Using MBSE for the Enhancement of Consistency and Continuity in Modular Product-Service-System Architectures

: Within emerging markets, ensuring the competitiveness of manufacturing companies is crucial to their success. The integration of new business possibilities, such as Product-Service-Systems (PSS) can provide one suitable solution. Especially within the architecture development process, large amounts of interconnected data and data types need to be processed and versioned. This leads to a signiﬁcant lack of data consistency and continuity along the development process of modular PSS architectures. This lack of consistency and continuity leads to a process being prone to errors, representing a signiﬁcant negative impact onto the company’s value-added stream. We provide one possible solution to these issues by presenting a PSS architecture modularization approach based upon the modularization methods of the Integrated PKT-Approach. Using concepts of Model-Based Systems Engineering (MBSE) for modelling these architectures, automated and dynamic analyses of the architecture for the iteration and harmonization of the PSS architecture under development are enabled. The at ﬁrst generically described approach is further detailed in the second part of this contribution by applying it to an industry case study for mobile laser welding systems. As a result, a clear support for the visualization of architecture iteration aspects as well as for the enhancement of data consistency and continuity is given.


Introduction and Motivation
Within nowadays business environments with the growing demands of the common megatrends, such as, e.g., globalization, manufacturing companies are confronted with an increasing market pressure [1]. This often leads to a significant increase of the demand for customer-specific solutions [2,3]. From the perspective of the manufacturing company, this enlarged scope of customer-specific requests can also be described as the external variety, which is subsequently targeted by producing more adapted product variants, leading to a substantially growing variety within the company, the so-called internal variety [2,3]. Common results of this process are an increase in product architecture complexity, diversity and variety, which inevitably come with an increase in complexity cost, manufacturing cost and are prone to errors within the whole product life cycle [4,5]. One proven solution to these issues is presented by the approach of modularization. By implementing modular product architectures (MPA) into the business structure of a manufacturing company, the described perceived external variety towards the customer can be kept at a high level, whereas the internal variety can be significantly decreased [4]. Within the literature, multiple modularization approaches can be found, with each of them aiming at different modularization goals. These approaches all differ significantly in various aspects, but one key aspect most of them have in common is their purely product-oriented view, as they are generally used for the modularization of pure product architectures.
Nevertheless, innovative business models enabling more creative solutions than purely product-based portfolios. Especially so-called Product-Service Systems (PSS), an integral offering of products and services [6][7][8], can be used in order to extend the offer and therefore the external variety of a producing company [9,10]. By integrating service elements in a traditional product architecture, on the one hand additional functions targeted by newly implemented services can be achieved, because the solutions space is extended [11]. On the other hand, by altering the underlying business model, already existing product functions can be changed into functions addressed by integrating the service mindset. This already shows the number of possibilities enabled by enhancing traditional product architectures to PSS architectures. A possible disadvantage is that the service part of PSS architectures is intangible and could be complex to be implemented strategically without increasing the PSS architecture's complexity significantly [10].
This leads directly to the conclusion, that not only product architectures but PSS architectures as well need to be modularized [12,13]. This has already been targeted in partial aspects in the research prior to this contribution by adapting the Integrated PKT-Approach [2,3] to develop a method set for the modularization of PSS architectures [14]. This approach is further described in Section 2 of this contribution, but similar to most of the available modularization methods, almost all methods represent mainly documentcentered approach relying on a large amount of interconnected data, which furthermore need be versioned within the individual steps of the method.
It is an obvious problem with these traditional document-centered approaches that they provide a large potential for data inconsistency across the models [15]-which we define as the vertical view-and a severe lack of data continuity along the modularization timeline-which we define as the horizontal view. Additionally, as PSS architectures are generally not created and supplied by a single company [7,16], but originate from a complex multi-stage network, an efficient information flow within this network is crucial in order to avoid data inconsistencies and maintain the architectures data continuity as well as its maintainability [17].
With these aspects in mind, we formulate the main challenge within the modularization of PSS architectures as a consistent data structuring, enabling a consistent vertical view and an enhanced data continuity along the horizontal view of the modularization process. Additionally, a clear possibility for the data traceability and maintainability needs to be derived, allowing for an efficient and holistic knowledge transfer within the PSS architecture modularization network. The practical relevance of this issue becomes even clearer when large, complex product architectures are considered. For example, when analyzing laser welding systems, the complete architecture consists of more than 100,000 individual components which are all linked together differently in five different product life phases. When using a document-based approach for developing a modular product architecture, it is obvious that with this multitude of individual components, objects and data interconnections, a consistency and continuity problem is inevitable. This needs to be addressed using computational support for the enhancement of these two stated issues.
