EXPLAINING THE INFLUENCE OF THE INDUSTRIAL CONTEXT ON THE BENEFITS OF SIMULATION TECHNIQUES IN ASSEMBLY PROCESS PLANNING

(1)

EXPLAINING THE INFLUENCE OF THE

INDUSTRIAL CONTEXT ON THE BENEFITS

OF SIMULATION TECHNIQUES IN

ASSEMBLY PROCESS PLANNING

A Multiple Case Study In Two Industrial Contexts

Master Thesis

by

(2)

2 Master Thesis TOM DD EBM028A30.2014-2016

Explaining The Influence of the Industrial Context on the Benefits of Simulation in Assembly Process Planning: a Multiple Case Study in Two Industrial Contexts

Word count: 14.461

Student

Dirk Weijers S1915622 (RUG) B4062425 (NUBS) E-mail

d.weijers@student.rug.nl

DD-MSc. Technology & Operations Management Supervisors: Dr. J.A.C. Bokhorst

Dr. G. Heron Universities

University of Groningen, Faculty of Economics and Business Nettelbosje 2, 9747 AE Groningen

Newcastle University Business School

5 Barrack Road, Newcastle upon Tyne, NE1 4SE

(3)

3

ACKNOWLEDGEMENT

(4)

4

ABSTRACT

Current developments in digital tools, that are rooted in the upcoming trend of smart manufacturing, offer new possibilities for manufacturers in the design of future assembly processes. A digital tool that is frequently discussed in the literature is simulation software. However, the existing body of literature indicates that simulation does not always offer the same benefits in each industrial environment. Therefore, this study aimed at identifying factors that cause differences in the design of assembly processes between two industrial environments that could influence the benefits of different simulation techniques. For that reason, a multiple case study was conducted in two environments: one environment with high production volumes and low product variety, and one environment in which companies faced low production volumes and high product variety. It was found that standardisation of

assembly sequences, reliability of acquired assembly task data, time-related granularity of planning, emphasis in resource planning, type of constraints, the assembly layout, the

assembly sequence task repetitiveness, learning effects and dynamics of the environment were determining factors that caused differences between the industrial in the design of assembly process. Future research has to be conducted to prove how these factors benefit simulation techniques.

Keywords: Assembly process planning, digital manufacturing, high volume low variety

(5)

5

1 INTRODUCTION

Current developments in Information Technology (IT) offer a broad range of applications, not only in the consumer market but also for industrial practices. Accordingly, IT has become one of the cornerstones of modern manufacturing (Mourtzis et al. 2015). Suppliers of software offer comprehensive system development, planning and control tools to integrate planning, simulation, operations and the enterprise resource planning function across the complete product lifecycle (Zuehlke 2010). As a result, smart manufacturing systems arise that have the ability to adapt to the dynamic market by using real-time data for intelligent decision-making, as well as predicting and preventing failures proactively by following a unified enterprise view (Kumaraguru et al. 2014).

In this, the term “smart” refers to the fact that even the smallest piece of equipment on the factory floor must have a certain degree of built-in intelligence (Zuehlke 2010). The intensification of this manufacturing intelligence has the purpose to obtain a real-time understanding, reasoning, planning and management of all aspects of the production process (Davis et al. 2012). As a result, the concept of digital manufacturing emerged. According to this concept, simulation technologies and data management systems are jointly used for optimizing the manufacturing process, before it is initiated, by reducing product development time and cost (Chryssolouris et al. 2009).

An important stage of the manufacturing process is the assembly process. Assembly consumes up to 50% of the total production time and accounts for more than 20% of the total manufacturing costs in traditional manufacturing industries (Rashid et al. 2012). This implies that the Assembly Process Planning (APP) is a crucial stage in the production process (Wang et al. 2008). APP includes determining the sequence in which components are assembled and the planning of resources, tools and fixtures. A good APP can increase the efficiency, quality and productivity of the assembly process (Xu et al. 2012).

(7)

7

production and supporting ramp-up phases in order to examine the performance under different configurations.

An industry that contrasts the HVLV industry is an industry that is characterised by low production volumes and high product variety (LVHV). In the LVHV industry, one-of-a-kind, frequently high-tech, capital intensive products that consist of many interrelated components are produced (Veldman & Alblas 2012). Companies that operate in this industry play an important role in the economy as they frequently deliver systems that are used for further production (Acha et al. 2004). In this environment, APP seems to differ considerably in comparison with a HVLV environment as APP in this industry shows similarities with project planning (Hicks & Braiden 2000). It seems that these companies cannot benefit from generalized assembly-line-balancing solutions since products are not produced in assembly lines (ALs) as the product variety is high and the production volumes low (Xu et al. 2014; Adrodegari et al. 2015).

There is literature that focusses on the use and advantages of simulation tooling for APP for complex products (e.g. Xu et al. 2012; Zhang et al. 2015). However, these studies were executed in a high volume environment where assembly activities were performed on mixed-model ALs So, despite the fact that recent developments in the field of digital manufacturing offer substantial benefits for APP, the literature remains scarce on the use of simulation software in APP for companies that operate in LVHV environments. Therefore, it remains unclear if simulation for APP can be used in a similar way as in the HVLV environment. The following question arises based on these findings: To what extent can simulation software for APP in a

LVHV context be used in the same fashion as in the HVLV context?

The theoretical contribution of investigating this question is an exploration on how recent developments in digital manufacturing also offer benefits for APP in an uncertain environment where companies face low production volumes and high product variety. In this, simulation is selected as a tool that is embedded in the concept of digital manufacturing. For practitioners, the value of this research can be found in an enhancement in decision-making for applying simulation software in their APP.

(8)

8

simulation are. Subsequently, a multiple case study at four companies (2 HVLV companies, 2 LVHV companies) in the Netherlands and Belgium is conducted, with the aim of obtaining empirical evidence to support or contradict the findings in the literature. The empirical findings are addressed by means of a within-case analysis and a cross-case analysis. The final aim of this research is to develop a theoretical framework in which the relationship between the differences in APP between the two industrial contexts and the benefits of that different simulation technique offer is captured.

2 LITERATURE REVIEW

Although the possibilities of simulation in APP received some attention in literature, it has never been explored how the use of simulation differs per industrial context. Therefore, the purpose of this literature review is to acquire a deeper understanding of APP and the purposes and applications of simulation in two industrial contexts: the HVLV context and the LVHV context. First, APP will be addressed by means of reviewing APP literature in general. Accordingly, methods and techniques that are used in APP and current developments in (digital) tools will be addressed. Hereafter, two industrial contexts will be described and company characteristics from which companies can be allocated to one of the two contexts will be identified. Finally, APP in both environments will be compared and the focus in this research will be sharpened by the means of reformulating the RQ.

