A Design Approach Of Mechanical Assemblies Based On MBSE And CAD Models Interoperability

The technological development of the last decades have been able to push human to develop their needs, so a way to new demands were opened and this can lead to a complexity problem. Thereby, a good interoperability between the product design activities can lead to the possibility of ensuring a promising satisfaction to all requirements. However, the major problem is the enormous discontinuity between them. Indeed, each one treats the product from its point of view without recourse to the requirements dened by others. This paper is interested in the collaborative work that brings together the system engineer, who deals with the system from a global view, and the designer, who is a specialist in the detailed design, in order to validate requirements. A new methodology has been proposed to dene the role of each one in the design process. This methodology focuses on the product development cycle from the analysis of needs to the validation phase. This obviously requires interoperability between the two domains of Model Based System Engineering (MBSE) and Computer Aided Design (CAD). Based on a pedal bicycle case study which is an industrial mechatronic product, the proposed methodology will be illustrated for validation and highlighting its advantages and limitations.

other actors, which generates an heterogeneity problem [5]. The heterogeneity problem between MBSE and CAD models has led to the development of research works. In the following section, a literature review on CAD and MBSE approaches and the interoperability between them will be presented.

Literature review on CAD approaches
To face the complexity problem of mechanical systems, several research works were developed. This will undoubtedly help to have simpler products not only in their understanding but also in their handling. For example, to deal with the problem of data transfer between physically distant designers, Shyamsundar et al. developed a tool to exchange assembly models facilitating data modi cations [6]. The transfer of data between the different actors is interesting because it will help to analyze the product properly by ensuring the right compromise between cost, quality and time. Other researchers addressed the problem of determining the assembly steps in order to reduce assembly time. Melckenbeeck et al. developed a new approach to calculate feasibility scores for each assembly step in order to deduct the optimal Assembly or Disassembly Plan (AP/DP) for the nal product [7]. In CAD domain, researchers also attach a great importance to assembly analysis because it has a great effect on the validation of product requirements.
For example, the automatic estimation of the manufacturing costs of a part from the early stages of design is a trend that is beginning to integrate CAD software. This application is interesting because it makes it possible to optimize and reduce the manufacturing costs even before moving to the production phase. For certain CAD software, this approach has been recently integrated but still underdeveloped. It only applies to mechanical parts made from sheet metal or machined mechanical parts [8]. In the same context, Ghali et al. developed a new methodology for the allocation of tolerances. The objective is to integrate tolerance allocation into CAD models taking into account the functional requirements of the product which can facilitate its analyzing. In the proposed approach, the Failure Modes, Effects and Criticality Analysis method (FMECA) is used to allocate tolerances based on the manufacturing di culties [9].
Besides the complexity of the 3D CAD design, there is also the complexity related to the generation of AP/DP which can also cause other di culties in the understanding of designers work by other disciplines. From a CAD model, the AP/DP of a mechanism is determined in order to save time, validate or approve the design solution. More recent works were developed in this theme and has focused on the implementation of AP integrated into CAD systems. Ben Hadj et al. were interested in the implementation of a method to generate APs integrated to Solidworks© software as a Plug-in. In this work, the authors began by extracting data from the CAD model, and then, collision tests were carried out to automatically generate feasible APs. The developed method were tested on several industrial products and has shown its e ciency in generating feasible APs [10]. In the same context, Belhadj et al. developed a new approach based on the elimination of connection elements to have simpler APs. Then for each part a score is calculated in order to know which part should be assembled rst. Finally a case-based reasoning algorithm is used for the insertion of all parts in the global AP [11]. Dini et al. developed an approach to nd subassemblies and APs from a mathematical formulation of a mechanical product. The proposed approach is based on three matrices (interference, contact and connection) [12]. For heavy machines and in order to better analyze the operator safety and the mounting tools workspaces, Bedeoui et al. proposed an approach to generate the most optimal APs based on the stability of components during assembly operations [13].
As for the AP generation, other researchers were interested in the DP generation to facilitate the complexity of mechanical systems. Li et al. proposed an integrated approach to generate disassembly constraints by directly using information from the CAD model which allows designers to assess the current design for its disassembly properties [14]. In the same context, Belhadj et al. were interested in the determination of sub-assemblies, especially for mechanisms with a high number of parts. The proposed approach automatically identi es subassemblies from a geometric model of a mechanical assembly and determines feasible DPs [15,16]. The authors developed a new method for parallel disassembly in order to improve the design quality of the product by taking into account a set of evaluation parameters such as the recyclability rate and the total disassembly time of the product [17]. Morato et al. proposed an AP approach by generating the potential layers of disassembly. This approach is based on motion planning and spatial clustering to generate sets of parts that have mobility between them [18].

