The Concept of a Hierarchical Digital Twin

Abstract

The concept of a digital twin has become a key driver of industrial transformation, enabling a seamless connection between physical systems and their virtual counterparts. The growing need for adaptability has accelerated the use of advanced technologies and tools to maintain competitiveness. In this context, the article introduces the concept of a hierarchical digital twin and illustrates its operation through a practical example. Production resource structures and timing data were generated in the KbRS (Knowledge-based Rescheduling System), which will serve as the Level II digital twin in this article. The acquired data is transferred via Excel to the FlexSim simulation environment, which represents the Level I digital twin responsible for modeling the flow of production processes. Because a digital twin must accurately reflect a specific production system, the study begins by formulating a general mathematical model. Algorithms for product ordering and for constructing the digital twin of the production processes were developed. Furthermore, three implementation scenarios for the hierarchical digital twin were proposed using the KbRS and FlexSim tools. The implementation of the hierarchical digital twin concept facilitated the development of the more comprehensive virtual model. At the same time, the integration of data between the two software environments enabled the generation of more detailed and precise results. Traditionally, a digital twin created solely within a single simulation platform is unable to represent all the structural components of a production system—an issue addressed by the hierarchical approach presented in this study.

1. Introduction

Nowadays, the digital world influences not only the lives of end users but also has a growing impact on the very functioning of businesses. By integrating numerous devices and sensors, supported by various types of software, it is possible to streamline operations across many enterprise levels, including production. Currently, tablets and smartphones can be used to communicate with machines and devices, and even control cars in everyday life (for example, by heating the car before we enter it). Furthermore, tablet and smartphone technology is also being used in the automotive industry, for example, in barcodes and Bluetooth. Smart manufacturing optimizes resource allocation and is characterized by real-time analysis, intelligence, continuous improvement in production and services, and perfect adaptation to market requirements in real time [1].

The introduction of the concept of a cyber–physical system (CPS) is a crucial element of information processing technology [2]. This new technology supports diagnostics, is service-oriented, introduces alarm management, and enables cloud data storage. It involves building a precise and realistic digital model of a device or machine using digital twin technology.

The concept of a digital twin has become a key element of digital transformation in the industry, providing a bridge between the physical and virtual worlds [3]. As a virtual representation of a physical object, system, or process, a digital twin enables real-time monitoring, analysis, and optimization. In the context of Industry 4.0, digital twins are the foundation for smart factories that integrate advanced technologies such as the Internet of Things (IoT), artificial intelligence (AI), big data analytics, and cloud computing [4,5,6,7].

Industry 4.0 is a paradigm for the digital transformation of production, originating from the German strategic initiative of 2011–2013: the ACATECH report [8]. It has demonstrated that intelligent, networked cyber–physical systems, integrating operational and business levels, enable flexible, personalized, and highly efficient manufacturing.

Within Industry 4.0, a key construct is the “smart factory”, which is a production facility operating as a cyber–physical system, capable of self-monitoring, learning, and adapting in (near) real time; it also involves integration with a digital supply network.

Data processing in the case of digital twin means the need for technology supporting huge amounts of data collected from various physical and virtual objects (machines, historical databases, etc.) [9]. Tao et al. proposed technologies such as data generation, modeling, cleaning, clustering, mining, and evolution [10]. To simplify the data collection process itself, it is necessary to reduce the data size. Ricks et al. proposed a rank reduction technique for data processing systems in this regard [11].

Data integration is a significant challenge. In this case, Cai et al. proposed a method for integrating data collected from sensors and production data as a basis for building a DT data processing system for a vertical milling machine [12]. Despite numerous studies on data fusion between physical and virtual objects, many aspects still remain to be investigated [9].

According to Rosen et al., DT systems can protect the production system during dynamic changes in the physical space. DT collects all data from physical objects and then uses it to simulate operational procedures in the virtual layer. The production system automatically executes orders according to the simulation results [13]. Vachálek et al. assume that thanks to constant interactions between physical and virtual spaces, DT systems will respond more rapidly (significantly faster) to unexpected changes in production processes [14].

However, the literature raises the issue of large discrepancies in the definitions, architectures, and levels of DT advancement: from simple 3D models to complex cyber–physical systems synchronizing the physical and virtual world [15,16]. In particular, research indicates that many implementations are limited to digital models (digital models) or digital shadows (digital shadows), lacking the full bidirectional integration and automation necessary for a true production twin [17,18].

