- Email: [email protected]
Design and operation of manufacturing networks for mass customisation
CIRP Annals - Manufacturing Technology 62 (2013) 467–470
Contents lists available at SciVerse ScienceDirect
CIRP Annals - Manufacturing Technology jou rnal homep age : ht t p: // ees .e lse vi er . com /ci r p/ def a ult . asp
Design and operation of manufacturing networks for mass customisation Dimitris Mourtzis (2)*, Michalis Doukas, Foivos Psarommatis Laboratory for Manufacturing Systems and Automation, Department of Mechanical Engineering and Aeronautics, University of Patras, Patras 26500, Greece
A R T I C L E I N F O
A B S T R A C T
Keywords: Manufacturing system Complexity Mass customisation
The mass customisation paradigm, in combination with the volatility of globalised heterogeneous markets, directly affects industries towards realising efficient manufacturing network configurations. This research work aims to support the design and operation of manufacturing networks based on a multiobjective decision-making and simulation approach. The alternative network designs are evaluated through a set of multiple conflicting criteria including dynamic complexity, reliability, cost, time, quality and environmental footprint. Moreover, the impact of demand volatility to the operational performance of these networks is investigated through simulation. The proposed approach is validated through a real life case acquired from the CNC machine building industry. ß 2013 CIRP.
1. Introduction Original Equipment Manufacturers (OEMs) operate in highly competitive, volatile markets, with fluctuating demand, increasing labour costs in developing countries, and new environmental regulations [1]. Driven by the ever increasing need to reduce cost and delivery times, OEMs are called to efficiently overcome these issues by designing and operating sustainable and efficient manufacturing networks. The selection of optimum manufacturing network configurations that satisfy these challenging objectives however, is a proven data-intensive, NP-hard problem [2]. Therefore, strategic level decision-making cannot be accurately performed solely based on the experience and past knowledge of a supply chain manager [3]. To support the decision process, this research work proposes a method for the design and operation of highly efficient modern manufacturing networks operating under demand fluctuations, economic and environmental constraints. 2. State of the art The manufacturing landscape is nowadays more complex and dynamic than ever, due to the consequences of globalisation and recent economic recession [4] among other. The design of the manufacturing network of cooperating companies and its operation are key decisions for companies in order to endure competition. The push-pull model followed by many modern industries in order to address the product personalisation requirements [1] has driven the development of postponement strategies [5]. Product and process variety has also been identified as an enabler towards improving the customer-perceived value [6]. Likewise, recent environmental directives consist of additional constraints when designing and managing supply chains [7]. Moreover, a large number of recent publications deals with the emerging aspects of increasing complexity of manufacturing activities and the dynamic nature of supply chains [8]. The * Corresponding author. 0007-8506/$ – see front matter ß 2013 CIRP. http://dx.doi.org/10.1016/j.cirp.2013.03.126
importance of managing the complexity in supply chains is indicated in [9] as the study depicted that lower manufacturing network complexity is associated with reduced costs and overall network performance [10]. Thus, complexity should be considered as a cost criterion that has to be minimised. Complexity has been studied through approaches based on information theory, time series analysis, axiomatic theory, coding systems for machines and products, and through methods inspired by fluid dynamics [11]. In this landscape, the industrial sector of CNC machine building is facing challenges. As far as Europe is concerned, the CNC sector accounts for 158,000 jobs spread over 1474 companies with a worth of 17,512 billion s. Thus, it makes a key contribution to the economy and balance of payments [12]. For producing highly customised products, flexible and configurable machines are required for providing the capability to perform diversified manufacturing tasks in the concept of mass customisation. The proposed method aims to the timely identification of optimum or near optimum manufacturing network configurations for the production of heavily customised CNC machines. Both static and dynamic characteristics of the system under diversified market demands are captured, when examining the performance of different manufacturing network configurations. A complexity measure is incorporated in the decisionmaking process. Moreover, the network reliability is considered during the manufacturing network design, along with criteria of cost, time, quality and environmental footprint. Finally, the applicability of the approach is validated via a real life case study with data acquired from a CNC machine building industry. 3. Design and operation of manufacturing network method 3.1. Modelling of the manufacturing network A modern manufacturing network is composed of cooperating OEM plants, suppliers and dealers that produce and deliver final products to the market. In this research work, Decentralised Manufacturing Network (DMN) and Centralised (CMN)
D. Mourtzis et al. / CIRP Annals - Manufacturing Technology 62 (2013) 467–470
3.2. Description of the design and operation method
Weeks
Volatile 6 Normal ( =2.5, =1.5) 5 4 3 2 1 0 Weeks
Orders
The decision-making procedure used for the identification of optimum manufacturing network configurations is based on resource-task assignment decisions. The process is based on 6 steps: (i) formation of alternatives, (ii) determination of criteria to satisfy objectives, (iii) definition of criteria weights, (iv) calculation and normalisation of criteria values, (v) calculation of utility value, and (vi) selection of alternative with the highest utility value [19]. A manufacturing network configuration alternative is defined as a set of partner-task assignments, capable to manufacture a customised product within a manufacturing network structure. Two methodologies are used in the decision-making process for the generation and evaluation of manufacturing network alternatives, namely the Exhaustive Search Algorithm (EXS) and the Intelligent Search Algorithm (ISA). The EXS is an enumerative method, whereas ISA generates a subset of the Total Number of Alternative (TNA) manufacturing network configurations through 3 adjustable control parameters (Fig. 1A) [13,14]. The control parameters are: the Selected Number of Alternatives (SNA) that defines the breadth of the search, the Decision Horizon (DH) that controls the depth of the search and the Sampling Rate (SR) that guides the search to high quality branches in the tree of alternatives. The required number of experiments (obtained through a suitable orthogonal array), the optimum values for these three factors, as well as their influence on the utility value, are obtained through a Statistical Design of Experiments (SDoE) [15] and were identified at SNA = 100, DH = 3 and SR = 10. The workflow of the method is depicted in Fig. 1.