During this contribution, we address these challenges by developing a model-based approach using the concepts of Model-Based-Systems-Engineering (MBSE) for the development of modular PSS architectures. The main goal is to derive a vertically consistent and horizontally continuous data basis for the interdisciplinary modularization of PSS architectures using the Integrated PKT-Approach. We show a possible implementation and visualize the modular PSS Architecture in different consistent and continuous versions. This allows a proper evaluation of target points for the modularization process itself with a clear support for the change effort estimation when it comes to the alternation of existing architectures.
The presented gap which this contribution intends to close is unique to the pertinent literature. First of all, the conjoint consideration of the data continuity as well as data consistency level within the development of modular PSS architectures by using the concepts of MBSE has not been researched yet, as the following Section 2 shows. Furthermore, using the tools provided by the MBSE approach for the automated and dynamic identification and visualization of product architecture improvement target points contains a significant degree of novelty as we suggest a different use of MBSE tools as opposed to their initially intended purpose.
The contribution begins with a section considering the state of the art relevant for PSS architectures and their modularization approaches as well as the pertinent research concerning MBSE with respect to the above-mentioned issues for modular architecture development. In Section 3, the generic developed MBSE-data model for the modularization of PSS architectures is displayed, followed by a detailed explanation based on a case study using a small laser welding system in Section 4. The results are discussed in Section 5, before the contribution concludes with a summary and an outlook for further research in Section 6.

State of the Art
Within the pertinent literature, there exist several different modularization approaches with different modularization mindsets [18], such as, e.g., product strategic methods such as the Modular Function Deployment (MFD) by [19] or technical functional approaches such as the Design Structure Matrix (DSM) by Pimmler and Eppinger [20]. As a third category, there are as well integrated approaches implementing both of these mindsets, such as the Product Family Master Plan (PFMP) [21] or the Integrated PKT-Approach [2,3], which is used as the methodical basis for this contribution This approach is based on a multi-step method with individual methods for the following main steps as shown in the following Figure 1. of MBSE has not been researched yet, as the following Section 2 shows. Furthermore, using the tools provided by the MBSE approach for the automated and dynamic identification and visualization of product architecture improvement target points contains a significant degree of novelty as we suggest a different use of MBSE tools as opposed to their initially intended purpose. The contribution begins with a section considering the state of the art relevant for PSS architectures and their modularization approaches as well as the pertinent research concerning MBSE with respect to the above-mentioned issues for modular architecture development. In Section 3, the generic developed MBSE-data model for the modularization of PSS architectures is displayed, followed by a detailed explanation based on a case study using a small laser welding system in Section 4. The results are discussed in Section 5, before the contribution concludes with a summary and an outlook for further research in Section 6.

State of the Art
Within the pertinent literature, there exist several different modularization approaches with different modularization mindsets [18], such as, e.g., product strategic methods such as the Modular Function Deployment (MFD) by [19] or technical functional approaches such as the Design Structure Matrix (DSM) by Pimmler and Eppinger [20]. As a third category, there are as well integrated approaches implementing both of these mindsets, such as the Product Family Master Plan (PFMP) [21] or the Integrated PKT-Approach [2,3], which is used as the methodical basis for this contribution This approach is based on a multi-step method with individual methods for the following main steps as shown in the following Figure 1. After having analyzed the status quo, the PSS architecture undergoes the first step described by the Design for Variety [3] according to Kipp and Krause, where the complete architecture is strategically altered on its component level in order to enable the second step, which is the actual modularization itself. Within the Integrated PKT-Approach, the architecture is modularized for each life phase from Product Development to Recycling individually, leading to individual and therefore in some aspects differing modular concepts for each of these life phases [2,22,23]. In the following step, these differences are solved by an interactive discussion of the corresponding interdisciplinary expert team from each life phase. This process is called Life phases Harmonization and leads to a final modular architecture concept. This final concept is considered as valid for each individual life phase and needs therefore to be re-circulated back from the top-level, after the harmonization generalized view, into the sub-level individual life phases view [3]. This brief summary of the Integrated PKT-Approach already displays a large amount of subsequent modularization steps based on a significantly large data basis containing customer requirements, customer relevant properties, functions, working principles, components, modules and the underlying set of dependencies and constraints linking each of these individual data. Furthermore, these data are composed differently within the timeline of the in Figure 1 displayed modularization approach as well as within the individual methods used within the method steps as well, especially when it comes to the Life Phases After having analyzed the status quo, the PSS architecture undergoes the first step described by the Design for Variety [3] according to Kipp and Krause, where the complete architecture is strategically altered on its component level in order to enable the second step, which is the actual modularization itself. Within the Integrated PKT-Approach, the architecture is modularized for each life phase from Product Development to Recycling individually, leading to individual and therefore in some aspects differing modular concepts for each of these life phases [2,22,23]. In the following step, these differences are solved by an interactive discussion of the corresponding interdisciplinary expert team from each life phase. This process is called Life phases Harmonization and leads to a final modular architecture concept. This final concept is considered as valid for each individual life phase and needs therefore to be re-circulated back from the top-level, after the harmonization generalized view, into the sub-level individual life phases view [3]. This brief summary of the Integrated PKT-Approach already displays a large amount of subsequent modularization steps based on a significantly large data basis containing customer requirements, customer relevant properties, functions, working principles, components, modules and the underlying set of dependencies and constraints linking each of these individual data. Furthermore, these data are composed differently within the timeline of the in Figure 1 displayed modularization approach as well as within the individual methods used within the method steps as well, especially when it comes to the Life Phases Modularization [23] according to Blees and Krause. These approaches all are just developed with pure product architectures in mind. As for this contribution, PSS architectures are relevant, the current state of the art considering PSS systems is described.