After analyzing the available literature on DT implementation, it turns out that the key and problematic element of this process is DT modeling itself [12,13,14,19,20,21,22,23,24,25,26,27,28,29,30,31]. The cited articles lack a common perspective on DT model construction. Some articles mention five-layer, three-layer, or five-dimensional modeling [24,32,33]. Given these discrepancies, it seems necessary to standardize modeling frameworks and develop more tools for DT construction.

Moreover, the latest reviews show that the use of DT in production scheduling and management of dynamic processes (e.g., variant production, just-in-time, production to order) is only just gaining attention—there is still a lack of structured methodologies that would enable the adaptation of DT to the needs of industrial plants [34,35].

Standards and reference frameworks such as ISO 23247 [36]—Digital Twin Framework for Manufacturing and RAMI 4.0 significantly contribute to the unification of terminology, architecture, and data exchange in DT which promotes the interoperability and scalability of solutions [15,16]. However, despite their existence, the empirical literature requires further research: there is a lack of case studies demonstrating the integration of DT with scheduling systems, simulation, and comprehensive management of variant production. Particularly in the automotive sector, where production must accommodate multiple vehicle versions and variable orders, the practical implementation of such complex DT remains rare. This gap is precisely the motivation for this work.

The rapid development of digital manufacturing processes in the automotive industry creates another opportunity to improve and accelerate the production of the final product. By “virtually launching” production in a simulation program, it is possible to verify most processes before production begins. This is important not only for the supplier and integrator but also for the manufacturer. For several years, manufacturers have also been using computer software to digitize the actual production process, known as a digital twin. The digital twin, thanks to its virtual commissioning capabilities, has also changed the pre-production process, for example, by simulating the assembly of pre-assembled parts on trolleys in the warehouse [37,38,39].

Simulations play a crucial role in the analysis of complex manufacturing systems, enabling the modeling of processes from the shop floor to the supply chain, both as-is and as-to-be [40,41,42,43]. However, traditional approaches require highly detailed models and often prove insufficient, especially when there is no consistent link between the virtual model and the actual object [41,42].

In environments utilizing IIoT and advanced data collection systems, digital twin (DT) technology addresses these limitations [1,40,41,42,43]. Each physical element can be represented as a virtual model fed with real-time data, allowing not only process analysis but also dynamic forecasting, optimization, and adaptation [44,45,46]. In systems requiring frequent reconfiguration, such as RMS [47], DT offers a distinct advantage over traditional simulation: it enables rapid, realistic assessments of the impact of changes, supporting better operational and strategic decisions.

Also, authors of the article [48] propose the use of a digital twin and a cobot to improve the work of a pathomorphological laboratory.

Next, in the paper [49], authors propose a predictive maintenance model for switch machines based on digital twins (DTs). Also, the DT model was created for a switch machine, which contains a behavior model and a rule model.

On production lines, components are picked from carts by robots or workers and assembled in the appropriate locations on industrial tables. Components placed in the cart are picked (selected) individually, which is a simple process if only one type of component is to be assembled at a single production station. However, different vehicle variants are created, even on the same production line. Therefore, different types of components are required depending on the vehicle’s intended use. One engine or body version will be assembled for an off-road version, and another for a city version. However, before the components reach the production hall, the warehouse is responsible for placing the components in the cart. The components must be arranged in the correct order. In the event of a mistake, production will be halted because the employee will be unable to assemble the correct part, or the picking robot will malfunction or collide with the part it is picking, for example, from a cart. The production sequence for each cart depends on the production station, and the sequence at the production station depends on the production order, which in turn depends on the customer order.

To analyze production processes using the example discussed, a digital twin was developed for a section of an automotive production line. Components are placed on carts in a warehouse, then retrieved from the carts by robots or workers and assembled in the appropriate locations on industrial tables at a selected station. However, this article focuses on the functioning of the digital twin as a hierarchical operational concept based on the general example of a final product production order. The focus is on the implementation of operations within a narrower scope, i.e., at the station where products are assembled. Simulations of a process fragment were conducted in FlexSim (ver. 2025), as the primary of Level I of digital twin tool, and in KbRS (ver. 20250524), which is a Level II of the digital twin tool, where data regarding production processes, durations of unit operations on resources in the processes, and their routes are entered and then ultimately transferred to the FlexSim software. Furthermore, based on the acquired data, KbRS also generates schedules for resource operations. The developed hierarchical digital twin concept aims to streamline the production data entry process by downloading data from the KbRS database into the FlexSim simulation program, rather than entering it sequentially in each program. Scenarios for problem-solving operations at the station discussed were also developed.