Seasonal 4+3sin(0.5x+ /3)
10 9 8 7 6 5 4 3 2 1 0
Orders
Orders
configurations are modelled and their performance is compared. As opposed to traditional CMNs, a DMN is structured upon more flexible relations between these manufacturing entities. Assembly tasks in DMN can be performed by suppliers and dealers, in proximity to the final customer, thus leading to decreased transportation and environmental costs, as depicted in [13].
Orders
468
30 25 20 15 10 5 0
7 6 5 4 3 2 1 0
Rapidincrease 0.01x2
Weeks
Benchmarking Uniform (a=0, b=5)
Weeks
Fig. 2. Demand profiles used in the simulation experiments.
be noted here that the number of orders controlled by these statistical distributions respects the capabilities of the system. The ratio p = demand requirements/system manufacturing capacity is maintained below 1, in order not to lead the system to unstable behaviour [16]. 3.3. Criteria of the multi-objective decision-making process The criteria for determining the performance of an alternative manufacturing network configuration are the following: Production and Transportation Cost (PTC): the sum of the production cost (PC) for manufacturing network partner i to perform task k and of the transportation cost (TC) from partner i to partner j, where i, j, k 2 N, i = 0,1,. . .,I, j = 0,1,. . .,J and k = 0,1,. . .,K [13]. PTC ¼
J I X K I X X X PCik þ TCi j i
i
k
ðsÞ
j
Lead Time (LT): the sum of processing and setup time (PT) for partner i to perform task k and of the transportation time (TT) from partner i to partner j [17,20]. LT ¼
J I X K I X X X PTik þ TTi j ðdaysÞ i
i
k
j
Energy Consumption (EC): sum of energy consumption (EP) for partner i to perform task k and of the transportation energy (ET) required from partner i to partner j [18].
EC ¼
J I X K I X X X EPik þ ETi j ðJÞ i
i
k
j
CO2 emissions (CO): the emitted tonnes of CO2 for the transportation (CE) required from partner i to partner j [18]. CO ¼
J I X X CEi j ðtonnes CO2 Þ i
Fig. 1. Workflow of the proposed design and operation method.