In the literature, Product-Service Systems (PSS) are described by different definitions see [7,8,14,24]. Even though the definitions have different emphases, the essence of all definitions describes PSS as a combination of tangible products and intangible services that provides value to the customer [10].
By combining product and service components, the solution space to fulfill customer requirements is increased, customer loyalty is strengthened and providers can differentiate themselves from competitors [9,11,16]. However, the combination of product and service components in a PSS increases complexity, which is a major challenge for many companies [10]. In order to master the complexity, the modularization of PSS is referred to in the literature [10,12].
According to Larsen 2018, methods for developing modular PSS consist of four generic method steps: identification of customer requirements, translation of customer requirements into service specifications, service modularization and definition of a configuration model [25]. In this contribution, modularization itself is of particular interest. The basic procedure of the actual modularization can be described on the basis of three steps [3]. Initially, the existing hierarchical (product) structure is divided into components. These represent the smallest unit, which is considered in the following [3]. A product component can be a single part, such as a sensor, or an assembly or a wheel mechanism. Similarly, on the service side, components can be entire service processes or individual service activities.
In the second step of the modularization the identified components are analyzed and in particular the relations of components among each other are examined [3]. In the final step, the components are clustered into modules and transferred into a modular (product) structure [3]. For clustering, the literature distinguishes between technical-functional and product-strategic approaches [18,24].
For the modularization of PSS exist different approaches in the literature [14,25,26], which can be divided into two groups. Methods of the first group modularize products and services at the same time, while methods of the second group focus on services [26]. In both groups, module development mainly results from functional dependencies, which corresponds to a technical-functional approach. Product-strategic module drivers are only addressed in the context of sustainability, but not used [14,26]. However, product-strategic module drivers in particular can open up great advantages [2].
To address this issue, Rennpferdt and Krause [26] propose a method for modularizing PSS that combines technical-functional and product-strategic approaches. The basis for this is the Life Phases Modularization according to Blees and Krause [2,3,23], which has been successfully applied in various industrial projects for the modularization of physical products [20,27]. The special feature of life phases modularization is the module formation with the help of network diagrams. The network diagrams link module drivers, module driver specifications and components with each other and thus enable module formation on the basis of product strategy aspects which originate amongst other from [28], see Figure 2.
For the modularization of PSS, additional module drivers for PSS were added to the module drivers for product modules according to [18,26]. Modularization can result in three different types of modules: pure product modules consisting only of product components, pure service modules consisting only of service components, and mixed modules consisting of product and service components. Rennpferdt and Krause 2021 give recommendations for clustering components into modules [26]. For example, on the product side, modules should be standardized and be as large as possible, since changes to physical products usually require a great amount of time and effort. In contrast, the service-side modular structure should be as fragmented as possible so that the PSS can be configured as flexibly as possible [26]. From these recommendations, it can be deduced that the modular structure should have as few mixed modules as possible, which also reduces general complexity, because fewer coordination efforts are required between departments.
The network diagrams of the individual life phases are combined in the next step of the life phases modularization for PSS to form a so-called Module Process Chart (MPC). This represents the second special feature of the approach. The representation in the MPC makes it possible to compare the modular structures of individual life phases with each other and to identify critical points and resolve conflicts [22,23,26] as shown in Figure 3.    general complexity, because fewer coordination efforts are required between departments.   The life phases in the MPC can be ordered in a chronological sequence for traditional physical products [3]. In the case of PSS, PSS-based business models can result in a PSS not being sold as before, but only leased. For example, this may lead to a PSS being returned from the usage phase to sales and being rented out again. This applies in particular to capital goods, such as machine tools, which usually have a very long service life This example illustrates that the change from a product to a PSS provider also changes the life cycle [9,10]. The MPC in Figure 6 shows the sequence of life phases for the mentioned example of machine tool that moves from the usage phase back to the sales phase.
