Scholarly article on topic 'The Transformable Factory: Adapting Automotive Production Capacities'

The Transformable Factory: Adapting Automotive Production Capacities Academic research paper on "Agriculture, forestry, and fisheries"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia CIRP
OECD Field of science
Keywords
{Factory / Flexibility / Simulation / "Production capacity"}

Abstract of research paper on Agriculture, forestry, and fisheries, author of scientific article — Thomas Creutznacher, Ulrich Berger, Raffaello Lepratti, Steffen Lamparter

Abstract In the complex field of fast changing market conditions and decreasing predictability of market development, a factory has to adapt its production capacities quickly and with minimal effort. This paper introduces a novel methodology, recommending the best method of capacity planning to change production volumes and to ensure optimal operation in a car factory with respect to several Key Performance Indicators (KPIs), e.g. OEE or energy efficiency. The plant's volume transformability will be increased by not only one, but rather many different possible responses to be analyzed by simulation: the so-called production variants. An algorithm determines the most dominant, pre-analyzed production variants from a data base and visualizes the responding strategies based on the predicted KPIs. After a user- and scenario-based weighting by a so-called Balanced Performance Indicator (BPI), an optimal production variant will be selected. Above all, this assistance system provides more variety and transparency for adapting capacities and supports the factory planning both for greenfields by improved testing und comparability of production variants and for brownfields by giving recommendations for action during run-time. Additionally, real measured data can be retrieved in a closed-loop into the data base to design a continuous learning system. The concept has been methodically implemented and validated in a virtual material flow simulation of an automotive body shop cell and is linked to a real demonstration facility at the Brandenburg University of Cottbus-Senftenberg. Test runs confirm significant saving potential in energy consumption and OEE losses for the given target capacity.

Academic research paper on topic "The Transformable Factory: Adapting Automotive Production Capacities"

Available online at www.sciencedirect.com

ScienceDirect

Procedia CIRP 41 (2016) 171 - 176

www.elsevier.com/looate/procedia

48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015

The transformable factory: adapting automotive production capacities

Thomas Creutznachera,b,*5 Ulrich Bergerb, Raffaello Leprattic, Steffen Lampartera

a Siemens AG, Corporate Technology, 80200 Munich, Germany b Brandenburg University Cottbus-Senftenberg, Chair of Automation Technology, Postbox 101344, 03013 Cottbus, Germany c Siemens AG, Divison Digital Factory, Postbox 4848, 90026 Nuremberg, Germany

* Corresponding author. Tel.: +49-911-895-3750; fax: +49-911-895-2151. E-mail address: thomas.creutznacher@siemens.com

Abstract

In the complex field of fast changing market conditions and decreasing predictability of market development, a factory has to adapt its production capacities quickly and with minimal effort. This paper introduces a novel methodology, recommending the best method of capacity planning to change production volumes and to ensure optimal operation in a car factory with respect to several Key Performance Indicators (KPIs), e.g. OEE or energy efficiency. The plant's volume transformability will be increased by not only one, but rather many different possible responses to be analyzed by simulation: the so-called production variants. An algorithm determines the most dominant, pre-analyzed production variants from a data base and visualizes the responding strategies based on the predicted KPIs. After a user- and scenario-based weighting by a so-called Balanced Performance Indicator (BPI), an optimal production variant will be selected. Above all, this assistance system provides more variety and transparency for adapting capacities and supports the factory planning both for greenfields by improved testing und comparability of production variants and for brownfields by giving recommendations for action during run-time. Additionally, real measured data can be retrieved in a closed-loop into the data base to design a continuous learning system. The concept has been methodically implemented and validated in a virtual material flow simulation of an automotive body shop cell and is linked to a real demonstration facility at the Brandenburg University of Cottbus-Senftenberg. Test runs confirm significant saving potential in energy consumption and OEE losses for the given target capacity.

© 2015 Publishedby ElsevierB.V. Thisis anopen access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the scientific committee of 48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015

Keywords: Factory; Flexibility; Simulation; Production capacity

1. Introduction

The technological demands of car manufacturing exist in the context of a complex area of rapidly changing market requirements and, therefore, a strong pressure to adapt. The numerous complex and dynamic impacts or turbulences affecting the company's capability are overlapping interdependent influences, which are classified into so-called internal and external change drivers [1]. Concerning the decreasing predictability of market development in terms of time and intensity of change, it is necessary to respond effectively and efficiently to both predicted and unpredicted events [2], which also implies more frequent adaptations of the production system. The challenge is to identify the requirements and to adjust quickly and with minimal effort,

which is illustrated particularly well by the example of transformable production capacities.