The above procedure is followed for both the DMN and the CMN configurations. The 10 best identified alternatives from the EXS (based on their utility value Ui) (Fig. 1B) for these two manufacturing network types are automatically modelled and simulated against a set of demand profiles. Totally, 80 simulation experiments were conducted (20 alternative network configurations 4 demand profiles) (Fig. 1C). The simulation period was one year and the profile used for the generation of the number of orders per week are: (i) seasonal demand (sinusoidal), (ii) rapid demand increase (parabolic), (iii) volatile demand (Gaussian) and (iv) benchmarking (uniform) (Fig. 2). The results of simulation are used for the calculation of dynamic complexity (Fig. 1D). It should
j
Quality (Q): the mean quality of the partners of an alternative manufacturing network configuration [19]. PI QLi Q¼ i I Reliability (R): total reliability, where s represents a serial and p a parallel resource [20], s, p 2 N, s = 0,1,. . .,S and p = 0,1,. . .,P. Rstot ¼
S Y
Rs ; for serial resources
s
Rptot ¼ 1
R Y
ð1 R p Þ; for serial resources; for parallel resources
r
Dynamic Complexity (CLZ): expressed as the unpredictability of the flowtime timeseries (Fig. 3A). CLZ is calculated through a
D. Mourtzis et al. / CIRP Annals - Manufacturing Technology 62 (2013) 467–470
469
assembled and transferred between different partners, eventually delivering the customised CNC machine to the customer. The machine consists of 7 components (4 standard and 3 customisable), each one consisting of a set of subcomponents. The TNA for this customised product configuration is calculated at 83,980,800 for the DMN and at 3,110,400 for the CMN. 6. Results and discussion
4. Software tool implementation The proposed methodology is implemented into a web-based software tool for the Design and Planning of Manufacturing Networks, namely the DPMNß, where a digital simulator is also integrated. The algorithms for the generation and evaluation of the alternative manufacturing network configurations are programmed in the JAVATM framework, following the Software as a Service (SaaS) architectural pattern. Moreover, the integrated simulator exchanges data through customised XML files, generated on the fly. The data model is implemented in the Oracle 11i relational database. The experiments were performed on a typical IntelTM i7 3.4 GHz powered workstation, with 8 GB of RAM. 5. Industrial case from the CNC machine building sector The network of the industrial case study comprises a number of partner facilities (OEM, Supplier and Distribution Centre plants) that cooperate, in order to manufacture and deliver the customised CNC machine to the customer. The modelling of the networks follows the CMN and DMN approach, as described in [13]. The total number of cooperating partners is 19, and the total number of individual resources is 76. The customised Bill of Materials (BoM) of the CNC machine is depicted in Fig. 4, together with the Bill of Processes (BoP) of the case study. The BoP represents the sequence of operations for the production of a CNC machine. Components are
Fig. 4. BoM and BoP of the CNC machine building industrial case study.
4,435,531
EXS DMN
184
0.856
0.853
290
0.868
166,515
1E+7 1E+6 1E+5 1E+4 1E+3 1E+2 1E+1 1E+0
Computation Time (ms)
0,88 0,88 0,87 0,87 0,86 0,86 0,85 0,85 0,84
0.874
Lempel-Ziv analysis of the timeseries obtained through simulation (Fig. 3B). Firstly, the average value of the flowtime values of the timeseries obtained through simulation, is calculated. The values of the flowtime that are equal or exceed the average value are encoded with the value 1, and with 0 otherwise (Fig. 3C). The string of 0 s and 1 s is used for the analysis of the complexity (Fig. 3D). LZ has been proven an effective means for the measurement of the Kolmogorov complexity. According to Kolmogorov [21], the complexity of a given string of 0 s and 1 s is given by the number of bits of the shortest computer program which can generate this string. The Lempel-Ziv algorithm identifies the patterns in the 0–1 string, in order to estimate the complexity [22]. The dynamic complexity CLZ of the manufacturing network is affected by the following parameters: number and size of buffers, resource cycle time, Mean Time Between Failures (MTBF), Mean Time To Repair (MTTR) and from the demand to manufacturing capacity ratio (p) described in Section 3.2. Therefore, the dynamic complexity CLZ is a 5-tuple: CLZ = f(buffers, cycle time, MTTR, MTBF, p ratio).
Utility Value
Fig. 3. Dynamic complexity (CLZ) assessment methodology.
Based on the utility values obtained from the computer experiments for the EXS and ISA, DMN depicts a clear advantage over the CMN (Fig. 5). The calculated criteria values for the DMN and CMN are depicted in Table 1.
ISA EXS ISA DMN CMN CMN
Fig. 5. Comparison of the utility value and computation time of EXS vs. ISA for DMN and CMN. Table 1 Criteria values for DMN and CMN obtained through EXS and ISA. Criteria
Units
DMN EXS
DMN ISA
CMN EXS
CMN ISA
PTC LT EC CO R Q
(s) (days) (MJ) (tonne CO2) – –
168,238 53.79 3.14080 942,979 0.8909 76.73
168,326 54.38 3.07968 953,512 0.8890 74.38
168,215 53.71 3.13800 941,712 0.8815 77.6
167,505 54.04 3.13800 947,338 0.8875 76.8
The ISA results are 2.4% worse that the EXS results for the DMN case requiring however, 15,295 times less computation time. The ISA results are of high quality when compared to the EXS results, belonging to the top 1% of the highest utility value solutions (Fig. 6). Thus, ISA has an evident advantage over the EXS. Analogous is the trend for individual criteria values (Fig. 7). DMN values are better than CMN, in most cases, because DMN allows the formation of alternatives where suppliers can perform assembly tasks, thus, reducing the total transportation distance, which leads to lower costs, CO2 emissions and energy needs. Furthermore, the operational performance of these networks is investigated through their simulation over the demand profiles presented in section 3.2. The simulation is performed for 52 weeks, which is 6240 working hours (three 8 h shifts/day, 5 days per week). In Fig. 8, the complexity values calculated from processing the simulation results are illustrated. The experiments were performed for the 10 best CMN and DMN manufacturing network configurations identified by the EXS and for 10 runs of the ISA using the parameters indicated by SDoE. The complexity values of the DMN are lower than those of CMN, the latter being 0.062 lower at handling predefined market demand (seasonal). In a DMN, the number of transportation tasks, which act as buffers with processing time, is reduced due to the fact that assembly tasks may be performed at supplier or dealer sites. As the buffer characteristics is a variable affecting complexity, this leads to reduced values of CLZ. The ISA results moreover, are of high quality when compared to the EXS results. Furthermore, the same trend is observed in the complexity values for all three cases. Complexity increases from 0.2 to 0.4 for deterministic demands (seasonal and rapid increase) to 1.0–1.2 for the volatile demands (Gauss and uniform distributions).