One challenge in the application of life phases modularization of PSS is to define the optimal modular structure for very extensive PSS families with many components and interactions between the components [26]. At the same time, the complexity itself is a challenge in the modularization of PSS. In addition to the high variety mentioned at the beginning as a complexity driver [10], further complexity drivers can be found in  [29] in their work cite, e.g., the dynamics of the service offering [30], dynamics due to technological progress [31], dynamics in decisionmaking processes [32] and changes in service elements and their interactions [33] as further complexity drivers for PSS. To mitigate the impact of these drivers, the effects of changes must be documented in a way that is understandable to all stakeholders. It is also important to be able to estimate the effects of changes as early as possible.
Both the development of PSS and the development of modular product families, as through the Integrated PKT-Approach, thus require a high degree of consistency and continuity in order to be applied effectively. However, since both areas are often documentbased, inconsistencies of data can occur. Inconsistencies are often a source of contradictions in the models, which are caused by a lack of consistency management as well as knowledge that is not explicitly documented and can lead to errors [34,35].
For example, components are used in different methodological tools, as shown above in MPC and network. However, these are not linked to each other in terms of data, so that a change in one element does not lead to a change in the other element. The effects of different drivers for PSS cannot be identified and analyzed by static product development [36].
In addition, versioning is only partially present in the various tools, but is not consistent. In particular, the dynamics required for the development of PSS within the methodical development are not met with a document-based product development. The iteration in the life phases of use and development within a PSS business process described above requires a consistent data transfer that also takes temporal change into account [37,38].
In order to increase both consistency and continuity, software support by means of Model-Based Systems Engineering (MBSE) lends itself. With the MBSE approach, system elements can be modeled and information can be linked so that information can be stored consistently and used in a networked model. The graphical modeling language SysML was developed for MBSE. The modeling software used, Cameo Systems Modeler (CSM), uses diagrams and tables in addition to SysML notation. It is based on nine diagrams, five of which are used to create the structure to enable consistency and four of which are used to create the behavior to enable consistency [39]. Figure 4 shows the two points of horizontal (continuity) and vertical (consistency) view that are relevant for the development of modular product families in the context of the technical implementation of PSS.  Different tools are used for the four relevant steps. Within each step, there are overlaps of data information between the tools, such as the components that are present in the Variety Allocation Model and the Module Interface Graph. Therefore, consistency across different models is desired. Over the time steps the information is changed, so data continuity is also desired, which is represented by the horizontal view. Parts of this continuity can also be found within the individual process steps, as the individual methods are set up step-by-step on each other and contain therefore continuity-critical aspects. Both can be supported by MBSE as shown in the following section.

Generic MBSE Data Modelling Life Phases Modularization for PSS
During this section, we display the generic data structuring model for the enhancement of the vertical view and the horizontal view as described during Figure 5. As a general baseline, the during Section 2 introduced PSS-Network Diagram with the four columns for module drivers, module driver specifications, components and modules is used. As this PSS Network Diagram (compare Figure 2) needs to be archived in individual versions within the modularization process, with the first versions originating from the Individual Life phases Modularization before the Life phases Harmonization process and then Different tools are used for the four relevant steps. Within each step, there are overlaps of data information between the tools, such as the components that are present in the Variety Allocation Model and the Module Interface Graph. Therefore, consistency across different models is desired. Over the time steps the information is changed, so data continuity is also desired, which is represented by the horizontal view. Parts of this continuity can also be found within the individual process steps, as the individual methods are set up step-by-step on each other and contain therefore continuity-critical aspects. Both can be supported by MBSE as shown in the following section.

Generic MBSE Data Modelling Life Phases Modularization for PSS
During this section, we display the generic data structuring model for the enhancement of the vertical view and the horizontal view as described during Figure 5. As a general baseline, the during Section 2 introduced PSS-Network Diagram with the four columns for module drivers, module driver specifications, components and modules is used. As this PSS Network Diagram (compare Figure 2) needs to be archived in individual versions within the modularization process, with the first versions originating from the Individual Life phases Modularization before the Life phases Harmonization process and then again as versions after the Re-Circulation as a result of the overall Harmonization, the generic data model needs to allow for a traceable and maintainable implementation of these individual PSS Network diagram versions.  It can be seen that the modelling of the PSS Network Diagram is performed by implementing individual instances of Block-Definition Diagrams (BDD) in CSM [40]. The individual objects in the four columns are assigned with individual stereotypes, attributing a machine-readable format of the modular PSS architecture and its base parts. In this case, the module drivers are assigned with a Life-Phase specific stereotype, such as  In the center of Figure 5, the Containment tree of the modelled architecture in CSM is displayed, which serves as the data backbone for the enhancement of the vertical continuity as well as for the horizontal consistency. The originally within the document-based approach in PowerPoint not interconnected data items, such as components and modules can be dynamically and backtraceable connected by implementing them into the individual parts of the containment tree and adding them from this origin into different models.