In order to meet current demand within a car's life cycle, the optimal operating mode of the production is at the upper limit of capacity utilization. Volatile and unpredictable capacity demand leads to a non-optimal capacity utilization, causing capacity control countermeasures to have limited effect. Low production utilization due to a bad order situation or disruptions in material flow is tantamount to economic losses, whereas a permanently overloaded production needs to be expanded. When attempting to optimize and adapt the maximum capacity to the required capacity, evaluating and comparing the factory's various options for adapting production volume are challenging tasks and always related to high engineering effort. In this respect, classic production targets have changed to a more sustainable evaluation through

2212-8271 © 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of the scientific committee of 48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015 doi:10.1016/j.procir.2015.12.138

resource, work and energy productivity. Due to a lack of tools, knowledge, experts and associated methodologies, this new targets are still not considered fully in production planning [3]. Therefore, new decision making tools are required to support the complex evaluation within production planning.

After discussing existing methods for adapting, planning and evaluating production capacities of an automotive flow shop (Chapter 2), this paper proposes a novel optimization approach that supports factory planning by introducing more transparency into the process of adapting capacities for both greenfields and brownfields. The spectrum of potential courses of action, the so-called planning, design or production variants, is typically generated and analyzed within a material flow simulation. This paper proposes a simulation-based and consistent evaluation of these variants with respect to several Key Performance Indicators (KPIs) and the identification of the "best" variant from the set of pre-analyzed variants to ensure optimal operation in a car factory (Chapter 3). This concept has been methodically implemented and validated by virtual material flow simulation of an automotive body shop cell and by a real demonstration facility at the Brandenburg University of Cottbus-Senftenberg (Chapter 4). Finally, the paper is concluded with a short summary and outlook (Chapter 5).

2. Capacity planning

Capacity planning is the study of transformable adaptations of capacity (Chapter 2.1). In the example of a body shop, the car factory can pursue different methods for implementing a long-term response to volatile capacity demand, and for adapting its capacity supply (Chapter 2.2). Within capacity planning, material flow simulation plays an essential role in generating various production variants (Chapter 2.3).

2.1. Transformability by capacity planning

Flexibility characterizes production systems adapting quickly and with minimal effort within the limits of a predefined area in order to adjust to new conditions, whereas transformability is a factory's potential to initiate and execute reactive or proactive changes outside a so-called flexibility corridor [1]. According to [4], these definitions affect the following three change dimensions: product, operation and capacity. The scientific and technical focus is usually on the product and operation dimension. For instance, the European project e-Custom deals with alternative manufacturing methods of customized cars [3]. This paper focuses on the less thoroughly explored capacity dimension, meaning a variation in the production volumes of different products to accommodate changes in demand, while remaining profitable [4]. Volume flexibility and transformability are affected at a fundamental level by both production planning and control; more specifically production control only optimizes the flexible capacity utilization during run-time [4], whereas the actual maximum capacity and transformability is affected by the production planning. Production planning is part of the so-called Product Lifecycle Management (PLM) or Digital Factory, which integrates all engineering processes and

applications of a car's product lifecycle in an integrated and holistic software solution. Capacity planning is an essential task of production planning and serves the first design of the maximum capacity (greenfield) and its future adaptions (brownfield) as well.

2.2. Methods of capacity planning

The ideas behind capacity adaptation can be effectively illustrated using the example of a body shop, which, organizationally, is located between the press and paint shop, grouping together all of the joining processes involved in producing a chassis, mostly by welding. In a modern body shop nearly all process steps are fully automated by industrial robots and organized into a cycled flow shop. Underbody and sides are typically produced by cell production and delivered to a line production for the chassis to be completed with many other supplied parts, followed by the so-called body-in-white assembly. The maximum possible quantitative capacity can be calculated as the product of the following factors [5]:

• Production time

• Intensity

• Capacitance

First, production time is related to the number of shifts, which can easily be adapted by e.g. weekend or night shifts. Second, the intensity is related to the inverse cycle time, which is constant for each station within each cell or line production. Facilities are always designed for minimum cycle time [6]. The cycle time is also influenced by the model mix, because of the amount of tool changeovers. Third, the capacitance is related to the number of deployed production units of different hierarchical levels (resource, cell, line or shop level). Adaptations that extend or reduce lines or that hold stations empty to be available for future use are of limited applicability and are followed by a so-called line balancing. A more often used way of adapting the capacitance is the duplication of entire cell or line productions. Totally independent cycle times and duplications across the production areas become possible when the cell production is decoupled from the line production with a flexible logistics system; this also supports platform-design approaches.

2.3. Discrete event simulation

Discrete event simulation is the standard tool for simulating dynamic material flows in discrete industries [7]. Conventional analytical methods are insufficient for multiple time-dependent stochastic influences; these can however be satisfactorily modeled and iteratively optimized within a simulation [8]. A simulation is examining systems that do not (yet) exist or real, existing systems without direct intervention into operation [9]. A simulation expert is still essential for even partially automated model generation [10]. The standard process used in simulation studies consists of preparation, execution und evaluation. In this process, multiple production variants are modeled, parametrized and tested until the simulation targets are achieved, e.g. the desired production

capacity. Transformability can be conceptually integrated into the evaluation of the production variants. Common standardized KPIs include the throughput, which describes the effectiveness of a production variant; efficiency or productivity on the other hand can be determined by the energy consumption per unit produced and the so-called Overall Equipment Effectiveness (OEE), which includes losses in availability, performance and quality [11].

3. Concept

This chapter outlines a functional and mathematical concept for supporting production planners in decision making during capacity planning in both greenfield and brownfield projects. A selection of production variants for capacity planning generated by simulation is taken as a starting point, combining the various different methods of capacity adaptation (Chapter 3.1). The variants are evaluated with respect to appropriate KPIs (Chapter 3.2) and compared to identify the "best" option (Chapter 3.3). The functional architecture developed to achieve this is shown in Figure 1 and links the process operations of:

• Discrete event simulation (1-4)

• Recommendation system (I-VI)

• Production (A-B)

In production environments, IT systems are typically used to monitor the state of production (A) and to relay instructions to the production processes (B).

The Graphical User Interface (GUI) displays the current state of production and the current KPIs to the production planner (I). If these indicators are unsatisfactory, the user can specify the desired KPI weighting or a new target capacity. For greenfield projects, (I) is omitted. The comparison module queries the data base for production variants that match the target capacity (III). These are relayed back to the module (IV) and subsequently compared with respect to the chosen KPIs and weighting factors (see Chapter 3.4). The most dominant variants are displayed to the user and visualized (V). The user will be able to use this information as a basis for studying the various variants and for performing an external cost analysis. Subsequently, the user can request additional simulation experiments, for example to obtain further details regarding a promising variant, to conceptualize new, unconsidered variants, or to make a final decision for production adjustment (VI). To facilitate the design of a self-learning system, measured KPI values that deviate from the predicted values in the current production variant can be transmitted to the data base via (A), (I) and (III) for future reference and comparison.

3.1. Production variants

When preparing the simulation, the simulation expert prepares a diverse selection of production variants Vi where i £ N (see Chapter 3.2) based on the current state of production. All production variants are described by a single model through various parametrizations. They are then delivered to the simulation (1) and statistically evaluated using KPIs (2) (see Chapter 3.3). If the simulation results are not satisfactory, the process can be iteratively repeated until the target parameters have been optimized (3). Previous approaches to this problem only designed one single production variant by this method, but with this approach, a diverse selection of alternative variants from various stages of the iterative processes can be retrieved and stored in a data base, together with data concerning their evaluation (4).

The simulation is initiated with data from production planning, and so for brownfield projects corresponds to the current state of production. These variants are described by parametrizations in a single model and constructed by mapping the presented methods onto the simulation data base [9] and the various different levels in the production system hierarchy [12]. This produces a three-dimensional space (see Figure 2), whose content uniquely describes the parametrization of one of the individual production variants, and which is subsequently relayed to the simulation environment. For example, the capacitance (method) of a production line (hierarchy) can be saved as two parallel lines (content) in the material flow data (data mapping) and cycle times of 50 and 25 seconds (content) in the production data (data mapping).

Fig. 1. Functional architecture.

Fig. 2. Dimensions of variants.