Literature review on MBSE approaches
According to the literature on MBSE approaches, new research works have investigated the development of this domain to facilitate the conceptual phase of complex systems. In order to facilitate the understanding of highly complex systems, Mhenni et al. developed a SysML methodology dedicated to mechatronic systems which is divided into two parts: the rst deals with the system from an external point of view by providing all the requirements to be satis ed, and the other illustrates in a detailed way the internal architecture of the system [19]. As part of the MBSE development and especially to improve languages used by system engineers, Noten et al. presented some experiences and limitations encountered when applying and customizing SysML. An industrial use case is designed by a multidisciplinary team through several companies [20]. These developments can lead to the knowledge expansion which offers the opportunity to ful ll the steps of veri cation and validation of the product requirements. In the same topic, Gardan et al. provided conceptual models that improve the MBSE to take into account Knowledge Management (KM). Using existing SysML diagrams, the proposed approach formulates a decision model that complements existing methods by offering KM solutions focused on organizational approaches. The ful llment of requirements relies on this conceptual approach to aid in the decision-making process [21]. Considering its important role, several research works were based on the application of system engineering to obtain a better quality of products. Based on the application of MBSE using MADe and MagicDraw Modeling Tools, Odita et al. veri ed the reliability, risk and safety of aerospace components [22]. Other researchers use pro les to model speci c aspects which aim to enrich SysML with external information and it can give the opportunity to model external problems in the MBSE environment. Pop et al. proposed a Uni ed Modeling Language (UML) pro le for Modelica language [23]. Moreover, Paredis et al. suggested another pro le that connect SysML and Modelica [24]. A new pro le named "SafeSysE" was proposed to facilitate the extension of system models with some safety properties [25]. Moreover, Brahmi et al. developed a SysML pro le for the description and the analysis of mechanical assemblies which can be used in the MBSE environment in order to describe the designer's work. The proposed pro le can enhance the continuity between the data generated by the two actors by checking and validating the design solution [2].

Literature review on interoperability and collaborative works
In recent years, researchers have been able to feel the real need for communication between the different disciplines in order to ensure all the requirements. They have been able to know that to ensure this exchange and interoperability of data, standard means of communication are necessary. For this reason, researchers worked on extracting and collecting data based on CAD models of assemblies in order to have information related to the mechanical parts allowing its reuse in other elds. Gouta et al. proposed a tool for extracting and collecting data from a CAD model of an assembly. These data are saved in Excel or Matlab format. The developed tool, called Computer Aided Design Laboratory (CADLAB), constitutes a link between the CAD system and Computer Aided Engineering (CAE) applications [26]. Kleiner et al. combined the functional, logical and physical requirements in a single tool integrated in CATIA-V6©. This approach allows signi cant collaboration between different engineering disciplines [27]. Other studies were carried out to highlight the importance of integrating different elds. Mauborgne et al. developed an MBSE approach to allow the evolution of the preliminary risk analysis method in order to meet the requirements for the automotive sector [28]. The problem of interoperability affects several elds such as the conceptual modeling. In this context, Tolk and Muguira presented a general model dealing with different levels of conceptual interoperability [29]. Zeng discussed a new approach that enables to achieve interoperability, as demonstrated by standards and best practices, projects and products in the broad area of knowledge organization [30].

Outcomes of the literature review
According to the literature review discussed previously, it can be noted that: Several research works are concerned in the development of CAD domain, 3D design and the implementation of computer tools to reduce the complexity of mechanical assemblies. Researchers in the system engineering environment developed new methodologies in order to facilitate the use of system engineering tools for the bene t of other domains.
The two elds which are MBSE and CAD evolve but each one apart.
A problem of discontinuity of communication between the two domains appears which favors the risk of nonconformities of the product design with the requirements of the customer.
In this paper, the proposed approach aims to solve the problem of collaboration between the system engineer who has a global view on the product and the designer who is a CAD specialist. The objective is to ensure the satisfaction of all the product requirements and to analyze and validate the proposed design solution. The proposed methodology is detailed and validated using an illustrative industrial example which is a pedal bicycle.
To achieve all the user needs, a compromise between the system engineer and the designer is required. Figure 1 presents the proposed methodology which illustrates the collaborative work between them. The outline of the proposed approach is decomposed into three main parts as follows: An MBSE preliminary analysis: in this step, a global study of the product from a functional point of view is performed by the system engineer.
CAD development: in this step, the designer proposes and analyzes the 3D CAD design of the proposed solution.
MBSE enrichment and validation: in this step a checking if the proposed design solution is satisfactory is done by testing the concordance of the collaborative CAD-MBSE work to the customer requirements.
For the validation of the collaborative methodology, an illustrative industrial example designed and manufactured by the company Pedalite. The treated example is a luminous pedal named KPL200 and presented in Fig. 2. The principle is simple, e cient and innovative: while pedaling, a small generator integrated into each pedal produces current and powers the ashing lights. The produced excess energy is stored and will be restored when the cyclist is not pedaling (downhill, at a stop, etc.). Battery life is around 5 minutes. The KPL200 pedal increases the visibility of cyclists and is with free maintenance: no batteries, no wires, no moving parts and no risk of breakdown. They are therefore particularly reassuring for parents who know that their children remain visible, even if their dynamo light fails. In this paper, the design process of the KPL200 pedal is detailed using the proposed approach.