D. Mourtzis et al. / CIRP Annals - Manufacturing Technology 62 (2013) 467–470
470
% of the utility value of ISA results compared to the EXS 0.75- 1 0- 0.45 1% 9% 0.650.75 0.45- 0.55 17% 31% 0.55 - 0.65 42%
Fig. 6. % of Utility value of the ISA results vs. EXS results.
DMNandCMN Leadtime (days) 51-55
4% 8% 38% 50%
55-59
results 11% 13%
24%
0.80-0.85
52%
59-62.5
Reliability 0.74-0.80
7%6%
CO2 emissions (tonnesCO 2) 1-6
35%
0.85-0.88
6-9
52%
9-13
0.89-1.00
62.5-68
the networks. Furthermore, DMN structures are more suitable than traditional CMNs for the needs of the demanding product personalisation environment. DMN configuration handles unexpected market requests more efficiently, as it functions on a more flexible manner, facilitating possible disturbances in excitations. In addition, the aggregated utility, which encapsulates the static and dynamic performance of the manufacturing network, verified that DMN is more suitable for today’s volatile environment. Finally, the ISA results were of high quality when compared to the EXS for all examined cases. Future work will focus on channelling order batches within the manufacturing network through the incorporation of production related constraints, effectively reducing the search space. Moreover, measures will be devised for estimating the structural manufacturing network complexity, with regards to the product characteristics. The relation between static and dynamic complexity will be thus investigated. Finally, a legacy Product Data Management system will be integrated for the automated extraction of the BoM of the product under investigation, effectively limiting the required human operator data entry task.
13-21
Acknowledgement Fig. 7. % of ISA vs. EXS results for individual criteria. Complexity vs demand proiles for DMN and CMN conigurations DMNEXS DMNISA CMNEXS DMNEXSmeanvalue DMNISAmeanvalue CMNEXSmeanvalue
The work reported in this paper has been partially supported by the EC funded project ‘‘e-CUSTOM-A web-based collaboration system for mass customization’’ (260067). References
1,2
Complexity
1,0 0,8 0,6 0,4 0,2 0,0
Seasonal
Rapi din crease Volatile Demand Profiles
Uniform
Fig. 8. Complexity for DMN and CMN for different demand excitations. Aggregated utility value vs. different demand proiles for CMN rapid seasonal uniform volatile
Utility Value (Uai)
1,0
Wu=0.5 Wc=0.5
0,8 0,6 0,4 0,2 0,0
Aggregated utility value vs. different demand proiles for DMN
Utility Value (Uai)
1,0 0,8 0,6 0,4
Wu=0.5 Wc=0.5
0,2 0,0 1 Best
2 3 4 5 6 7 8 10 bests alternatives from the EXS experiments
9
10
Fig. 9. Uai for different demand excitations, for CMN and DMN.
Additionally, the fusion of complexity together with the utility value (Ui) (Fig. 1D) yielded the aggregated utility (Uai) diagram presented for the CMN and DMN (Fig. 9). The weight values assigned to the Ui (Wu) and complexity (Wc) are 0.5. The results for Uai further validate the superiority of DMN over CMN. 7. Conclusions and outlook The presented method and tool can support strategic level decisions related to the design of efficient manufacturing network configurations. The incorporation of criteria of cost, time, environmental impact and quality encapsulated some of the most significant objectives that manufacturing industries are striving to achieve nowadays. Moreover, the inclusion of complexity as a decision making criterion depicted the operational characteristics of
[1] Simchi-Levi D (2010) Operation Rules, MIT Press Ltd, Cambridge, Massachusetts, USA. [2] Zhou G, Min H, Gen M (2002) The Balanced Allocation of Customers to Multiple Distribution Centers in the Supply Chain Network: A Genetic Algorithm Approach. Computers & Industrial Engineering 43(1–2):251–261. [3] Shapiro J (2001) Modelling the Supply Chain, Thomson Learning Inc, Duxbury. [4] European Commission, 2010, Impact of the Economic Crisis on Key Sectors of the EU: The Case of Manufacturing and Construction Industries. [5] Hu SJ, Ko J, