absolute resource consumption [3]. For a scenario of consistent implementation of circular economy patterns, the Circular Economy Initiative Germany predicts significant savings of over 50 % of primary raw materials by 2040 [4]. This projection is not solely contingent on abstaining from resource use but also hinges on the strategic incorporation of secondary materials into production processes.

The concept of a circular economy, as a solution, is defined as an industrial model that decouples production performance from resource use, making it inherently restorative and regenerative [5]. Departing from the traditional linear production model, the end-of-life phase of a used product becomes the starting point for a new lifecycle. Various circular economy methods are distinguished in the literature (e.g., recycle, remanufacture, recondition, repair, cf. Figure 1) [5–9]. Remanufacturing stands out as the method with the highest standards in terms of quality and warranty for reprocessed products compared to alternative methods. It is the sole circular process that can compete with a new product in terms of quality and guarantee [10], [11]. Remanufacturing is defined as a value-retaining standardized process in which a remanufacturing product with at least the functionality and performance of the original product is created from reprocessed components of one or more used parts as well as new components [12]. This process adheres to predefined technical specifications, including quality standards, resulting in products with the same warranty as a new product [13–18]. The circular factory goes above and beyond remanufacturing, enabling maximum value retention within the product.

Figure 1:

Circular economic patterns and the focus of the circular factory [5–9].

Despite the considerable potential (example of a starter motor) for the resource (−88 %) and energy (−56 %) savings, as well as reduced CO₂ emissions (−53 %) offered by this approach in general and remanufacturing in particular, the implementation of these concepts is still in its early stages [19]. In 2021, the ratio of remanufactured products to new products in the EU was estimated to be only 5 % in representative economic sectors, except for the aviation industry (approximately 10 %) [20].

The “Circular Factory for the Perpetual Innovative Product” initiative is primarily concerned with the problem of achieving maximum value retention in products. This initiative addresses several key problems in modern industrial production, including increasing resource consumption, obstacles in implementing remanufacturing practices, early-stage development of necessary technologies, gaps in knowledge about disassembly processes, uncertainties in production processes and product conditions, and the lack of integration between linear and circular production methods.

Its goal is to achieve a breakthrough in circular production on an unprecedented industrial scale with a focus on value-added stages (manufacturing and assembly) where the most value is added to the product (cf. Figure 1). Therefore product functions need to be analyzed and predicted and uncertainty in product conditions and production processes needs to be manageable. To get knowledge from humans, the production technology needs to learn from them and the knowledge and further information need to be stored and made accessible at all times. Lastly, a factory combining linear and circular production processes must be designed.

The concept of circular economy originated in the 1970s intending to reduce the use of inputs in industrial production but has since shown its potential applicability to any resource [21]. In their 1990 work, Pearce and Turner describe a transition from a traditional linear economic system to a circular economic model. This circular model emphasizes the reintegration of waste from extraction, production, and consumption processes back into the economic cycle as valuable inputs [22]. The shift towards a circular economy is complex, given that our current economic system is heavily based on a linear model [5]. Originally, the ‘3R’ principles – reduction, reuse, and recycling – were the key strategies promoted for implementing a circular economy [23], [24]. However, with the increasing focus on sustainable innovation in recent years, the circular economy has evolved to embrace a 6R framework. This new approach incorporates recovery, redesign, and remanufacturing alongside the original 3Rs. The adoption of these 6R practices has led to improved outcomes globally [24]. Further principles evolved and extended the understanding of circular economy [25]. Although there are legal frameworks and initiatives in place, such as those in the USA and Asia [26], the European Green Deal [27], and the Circular Economy Act [28], there is still ample opportunity for further research and innovation [29]. Existing approaches in the context of the circular factory have so far only been considered as individual aspects. There were different aspects such as product-based data recording [30], sustainable product development [31], product design for easier reprocessing [32], approaches to new business models [33], part identification [34], and basics for industrial disassembly processes have been researched [35], [36], [37], [38], but there is a lack of control over product and process-related uncertainties as well as an approach for integrated linear and circular value creation. Other approaches deal with automated (dis)assembly technologies [39], remanufacturing processes [40], re-assembly processes [41], and sustainable product development [42], but what remains is a human-learning production technology and an instance-specific functional model for forecasting system reliability is not taken into account.