When applying changes to this central root data core, these changes become consistently and continuously integrated into all existing models. The following Figure 5 displays this modelling trace by color-coding the individual objects. For example, the document-based components symbolized in green become integrated into the root data core and from there into the Network Diagram. These classifications are implemented using different CMS-stereotypes.
It can be seen that the modelling of the PSS Network Diagram is performed by implementing individual instances of Block-Definition Diagrams (BDD) in CSM [40]. The individual objects in the four columns are assigned with individual stereotypes, attributing a machine-readable format of the modular PSS architecture and its base parts. In this case, the module drivers are assigned with a Life-Phase specific stereotype, such as PD_module_driver_2, defining this module driver as part of the life phase Product Development. This process is carried out for both the module driver specifications as well as the modules. Only the components are not assigned as part of an individual life phase, as they are the same in the entire modularization process. In this case, only a differentiation between service components and product components is made, which is also assigned to the component blocks via specific stereotypes.
One important fact to state is that the individual modules are not classified via a stereotype whether they are pure product or service modules or even mixed modules. This enables a dynamic modification ability and thus the derivation of the exact PSS architecture.
This determination of the module type and their classification into one of the three categories product-, service-or mixed module is crucial for the harmonization process, as one major objective of the harmonization is the reduction of mixed modules, see [26]. For the realization of such mixed modules, an interdisciplinary team being composed of experts from both the product and the service department needs to collaborate, which is possible, but still time-consuming and therefore cost-intensive as well as prone to errors. This leads to the conclusion, that as far as possible such modules should be identified in order to find a solution for an architecture re-arrangement.
This identification is one of the key parts we intend to present during this contribution, as it allows for a dynamic derivation of the module category based on the component composition of each module. For the implementation in CSM, we developed a specific Validation Suite package, containing three OCL-programming based constraints allocated to the module stereotypes. One majorly important condition necessary for the Validation Suite to function is the set-up of the BDD with Allocation-Links. Based on these links, the Validation Suite code checks for each individual object classified by the stereotype Module which individual components are allocated to this module. The classification contains three different possibilities: For this implementation, we used OCL-Code combined in different variants for the three individual cases. The following code derives the mixed module classification: not self.encoded5FrelationshipOfRelatedElement -> exists (a|a.relatedElement -> exists(t|t.oclIsTypeOf(Z_Test::Stereotypen::product_component))) or not self.encoded5FrelationshipOfRelatedElement -> exists (a|a.relatedElement -> exists(t|t.oclIsTypeOf(Z_Test::Stereotypen::service_component))) Depending on which case is relevant for the analyzed module, CSM marks the module as service, product or mixed module and colors them differently. As CSM does not allow for a completely free programming, we used different alert levels integrated within the Validation Suite page of CSM. In this case, Info, Alert and Error are used to mark the individual module categories in blue, yellow and red. The results are displayed in the PSS Network Diagram and can also be exported as an .csv data sheet for further external use.
For small data structures as the one presented during this section, the display of the categorization is sufficient in the PSS Network Diagram. With larger structures, in order to keep the visual support and the traceability of information as high as possible, we implemented a specific Dependency Matrix (DM), displaying all relevant objects, their dependencies as well as their Validation Suite result, as the following Figure 6 displays. The above displayed Figure 6 shows the components allocated towards individual modules, module driver specifications and module drivers. The original allocations which have been set during the modularization process are displayed in blue, the orange arrows symbolize implicit connections, which are automatically derived through the data model by CSM. This is enabled by the classifier of all specifying elements as a rule definition of The above displayed Figure 6 shows the components allocated towards individual modules, module driver specifications and module drivers. The original allocations which have been set during the modularization process are displayed in blue, the orange arrows symbolize implicit connections, which are automatically derived through the data model by CSM. This is enabled by the classifier of all specifying elements as a rule definition of the DM.