The aim of this article is to propose and test the concept of a hierarchical digital twin—integrating planning, simulation and scheduling tools—and to demonstrate its usefulness in the context of variant production in an automotive plant.

2. Digital Twin Concept

In contemporary Industry 4.0 research and practice, the digital twin (DT) is increasingly recognized as a key mechanism for achieving cyber–physical integration. According to Grieves and Vickers, a DT [50] is a coherent set of virtual information structures describing a physical object, its behavior, and state throughout its lifecycle, coupled with a flow of measurement data and capable of powering simulation, analysis, and prediction. A triad has also been established in the literature, namely a digital model (digital model), a digital shadow (digital shadow), and a digital twin (digital twin), varying in the direction and automation of data flow (from unidirectional to bidirectional, controlling).

A digital twin is a digital representation of real-world objects, such as a production plant, its products and processes, and its management and production system. It represents the real world in a virtual system. Therefore, it is created using real-time data and modeled in 2D/3D to replicate and simulate the behavior and performance of its real-world twin. A digital twin is, in most cases, a very complex system, composed of interconnected layers, each of which plays a key role [51]:

• Physical layer—the actual object or system;
• Data layer—collects raw data from sensors and devices;
• Connectivity layer—transmits data to a centralized system;
• Data processing layer—cleanses and organizes data for usability;
• Modeling and simulation layer—creates digital replicas using models and simulations;
• Data analysis and intelligence layer—uses artificial intelligence and analytics to obtain actionable insights;
• Control and automation layer—provides feedback and control of the physical system;
• User interaction layer—interfaces for engaging users;
• Visual layer—presents data and models visually.

2.1. Frames of Reference and Standardization

To ensure interoperability and consistency in digital twin implementation, a number of standards and reference frameworks have been developed:

• RAMI 4.0. (Reference Architectural Model of Industry 4.0) is a reference architecture model for Industry 4.0 that defines the structures and relationships between various system components. Its task is to organize interoperability and integration along three axes: (1) layers (from resource to business), (2) IEC 62264 [52]/IEC 61512 [53] hierarchies (from field to enterprise), and (3) the product life cycle according to IEC 62890 [54]. RAMI 4.0 combines existing standards (including OPC UA, IEC/ISO) and defines an I4.0 component as a functional “wrapper” of a resource with semantics and interfaces [3].
• ISO 23247 (Digital Twin Framework for Manufacturing) is a series of standards (from 2021) defining a framework for the creation and operation of manufacturing twins, covering terminology, stakeholder roles, logical and component architecture, mechanisms for identifying “observable elements” (product/process/resource), and data flows and synchronization. Additionally, NIST literature presents an analysis of the 23247 series and how it can be specialized for discrete, batch, and continuous manufacturing [55].

2.2. Digital Twin in Production—Overview of Concepts and Taxonomies

Literature reviews indicate the lack of a single, universal definition of DT and a high heterogeneity of applications. Kritzinger et al. [56] propose the DM–DS–DT taxonomy, which allows for classifying solutions according to the level of feedback and autonomy. Review papers from 2018 to 2025 [56,57,58,59,60] catalog applications along the following dimensions:

• Representation object (product, process, resource, production system);
• Lifecycle stage (design, planning, commissioning, operation, service);
• Functions (monitoring, diagnostics, optimization, prediction, control);
• Techniques—multi-scale simulation, analytics/ML, model calibration, anomaly detection. Recent work emphasizes the need for a common taxonomy of DT applications in manufacturing, combining functional aspects and implementation maturity.

The concept of a digital twin has evolved from simple 3D models to advanced systems integrating data in real time. Kritzinger et al. [56] propose the classification of digital twins into three levels, namely digital model (DM), digital shadow (DS), and digital twin (DT), depending on the degree of integration and interaction with the physical object.