The MBSE preliminary analysis step i. Need analysis
With the technological development, customer requirements still increase. Indeed, in order to develop competitive products, understanding the entire customer's needs is a decisive matter for the rest of engineers work. So before embarking on product design which can lead to sometimes expensive and unwise solutions, it is necessary to dissect and analyze all of these needs. This is the stage where the system engineer is invited to study needs and to translate them into algorithms or in textual format that enables to continuously improve the understanding of the system by other actors. This is the rst step that opens the chance to decipher all the necessities. For the treated example, the primary need from the pedal is to improve road safety by increasing the visibility of cyclists at night even under di cult weather conditions.
ii. Functional analysis After analyzing customer needs, the second fundamental step is the de nition of the functional requirements. These requirements are the result of a customer request that comes forward with imperfect information. He tries to explain his needs to the system engineer which is responsible for dissecting them in a speci cation and describing them in forms of actions and conditions to be satis ed by the nal product. This description need to be complete, consistent and have a very clear structure in order to be understandable during the validation phase. This step is one of the most di cult one since it brings together the different actions to be satis ed, especially since the problem is not well formalized yet. For this, the system engineer uses an interface dedicated to de ning requirements (SysML requirement diagram) to summarize the in-depth of his work. Each requirement block contains a unique name, a persistent identi er, and a text that brie y and generally describes the purpose of the requirement. It is mentioned that during this step there is a part of the system requirements which is devoted to the analysis of the product design. For the treated example, the functional requirement diagram is presented in Fig. 3. It details the requirements related to the transmission of the mechanical energy supplied by the cyclist to the crank-set of the bicycle.
iii. Initial assembly structure The system engineer plays the role of an orchestra conductor ensuring multidisciplinary work between several jobs. Considering that he is not a specialist in a particular eld, including mechanics, and that he only has a global vision of the system to be designed, the design of the mechanical assembly is allocated to a specialist in CAD. Therefore, during this step and after studying the system from a functional point of view, the system engineer can propose the initial structure of the assembly. This structure basically contains the parts which are essential in the functioning of the system and it can be modeled in a block de nition diagram "bdd" from SysML. At this stage this diagram is used to describe the external architecture of the mechanism in a rough way. The building blocks of the diagram contain the name, attributes of the components, and are linked together by composition links. As represented in Fig. 4, the initial structure of the treated example can be composed of ve essential parts which are: a pedal body and a pedal axis used to transmit the mechanical energy supplied by the cyclist during pedaling; a double intermediate wheel which multiply axis rotation; a generator for converting kinetic energy into electrical energy to have light ashes and an integrated circuit used to store the surplus of energy produced.

iv. Allocation and traceability
The traceability step has a major importance in improving the quality of the assembly. Through this step, the system engineer is invited to allocate each block of the assembly structure to the requirement to meet. The modi cations that will be executed on any of the requirements will automatically in uence the assembly blocks. Thus, the link between requirements and assembly components will be forged at this stage. This step is used to ensure the consistency of the solutions and to check not only that the proposed solution is good but that it is also the best while taking into account the good compromise between the proposed solution and the suggested requirements. For the treated example, Fig. 5 represents the allocations related to the part « pedal body ».

v. MBSE data generation
Once the preliminary analysis of the different MBSE steps is developed, the MBSE data generation phase can begin. The system engineer needs to represent his work through a structured model for the bene t of the CAD designer. To meet this need, the Extensible Markup Language format (XML), which is a simple, intelligible and coherent le, can be used. In this work, the XML le is generated automatically using an implemented macro which is developed in order to allow the standard exchange between the two elds. The resulting le is composed of three blocks. The rst block contains all the functional requirements, ordered by the system engineer during the functional analysis stage. The second block includes the initial structure of the assembly. The third block contains allocations between components and requirements. Figure 6 presents the XML le of the treated example resulting from the generation of the MBSE data.