This article describes the concept of a circular factory for the perpetual innovative product and its subdisciplines for the research agenda. The research on the circular factory is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) as part of the CRC 1574 (https://www.sfb1574.kit.edu/index.php) and implemented by an interdisciplinary team from various fields, including production science, product development, material science, human factors, robotics, computer science, and knowledge modeling. The scientists will explore scientific questions. Once the questions have been clarified, the vision can become a reality. Following the introduction in the current chapter, Section 2 presents the vision of the circular factory for the perpetual innovative product and the description of the overall system. Furthermore, it defines the core aspects and research questions. Section 3 delves into the research area of predicting the product function, followed by Section 4, which addresses managing uncertainty. Section 5 discusses production technology that learns from human actions. Section 6 explores the changeable, autonomous production system, and Section 7 covers knowledge modeling. These chapters correspond to the working groups of the CRC and can be seen as subdisciplines involved. Finally, Section 8 provides a summary and outlook.

2 Circular factory for the perpetual innovative product

The vision of the circular factory is based on the general understanding of remanufacturing (cf. Section 1), but goes far beyond its implementation. Unlike traditional circular economy approaches that primarily focus on recycling or reusing materials at the end of a product’s life cycle, the circular factory emphasizes the maximum preservation of added value by material cohesion. Used product instances (their subsystems, components) will constantly be transferred to the same or a new product generation.

The primary goal is to ensure maximum retention of added value by minimizing the disassembly of used products and updating their technology so that the reprocessed products from the circular factory can be established on the primary market. While products are usually developed in so-called product generations, i.e., they follow the further development of a reference product [43], through cross-generational product development, product instances (or their subsystems and components) can be continuously transferred to the product generation currently on the market (G _n−1) or a new product generation under development (G _n ) [44]. This will enable theoretically perpetual product use (cf. Figure 2). Even if this does not appear to be feasible in practice, the vision of the perpetual innovative product should be introduced in a manner comparable to the Total Quality Management (TQM) goal of “zero defects”.

Figure 2:

Vision of the perpetual innovative product [45].

In the circular factory, the products will be qualified for the primary market, as they have an equivalent performance compared to purely linearly produced products and innovative functions are also integrated using technological updates. A product leaving the circular factory is always a new product of the original (G _n−1) or a newer (G _n ) product generation with an individual share of reprocessed components or subsystems, which can be produced in large series but meet the requirements of a new product in terms of reliability, function, warranty, etc. The aim is to minimize the number of new, purely linearly produced products. Only a few purely linearly produced subsystems and components should be added or manufactured in the circular factory to meet current market requirements. The transfer of a used product into newer product generations is based on the integrated design of the product (generations) and production systems (Product-Production-CoDesign) [44]. The focus here is on interlinking product description, starting with product development and production, including data from usage, sales, and service to diagnosis and refurbishment in the circular factory. This involves the development and integration of a methodical procedure for product development, an approach for a design for circular factory for products that are tailored to circular processes.

In order to make the vision of the perpetual product possible in terms of production technology, it is necessary to combine circular production processes with existing linear production processes in an integrated production system (cf. Figure 3). This integrated production system will be capable of both the linear production of new products without reprocessing and the circular process chain, using the same production resources to enable a highly efficient circular production system. This will make it possible to achieve high-performance, highly efficient high-volume production as known from linear production. New products with individually reprocessed components will be manufactured for the first time on an industrial scale and, therefore, with low unit costs and sold economically on the market. Circular production processes will be established at high-wage locations in large-scale production, and the added value can be maintained. This is demonstrated using the example of the angle grinder, a hand-held high-tech product manufactured in Germany. The angle grinder, its subsystems or individual components are used to test the various processes in the circular factory. It is suitable because the generations of the product are short (18–30 months), there is a high number of variants per generation (more than 10) and there is a high degree of uncertainty due to the wide range of applications.

Figure 3:

Integrated production system for linear and circular production [45] (green: focused process steps of circular production, white: focused process steps of linear production, based on [2]).

Within the circular factory, the following five core aspects can be derived and will be the subject of working groups in the CRC:

1. Prediction of the product function: How can subsystems and components be identified, remanufactured, and technologically updated in order to establish value-added products from the circular economy in the primary market?

2. Managing uncertainty: How can uncertainty that arises from uncertain states of used products and derived process sequences in the circular factory be managed?

3. Production technology that learns from human actions: How can human action knowledge in the execution of complex manipulation tasks be learned from human observation and transferred to industrial automated production technology?

4. Changeable, autonomous production system: How can self-learning, autonomously adapting production resources be used to realize circular production in large-scale productions?

5. Knowledge modeling: How can a unified semantic representation of circular factory knowledge be enabled to integrate information across all domains?

The core components of the circular factory and the guiding research questions it addresses are presented as examples in the following sections as well as in more detail in the specific articles in this special issue:

1. Enabling the Vision of a Perpetual Innovative Product

2. Managing Uncertainty in Product and Process Design for the Circular Factory

3. Learning Human Actions from Complex Manipulation Tasks and Their Transfer to Robots in the Circular Factory

4. Self-learning and Autonomously Adapting Manufacturing Equipment for the Circular Factory

5. The Role of an Ontology-based Knowledge Backbone in a Circular Factory.

3 Prediction of the product function

In product development, innovations often progress through distinct generations, each building upon a reference product [46]. Production volume follows the conventional life cycle stages of introduction, growth, maturity, saturation, and decline. In transitioning from linear to circular production, the challenge arises of predicting and implementing functions in new product generations using components from older ones.

Circular production traditionally focuses on remanufacturing products within the same generation, creating a constrained market due to customer preferences for perpetual innovative new products. A more intensive interweaving of product development and production is required to address this, aiming for a perpetual innovative product [44]. To tackle the resulting already raised research question in Section 2, three key challenges are identified:

1. Inspection and diagnostics of circular products and subsystems,

2. Specifications for rework of the embodiment,

3. Design for circular factory.

Ensuring the functionality and reliability of the new products from the circular factory necessitates accounting for diverse information, such as usage data, inspections and diagnostics, and material analyses. Automated inspection of used products and component inspection face difficulties due to the wide range of variants and generations, prompting the development of an over-instrumentalized measuring and inspection station. Autonomous measurement technology concepts are also explored, incorporating the diversity of product variants and generations.

To overcome the mentioned challenges regarding the specification for rework of the embodiment necessary for technological updates, it becomes necessary to introduce a reliability model derived from the functional model, providing insights into the product’s reliability and remaining service life. A product instance and its subsystems might need to be reprocessed and therefore reevaluated. The models are introduced in Figure 4.

Figure 4:

Models for analyzing, reworking, and technologically updating subsystems and components in a circular factory.

Functional Model: In the context of circular factory design, the functional model assumes significance by delineating the intricate relations between a product’s embodiment and its functionalities. Providing both qualitative and quantitative modeling elements, this model will offer essential input for inspection procedures, contributing to a deeper understanding of the product’s behavior as basis for the prediction of the functions of new products which are built based on the used products, subsystems and components.

Reliability Model: Integral to the planned circular production, the reliability model will extend the predictive capabilities to anticipate the degree of function fulfillment over time. Validated at subsystem and system levels, this model forms the basis for decision support during the processes necessary for technological updates of the used products to new ones.

Instance-Individual Tolerance Schemes: To facilitate concrete decision-making at the instance level, the establishment of instance-individual tolerance schemes becomes paramount. These schemes, delineating compliance ranges for attributes of each instance, operate in tandem with the digital twin. There are two distinct types of tolerance schemes, namely function-related and reliability-related, which are indispensable for effective inspection and decision support during the rework processes.

The vision of the circular factory ensures that every product leaving it is essentially a new one, presenting a unique challenge of transforming used product generations into renewed ones. Addressing this challenge involves anticipating the reusability of subsystems and components in future generations during product development. The Design for Circular Factory is dedicated to creating products tailored for circular processes, aspiring to generate new products with innovative properties while avoiding functional restrictions. This approach involves cross-product generation and cross-variant considerations, developing functional models to maximize the reuse potential of used products, subsystems, and components.

The proposed framework contributes to the realization of a perpetual innovative product within a circular factory, aligning with the vision of a sustainable and efficient production system.

4 Managing uncertainty

The circular factory is influenced by multidimensional uncertainties on different levels. Therefore, one needs to answer how these uncertainties are managed that arise from uncertain states of used products and derived process sequences in the circular factory.