2.3. Applications in a Smart Factory

The concept of a digital factory encompasses the broader context of industrial transformation, in which digital models, simulations, and tools integrate all stages of the product lifecycle—from design, through production, to maintenance [3]. The digital factory is the basis for implementing innovative solutions such as process automation, intelligent management systems, and collaborative robots (cobots). According to Grieves and Vickers [50], the digital factory is the foundation for the practical application of digital twins in the industrial environment in the following areas:

• Predictive maintenance and reliability: Real-time twins of machines, lines, and systems (e.g., robots, CNC machine tools) allow for the prediction of degradation, the planning of condition-based maintenance, and the minimization of downtime. Combining physical models (model-based) with data (data-driven) increases prediction accuracy.
• Flow and planning optimization: DT for production cells and lines supports scheduling, load balancing, layout configuration, and “what-if” scenarios (including reconfiguration of small-batch production/mass customization). In flexible production cells, DT reduces commissioning effort and shortens changeover times.
• Quality and production startup: Real-time twins of processes (e.g., welding, 3D printing, molding) enable parametric control, automatic calibration, and closing the quality loop (in-line metrology → setpoint correction).
• Safety and ergonomics: Virtual commissioning and HRI (human–robot interaction) simulations reduce the risk of accidents and enable the design of workstations with operator safety and workload in mind.
• Energy management and sustainability: Utility twins (UTs) model utility consumption, emissions, and costs, supporting ESG decisions and certifications (e.g., ISO 50001 [61]), as well as demand response management.

In summary, the digital twin and the digital factory are interconnected concepts that, in synergy, shape the foundations of Industry 4.0. Their implementation represents not only technological advancement but also a strategic shift in the approach to product lifecycle management and production processes, through increased OEE, reduced commissioning time, improved quality, and lower costs of unplanned downtime. Strategically, DT and smart factories enhance supply chain flexibility, resilience, and the personalization of production.

The lack of widespread AAS implementation and ambiguous semantics of process data result in integration costs:

• Data management and security: Legal challenges (data ownership, NDAs), OT/IT cyber-security, and business continuity requirements.
• Model validation and trust: DT reliability requires continuous validation/recalibration; black-box ML applications limit auditability.
• Scalability in SMEs: Competency and investment barriers, as well as the lack of ready-made domain “templates”, hinder adoption.

Despite the numerous benefits, implementing digital twins comes with some challenges:

• Integration complexity: Combining disparate systems and technologies into a single, coherent structure can be difficult and costly.
• Security and privacy: Storing and processing large amounts of data require appropriate security measures against cyber threats [62].
• Data management: Effective data management, quality, and timeliness are crucial to the effectiveness of digital twins.

The future of digital twins involves their integration with artificial intelligence, augmented reality (AR), and cloud computing technologies, which will enable even more advanced analyses and optimization of industrial processes.

3. Materials and Methods

The solutions available today enable the production of non-standard, distinctive, personalized products in single quantities and small batches, tailored to the needs of individual customers—so-called customization. However, this has created greater challenges for manufacturers in properly planning and manufacturing “tailor-made” products. Fulfilling orders and allocating resources to the required task is a complex issue, as it requires coordination of activities for individual products, which vary in variants. This problem is also evident in the implementation of these activities on the production floor, where tasks are performed at the appropriate time using designated resources. At the same time, this manufacturing system is an increasingly complex entity composed of modern technology solutions and Industry 4.0.

When companies execute production orders, it is crucial to properly plan and organize tasks to complete them within the assumed timeframe under specific conditions. These activities are supported by various technologies designed to streamline task execution, such as a digital twin. In this context, this article presents the concept of a hierarchical digital twin. Its goal is to integrate auxiliary simulation programs.

This work uses a multi-method approach, combining the following:

• Production scheduling and resource planning using a proprietary tool (KbRS (ver. 20250524);
• Production flow simulation using simulation software (FlexSim (ver. 2025);
• Mathematical modeling of the production system (description of resources, stations, buffers, AGV trolleys, routes);
• Scenario-based disruption analysis, taking into account various potential problems (missortment, delays, transport conflicts);

Validation of simulation results by comparing metrics (e.g., production completion time, delays, throughput) for various scenarios.

Procedure:

• Collecting input data: Production orders, resource structure, operation times, transport routes—in the KbRS tool;
• Automatic generation of production schedules and data structures;
• Data transfer to the simulation model (FlexSim)—hierarchical integration of tools;
• Building the simulation model (stations, flows, buffers, AGV trolleys, assembly stations);
• Simulations for various scenarios (baseline, with errors, with delays, with process reorganization);
• Result analysis: assessing the impact on production lead time (Cmax), identifying bottlenecks, and assessing the cost-effectiveness of interventions (e.g., purchasing a robot, changing procedures).

This combination of methods—scheduling, simulation, and scenario analysis—constitutes a comprehensive research methodology.

Based on the adopted methodology, the following hierarchical digital twin architecture was proposed (Figure 1):

• Level II—planning and engineering tools (KbRS, CAx/CAD): Generation of production data, resource plans, AGV routes, order structures, BOMs, and variants.
• Level I—simulation tool (FlexSim): Simulation of production, transportation, assembly, and internal logistics flows using data from Level II.
• Integration layer: A data exchange mechanism (e.g., Excel files, databases) between tools, and automating data transfer without repeated manual data entry.

The framework enables the rapid creation of a digital twin of a production plant that is flexible (possible to change orders/configurations), scalable (subsequent batches/orders/variants), realistic (based on actual production data), and capable of scenario analysis and operational decision support. Moreover, the framework enables the simulation of disruptions and the assessment of their impact on production indicators—which is crucial in a variant and just-in-time production environment.

The required production data feeding the integrator will be retrieved/generated from programs such as KbRS, which will be used to develop production process schedules, or from CAx and CAD programs, which are engineering programs. KbRS is proprietary software developed by the Department of Engineering Processes Automation and Integrated Manufacturing Systems, Faculty of Mechanical Engineering, Silesian University of Technology, where the authors work. This data will then be transferred to a first-level tool—a production flow simulation program, e.g., FlexSim, which also reflects a real-world object. Integration between them will be based on data exchange files, such as Excel or other files. Creating a digital twin, despite being extremely time-consuming, is insufficient in the context of simulating the flow and complexity of production processes. To support the creation of a digital twin, a hierarchical digital twin concept was developed through integration with other supporting tools, from which data will be obtained for mapping the complexity of production processes in the virtual world. This concept is illustrated in the article using an example where data regarding resources and task execution were captured in the KbRS program (Level II), while production process flows were simulated in the FlexSim simulation program (Level I).

However, the actual scope of our work described in this article is much broader than the digital shadow (DS) stage. In this article, we address the issue of inter-program communication (data transfer) within the digital twin (DT), and between the FlexSIM software and KbRS (software available at https://kbrs.pl/). Communication with the real object/world and DT is achieved using a Siemens S7 controller (Siemens AG, Munich, Germany).

In accordance with RAMI 4.0 (Figure 2), this article describes only the first two lower layers of this model for data transfer within DT. The subsequent layers, i.e., the communication and information layer, concern related work, and communication between the real object/world and DT is based on communication with a real PLC controller via the Ethernet/IP protocol [63].

Since the digital twin represents a specific production system, a general mathematical model was first developed. The analyzed production system consists of various stations M1–Mj, where various operations are performed. The system also includes AGV trolleys used for transport tasks in the warehouse (loading station) and the production hall to which these trucks are to reach (a set of specific stations/machines).

Therefore, the system (S) can be described as the following components:

S = {PP_i, i = 1…j; M_j, j = 1…o; BP_p, p = 1…s; AG_k, k = 1…n; TJ_m, m = 1…t}

(1)

where

• PP_i—number of production processes;
• BP_p—number of buffers at the stands;
• M_j—number of resources;
• AG_k—number of AGV trolleys;
• TJ_m—number of routes.

and wherein,

St = (iDost, iKos, iPrac, iKal)

(2)

where

• St—stand;
• iDost—resource availability;
• iKos—unit labor cost;
• iPrac—list of employees in resources;
• iKal—working time calendar.

This system, based on specific orders, also requires components that are delivered to the company in a timely manner. A challenge that complicates the management of such a system is the required quantity of semi-finished products of the appropriate quality at a given time, often delivered on a just-in-time (JIT) basis.