Another relevant fact to be stated is the positioning of the individual life phase modularization results next to each other, compare Figure 6 (Box 1). The definition of the life phases is marked by, e.g., the module stereotypes beginning with abbreviations for the life phase such as PD for Product Development. This set-up of the life phases next to each other allows for a fast and traceable support for the conflict analysis when it comes to the harmonization process. Figure 6 displays the marked modules as well as their composition and origination with respect to the three different categories, compare Figure 6 (Box 2). The mixed modules, which are considered as the least desirables are marked in red. During the harmonization process, these red modules are to be targeted as first in order to reduce the complexity resulting from the interdisciplinary work. This leads to a direct derivation of the critical modules.
With this visualization, we are able to mark and display the relevant issues within the initially modularized PSS architecture, which is to first step to the improvement and iteration of the final modularization result. In this case, we understand the term iteration as follows: As we cannot determine whether an optimal solution has been achieved in the underlying multi-dimensional framework of market-, customer-and company parameters, one can only support the iterative improvement of the PSS architecture. Based upon lessonslearned as well as known effects and impacts from the applied methodical approach, a continuous improvement process can be integrated based on the existing architecture. As this improvement process is to be considered as a recursive process, the iteration nature of the process is described.
Furthermore, the traceability is enhanced, as the module drivers and their specifications can be backtracked dynamically from the individual product or service module. Therefore, when analyzing several alternatives, experience-based knowledge about critical module drivers can be generated.

Case Study of an MBSE Supported PSS Architecture Modularization for Laser Processing Systems
In this section, we show the application of the in Section 3 generically described approach for the MBSE-based modularization for PSS architectures. As a descriptive example, we apply the systematic to the case study of a mobile laser welding system. This laser welding system can be sent out to different customers with varying extents of service modules. At the customer's site, the machine system can be used for production and will be returned to the provider after the finished production. The following Figure 7 displays the Module Interface Graph (MIG) of the laser welding system.
A MIG is used as a visualization method according to [2] and displays the product similar to a technical drawing with the components' rough size and position. The color coding shows the degree of variety within the components, where grey symbolizes variant components and white components represent standard components. The differently colored arrows between individual components show different flows, such as, e.g., electrical or information flow [41]. Additionally, to the MIG we visualized the service components, as the classical MIG is generally only used for product components. These service components are listed on the right of Figure 7. The service components are shown in a Service-Blueprintbased visualization. A Service Blueprint contains the activities that are needed to offer a service [42] and is used in this example to show which service components are visible for the customer and which are not.
In this section, we show the application of the in Section 3 generically described ap-proach for the MBSE-based modularization for PSS architectures. As a descriptive example, we apply the systematic to the case study of a mobile laser welding system. This laser welding system can be sent out to different customers with varying extents of service modules. At the customer's site, the machine system can be used for production and will be returned to the provider after the finished production. The following Figure 7 displays the Module Interface Graph (MIG) of the laser welding system.  The PSS example consists of a three-axis laser welding system with a beam source, an optical fiber and an optic module. The three axes, which are two linear and one rotational axis are numerically controlled. The complete system is moveable on a wheel basis carrying the whole system. For this contribution the mainly interesting parts are the service components. The Operator's Training is considered as variant, as it is offered with varying extents and has a close interconnection with the Control component. This is due to the reason, that the more possibilities the HMI offers for process adjustments, the more training the operator needs. The Process support is considered as variant, as a variable package of already known process parameters can be added to the product itself.
Alternatively, the largest extent can be a process support technician at the customer's suite. The third service component, Logistics, is a standard component, as the machine is delivered, set-up and returned to the owner after the finished production process. The General Support component describes a varying extent of customer support, with expressions ranging from phone support to a service technician at place. The last component describes the machine maintenance, which is especially relevant for long production cycles at the customer, as wear and spear part replacement can be crucial for a manufacturing company. Based on this product and service portfolio, the MBSE-based modularization for PSS architectures has been carried out. In the following, we display the result of two individual life phase modularizations for the life phases Product Development and the Sales with respect to their implementation and the application of the developed Validation Suite for the Harmonization support.
The following Figure 8 displays the modelled PSS Network Diagram for the laser welding system with the two before mentioned life phases Product Development and Sales. Systems 2021, 9, x FOR PEER REVIEW 13 of 18 For visualization reasons, the Network Diagram is displayed as a Dependency Matrix (DM). Additionally, the Validation Suite Package for the module classification has already been applied, displaying the three different module types in blue, yellow and red. As this contribution is furthermore about the vertical consistency and horizontal continuity support for the modularization process, we display two specific critical aspects within the in Figure 8 displayed unharmonized modularization.
Within the left part of Figure 8, an excerpt of the modelled PSS architecture DM can be seen. The left column displays all relevant modules from both the Product Development as well as the sales life phase, which can be seen by the prefix PD_ or Sales_. The rows display the implemented components, where only three of all modelled components are displayed. The color coding displays the automated result of the Validation Suite Package, which categorizes the modules. The blue markings represent pure service modules, whereas the yellow markings display pure product modules. The main target points are symbolized by the red parts, as they represent mixed modules.