For efficient circular production, these product, process and model uncertainties must be recorded, modeled and controlled. In the literature, a distinction is generally made between uncertainties with regard to quality (product condition) and the number of returned used products as well as the time of return of these used products [17], [18], [47], [48]. The consideration of the perpetual product adds further uncertainties concerning the diversity of generations and the variants contained therein. These product-related uncertainties cannot be controlled by a predefined process sequence, as in linear production [49]. Therefore, process sequences must be derived, which must be individually designed depending on the product characteristics and thus lead to further process-related uncertainties [17]. In each step, a decision must be made individually and depending on the result of the previous process step as well as the expected result and the characteristics of potential subsequent processes (e.g., duration, necessary/available resources, skills, capacities) as to how to proceed with the used product and its subsystems and components. Furthermore, due to the comprehensive information technology in the circular factory, it must be assumed that uncertainties regarding the models, operations, processes, etc. used can also influence the final product. Such uncertainties arise, e.g., when processes are learned from human workers for automation. Also for connecting product-based and process-related uncertainties, models must be established, which may be affected by imperfections, leading to additional uncertainties. These model-related uncertainties must also be taken into account.

In the circular factory, the described multicausal uncertainties will be considered as a basis to specify the quality of the derived conclusions in the further use of the information. It is, therefore, essential that uncertainties of the respective objects of interest (variables, attributes, models, functions, sequences, operations, processes, etc.) are consistently and integratively considered, adequately expressed, transferable, and usable. The aim is to achieve consistent uncertainty modeling with which uncertainty can be uniformly described, propagated, merged, utilized, communicated and interpreted. If necessary, this should be supplemented by meta-information, for example concerning the origin of information, to exclude multiple considerations. Uncertainties, both epistemic (i.e., caused by a lack of knowledge about products, processes, and models) and aleatory (i.e., caused by the stochastic nature of measurements), are treated uniformly probabilistically in accordance with the Guide to the Expression of Uncertainty in Measurement (GUM) [50] and interpreted as probability in the Bayesian sense as Degree of Belief (DoB) [50], [51], [52], [53]. In this context, a probability distribution will be used to describe the possible values of the object of interest and to specify normalized weights for these values. In consequence, a stringent mathematical description of uncertainty is available, which can then be used for fusing different information contributions. The fusion step itself will be performed using Bayesian fusion, thus enabling information contributions from different (heterogeneous) sources and prior knowledge to be taken into account consistently [54]. Here, all probability distributions can be expressed as conditional distributions given the current information status. If new data or information is available, the fusion is iteratively repeated, leading to a change of the level of information and a reduction of uncertainty. To obtain a certain (possibly low) level of uncertainty, suitable sensor systems that are able to collect information about the object of interest need to be identified and applied in a superordinate sensor scheduling system.

In summary, multicausal uncertainties will be controlled, which improves the basis for decisions in the circular factory. Scientific issues include the generic bias-free transformation of arbitrary heterogeneous facts and the efficient implementation of Bayesian fusion for high-dimensional probability distributions using approximate methods in an agent-based architecture.

5 Production technology that learns from human actions

In the vision of the circular factory, used products undergo a process of disassembly, reprocessing, and transition into either their original generation or a newer product generation. These process steps entail complex requirements, including dealing with uncertain product specifications such as corrosion or contamination, as well as encountering unknown conditions like damaged or partially disassembled products [55]. The existing uncertainties make the automation of these tasks challenging [56], leading current processes for remanufacturing used products to predominantly rely on manual execution (cf. Figure 5). Human abilities and the experiential knowledge of workers play a crucial role in these processes [57]. However, externalizing this knowledge proves difficult, as the activities and procedures are highly individualized for each unique product instance due to the aforementioned uncertainties, making them inadequately represented in traditional work step descriptions [58]. Consequently, circular value creation approaches at high-wage locations are often economically unfeasible [4].

Figure 5:

Production technology that learns from humans [59].

To implement such approaches on an industrial scale and in mass production, it is essential to externalize existing knowledge and transfer it to standardized, automated production facilities. Building on the groundwork laid by AgiProbot [60], [61], the integration of human uptake, learning, and transfer processes can be efficiently incorporated into automated production processes.