2.1

Assembly process planning

Assembly is the activity of creating connections between components or sub-assemblies to form to end product (Abdullah et al. 2003). In traditional industrial manufacturing (e.g. automotive, steel or machine-building), assembly is considered as an important process step in manufacturing since assembly work accounts for up to 50% of the total production time (Rashid et al. 2012). Consequently, assembly costs form a significant proportion of the overall production cost, accounting for more than half of all the total direct costs (Riley 1996). The assembly process, and APP in particular, is therefore an interesting area to consider for possible improvements and cost savings.

(9)

9

the basis of their rules or heuristics in the assembly of all parts of a product, arrange a specific assembly sequence according to the product design description (Rashid et al. 2012). In addition, the resource plan includes the planning of facilities, tools and fixtures that are required for the assembly activities.

According to Marian et al. (2003) and Li et al. (2014), ASP is the first and most important part of an assembly plan as it affects other aspects of the assembly process such as resources, the assembly layout, efficiency and costs. However, it should be noted that ASP and resource planning cannot be considered separately; although the assembly sequence can determine the resources required at a specific time, the availability of resources (e.g. available floor space) can limit the possible sequences as well. Therefore, although ASP is performed first, the relationship between ASP and facility, tool and fixture planning is reciprocal (see Figure 1). After the assembly sequences and required resources are determined, assembly instructions can be generated, which is considered as the third main aspect of APP (Su & Smith 2003).

Figure 1 Overview APP (adopted from Su & Smith, 2003)

(10)

10

the many possible sequences. The most frequently discussed methods in the literature are outlined in section 2.2.

2.2

Methods and techniques used for APP

A well-designed assembly process consist of a more optimal assembly sequence, less changing resources, more reasonable assembly paths, and more reliable operations than others (Li et al. 2014). Therefore, methods and techniques that are used in APP aim at finding optimal assembly sequences and resource plans by planning which assembly activity takes place where, when, and how.

According to Wang et al. (2008) methods for determining the optimal assembly sequence can be classified in exact methods and heuristic methods. Exact (enumerative) methods consist of techniques that aim at exactly finding the optimal results, where heuristic methods strive to find results more efficiently. Li et al. (2014) define methods for ASP more specifically and identify four different types of methods: graph theories, knowledge-based approaches, intelligent

heuristic search-based algorithms and virtual reality based methods. An overview of the

different methods can be found in Table 1.

Approach Method Description

Graph theory (exact)

Neural network-based computational schemes

(Chen et al. 2008)

AND/OR precedence constraints and costs for the connections in the network are displayed

Geometric reasoning approach (Lee & Shin 1990)

An liaison graph is generated wherein each cluster of mutually inseparable parts is displayed as a node.

Knowledge-based (heuristic)

Connection semantics based assembly tree (Dong et al. 2007)

To reduce the computational complexity of exact methods, non-geometric solutions are with geometric solutions in a framework

Disassembly approach

(McGovern & Gupta 2007; Hu et al. 2011)

An existing product is disassembled into its consisting parts to balance the AL.

Intelligent search-based

algorithms (meta-heuristic)

(11)

11

SA (Seyed-Alagheband et al. 2011) Early introduced method that focuses on balancing the AL by using part-by-part search to all feasible assembly sequences.

AC (Yu & Wang 2013) AC is an innovative technique that uses an assembly by disassembly technique as a graph-based meta-heuristic search algorithm. Virtual-reality based

Evaluation of sequences (Ye et al. 1999) Evaluating assembly sequences by simulation in a virtually natural but more cost-effective setting, through viewing and manipulating virtually assembly of parts

Training and instruction generation (Brough et al. 2007)

Training for supervisors and employees before the actual assembly process is started.

Table 1 Overview techniques for APP

The exact methods defined by Wang et al. (2008) show similarities with the graph theory category defined by Li et al. (2014) since both these categorisations require all the assembly components, relations and assembly operations to be known and strive to find the optimal result exactly. Examples of graph theories are a geometric reasoning approach (Lee & Shin 1990) and neural network-based computational schemes (Chen et al. 2008). Further, to reduce the computational complexities, knowledge-based approaches were introduced by using people’s experience and knowledge (Li et al. 2014). For example, Dong et al. (2007) introduced the connection-semantics-based assembly tree which provides a way to consider both geometric information and non-geometric knowledge. Other examples of knowledge-based approaches for ASP are disassembly approaches in which the sequence is based on the disassembly of an existing product (McGovern & Gupta 2007; Hu et al. 2011). These knowledge-based approaches show similarities with the heuristic methods mentioned by Wang et al. (2008), and aim at finding an optimal sequence more efficiently. Nevertheless, a major drawback of these knowledge-based methods is that they tend to stick in local optima and therefore do not find the optimal sequence.

(12)

12

simulated annealing (SA) (e.g. Seyed-Alagheband et al. (2011)). The focus in GA and AC is solely on determining the optimal sequence, where the focus in SA is more on balancing an AL in particular. These methods can intelligently and efficiently compute optimised assembly sequences after information on the assembly model is generated in an effective and reasonable way. This suggests that these intelligent methods require graph-based methods and knowledge based methods to be performed upfront.

Hence, the methods and techniques as described above are frequently exploited to find a starting solution and serve as an initial approximation of the assembly system. In order to evaluate and verify sequences and resources plans, virtual-reality based methods can be deployed. The aim of these methods is to evaluate assembly sequences by simulation in a virtually natural but more cost-effective setting, through viewing and manipulating virtually assembling parts (Ye et al. 1999). In addition, virtual-reality techniques can be used for training of supervisors and employees to generate work instructions and prepare the employees for operations (Brough et al. 2007). An overview of all the different methods and techniques can be found in Table 1. Conclusively, it can be remarked that the described methods and techniques are mainly used for balancing ALs. Here, AL balancing consists of assigning tasks to an ordered sequence of stations such that the precedence constraints among the tasks are satisfied according to pre-defined goals (Erel & Sarin 1998; Hu et al. 2011). Secondly, it must be remarked that the four different types of methods for APP not mutually exclusive in use. Even more, it seems that the different types of methods are used in combination. The initial sequences are calculated by a combination of optimal graph theories and rational knowledge-based approaches. Hereafter, the sequence is optimised by using intelligent meta-heuristic approaches which are evaluated by virtual-reality (i.e. simulation) based methods. Especially the latter method type is interesting for further investigation since it aims at increasing the predictability and transparency of the assembly process which shows similarities with the concept of digital manufacturing. The next section elaborates on this concept by first addressing the nature of digital manufacturing and by then focusing on simulation in APP in particular.