3.2. Simulation evaluation oof variants

Based on [11], the n production variants V where i E [1;n] prepared by the simulation expert are evaluated in a Monte-Carlo simulation by fevaiuaiton according to the following KPIs (Equation 1):

• Availability Asim

• Performance Psim

• Quality QSim

• Energy Consumption ESim

• Throughput Tsim

The first innovative aspect of this procedure is the consideration of the OEE as a time-dependent parameter for the evaluation of entire factory units through summation of the respective time components of the m resources Rj where j G [1;m]. Energy considerations in the planning phase are increasingly the object of scientific research, and so integrating these into the consideration of various KPIs is also a novel aspect of this process. Excluding the throughput Tsim, all of the above quantities are normalized (Equation 2).

fevaluation (Vi ) KPIsim(Vi )

= {Asim (Vi ), Psim (Vi X Qsim (Vi X Esim (Vi X Tsim (Vi )} AimV ), PsimV ), QsimCK ), E^ ) e[0;1]

(1) (2)

Tsim(Vi ) = AOsink (Vi )-{TT(Vi ))"1

OEEsimVi ) = Asim(V ) • PsimVi ) QmV )

Total time (365 days x 24 hrs)

Total operations time not scheduled

Running time (Production time) Availability loses

Theoretical output

Actual output Performance loses

Good Quality Quality loses

Fig. 3. OEE loses.

Equation 5 allows the availability factor Asim to be calculated. The total operation time TOT is specified in the simulation and already includes planned downtime. The running time RT is the difference between the TOT and the availability losses. Availability losses arise from unplanned maintenance, cleaning, line constraints, organizational/idle periods and technical disruptions. For simplicity, disruptions greater-than-or-equal 1 minute are considered technical disruptions, whereas disruptions less than 1 minute are included as performance losses.

AsimVi ) = SRTÄj(Vi ) -I ZTOTr (Vi )

1=1 1 (m 1

Equation 6 can be used to calculate the performance factor Psim. The actual output AO is the difference between the theoretical output TO and the performance losses. The theoretical output TO is considered over the running time RT. Performance losses can arise due to reduced speeds of operation, minor stoppages and changeovers. Only disruptions less than 1 minute are included in this category of losses.

P (V ) =

simy i >

S ÄOR (Vi ). 1=1 1

ZTOSj(Vi ) 1=1 1

The simulated throughput Tsim describes the effectiveness of the production variant at the actual capacity achieved (Equation 3). The total time TT is the total duration of the simulation and is constant for all variants, set to 1 year. The actual output AO is obtained as a measurement from a counter placed at the sink of material flow.

The quality factor Qsim is calculated with Equation 7. Measurements of the good quality GQ are taken at so-called quality gates at the end of a production line. For each station, appropriate normally-distributed quotas need to be determined for both scrap and rework. The products selected for reworking are repaired away from the main production chain and then reintroduced into the sequence of processes. Rework is removed completely from the production. The number of quality gates and the duration of the quality checks are related to the throughput.

The KPIs availability, performance and quality describe various types of loss and thus the efficiency of a production variant (see Figure 3). They are included as multiplicative factors into the OEE^m where OEEsim(Vi) £ [0;1] (Equation 4).

Qsim(Vi ) = ZGQr V ) -I S AOr (Vi ) j=i 1 j 1

Energy consumption can be modeled using either data provided by the machine manufacturer, or using in-house measurements. For each station Rj, depending on the production variant with a given set of machine parameters (e.g. cycle time), the energy load profile at an effective power PR_j(Vi) can be established in the simulation. Integrating over time, Equation 8 yields the absolute energy consumption Eabsoiute per unit for a given production variant. This value is normalized with respect to the most energy-intensive variant as shown in Equation 9, giving E^.

Eabsolute(Vi ) = 2 J Pr (Vi ) dt ^J^i ^ 1=1 t=0 1

Esim (Vi ) = Eabsolute (Vi ) ' (maX(E

absolute (Vi ))

3.3. Selection and comparison of variants

4.1. Real experimental setup

The recommendation system for supporting simulation-based production planning can be manually triggered in the GUI if the KPIs of the current capacity are not satisfactory, or to adjust production to meet new capacity requirements. The system performs the two primary functions of variant selection fselection and variant comparison fompariso„.

Variant selection fselection begins as shown in Equation 10 with the transmission and mapping of the current production variant Vc from the production area onto a pre-existing variant in the data base. If there is no such variant available in the form defined in Chapter 3.1, it will be inserted as a new entry. To facilitate the design of a self-learning system, the real normalized KPIs KPI£Vc) of the actual variant are written into the data base (Equation 11). Next, the wanted set of variants {Vw} that achieve the user-defined target throughput TGUI within a deviation of 5 % are returned from the data base to the recommendation system, together with their simulated or real KPIs {KPIim/AVw)}.

f selection (Vc , KPIr (V ), Tguj ) = Vw },{KPIsim, „ (V„ )} (10)

KPir (V- ) = {Ar (V- ), Pr (V- ), Qr (V- ), Er (V- ), Tr (V- )}

During variant comparison fco

the set of selected

({VJ,{KPIsim! r (Vw )}, Wa, Wp , Wq , We ) = V (12)

WA + wp + Wq + WE = 1

BPI (Vw) = wA ■ Asm r (Vw ) + Wp ■ Psim/r (Vw )

+ WQ ■ Qsim/ r (Vw ) + WE ■ (1 - Esim/r (Vw ))

4. Implementation and validation

In order to validate the concept and to illustrate the displayed methods of capacity adaption, a real demonstration facility has been implemented at the Brandenburg University of Cottbus-Senftenberg (Chapter 4.1), which is linked to a discrete event simulation (Chapter 4.2). The results of the prototypic realization are depicted within a recommendation system (Chapter 4.3).

A real experimental setup has been developed with which the automated assembly of a body shop from the automotive industry can be well depicted. An exemplary semi-finished product is assembled from four parts. The setup mainly consist of two industrial robots and an automated-turntable (see Figure 4). The turntable is manually loaded and supplies the robots. An OPC interface transfers the current state of the production cell and real-time monitored KPIs to the recommendation system. In reverse, control instructions can be reported back to the PLC (Hardware-in-the-Loop).

variants {Vw} are weighted according to Equation 12 and their real and simulated KPIs {KPIsim/r(Vw)}. The weighting factors are chosen by pre-defined scenarios (e.g. energy-saving or loss-free), subject to the constraint shown in Equation 13. The variants are weighted using the so-called Balanced Performance Indicator BPI where BPI(Vw) £ [0,1] (Equation 14) [13]. The variants with the highest BPI-values are visualized with the expected values of the KPIs and their standard deviations. The user then chooses the best variant Vb; one possible choice is for example to select max(BPI(Vw)).

Fig. 4. Real experimental setup.

The collection and analysis of energy data is done by a energy meter. A visualization application allows the choice between the different scenarios and shows energy data as well as further KPIs such as OEE or throughput. In total, there are four different production scenarios that are realized by different PLC and robot control programs, showing normal operation state (V1), higher operational production speed (V2), two shift operation (V3) and simultaneous assembling (V4). The starting condition of the demonstrator is V1 with a cycle time of 87 seconds at reduced machine speed. V2, V3 and V4 duplicates the capacity and throughput.

4.2. Simulation

The program Siemens Plant Simulation V11 has been used to rebuild the presented facility and production variants in a material flow simulation. In the used model, two robots are supplied by a material source and buffer representing the turntable (see Figure 5). Afterwards, the assembled parts end up in a second buffer and the material sink. Data concerning disruptions, quality and energy are given to each station. The displayed methods and scenarios are build in a single model and corresponds to different parametrizations (see Table 1).

h} • Source Buffer 1

Turntable

Fig. 5. Simulation model. Table 1. Parameter of production variants.

1 Variant description Robots Cycle time Shifts

1 Normal operation 1 87 s 1

2 Half cycle time 1 43.5 s 1

3 Two shifts 1 87 s 2

4 Second robot 2 87 s 1

4.3. Results and recommendation

Table 2 shows the measured KPIs of V1, representing the current state of production, and the simulated KPIs of V2 to V4. Therefore, the target throughput is 662 parts per day with a deviation of 5 %. V2 achieves higher performance to gain half cycle time though higher machine speed. The high energy consumption is due to the exponential relationship between the robot's machine speed and it's energy consumption [14]. The quality is slightly lower because of inaccurate positioning at higher speeds. While V3 is not showing abnormalities, V4 is characterized by a lower energy consumption, which results from the common use of the turntable. Availability and performance are decreased because of the common protection area and the higher coordination effort for two robots. Due to those losses, the total throughput does not achieve the target throughput.

Table 2. Results.

i KPI,im/r T [parts/d] A [%] P [%] Q [%] E [%]

1 KPIr 331 92 30 90 74

2 KPI,im 662 92 72 87 100

3 KPI,im 662 92 30 90 74

4 KPI,im 568 85 28 90 69

By the choice of a suitable weighting, the user is recommended to select a variant from the set {Vw} = {V2,V3,V4}. Using Equation 13, an energy-saving weighting where wA = 0.2, wP = 0.2, Wq = 0.2, wE = 0.4 gives max(BPI(Vw)) = BPI(V3) = 53 % with a 26 % lower energy consumption than V2, whereas a loss-free weighting where wA = 0.3, wP = 0.3, Wq = 0.3, wE = 0.1 gives max(BPI(Vw)) = BPI(V2) = 75.3 % with a more than doubled performance compared to V3 or V4. In summary, saving potentials regarding energy consumption and OEE losses should always be seen in the context of the user-based weighting. The displayed evaluation of variants enables an optional effort examination, which should consider costs for energy, human resources or additional machines. After choosing a variant, it can be executed at the demonstrator.

5. Summary and outlook

After a scientific classification of the topic, a conceptual planning and optimization system has been introduced in this paper, that enables a comparison of various capacitative adaptions in the production planning to change production volumes. Therein, a discrete event simulation evaluates production variants by the following KPIs: throughput, energy consumption, availability, performance and quality. For a target capacity, an integrated recommendation system chooses suitable variants from a data base and runs a comparison with a user- and scenario-based weighting. To design a self-learning system, the preferred variant can be executed and the measured KPIs will be transmitted back to the data base. The concept is practicable both for greenfields and brownfields and increases the variety and transparency for adapting capacities. It has been implemented and validated in a real demonstration facility and a material flow simulation. Test runs confirm significant saving potential in energy consumption and OEE losses for the given target capacity.

The next steps will be to extend the simulation to an entire body shop with more variants, more hierarchical levels and combinable methods of capacity adaption. A substantial progress will also be achieved in an automatic variant design from the simulation's degrees of freedom to fill the data base.

References

[1] Nyhuis P, Fronia P, Pachow-Frauenhofer J, Wulf S. Wandlungsfähige Produktionssysteme. Ergebnisse der BMBF-Vorstudie "Wandlungsfähige Produktionssysteme"; 2009.

[2] Reinhart G. Flexibilität und Wandlungsfähigkeit von Fabriken im globalen Wettbewerb. 5. Deutsche Fachkonferenz Fabrikplanung 2004.

[3] Minhas SUH, Berger U. A web-based smart interface system to support automated generation of LCA simulation models, 47th Hawaii International Conference on System Science (HICSS); 2014.

[4] Wiendahl HP, et al. Changeable manufacturing classification, design and operation. Annals of the CIRP 56; Vol. 2; 2007. p. 783-809.

[5] Fischäder H. Störungsmanagement in netzwerkförmigen Produktionssystemen. Wiesbaden: Gabler; 2007.

[6] Lepratti R, Berger U, Creutznacher T, Minhas SUH. Dynamic cycle times for adaptive manufacturing control in automotive flow shops. XXIV International Conference on Information, Communication and Automation Technologies (ICAT); 2013.

[7] Banks J, Carson JS, Nelson BL, Nicol DM. Discrete event system simulation, 5th ed. Prentice Hall: Upper Saddle River; 2010.

[8] Kuhn A, Reinhart G, Wiendahl HP. Handbuch Simulationsanwendungen in Produktion und Logistik. Munich: Carl Hanser; 1997.

[9] Verein Deutscher Ingenieure. Simulation of systems in materials handling, logistics and production: Fundamentals; Düsseldorf: Beuth; 2010.

[10] Straßburger S, Bergmann S, Müller-Sommer H. Modellgenerierung im Kontext der Digitalen Fabrik: Stand der Technik und Herausforderungen. Karlsruhe: KIT Scientific Publishing; 2010.

[11] ISO/WD 22400-2. Manufacturing operations management: Key performance indicators: Part 2: Definitions and descriptions of KPIs; 2012.

[12] IEC CEI 62264-3. Enterprise-control system integration: Part 3: Activity models of manufacturing operations management; 2007.

[13] Creutznacher T, Berger U, Lepratti R, Lamparter S. Situationsadaptive Fertigung Steigerung der Flexibilität automatisierter Fertigungsprozesse in der Automobilindustrie. VDI Automation; 2014.

[14] Creutznacher T, Lepratti R, Lamparter S, Heinecke G. Performance adaptive manufacturing processes in an energy efficient car production. 11th Global Conference on Sustainable Manufacturing (GCSM); 2013.