Furthermore, the production system contains production orders that define customer expectations, defined by production volume, order completion date, and proposed costs. This information can be defined as customer requirements, which can be recorded as a production order (PO):

WK = ∑ZP_1…i, 1…i—number of orders

(3)

WK = {WS, PP, T, BS, TP, CZ}

(4)

where

• WS—series size (number of elements of the i-th order);
• PP—production process, (PP = 1…j—number of production processes);
• T—execution time of the i-th order;
• BS—number of elements in the production batch of the i-th order;
• TP—the period of introducing production batches of the i-th order;
• CZ—cost of executing of the i-th order.

In connection with the recorded customer requirement model, a general final product purchasing algorithm is also presented below (Figure 3), a fragment of which will be used for further analysis. This will be the stage of component assembly at the production station and its delivery to the specified station.

Before a production order reaches the manufacturer, the dealership collects customer data regarding the selected variants of final model, such as color, dimensions, type, etc. The customer then makes the decision to purchase the final product within a specified timeframe. If the customer decides to purchase the final product, the dealership sends all data to the manufacturer, where the order is processed. The finished product is then shipped to the dealership and sold to the customer. If the customer decides not to purchase after placing the initial order with the dealer, the entire process ends. However, if the customer wishes to modify the order, all the order details are reconciled at this stage, and the order is then sent to the manufacturer.

Focusing on production, the manufacturer creates a list of required parts and semi-finished products, which are delivered from the warehouse to the production lines. This list includes the types of parts needed and their quantities, so that they can be delivered to the appropriate assembly station in the next step. In this case, this list is generated by the system, while warehouse employees are responsible for collecting the required parts according to the generated list. The collected parts are then placed on AGV carts, which transport them to designated production zones/areas.

The implementation of these tasks is complex and requires multiple simultaneous actions to ensure that the components reach the appropriate stations at the specified time, which also translates into achieving the appropriate production metrics. Therefore, based on available technological solutions and the assumptions of Industry 4.0, the article describes the concept of a hierarchical digital twin.

3.1. The Concept of Creating a Hierarchical Digital Twin

For this purpose, an algorithm for creating a traditional digital twin was first developed (Figure 4).

In the first step, the manufacturer must have the appropriate data regarding production orders. If insufficient data is available, the manufacturer waits until all the required data is available. Simultaneously, if all order data is available, detailed production data is generated regarding production flow, employees, operation times, etc. This data is available in digital form and transferred to the appropriate database. Then, based on the obtained data, a digital twin is created in the simulation program. At this stage, if new data regarding processes or times becomes available, the data is updated and the model is modified. If there are no changes in the data, simulation experiments are performed, the best solution is selected for the current production, and the results are transferred to the database. If a new solution emerges, the experiments are repeated, and after approval, the solutions are transferred to the database. Access to transparent data and results allows not only better decision-making but also the operation of the entire enterprise.

Due to the construction of the process simulation model (digital model—DM), this model can be written as a set of elements, namely

DM = {I, SO, C, CT, T, O}

(5)

where

• I—inputs—energy, raw materials, information;
• SO—simulation objects—machines, trolleys AGV, robots, employees, stores, conveyors;
• C—connection between simulation objects;
• CT—constraints—parameters of processes;
• T—time;
• O—outputs, finished products.

With the required data on the production process, the next step was to build a simulation model in FlexSim as a fragment of the production line. A view of the model is shown in Figure 5.

This model was also used to develop the concept of a hierarchical digital twin, described according to several scenarios. These scenarios differ in their operational concepts. The FlexSim software was used to demonstrate the digital twin model, acting as the simulation software, which in this case serves as the master tool for the digital twin. The entire concept, as well as the scenarios created, is based on the use of a scheduling software, KbRS, which in this case is a second-level tool in the hierarchy used to acquire data on the structure of production orders and processes, and the duration of individual operations on resources and their routes, as well as to generate schedules and transmit them to FlexSim. This data will be the basis for the operation of the entire production system as a hierarchical digital twin.

Therefore, the first step is to have at hand the data on production processes, unit times on resources and routes, and the required data that was supplemented, based on which sample schedules were generated (Figure 6).

The KbRS software is used to integrate activities between the two programs, eliminating the need to enter data into separate programs. This can be performed only once in the KbRS program, from which the data will be retrieved by the FlexSim simulation program. The KbRS software accelerates the task planning process and contributes to greater automation in creating a digital twin of the production system. This software allows for visualization of results using graphs, tables, and reports. A view of the KbRS–FlexSim integration concept is shown in Figure 7.

As part of the integration concept, aimed at creating a digital twin of the production system, operational scenarios were developed for a selected station that had previously experienced problems. These scenarios are linked to a section of the production system and are based on previous cases where semi-finished products are assembled. This article contains an experiment to test the operation of the digital twin and the resulting integration of the two program levels.

3.2. Scenario 1

In scenario 1, the problem that arose concerns the AGV trolleys’ access to the station. After loading components in the warehouse, the AGV must reach a designated position (Figure 8).

Because the system employs a collection of several AGV trolleys (also a specific set of machines/robots) to deliver components to specific stations, errors can occur. A truck either does not reach a specific station or does not reach it at all.

3.3. Scenario 2

The second scenario concerns delays with machine M6. This is a significant problem due to the just-in-time (JIT) principle implemented in the company (Figure 9).

This delay is due to problems occurring on the production floor. However, it also impacts the company’s costs. The costs incurred, including those related to delays, will be discussed in subsequent articles.

3.4. Scenario 3

The third scenario involves confusing the order in which the glass is delivered to the station. Due to the required order in which the glass is mounted on the AGV trolleys, errors often occur. It is advisable to mount the glass in the order in which it is mounted at the station (Figure 10).

In this case, the robot handling the glass must have it arranged in a standardized manner, maintaining the order and specifying which side should face up. Due to the use of robots in the station, it is also crucial to properly feed the glass to the robot to prevent damage during assembly.

4. Results and Discussion

The described scenarios differ in the way the processes are carried out. Scenarios 1 and 2 were developed for two-shift operation, but they differ as in Scenario 2, delays occurred at some stations (machines), which delayed the completion of the entire production order. The simulation program shows the production flow between stations and the timeframe for producing fewer finished products, while in the KbRS system, this is reflected in the schedules and data regarding the completion time (Cmax counted in minutes).

Scenario 3 shows the problem of incorrectly sequencing the glass panes into the station, which causes the station to stop working. In this case, this is not visible in either the simulation or KbRS programs, as it is unknown how many such errors there will be or whether they will occur at all within a given time. However, it is possible to introduce such an error in both programs and analyze the system.

In general, the concept of a hierarchical digital twin was applied to the example discussed in this article. On the one hand, a model was built in FlexSim, which reflects a real-world example of production in the automotive industry and shows production flows. KbRS, on the other hand, displays the structures of production orders and processes for which schedules were generated for the discussed example. This data is intended to serve as the basis for data acquisition and generation in the digital twin model (FlexSim). Furthermore, using KbRS, it is possible to obtain schedules for various rules and problems that arise during production, enabling better decision-making and verification of the current situation.

The article also presents an algorithm for creating a digital twin, which supports decision-makers/managers in its creation. Next, using the example of production in the automotive industry, three problem scenarios that occurred in the company were described. The work was carried out in two shifts. The analyses performed allowed us to identify certain problems and implement certain changes:

• Scenario 1—The problem was that the truck either did not reach a specific machine or did not reach it at all. In this scenario, additional interlocks and controls were introduced to minimize the number of trucks that did not reach a specific machine. Signal amplification was also implemented to prevent the truck’s signal from being lost in the hall. The results obtained for the Cmax completion time were 461 min.
• Scenario 2—Machine delays. This problem poses a high risk of not completing production on time. As indicated by the schedule, the completion of production tasks has been postponed and requires urgent action. Thanks to this analysis, we know when production will be completed. In line with the just-in-time principle, the decision was made to purchase a robot that will perform the task of feeding glass to the assembly station. The purchase of the robot will solve the problem of a shortage of the required number of workers. The results regarding the Cmax completion time were 511 min.
• Scenario 3—A problem with the order in which the glass panes were delivered to the station and their proper arrangement. To this end, standardization of work at the station and Poka–Yoke were implemented to eliminate human error during glass loading. The changes introduced at the station improved the glass loading process. In this scenario, the data is exactly as is in Scenario 1, but if the times associated with a station failure due to incorrect glass loading order are added, the Cmax production completion time will increase.