The central scientific question revolves around the concept of human-learning production technology, examining how human procedural knowledge in executing complex manipulation tasks can be learned from human observation and transferred to industrial, automated execution by robots. The primary objective is to initially acquire and interpret human behavior during process execution in a multimodal manner, enabling the description of assembly/disassembly-relevant strategies and actions. The challenges in human perception lie in investigating how human procedural knowledge from multimodal data (eye/gaze behavior, speech, body poses, forces, hand movements) can be externalized in a machine-executable format. Deep learning-based methods will be developed to automatically capture humans’ current state in a fine-grained manner without requiring a large amount of annotated training data. Additionally, techniques will be explored to automate the externalization and description of manifested implicit knowledge through the empirical analysis of human eye and gaze behavior. Previous findings from the Carl Zeiss Foundation research group AgiProbot indicate that eye and gaze movements, in addition to attention focus, provide insights into cognitive processes such as processing depth and problem-solving strategies [59]–[61]. For comprehensive learning from human observation, both explicit (motion demonstrations) and implicit procedural knowledge (cognitive information processing) must be externalized and integrated. As this represents a special situation, a specially designed learning cell as an experimental environment is used for this purpose. In addition, learning can also take place directly on the shop floor in a manual dis-/re-/assembly station.

In the next step, robotically processable task models will be created to reproduce and generalize the learned actions on robots. The challenges of robot programming predominantly involve determining how the necessary task models for two-armed manipulations can be learned from human observation and achieving task-specific, category- and experience-based generalization of task models. A library of task models is built using machine learning methods based on data from multimodal human capture and knowledge of robotic execution, and it is continuously expanded and improved. The task models represent procedural knowledge at symbolic and subsymbolic levels. The task models will be executed on a two-armed robot. Subsequently, the transfer to kinematics in the circular factory takes place. A physical and photorealistic simulation of production resources will be developed and used to generate various artificial data for different learning tasks. Subsequently, Sim-to-Real transfer methods can be applied to utilize the manipulation skills optimized in the simulation on real production resources and generalization occurs.

The central research question focuses on two main areas: the multimodal capturing of human behavior and the reproduction and generalization of learned actions on robots. The efforts are centered around a learning cell, set up with standardized environmental conditions adjacent to the productive area of the circular factory. For multimodal capturing, the learning cell is equipped with the necessary sensors (RGBD cameras, microphone, eye tracking system, sensory tools) to conduct research experiments. Participants are observed and recorded using the integrated sensors during complex assembly/disassembly processes. Additionally, a semi-humanoid two-armed robot is present in the learning cell to evaluate task models learned from human observation. Data from the execution of these task models are provided to the simulation, employing Meta- and Few-Shot Learning, facilitating adaptation of skills on the Transformer-Cell. Internal feedback loops for implementing captured human procedural knowledge into machine learning methods facilitate extensive knowledge exchange. This allows assessment of the effectiveness of learned task models on robots, identification of missing information from human observation, and refinement of developed methods.

6 Changeable, autonomous production system

To realize the manufacturing of a perpetual product, it is essential to integrate circular production processes seamlessly with existing linear production methods within a unified production framework. The question arises of how self-learning, autonomously adapting production resources, including self-learning and autonomously adjusting systems, as well as efficient and robust production planning and control, can be used to realize circular production at large-scale in high-wage countries.

The integrated system will empower both linear production and circular process chains, harnessing shared production resources synergistically to establish a highly efficient manufacturing environment. The incorporation of reprocessed subsystems and components into the linear value chain achieves a level of efficiency akin to traditional linear processes, facilitating high-volume mass production. This integrated approach can establish circular production processes not only at high-wage locations but also in mass-production scenarios, thereby preserving value creation. Although there are existing methodologies, as outlined in refs. [62], [63], aimed at enhancing the agility of production systems, their primary emphasis lies in adapting to flexible scaling. Introducing modularity to the intralogistics system [64] and devising decentralized control concepts [65], [66] can improve adaptability and flexibility. However, these strategies have not been investigated within remanufacturing or hybrid production structures. Circular factory production planning involves long-term capacity planning [67], dynamic scheduling [68], and sustainability-focused KPIs. Complexity arises from uncertainties related to returned products, arrival times, variants, and learning capabilities of production resources [48]. Existing approaches like [69], [70] lack the needed dynamism and intelligence for effective planning and control ensuring rapid responsiveness. The consideration of uncertainties concerning the condition and quantity of used products, particularly at the tactical and operational levels, presents a challenge that renders rigid a-priori planning and control unfeasible. The wide array of error states and wear phenomena of the products requires adaptive adjustments of production resources to new product instances and coordination in unforeseen situations, posing challenges beyond conventional approaches. This innovative approach will leverage the high flexibility, adaptability, and autonomy of all production resources on the shop floor, enabling them to independently adapt to changing conditions and challenges. Within the circular factory, all production resources will be equipped for the integrated execution of both linear and circular processes, maximizing flexibility at the shop floor level. This is achieved on a laboratory scale through integrated planning and control with dynamic uncertainty, incorporating the formal representation of highly adaptable production systems. Key components include an inspection station, a manual dis-/re-/assembly station, the Transformer-Cell with special machines such as additive and subtractive machines for highly demanding processes, and support for value creation through mobile assistants. The universal and configurable Transformer-Cell incorporates all process steps from both linear and circular production. It uses two industrial robots capable of performing a variety of handling and manufacturing tasks through modular end effectors and process modules, which can be optimally placed to reconfigure the cell for different tasks [71]. For manufacturing reprocessing steps, a combination of subtractive and additive processes ensures maximum value retention through adaptation or compensation mechanisms. The design of these processes, particularly concerning material transitions, introduces additional research questions. An autonomous and scalable intralogistics system is essential for the flexible interconnection of individual stations. This system also reconfigures the robot cell and expands to perform value-adding tasks, thereby necessitating the exploration of intelligent mobile assistants.

The construction, both physically and in software, of the Circular Factory at the laboratory scale stands out as a pivotal outcome of this comprehensive approach, making the vision reality (cf. Figure 6). The hardware on the shopfloor (Transformer-Cell, Manual dis-/re-/assembly station and mobile assistants) as well as the Learning Cell in the laboratory realize the linear and circular production processes.

Figure 6:

Circular factory on a laboratory scale [45].

7 Knowledge modelling

The vision of the Circular Factory for the perceptual innovative product as proposed in this paper brings together several domains such as production, product engineering and robotics. To solve the raised research question in Section 2 about how to bring all of them together the collaborative development of knowledge management for the circular factory comprises three core components: the Knowledge Base Management System (MS), Production Logic MS, and Runtime MS (cf. Figure 7).

Figure 7:

Concept for knowledge modeling in the circular factory based on [72].

This research and accompanying modeling involve the shared Knowledge Base MS semantically represents diverse models, including products, subsystems, components, and production resources, continually enriched with extracted knowledge (cf. Figure 7). It introduces novel features such as the storage and semantic querying of uncertain product-related data and the utilization of experiential knowledge (cf. Figure 7). The Runtime MS coordinates machine, process, and robot controls, ensuring integration within the circular factory through specified interfaces between software and hardware components. Additionally, real-time sensor data informs decision-making in control systems (cf. Figure 7).

Within the Data Infrastructure applications for overarching control tasks, such as order sequence formation get established. This research employs ontological modeling based on the circular factory’s underlying ontology and an ontology-based event-discrete process simulation [44]. This facilitates the modeling and representation of processes, resources, products, and their hierarchical connections, supporting traceability and serving as a runtime model for simulation studies. Concerning research data management, the imperative of handling digital research data in engineering is underscored. A unified research data management infrastructure is advocated, with open-source systems like Ckan, Dataverse, DSpace, and Invenio proposed as potential solutions [73–77]. Data heterogeneity gets addressed through an ontology-based knowledge graph and the hierarchical structuring of core and sub-ontologies [78]. The knowledge graph acts as a central repository, facilitating data exchange in the circular factory as examples in Bosch or Siemens already show [79], [80].

The use of ontologies and knowledge graphs, particularly RDF, OWL, and SHACL, is pivotal for semantic data description [81]. These stored data and ontologies enable interactive processing and analysis through query languages like SPARQL [82]. Interfaces to Triple Stores, databases, and dashboards are managed using the Kadi4Mat research data infrastructure [83].