The left part of Figure 8 furthermore shows at one glance relevant addressing points for the harmonization process. As described above, the same component may not be part of different modules in different life phases. In the case displayed in Figure 8, the Fiber component, which is the most right of the three displayed components is part of the Product Development specific module Laser, which furthermore is marked as a red mixed module. The Fiber component is also part of the Sales module Standard System, which represents a conflict and needs to be addressed within the harmonization process. The same issue comes with the Control component, which also occurs in not matching modules amongst individual life phases.
The right part of Figure 8 mainly displays the module category distribution and the clustering of the underlying component stereotypes. It can be seen that the lowest row shows a properly designed service module with no interdisciplinary interference. On the other hand, the laser module represents a significantly negative impact onto the PSS architecture, as it consists of a large number of critical components leading to a mixed module composition. Therefore, this module can be considered as one major target point for the harmonization and iteration process. As these two steps imply a version generation of For visualization reasons, the Network Diagram is displayed as a Dependency Matrix (DM). Additionally, the Validation Suite Package for the module classification has already been applied, displaying the three different module types in blue, yellow and red. As this contribution is furthermore about the vertical consistency and horizontal continuity support for the modularization process, we display two specific critical aspects within the in Figure 8 displayed unharmonized modularization.
Within the left part of Figure 8, an excerpt of the modelled PSS architecture DM can be seen. The left column displays all relevant modules from both the Product Development as well as the sales life phase, which can be seen by the prefix PD_ or Sales_. The rows display the implemented components, where only three of all modelled components are displayed. The color coding displays the automated result of the Validation Suite Package, which categorizes the modules. The blue markings represent pure service modules, whereas the yellow markings display pure product modules. The main target points are symbolized by the red parts, as they represent mixed modules.
The left part of Figure 8 furthermore shows at one glance relevant addressing points for the harmonization process. As described above, the same component may not be part of different modules in different life phases. In the case displayed in Figure 8, the Fiber component, which is the most right of the three displayed components is part of the Product Development specific module Laser, which furthermore is marked as a red mixed module. The Fiber component is also part of the Sales module Standard System, which represents a conflict and needs to be addressed within the harmonization process. The same issue comes with the Control component, which also occurs in not matching modules amongst individual life phases.
The right part of Figure 8 mainly displays the module category distribution and the clustering of the underlying component stereotypes. It can be seen that the lowest row shows a properly designed service module with no interdisciplinary interference. On the other hand, the laser module represents a significantly negative impact onto the PSS architecture, as it consists of a large number of critical components leading to a mixed module composition. Therefore, this module can be considered as one major target point for the harmonization and iteration process. As these two steps imply a version generation of the DM and the underlying individual life phases Network Diagrams, the use of MBSE provides a significant improvement for increasing the consistency and continuity improvement for versioning tasks.
As within Figure 8 the laser module has been identified as a critical module both for the harmonization as well as the module category iteration process, it is used in the following for the explanation of the version generation and for the resulting improvements. Figure 9 shows three of the initially described steps for the Life Phases Modularization from the Integrated PKT-Approach. The upper box of Figure 9 contains the model for the original unharmonized version, which is called Version n. As the displayed laser module is defined by two individual module drivers implying both service and product components, the laser module is categorized by the Validation Suite Package as a mixed module, leading to a red box marking of the laser module box. The upper part of Figure 9 displays for explanatory reasons only the Network Diagram of the Product Development life phase. As this red marking is an alert flag for the targeting during the iteration and harmonizing process, the second displayed step is to create a new Network Diagram version containing the improved PSS architecture. During this iteration step, the displayed laser module could be split into two individual modules. With these changes in mind, the original version n is altered and re-circulated into the new version n+1, which describes the third step of the Life Phases Modularization. the DM and the underlying individual life phases Network Diagrams, the use of MBSE provides a significant improvement for increasing the consistency and continuity improvement for versioning tasks. As within Figure 8 the laser module has been identified as a critical module both for the harmonization as well as the module category iteration process, it is used in the following for the explanation of the version generation and for the resulting improvements. Figure 9 shows three of the initially described steps for the Life Phases Modularization from the Integrated PKT-Approach. The upper box of Figure 9 contains the model for the original unharmonized version, which is called Version n. As the displayed laser module is defined by two individual module drivers implying both service and product components, the laser module is categorized by the Validation Suite Package as a mixed module, leading to a red box marking of the laser module box. The upper part of Figure 9 displays for explanatory reasons only the Network Diagram of the Product Development life phase. As this red marking is an alert flag for the targeting during the iteration and harmonizing process, the second displayed step is to create a new Network Diagram version containing the improved PSS architecture. During this iteration step, the displayed laser module could be split into two individual modules. With these changes in mind, the original version n is altered and re-circulated into the new version n+1, which describes the third step of the Life Phases Modularization.  The results are displayed in the lower part of Figure 9. It can be seen, that the formerly as red marked mixed module is now split into two individual modules, with one blue market service module containing only service components and one yellow marked product module, containing only product components.