CAD development step
i. 3D CAD design Once detailing the system from a functional and technical point of view and traducing the content of the work in an XML le, the 3D CAD design step can begin. The resulting le is sent to the designer in order to create the 3D CAD design solution. This will avoid design errors from the beginning when creating perfectly optimized parts, thanks in particular to the parametric functions and to the possibilities of simulations offered by CAD software. This will obviously help to avoid the need for multiple prototypes and tests. Thus, during this step, the designer is invited to create the solution based on the XML le generated by the system engineer. The designer details all the assembly components and constraints between them. The 3D CAD solution for the treated example, composed of 29 parts, is presented in Fig. 7. Table 1 illustrates the nomenclature of the proposed solution as well as the assembly time (At). ii. Assembly analysis Once the 3D design solution is performed and before returning an assembly design report, the proposed design solution should be evaluated to perceive if there would be any changes before continuing the collaborative work between CAD and MBSE engineers. In this work, a set of evaluation metrics are used to assess the assembly. These metrics are as follow: the maximum number of parts, the total assembly time (TAT), the assembly quality and the AP/DP plans which undoubtedly in uence the e ciency of the design solution. It is noticed that, the evaluation report, based on the previously cited metrics, of the design solution can help the system engineer in the validation of the proposed solution. These metrics are detailed in the following section.

Maximum number of parts
Using this metric, the designer need to analyze the assembly by checking if the total number of parts is optimal or can be reduced. The proposed metric is based on a procedure which tries to minimize the total number of parts. Consequently, the assembly design is simpli ed and the production costs are reduced. According to Boothroyd, the optimal number of parts in the assembly design can be calculated using a procedure based on the answer to four main questions [31]: 1. Is the part a base?
2. Is there a functional relative movement between the part and the others already assembled?
3. Does the part have to be made of different materials or be isolated from other parts already assembled?
4. Does the part need to be separated from other parts in order to allow assembly or disassembly?
For each part of the assembly, if the answer to the four questions is "no", the considered part is not required and can be eliminated from the assembly design. Else, the part must be retained and considered as a required one. Consequently, the sum of the required parts constitutes the minimum number (N min ) that the assembly design must have to satisfy requirements. For the treated example and based on this procedure, the Nmin of parts is 29 parts. Taking into account the requirements of the system engineer proposed in the functional analysis stage, this value can be accepted.

Total assembly time (TAT)
To assess the design quality, the assembly time is a signi cant metric. Indeed, a long assembly time minimizes its quality. According to Boothroyd, experimental tables are used for the calculation of the handling and the insertion time of each part during the assembly process [32]. For the treated example and based on this methodology, the TAT is calculated and estimated to 141 seconds. It is noticed that the system engineering requirement for the maximum value of the TAT is 200s which is higher than the TAT of the proposed solution. Consequently, based on the TAT metric, the proposed design solution is satisfactory.

Assembly e ciency (AE)
The assembly e ciency (AE) metric can estimate if the proposed design can be accepted or not according to a prede ned threshold. The AE can be calculated using the following formula: Where: T min : Minimum assembly time in the assembly line of the industry.
N min : Minimum theoretical number of parts.
TAT : Total Assembly Time of the product.
For the treated example, the AE is obtained as follows: For the system engineering requirement, the minimum value of the AE is equal to 50%. Consequently, the proposed design, having an AE = 61%, is considered as a satisfactory solution.
Generation of Assembly and Disassembly plans (AP/DP) Determining the assembly and disassembly plans (AP/DP) is an important indicator to evaluate a design solution. In fact, choosing the wrong AP/DP can lead to a signi cant loss of time and money. The generation of feasible AP/DP is performed according to Ben Hadj et al. approach [11]. First, and to have a simpler AP/DP, a procedure of elimination of connection elements is done. Then, for each part, a score is calculated in order to know which part should be assembled rst. Finally a case-based reasoning algorithm is used for the insertion of all parts in the global AP. By a similar reasoning the DP can be generated. The AP/DP plans of the treated example are shown in Fig. 8 and Fig. 9.
iii. CAD data generation Once the evaluation of the proposed design is performed using the previously detailed metrics, a design report is generated and exported to the system engineer using a simple and easy format. Indeed the exported report ensures the checking and validation of the design solution by testing the satisfaction of the requirements speci ed by the system engineer.
Similar to the XML-MBSE generation, CAD data coupled to design metrics are also generated in an XML format where the resulting le summarizes the designer work. The resulting XML le contains all the data related to the parts attributes, such as: name, size, volume, weight and At. As well as the data related to the assembly such as: optimal number of parts, TAT and AE.