2.3

Digital Manufacturing and APP

(13)

13

(Eastham et al. 2013). In other words, the PLM concept yields integration of product data from requirement, through product design, manufacturing, deployment and disposal. To facilitate this integration, PLM systems consist of a data management software that helps to create a common platform for product development, which aims at improving communication and collaboration in the organisation throughout the PLC (Ameri & Dutta 2005). The importance of this integration of data throughout the PLC was already mentioned by Su & Smith (2003), who argue that, from a concurrent engineering point of view, product design, APP and production simulation should not be considered separately.

A tool that is embedded in PLM software and that can be used to model the assembly process is simulation software (Xu et al. 2012; Zhang et al. 2015). Assembly process simulation visually simulates the process of forming an assembly and allow for the experimentation and validation of different assembly system configurations (Mourtzis et al. 2015). It assesses the operations of moving parts and can determine how the various parts interact and where failures will occur (Zhang et al. 2015). Currently, simulation capabilities are indispensable for evaluation purposes and frequently proved to be useful for safer conclusions for the design of assembly systems (Chryssolouris et al. 2009; Mourtzis et al. 2015; Edmondston & Redford 2003; Genikomsakis & Vassilios 2008). The benefits of different techniques in simulation for APP in the literature are diverse (see Table 2).

PLC phase Technique Benefit

Assembly system design

Line-balancing (Su & Smith 2003; Li et al. 2014; Michalos et al. 2010)

Improve the efficiency and productivity of the assembly system by comparing different assembly systems and AL configurations

Analysis for manual operations by a virtual environment (Brough et al.

2007; Ma et al. 2010; Chryssolouris et al. 2000)

Reduction of task times and increase ergonomic safety.

Collaborative virtual assembly (Mourtzis et al. 2015; Wu et al. 2012)

Increase collaboration and integration across the production process.

Assembly operations

Simulation-based scheduling (Pappert et al. 2010; Frantzén et al. 2011)

Schedule the dispatching for parts by using priority rules. Simulation and optimisation are executed during the assembly process and benefit the evaluate different scenarios.

(14)

14

The different simulation techniques for APP are categorised based on the classification of simulation techniques in manufacturing by Negahban & Smith (2014). To clarify, as Wang et al. (2008) argue, APP activities especially are performed during product design and manufacturing. These two stages in the PLC correspond to what Negahban & Smith (2014) describe as the systems design phase and systems operations phase respectively. Therefore, it is assumed that APP, and thus simulation for APP, is performed and used during these two stages. To illustrate, in assembly system design, line balancing simulation techniques and virtual reality techniques are used to evaluate (aspects of) system designs before production is started, so no materials are wasted (Hlupic et al. 1999). Simulation-based scheduling, on the other hand, can be used to during the operational phase (i.e. assembly operations) for rescheduling purposes. The next paragraph elaborates on the simulation techniques that are used in the assembly system design phase and during the assembly operations phase.

Firstly, simulation software can be used to compare different designs for assembly systems or sequences (Su & Smith, 2003). In this, simulation software is frequently used to balance the configuration of an AL (Li et al. 2014; Michalos et al. 2012). Especially companies that operate in industries where production volumes are high and product variety is low, achieve greater efficiency and productivity gains in their APP by applying this simulation technique. Secondly, analysis for manual operations can be performed by the use of virtual reality by verifying human-related factors (Ma et al. 2010; Chryssolouris et al. 2009; Brough et al. 2007). This can be particularly useful when the sequence in which an operator has to perform tasks can be determined arbitrarily or for estimating the ergonomic stress during an assembly task. Thirdly, another application of simulation is to enhance the interaction between companies or departments during the design phase and assembly planning phase (Wu et al, 2012). This is labelled as collaborative virtual assembly (CVA) and serves as a real-time assembly environment where designers and planners can exchange data, discuss and verify assembly schemes (Mourtzis et al. 2015; Wu et al. 2012). This technique is particularly useful when departments or companies that are jointly responsible for APP are geographically dispersed. Fourthly, simulated-based scheduling can be used to schedule the dispatching of parts by including priority rules (Pappert et al. 2010; Frantzén et al. 2011). More specifically, in simulation-based scheduling for APP, scenarios are repeatedly simulated, analysed and optimised during the assembly process (Frantzén et al. 2011).

(15)

15

times. This suggests that simulation is especially beneficial in an environment where companies assemble in large product volumes on ALs. However, the concept of AL balancing is starting to gain attention in the low volume production of customized products as well (Scholl & Becker 2006; Pappert et al. 2010). For example, in a LVHV production environment, scheduling workforces is often considered as a major difficulty since it is possible to work on a product with different amounts of workers with different amounts of time (Pappert et al. 2010). This might indicate that it is interesting to investigate the possible benefits of simulation as a decision support tool in this context.

However, a deeper understanding of this LVHV context is required before conclusions can be drawn. For that reason, the next section firstly addresses the this context and positions it afterwards by making a comparison with the HVLV environment.

2.4

Two Industrial Contexts

In literature, the products that are produced in the LVHV industry companies are frequently labelled as capital goods. Capital goods could range from rather simple capital-intensive products (e.g. roadwork) to very complex capital-intensive products (e.g. photolithography machines), consisting of numerous complex components (Acha et al. 2004). It must be remarked that both these types of products include high-costs but are considerably different in terms of complexity. In this research, the type of products that are produced in the LVHV industry deal with high complexity that complicates the APP. The remainder of this section will address characteristics of products in the LVHV industry and explain how some of these characteristics cause complexity for companies. Finally, a comparison with the HVLV environment will be drawn to highlight the aspects that might be interesting for investigating the influences on APP.

Low Volume High Variety

Complex products and systems (CoPS) that are produced in the LVHV industry can be considered as high cost, one-of-a-kind, engineering-intensive products, systems, networks and constructs (Hobday 1998; Veldman & Alblas 2012). The complexity of these type of products relies in the high number of customized components, the breadth of (new) knowledge and skills required for development and production of these products, and the low volume in which they are produced (Hobday 2000; Xie et al. 2014; Bertrand & Muntslag 1993).

(16)

time-16

dependent complexity and time-independent complexity. Time-dependent complexity relates to the changes and uncertainties of design requirements and manufacturing environments. For example, LVHV companies operate in a highly volatile environment: customer orders and product shipments can change by more than 50% in a year (Andeson et al. 2000). These changes can be explained by the fact that companies in these operate in high-technology environments, where state-of-the art technologies are replaced by newer technologies rapidly (Veldman & Alblas 2012; Xu et al. 2014). Time-independent complexity can be imaginary (i.e. a lack of knowledge about the underlying product structure) or real (i.e. arises from the random nature of the system). Time-independent complexity results from the ever-increasing number of parts and their interrelatedness (Xu et al. 2012; Hobday 1998), which cannot be eliminated. Also, not sufficient knowledge is available to understand the structural mappings of their design elements (Veldman & Alblas 2012). The imaginary complexity suggests that knowledge is difficult to obtain since products strongly differ from each other. This high product variety can be explained by the high degree of customisability.

In the LVHV each product is fully customized and one-of-a-kind; the product is linked to the customer order at the start of the design phase and finished product designs are not completely known before an order is placed (Veldman & Alblas 2012; Hicks & Braiden 2000; Adrodegari et al. 2015). A concept that relates the degree of customizability to the design of the production process, is the order penetration point (OPP), which is defined as the point in the value chain of a product where the product is linked to a specific customer order (Olhager 2003). The range of OPPs varies from make-to-stock to engineer-to-order (ETO). Companies in the LVHV environment are most often classified as ETO since each product is fully customised, starting at the design phase.

(17)

17

To summarize, the uncertainty in the LVHV industry mainly relies in high customisability of the products, low production volumes, highly fluctuating demand and the duration of the production process that pressurizes the time-to-market. This is in contrast with the HVLV environment, where companies attempt to obtain competitive advantage by adopting a strategy that benefit from the highly repetitive nature of their production process (Olhager 2003). In order to position the contexts of this research, the next section addresses this HVLV environment.

High volume low variety

Before describing the HVLV environment, it has to be noted that that the terms ‘high volume’ and ‘low variety’ should be interpreted relatively. To clarify, the HVLV industrial context should not be considered as a mass production environment where production volumes are extremely large and there is minimum variety between products. The term ‘HVLV’ is used to emphasize the inclusion of a degree of customisation in its products, and its contrast with the LVHV environment. More specifically, in this research, it is assumed that companies that operate in a HVLV adopt a assemble-to-order (ATO) strategy.

An ATO strategy yields that the production does not start until the customers places an order; standard parts are kept in stock and are assembled to (sub)components, the moment when an order arrives (Olhager 2003; Stevenson et al. 2005). Implicitly, the OPP in ATO companies located in a later production stage than in LVHV (ETO) companies (see Figure 2).

Figure 2 Overview ATO and ETO strategy (adopted from Olhager, 2003). Note: dotted lines represent forecast-driven activities where straight lines depict customer-order-driven activities.

(18)

18

perform them. Furthermore, the number and variety of components are considerably lower than in the LVHV industry where products are less complex and have a lower degree of customisation (Aslan et al. 2012). In addition, a summary of the comparison between the characteristics of the LVHV context and the HVLV context is displayed in Table 3.

Overview context characteristics

Characteristic LVHV HVLV

Degree of product customisability High Low/Medium

Number of product components High Low/Medium

Repetition of assembly tasks Little High

Range of knowledge and skills involved in assembly activities

Broad Narrow

Production Volumes Low High

Assembly Layout Projects AL

Production strategy ETO ATO

Table 3 Overview characteristics per context

Now that the LVHV and HVLV environments are positioned and more elaborately described, the characteristics can be used to explain how APP is organised in each industrial context.

2.5

APP in HVLV and LVHV industries

(19)

19

LVHV HVLV

ASSEBMLY SYSTEM DESIGN

ASP

Non-standard sequence by adopting a project scheduling approach

Standard sequence by using methods and techniques for ASP

ASP is based on the experience of planners and operators.

ASP is based on methods and techniques are used in a combinatorial way to determine the assembly sequence.

Resource planning

Deployment of workforces is of major importance

Resource planning mainly determined by the ASP

ASSEMBLY OPERATIONS Project assembly layout AL layout

Low task repetitiveness High task repetitiveness Little learning effects High learning effects

SIMULATION Use of simulation-based scheduling to increase

planning reliability

Used for evaluation of AL configurations Numerous manual involvements in the use of

software for APP

Software for APP can be efficiently implemented payoff is large.

Table 4 Overview differences in APP per context

Assembly system Design

ASP

The APP methods to determine assembly sequences (e.g. intelligent search-based algorithms) lead to major increases in efficiency and productivity gains in HVLV environments, while the benefits of these methods tend to have limited effects in the LVHV context. To explain, generalized solutions for standard assembly sequences will not work in this environment as it is difficult to experiment with designs in the LVHV context since a new tailor-made assembly system has to be designed for every new order (Adrodegari et al. 2015). As a consequence, ASP in the LVHV relies on the experience of skilled workers (Adrodegari et al. 2015). For example, many companies in LVHV environments still depend on scheduling by experienced staff which leads to unused capacities and due date violations (Pappert et al. 2010). This implies that knowledge on how to organize the APP is only available in few workers, which complicates knowledge-sharing and consequently collaborative APP (Wang et al. 2008).

Resource planning

(20)

20

2010). To refer back to the APP model in Figure 1, it can be concluded that the reciprocal relationship between resource planning and ASP differs per context. To be more specific, ASP is considered as more important in the HVLV context while planning resources is more determining the LVHV context.

Assembly Operations

Another aspect that is important to consider is the repetitiveness of assembly tasks and APP tasks. Assembly tasks refer to the tasks that have to be performed by an operator while assembling an product. Conversely, APP tasks refer to the activities of a planner during the APP process. In the HVLV environment, the tasks that have to be performed at the workstations on the AL are frequently highly repetitive (Abdullah et al. 2003). Since the assembly system’s performance relies on the worker’s ability to remember and perform tasks (Michalos et al. 2010), companies in the HVLV context strive to increase learning effects by increasing task repetition. In contrast, a consequence of the tailor-made (project layout) assembly systems in LVHV contexts is that there is a low repetitiveness of tasks for both planners and assembly workers when comparing two different projects. Consequently, learning effects of workers might be limited or even negligible in this context (Cho & Eppinger 2005).

Simulation

As described in section 2.3, simulation for APP in the HVLV context is used to evaluate different configurations of ALs. However, the use of simulation in the LVHV context remains somewhat vague in literature. Hicks & Braiden (2000) state that in LVHV situations, companies adopt project-planning type approaches for APP each time a new order arrives. Accordingly, APP in a LVHV environment shows similarities with resource-constrained project scheduling problems (RCPSPs) as described by Herroelen & Leus (2005). This concept aims at scheduling and planning project activities with uncertain durations in order to minimize the overall project time. However, in the LVHV environment, minimising the overall production time is not the only aim of project scheduling: storage costs and capital commitment are factors in completing the product at the right time (Pappert et al. 2010).

(21)

21

software tools (e.g. simulation software) for APP in the LVHV context require numerous manual involvements, which limits the added value (Adrodegari et al. 2015; Xu et al. 2012).

2.6

Research focus

After reviewing the literature, it can be concluded from section 2.3 that simulation techniques offer benefits for APP (see Table 2). Additionally, it was also concluded in section 2.5 that simulation techniques do not always offer similar benefits per industrial context. For example, benefits from simulation for line-balancing purposes do not apply to a LVHV context. Therefore, it can be concluded that the volume variety context has certain influence on the relationship between simulation techniques and the benefits of simulation in APP. This conclusion is captured in the conceptual model in Figure 3.

Figure 3 Conceptual Model. Note: All the variables in this Figure should be considered in the context of APP.

(22)

22

How do the differences in APP between the HVLV and LVHV environment influence the benefits of applying simulation in APP?

To answer the RQ, four sub questions (SQs) are established. The purpose of these questions is to systematically break up the RQ. SQ 1 focuses on obtaining empirical evidence on how APP is organised within a particular industrial context. SQ 2 has the specific aim to gather empirical evidence on the differences and similarities in APP between the two contexts. Subsequently, SQ 3 and SQ 4 focus on the development of a framework that connects differences in APP between the two environments to the benefits of simulation. The SQs are as follows:

1. To what extent is the organisation of APP within the case companies similar to APP as described in the literature regarding the industrial contexts the cases companies are operating in?

2. To what extent are the differences in APP between the two contexts as described in literature similar with the differences in APP between the case companies?

3. Which differences in APP between the LVHV context and the HVLV context benefit or hinder the deployment of specific benefits of simulation in APP?

4. How can the relationship between APP differences and the purposes of simulation be generalised?

3 METHODOLOGY

(23)

23

develop grounded relationships between the factors that cause differences in APP and the benefits of the simulation of simulation techniques (i.e. developing a theoretical framework). An overview of this process is displayed in Figure 4.

In the remainder of this section, the selection criteria for the four case companies are described and the data collection and data analysis methods are addressed. Finally, the four case companies are described according to their industrial context.

Figure 4 Overview research

3.1

Case selection

(24)

24

a deeper understanding how APP is organized in a specific environment and how this differed from what is argued in the literature. Furthermore, the cases were retrospective cases as the timeframe of this research is limited and retrospective cases allowed a more controllable selection of data (Karlsson 2009, p.171).

The criteria on which the cases were selected (Table 5) were related to the variables that influenced the APP. As mentioned before, the first selection criterion was the industrial context the company was operating in. For two cases this was the LVHV context, and two cases in the HVLV context. The reason why it was chosen to explore APP in two polarised contexts is to highlight the commonalities and differences in the observed phenomena by applying replication logic in order to develop a theory (McCutcheon & Meredith 1993; Voss et al. 2002). Secondly, two of the four case companies, should already have deployed simulation in the APP. To be more specific, at the most, one company in the LVHV used some form of simulation for APP, and at least one company in the HVLV industry used simulation in APP. This allowed the comparison between the use of simulation in two different industrial contexts. Thereby, the fact that there was one LVHV company that was not using any form of simulation in their APP, created possibilities for a test case for simulation to test the relationships developed framework. However, it must be mentioned here that the use of simulation for APP in the companies was different than initially expected, which limited the findings in this research (see section 5).

Cases

Selection criteria A B C D

Industrial context HVLV HVLV LVHV LVHV

Use of simulation Yes/No Yes No Yes

Assembly Layout Mixed-model line

Mixed-model

line Project Project

Production strategy ATO ATO ETO ETO

Table 5 Case selection

(25)

25

3.2

Data collection

In order to obtain information about the APP of the four cases, data from primary and secondary sources was collected to obtain the clearest, most objective view on APP (McCutcheon & Meredith 1993). Triangulation of data was ensured by collecting data from multiple types of sources (Karlsson 2009, p.176). Therefore, data was collected by conducting interviews with workers and managers, direct observations and informal conversations.

For the interviews, a semi-structured interview technique was used. This technique allowed the interviewer to elaborate on pre-determined questions during the interview. Accordingly, the interviewer was able to ask for clarification on answers or for additional information (Barriball & While 1994). By using open questions during the interview, interviewees were ‘steered’ towards a certain direction in their answers by the interviewer, which increased the reliability of the data (Karlsson 2009). The interview questions (Appendix A) were structured according the ‘funnel’ model; starting out with broad and explorative open questions about company characteristics and APP. Hereafter, the questions became more specific and focused on different aspects of APP derived from the literature. After the interviews, the transcript was sent to the interviewee to check the correctness of the answers and missing elements.

Furthermore, multiple sources of evidence were used by interviewing both workers and managers; these were people from different hierarchical layers and with different functions related to the APP. An overview of the interviewees can be found in Appendix B. Lastly, direct observations on the assembly floor were an additional source of data. During the observations, the assembly process was described in detail – usually by an operator – while walking around in the factory halls. Furthermore, informal talks with workers on the AL were carried out to examine how they experienced work on the assembly line and to obtain an understanding of the problems these workers encounter.

3.3

Data analysis method

(26)

26

rational manner (axial coding). Finally, selective coding was applied by linking the (sub)categories to find relationships among them.

Hereafter there were two steps for analysing the data in the cases; a within-case analysis and a cross-case analysis Eisenhardt (1989). In the within-case analysis, single case descriptions of how APP is organised in the case companies were provided. These descriptive write-ups were important as they formed the core of the creation of insights later (Gersick 1988; Pettigrew 1990). In this analysis, the empirical data was and summarised and ordered by tables and visualised by diagrams as this improved the comprehensiveness for the researcher (Miles & Huberman 1984; Eisenhardt & Graebner 2007; Voss et al. 2002). For example, the steps that are undertaken in the APP process were displayed in Business Process Modelling Notation (BPMN) to make the process readily understandable (Chinosi & Trombetta 2012). Also, patterns were identified in relation with the steps undertaken in APP and how this relates to the context.

The within-case analysis was followed by a cross-case analysis. The aim of the cross-case analysis was to compare the APP process within each environment to generalize aspects. For this, literal replication was applied by comparing case A and B and by comparing case C and D. Afterwards, the generalisations for each context were compared with what was suggested in the available literature to find similarities and contradictions. Hereafter, the aim was to identify factors that cause differences in APP between the two contexts. Therefore, theoretical replication was applied by comparing APP in case A and B with APP in case C and D. The replication was applied until theoretical saturation is reached; when similar instances are seen over and over again (Pandit 1996). Again, the findings here were compared with the existing body of literature.

3.4

Quality criteria

To ensure the quality of this research, validation and reliability of the collected data was of major importance. Yin (2009) identified four different quality tests: construct validity, internal validity, external validity and reliability. In this research, the construct validity was ensured by the use of multiple sources of evidence, the establishment of a chain of evidence (i.e. questions during the interviews are linked to variables and the RQ) and by letting the interviewees review the transcripts of the interview.

(27)

27

analysis and across-case analysis. For example, different patterns on variables that influence the differences in APP between the environments were recognised in each case. Next, external validity, which is regarded as the extent to which the findings can be generalized, was ensured by applying replication logic as mentioned in the previous section. Finally, the reliability of the data was ensured by using multiple sources of data, by asking the same questions multiple times in a different way, and by letting the interviewees review the transcript.

3.4.1 Ethics

Recently, it has become almost impossible to do research without risking criticism incurred by research conduct that is considered to be unethical (Bryman & Bell 2015 p.6). Therefore, it was crucial that this research was conducted as ethical as possible. In order to ensure this, the 10 principles defined by Easterby-Smith et al. (2008) for conducting research in an ethical way were applied (Table 6).

Key Principle (Easterby-Smith et al. (2008)) Application

1. Ensuring no harm to participants

Quotations in section 5 and throughout the report are never linked to personal information and names.

2. Respecting the dignity of research participants

An informal conversation is never started unless the interviewee initiates it. No private issues are discussed during the interviews (e.g. frustrations on the work floor)

3. Ensuring a full informed consent of research participants

Permission is asked to record the interview when the interview starts and explain what how the gathered data will be used.

4. Protecting the privacy of research subjects

Names and personal data are not mentioned throughout the report.

5. Ensuring the confidentiality of research data

Gathered data and information is never shown to persons other than the supervisors. Digital documents are kept in secured folders.

6. Protecting the anonymity of individuals or organisations

The names of interviewees and companies are masked (e.g. interviewee X or company A).

7. Avoiding deception about the nature of the research

Clearly state what the goal of the research is and what the deliverable is for the company. 8. Declarations of affiliations, funding

sources and conflicts of interest

(28)

28 9. Honesty and transparency in

communicating about the research

Giving a presentation at the companies in the beginning and at the end of the research. 10. Avoidance of any misleading or false

reporting of research findings

Let the interviewees review the transcript and final report

Table 6 Principles of research ethics

3.5

Framework Development

The plan was to develop a framework that connects the factors that cause empirical differences in APP between the HVLV and LVHV context, and the benefits that simulation techniques have to offer. The idea was to verify the relationship by conducting an example simulation study in the LVHV context Afterwards, the proposed relationships had to be tested by evaluating the results of the simulation study. This evaluation was performed by two techniques: conducting interviews with planners on the results of the simulation study and by comparing the results of the simulation study in the LVHV context with the application of simulation in the HVLV context. However, this evaluation was hindered, and thus the development of the framework, which is elaborated in section 5.

To conclude, a general overview of the complete methodology is displayed in Figure 5.

(29)

29

3.6

Introducing the case companies

The description of the case companies is provided according to the list of company characteristics that was presented in Table 3 in the theoretical background. An overview on how the case companies score on the characteristics can be found in Figure 6 and Table 7. A more detailed description of each company can be found in Appendix C.

Figure 6 Overview case company scores

In Figure 6, the scores of the case companies on the different company characteristics are scaled from 1-5. A more elaborate explanation on the calculation on the scores can be found in Appendix D. 0 1 2 3 4 5 Volume Product variety Production strategy Number of components Novelty technologies Tasks repetiteveness Assembly layout

Overview case companies

(30)

Company Industry Industrial Context Average Yearly Production Volume

Product Variety Production Strategy Number of product components Novelty technologies Repetitiveness Assembly/APP Tasks Assembly layout A Automotive HVLV 255.000 (high) 100 specific types, (medium variety) Assemble-to-order 100-150 (low) Medium. Sometimes new product options are introduced by the car manufacturer, which have to be incorporated in the assembly process Planners: High Operators: High Mixed model AL B Automotive HVLV 54.000 (high)

4 main types (low variety)

Assemble-to-order

1500 (low) Medium. New

emission standards pressurize R&D to improve the engine design and include new technologies Planners: High Operators: High Mixed model AL(s) C Offshore constructions

LVHV 3 (low) 2 main systems

(tensioners and winches). However, no system is the same (high variety). Engineer-to-order 30.000 (medium) Medium. Company C uses reliable technologies from which it’s working has been proved

Operators: medium. Planners: low

Dock layout

D Semiconductor LVHV 200 (low) 8 main machine

types. However, no machine is the same as a result of high customisation (high variety) Engineer-to-order 200.000 (high) High. Company D is operating in a technology driven market which challenges assembly planning. Operators: (within one work centre) high. Planners: low. Final assembly: dock layout. Work centres: line layout

(31)

4 FINDINGS

In this section, the analysis of the interviews (an overview of the interview protocol and interviewees can be found in Appendices A and B respectively) and field observations at the four case companies is provided. Only the cross-case analysis is presented since the SQs specifically focus on comparisons in, and between, the two environments. The findings for all individual cases are described in the within-case analyses in companies which can be found Appendix E.

4.1

How APP is organised in the HVLV environment

APP is the activity in which the part assembly sequence and resource usage are determined, with the aim of minimizing time and costs and increasing the quality and efficiency of the assembly system (Su & Smith 2003; Wang et al. 2008; Jones et al. 1998). Similar to section 2.5, the findings on differences in APP between the environments are categorised according to the PLC phases in which APP is performed (assembly system design and assembly operations). Furthermore, a third category is “simulation”, which focuses on the use of simulation for APP in the companies used in this research.

Assembly system design

ASP

For both HVLV cases, ASP was initiated when a new product design arrives from the main manufacturer. This design was considered as ‘frozen’ and did not change over time except for minor adjustments that were made after the assembly process was initiated. The accompanied assembly sequences were broadly predetermined by the main manufacturer, which restricted the possibilities to experiment with different sequences. Therefore, ASP in these companies was concerned with determining the sequence in between prescribed points in the AL. In company A, the sequence was determined by analysing time duration of assembly tasks with Methods Time Measurement (MTM) techniques (see Appendix F for an example). As one plant industrial engineer from company A stated: “We do not work with complex algorithms,

everything is based on time analysis and assembly sequences are determined in collaboration with experienced operators.” This shows similarities with company B, where the sequences

(32)

32

Nonetheless, the role and importance of experience of the planners with APP in this method was somewhat vague. As one respondent from company A stated: “Experience with the product

is useful but not required to perform APP”. Here, experience was related to product knowledge,

and knowledge on the ways of working during the assembly process. Product knowledge was especially considered as important during the design phase, prior to the APP. However, knowledge on the way of working (e.g. knowledge on constraints) was in both companies considered as important. As one respondent in company B argued: “Experience is crucial to

estimate ergonomic aspects and assess the workability”. This leads to the conclusion that

experience in APP is certainly important in this environment.

Similarly, shifting tasks between the stations to minimise the cycle time shows similarities with the principles of line-balancing. This is in line with what literature suggests about the focus of APP in these environments (e.g. Abdullah et al. 2003), as (re-)configuration of this AL is of critical importance for both companies to implement a cost-efficient system (Boysen et al. 2007). Furthermore, similar to what the literature suggested, the case companies worked with standard assembly sequences for each product. However, it was surprising that none of the methods or techniques to find optimal assembly sequences were used in the case companies. This diverges considerably from what the literature in section 2.2 stated. Especially since the number of components (and thus the amount of feasible assembly sequences) for the products were considerable in both companies, it was expected that at least some techniques were used to manage this complexity (Li et al. 2014). Furthermore, the role of experience of operators in this context seems to diverge from what is stated in the literature. The fact assembly sequences were determined in collaboration with operators, suggesting that experience from the floor was in fact important and sequences are not solely determined by planners that rely on data.

Resource planning

An interviewee from company A stated: “We strive to group assembly tasks that require

equivalent equipment and skills as much as possible in order to reduce investments in equipment.” This indicated that the most important aspect of resource planning in this

environment, was the planning of tools and fixtures on the AL. Furthermore, both companies deal with prior investments in the AL, which displacement would incur major costs. As one interviewee of company B stated: “We have only one valve adjustment machine, that is a

constraint in the AL that we have to deal with.” These prior investments are considered to be

(33)

33

considering the type of constraints that have to be incorporated is something that is not described in the literature. More specifically, in the literature it was stated that the resource planning finds place after, and was based solely on, the ASP. Here, nothing was mentioned about the type of constraints that needed to be taken into account.

Assembly operations

In both companies, assembly activities were performed at workstations on a AL, where different product types arrived in succession. This type of AL corresponded with what in the literature is labelled as a mixed-model line, where various models arrive in a random sequence (Manavizadeh et al. 2012). One process engineer from company A stated: “We attempt to let

the operators perform the same task as much as possible.” Implicitly, the task repetitiveness at

the workstations on both ALs was high and routine work was stimulated to increase standardisation and prevent errors in both companies, which was in line with the literature on ALs. For example, as Cho & Eppinger (2005) mentioned, this iteration of tasks leads to learning effects by the operators and will eventually reduce the time required to complete a task. Furthermore, the depth of time measurement was equal in both companies. The overall cycle times of the AL were in minutes, and task times at the stations were planned in seconds, which increased the comprehensiveness of the planning.

Simulation for APP

No simulation software or other computer-aided tools were used to evaluate the different sequences or to balance the AL. Instead, evaluation of the proposed sequences occurred in-between the steps of the process of APP and was based on the workability for the operators. The reasons for this absence of simulation software were different. A floor manager from company B stated: “In the past we simulated a few stations with Tecnomatix Plant Simulation

to evaluate assembly sequences. However, this was not appreciated by the operators as they were unwilling to give up their current way of working.” A respondent from company A

mentioned: “We try to keep it as simple as possible, evaluation is done in between the project

phases in collaboration with experienced operators.” This indicated that both companies might

(34)

34

complex process (Hu et al. 2011) and the benefits of simulation techniques might be difficult to explain to people that are not involved in the APP (Hlupic et al. 1999). Nevertheless, virtual reality (VR) might offer efficiencies here for evaluation as it can reduce the need for physical experimentation (e.g. offline assembly) by experimenting in a VR simulation-based environment (Chryssolouris et al. 2000).

4.2

How APP is organised in the LVHV environment

ASP

Both companies had a standard assembly sequence, which was based on task time analysis. For the estimation of task times, frequently historical data of comparable projects was used. However, as one respondent from company C stressed: “Using historical data for task time

estimation is a good initial approximation but experience from the workforce is required for a more detailed estimation since no project is the same.” This experience especially relied on

knowledge of the sequence in which a machine had to be assembled. In addition, it can be concluded that none of the methods and techniques for determining the optimal assembly sequence are used in this environment.

This can be explained by the fact that the standard assembly sequence changes constantly over time in this environment. To be more specific, a project engineer from company C mentioned:

“An assembly planning is established and fixed. However, frequently before the assembly process starts, adjustments already have to be made.” This indicates that the usefulness of

finding an optimal sequence prior to the start of the assembly process is limited as constraints (e.g. delivery times of suppliers) change rapidly. In addition, long lead items (LLIs) restrict the number of possible sequences in both companies.

Furthermore, another remarkable aspect of ASP in the case companies was that granularity in terms of time measurement is a difficulty in the LVHV environment. An interviewee of company D stated: “Specific process steps are calculated in minutes while the overall project

planning is planned in weeks or months.” This indicates that time efficiency in minutes or

seconds, obtained by the use of optimisation techniques for APP, may have a limited impact since the overall cycle time is measured in months.

(35)

35

time (Adrodegari et al. 2015; König et al. 2007). Secondly, no methods or techniques that were derived from literature are used in this context. Instead, companies in this environment trust on the experience of planners and operators (Pappert et al. 2010). Thirdly, in addition to the literature, the case companies each have difficulty estimating the benefits of greater temporal efficiency. This is due to the strongly differing time units (either minutes or days) used to measure the duration of assembly tasks.

Resource Planning

The emphasis in resource planning was mainly concerned with the planning of workforces and floor space. To illustrate, an one project engineer from company C argued: “Manpower is the

most important resource to consider. The planning of resources are based on the most efficient capacity of 25 operators.” This makes project scheduling an important task of the resource

planning. Additionally, as one senior design architect of company D stressed: “In order to

prevent that people get in each other’s way, we draw the walkways in Ms PowerPoint,” and

“[w]e attempt to let an operator perform as much tasks as possible at one location in order to

minimise walking times.” This indicates that estimating floor space is also an important resource

to consider. Other important resources in this context were transportation equipment (e.g. cranes), which were shared by multiple operations that are performed simultaneously. As a result, resources in terms of space on the construction floor were difficult to estimate in this environment.

When reflecting these findings on the literature, it is suggested that only human workforces are the most important resource to consider in this environment (Pappert et al. 2010). However, it can also be concluded from the cases that planning floor space is critical as well and should therefore be added to the literature.

Assembly operations

Both companies used a project assembly layout for final assembly and AL layouts for preassembly. Further, the overall task repetitiveness for assembly workers was low since the products were one-of-a-kind. However, as the lead logistics in company C mentioned: “The

specific actions that have to be performed are almost always screwing and mounting related tasks.” Or, as a design engineer in company D remarked: “Despite the fact that task iterations are high, the chance that tasks are performed in the same sequence is minimal, so the only thing you can do is train and standardise specific tasks.” In that sense, tasks can be considered to be

(36)

36

standardising mounting tasks). In addition, because of the low task repetitiveness, the learning effects in this context were limited.

Furthermore, another recurring aspect of the assembly operations in this environment were the dynamics of the assembly process. For both LVHV cases, time to market was crucial; a delivery that was not on time resulted in major contractual costs charged by the customer, as their production depended on the ordered systems. In addition, another reason why the time to the market was crucial for company D was that they strived to stay ahead of the competition by delivering the newest technologies to its customers.

These dynamics caused uncertainties and therefore companies in the LVHV context adopted business rules for prioritizing certain assembly tasks above others. Here, a business rule is defined as: “guidance that there is an obligation concerning conduct, action, practice or procedure within a particular activity or sphere” (Boyer & Mili 2011, p.6). For example, a business rule of company C that was provided by the leads system assembly: “Assembly tasks

are planned based on the arrival of LLIs.” Furthermore, as one manufacturing architect of

company D stressed: “The alignment between the departments is extreme; when changes are

made in design, it is immediately sent to the production department, where they are trained for it so they know what to do.” These business rules and alignment of departments thus helped

LVHV companies to manage the dynamics.

When reflecting on the findings in the literature, several conclusions can be made. Firstly, the assembly layout that the companies adopted is, in line with literature, a project layout. Secondly, it was argued in the literature that the task repetitiveness and subsequent learning effects are low in this context (Cho & Eppinger 2005). However, from the cases it can be concluded that task repetitiveness can be high at a very specific level. Nevertheless, the sequence in which tasks are performed, is almost never the same. Thirdly, the dynamics of the assembly operations seem to play an important role during APP and should therefore be considered when comparing APP in the two contexts. When reflecting on the literature, the dynamics can be related to the concept of time dependent complexity, where the underlying generic design of a product is continuously under revision during delivery of products (Veldman & Alblas 2012). Furthermore it seems that the concept of concurrent engineering is of major importance in this environment to integrate design and planning activities (Su & Smith 2003).

(37)

37

Simulation

The use of simulation in APP was different per company. Company C did not use simulation in their APP as they prioritise other APP activities. In contrast, company D used simulation for APP to increase the transparency and predictability of their assembly operations. As one senior design architect mentioned: “By virtually assembling the machines multiple times, the machines

are built more frequently than we would ever do under real circumstances.” Here, simulation

was especially used in the design phase in order to assess how many resources are required. A specific application was to analyse movements during assembly tasks to assess required space. This is in line with literature, which suggests that simulation in this environment can be used to analyse human motions (Ma et al. 2010). Furthermore, again similar to what is stated in the literature, company D benefits from the fact that with simulation, the assembly process can be performed without wasting any materials or resources being wasted (Hlupic et al. 1999).

4.3

Differences in APP between the two environments

Now that there is empirical evidence on how APP is organised per environment, the findings on the differences between the environment are compared with the differences derived from the literature. In Figure 7, it is summarised how the findings in literature are combined with, or adjusted to the findings in the cases, lead to an empirical model of factors that cause differences in APP between the HVLV and LVHV context.

ASP

(38)

38

Furthermore, the role of experience of planners in APP in the HVLV environment is more important than what was expected from literature. This experience was especially related to knowledge on the ways of working. However, it can also be concluded that companies in the LVHV context rely more on experienced planners and operators to determine sequences than in the HVLV context. An explanation for this was that there is a large difference between the environments in the acquisition of reliable data. In the HVLV context, historical data (e.g. for task times) can be considered as reliable as it was gathered over multiple repetitions. In contrast, in the LVHV environment, historical data cannot be considered as reliable as products strongly differ in design. The above-mentioned suggests that not the role of experience in APP should be considered, but the reliability of the acquired data.

Another important aspect in which ASP in both environments differ, is determining the granularity of planning. In the LVHV context, certain tasks are calculated in minutes or seconds, while the overall planning is in months. This implies that optimising an assembly sequence in minutes or seconds does not provide the same benefits as in the HVLV context. Despite the fact that both companies in the LVHV indicated that the planning of granularity is a difficult, this concept cannot be found in the literature and should therefore be added. Resource Planning

The emphasis of resource planning is different in both environments. In the HVLV context, tools and fixtures on the ALs are the most important resources to consider. In the LVHV context; however, the most important resources are the workforces and the estimation of required floor space.

(39)

EXPLAINING THE INFLUENCE OF THE INDUSTRIAL CONTEXT ON THE BENEFITS OF SIMULATION TECHNIQUES IN ASSEMBLY PROCESS PLANNING