Simulation is a key element in the analysis of various production systems, enabling not only the modeling of production processes, but also the processes beyond their range, e.g., logistics, supply chain, warehousing, etc. However, the problem that arises with this type of solution concerns very detailed models and large amounts of data, which often turn out to be insufficient, especially when there is no consistent connection between the virtual model and the real object, as their communication is disrupted. Moreover, in order to function properly, the virtual model should closely simulate the physical system to become a digital twin. The literature describes various concepts for creating a digital twin, e.g., by creating or using a program that simulates the assembly (production process) of pre-assembled parts on trolleys in a warehouse [37,38,39]. However, many implementations are limited to digital models or digital shadows, failing to provide the full two-way integration and automation necessary for a true production twin [17,18]. Therefore, the traditional approach to DT creation often proves to be insufficient, especially when there is no compatibility between the real and virtual models [41,42].

In the described concept, the authors integrate a professional process simulation application (Level I) with other professional applications used to generate other data necessary for the proper functioning of DT (Level II). Therefore, there is no need to build an individual DT for each analyzed production (technical) system, which requires an appropriate level of IT skills. In the case of the proposed solution, the only IT problem was the integration of various applications using an appropriate data exchange file. Thus, a similar range of necessary applications and skills for their integration is used to build any DT for any production (technical) process. In the case presented in the article, Level II was a process scheduling program (KbRS). Based on data from the simulation program (Level I), the KbRS program generates schedules for source operations. The developed concept of a hierarchical digital twin has the aim of improving the process of introducing production data by transferring data from the KbRS database to the FlexSim simulation program, instead of introducing it sequentially into each program. In addition, the data obtained after simulation can be saved to a database and used for further analysis. Another notable aspect is that the use of DT in production scheduling and dynamic process management (e.g., variant production, just-in-time, made-to-order production) is becoming popular, which means that there is still a lack of standardized methodologies that would enable DT to be adapted to the needs of industrial enterprises, and this is also the next stage in the development of the concept discussed in the article. The use of simulation in DT means that some objects in the virtual model still need to be created from the beginning, as there is a variety of technologies and production organizations in different enterprises and industries, and this is another element for further analysis and research. Furthermore, virtual models do not show process components such as human error, machine failure, or staff shortages and absences, which are difficult to predict but have a significant impact on production processes and are also elements of further analysis. Thus, there is also a concept of integrating different types of software to develop a model that includes emerging problems and tries to solve them.

The aim of this article is to present and test the concept of a hierarchical digital twin—an integrated planning, simulation and scheduling tool—and to demonstrate its applicability in the context of variant production in the manufacturing enterprise. Further research will include developing the described concept of a hierarchical digital twin, further parameterization of the DT model, changes resulting from the need to adapt data from real objects to simulation software and DT creation, as well as attempts to adapt other programs to build the described concept, and the development of common data models and their further development.

5. Conclusions

This article focuses on the use of digital twins (DTs) in the context of industrial production, with particular emphasis on the automotive industry. A digital twin is presented as a virtual representation of a real-world object, system, or process, enabling real-time monitoring, analysis, and optimization. The article describes the DT concept, its standards (RAMI 4.0, ISO 23247), and its applications in smart factories: from predictive maintenance and production flow optimization to safety and energy management. The practical section presents a hierarchical digital twin in which data from production support programs (e.g., KbRS, CAD/CAx) is integrated with the main simulation tool (FlexSim). This approach allows for better representation of the complexity of production processes and improved planning and scheduling of tasks in the production system. The article also describes three problem scenarios on an automotive production line:

• Problems with AGV trolleys’ access to station;
• Delays on selected machines in just-in-time mode;
• Incorrect sequencing of components (glass) in an assembly station.

Analysis of these scenarios demonstrated that integrating data from various levels and tools allows for the identification of problems, streamlining of processes, and making of better decisions. The article emphasizes that DT alone is not sufficient to faithfully represent the complexity of production—its hierarchical integration with other systems and tools is necessary.

The conclusions emphasize that the use of a hierarchical digital twin

• Improves order planning and execution;
• Enables rapid response to disruptions;
• Requires consistent data management and user training;
• Increases production flexibility and efficiency in the context of Industry 4.0.