MBSE enrichment and validation
i. Assembly architecture updated At this stage, all the assembly parts are known and detailed, which allows the system engineer to enrich and update the assembly architecture initially proposed. The generation of the new structure becomes easier thanks to the XML le received from the CAD environment. Consequently, a new enriched structure is modeled in a « bdd » which contains new information that comes to enrich the structure of the assembly initially proposed by the system engineer. For the treated example, the enriched structure of the assembly and a zoom on a part are presented in Fig. 10. In this gure, the pedal assembly is composed of ve sub-assemblies. For example, the « pedal axis » sub-assembly is composed of a part named "axis" which is de ned by a set of attributes such as the name, the surface, the handling time of the part, etc.
In order to detail the internal composition of each part which allows a better understanding of the assembly, an internal block diagram «ibd» is introduced. This diagram enriches the content of the «bdd» by adding the connection links between parts. For the validation example, the assembly architecture is presented in Fig. 11.
ii. Geometrical requirement de nition A mechanical assembly is a set of parts connected to each other by mechanical links. These connections can only be perfect if the parts in contact meet all the geometric requirements. After the design solution, new geometrical requirements will appear and must be satis ed. During this step, the system engineer will represent all the requirements that the mechanical parts must meet. The XML le received from CAD domain which details the relative information to subsequently test their satisfaction is used for this step.
Since mechanical assemblies contain a large number of standard parts whose number is generally more important compared to non-standard ones, their geometrical requirements, known in advance, are modeled in a SysML requirement diagram. This procedure can save time and reduce the work of the system engineer when representing the geometrical requirements of the assembly. The geometrical requirements of standard components and a zoom on only one part which is the pad are presented in Fig. 12. Thus, to model all the geometrical requirements of the parts constituting the assembly, the algorithm shown in Fig. 13 is proposed. The system engineer rstly identi es all the assembly parts and then selects a single component. If the chosen one is a standard component then its geometrical requirements are directly collected from the SysML requirement diagram previously illustrated. Otherwise, new requirements are to be de ned based on the designer work. This step needs to be repeated until all the parts of the assembly are treated.

iii. Validation
The validation stage is the nal step of the interoperability exchange between the systems engineer and the CAD designer. During this step, the system engineer is asked to test whether all of the predetermined requirements have been validated. Therefore, a documented proof is established in order to check if the designed solution is accepted or not. This step consists of grouping together the requirements collected from the MBSE environment with the mechanical parts collected from the CAD domain. The two aspects are structured in a SysML allocation matrix that helps providing an e cient way to communicate and interconnect elements from different domains. The role of the allocation matrix is to allocate relations of satisfaction between each requirement (from MBSE domain) with the assembly parts (from CAD domain).To deal with this, the system engineer is invited to check all the requirements one by one. He tests whether the requirement has been met by the appropriate mechanical part or not. Moreover, the required metrics of the design solution which are introduced in the "Functional analysis" phase will be compared to the values found by calculation from the proposed solution which will facilitate their satisfaction test. This will obviously make it possible to deduce whether the proposed design solution is optimal and adequate or not. Indeed, if all the system requirements have been veri ed and validated by the mechanical parts, the design solution is validated. Otherwise, a report will be sent back to the CAD designer that contains requirements which are not satis ed in order to make the appropriate modi cations on the assembly design. The allocation matrix of the 3D CAD design of the pedal mechanism and a zoom on the validation of the assembly design analysis indicators part are presented in Fig. 14.

Discussion
The problem of communication discontinuity between MBSE and CAD domains can lead to a nonconforming product with requirements. The proposed approach presents an interoperability process between the system engineer and the CAD designer. A standard exchange format is available for each of them in order to facilitate the comprehensibility of the developed work. This will avoid not only the discontinuity problems but also helps the system engineer to translate the assembly data into SysML format. The proposed approach considers some evaluating metrics such as the feasible AP/DP, the TAT and AE which allow the requirements validation step. Finally, the system engineer should judge if the design solution is adequate or not according to the system requirements.

Conclusion And Future Work
To achieve customer satisfaction, communication and exchange between different disciplines plays a key role in product validation. This exchange can sometimes be di cult given the lack of standard means of communication. This paper proposes a methodology that facilitates the MBSE-CAD collaboration. Between the two disciplines, each engineer develop his own work and the data exchange is allowed through les in XML format, which implies an easy interconnection. This involves a standard exchange language between the two domains that nally facilitate the validation step of the designed solution. Using a pedal bicycle example, the methodology was detailed.
As a future work, the developed approach can be improved by favoring the automated check of the validation step.