In ontology and knowledge graph development, the project draws on existing literature and experiences, incorporating concepts such as pattern-based ontology development [81]. This research explores ontologies for product modeling, resource-oriented ontologies, and specialized ontologies for multimedia and sensor data based on [72]. The development of the Runtime MS considers modeling of runtime systems, including hierarchical and non-hierarchical communication architectures, with a focus on standards like AutomationML and OPC UA [84], [85].

In conclusion, the collaborative knowledge management effort for the circular factory integrates advanced systems, including ontological models, knowledge graphs, and sophisticated data management strategies. His combined approach of research and modeling ensures the effective operation and integration of diverse subprojects within the broader research initiative.

8 Summary and outlook

The research project lays the groundwork for an integrated method of product development in future circular factories, which is a visionary concept. These factories will integrate linear and circular production, learn from humans, and continuously adapt to new products. With the scientific foundations of the circular factory, there is the potential to enable large-scale processes for the renewal of used products, offering diverse perspectives to strengthen the economic position of Germany and Europe. The following innovation potentials can be realized:

1. Production of Products without Resource Waste: The reuse of used subsystems and components in further usage phases and product generations will reduce resource consumption.

2. Reduction of Waste Volumes: Reusing used products and transitioning them into the latest product generation will result in lower quantities of non-usable products, which would otherwise need to be recycled.

3. Creation of Economic Potentials: Integrated linear and circular production with highly utilized production resources makes the circular factory economically attractive. Additionally, production costs can be reduced through the lower material and energy input of reprocessed components and subsystems, ultimately allowing for attractive profit margins.

4. Perpetual Product Increases Customer Value: The transition of used products into a current product generation with the latest innovations will provide customers with sustainable products without functional drawbacks compared to linearly produced items.

5. Value Creation in High-Wage Locations: Through human-learning production technology and the transfer of manual processes to machines, high-volume circular production becomes feasible, contributing to the long-term assurance of value creation in high-wage locations such as Germany and Europe. Simultaneously, humans remain central as a source of knowledge and enabler of human-centered production technology.

For this purpose, the conventional manual remanufacturing processes will be transitioned into an initial learning automation in this context. This involves developing targeted methods for learning complex manipulation strategies for disassembly, reassembly, and assembly processes through human observation. This process leads to a comprehensive understanding of unique products and processes as a short-term goal.

Simultaneously, the foundation for the subsequent renewal of products with diverse variations and large-scale production of handheld tools, will be established through the investigated Product-Production-CoDesign. In the long term, the perspective of perpetual product renewal emerges, where subsystems and components of a used product can be reused extensively across different types, variations, and products in the next product generations. Furthermore, various exemplary products are considered. Initially, the production technology, capable of learning from humans, concentrates on electric motors and handheld tools.

The ultimate goal of the circular factory is to empower production technology for maximum value preservation during the reprocessing and technological upgrading of used products within an adaptable, autonomous production system of the circular factory. This results in symbiosis, i.e., the fusion between planning and control.

However, it is crucial to acknowledge that the successful implementation of the circular factory as a dominant economic model relies heavily on the alignment of regulatory frameworks and business models in the right direction. This vision is contingent on regulatory support and business model innovation, ensuring that the circular factory can function independently of existing conventional or future business models. Business models are initially only considered from a product perspective in the CRC.

In the first period of the CRC, the focus is on known product types from two defined product generations with uncertain conditions. Methods for individualized, cross-generational product development will be created, uncertainties will be captured, modeled, and assessed, and the capability for manipulation in production technology will be established. Additionally, systematic approaches for production implementation, as well as for the systematic acquisition, storage, and exchange of data, will be developed.

Following the first period, the knowledge and results gained will be integrated and applied to products from multiple generations with uncertain conditions. This involves integrated planning and control of Product-Production-CoDesign, a value-based approach to dealing with uncertainties, and automation of functional assessment. Moreover, human skills and knowledge will be transferred to new problem scenarios, the systematic approach will be expanded, and a central, operational knowledge graph will be established. In the last period, the goal is to achieve autonomy in the circular factory. This involves developing autonomous function prognosis, strategies for resolving uncertainties and evaluating unknown subsystems, incremental autonomous learning, self-configuring production technology for maximum adaptability, and intelligent decision support.