Discussion
The general issue to be targeted by the presented approach can be split up into two main aspects.
First, the data consistency and continuity for the development of modular PSS architectures is to be enhanced on order to increase the data backtraceability, maintainability and versioning. By modelling each individual component, dependency and constraint of the individual modular PSS architecture versions within an interconnected meta-model, changes to the central root data core affect automatically all dependent models. This is crucial for the data consistency within the individual development process steps, as it reduces the error rate of missed connections or dependent constraints as far as possible. Additionally, when it comes to process steps building up on each other on a time-wise scale, this data modelling also enhances the horizontal view, which we described as the data continuity. This enables the clear identification of all changes made to a specific timestep in the development process. Furthermore, it dynamically addresses all relevant changes in the root data core to the individual continuous models. This significantly reduces the effort to derive in a later analysis phase, which changes have been made, and more specifically, why these changes have been made.
Secondly, the implementation in the MBSE environment enables several further dynamic analyses considering the generic set-up of the modular PSS architecture. By implementing the modular PSS architecture in an interconnected meta-model, a computerreadable and processable form of the architecture is available. This enables an in-the-loop check for data inconsistencies as well as for the presented differences between pure service modules, pure product modules and mixed modules. As specifically these mixed modules serve as a target point for the architecture iteration in order to increase its performance, the automated and complete identification of these target points is crucial. Especially for large and complex product architectures with several thousand components and interconnections, this dynamic and automated approach is highly relevant for the applying industry.
Through the Case Study, an added value for companies could be proven. Interviews confirmed that the additional, objective information supports and accelerates the analysis and the decision-making process. Especially the traceability of the modularization is a great advantage for the company, as decisions are documented in a way that is comprehensible for everyone. For the presented example of the laser welding systems, an error reduction rate due to this dynamic traceability of more than 80% can be stated.
The limitations to this research are, that it does not provide a solution for deriving an optimal solution, as it just detects issue points within the existing architecture as a basis for a further iteration step in order to increase the architecture's performance.

Conclusions and Outlook
For the background of the development of modular PSS architectures, especially for architectures containing several thousand components and dependencies, keeping data consistency and continuity at the highest possible level is of major importance. This directly leads to the research objective, which is the modelling of modular PSS architectures in an MBSE environment in order to allow for its dynamic and automated analysis and derivation of architecture iteration aspects. All in all, we consider the application of MBSE for the presented development process of modular PSS architectures with methodical approaches such as the Integrated PKT-Approach as a major benefit to this whole research area.
The presented approach significantly enhances the modeled architecture's documentation and therefore, its maintainability and backtraceability by applying the concepts of MBSE. The increased transparency and dynamic analysis of the modular PSS architecture in order to support the identification of critical aspects for the iteration and harmonization of the architecture is one major improvement this contribution supplies.
The main research issue is concentrated towards the two aspects of the vertical data consistency perspective and the horizontal data continuity issue. The presented approach can significantly improve both perspectives by relying on a dynamically adjustable, consistent and continuous data meta model. The data continuity is enhanced by the visualization of the modular architecture within the Dependency matrix in combination with the applied Validation Suite Package.
Additionally, this leads to a continuously backtraceable version evolution, simplifying the maintenance and adjustment of the existing modular PSS architectures.
As the concepts of MBSE not only allow for a pure data documentation and analysis but also specific visualizations, we visualize specific improvement and iteration target points. This allows for a clear, dynamic and holistic visualization of the improvement potentials, which can then be used for an iteration of the modular PSS architecture.
As an outlook to this contribution which supplies an identification and visualization of improvement potentials within modular PSS architectures, the next step could be to automatically and dynamically assess the performance of different alternatives of modular PSS architectures originating from these improvement potentials. For pure product architectures, such approaches already exist as presented in [36,43]. This performance simulation approach brings both customer-and company perspectives together in order to create one individual multi-dimensional performance criterion for each available alternative as described in [44].
With consideration of the during this contribution presented approach for the modelling of PSS architectures, this simulation approach can be extended to analyze the performance of PSS architectures, especially as the underlying data meta-model is consistent with the meta-model used during this contribution.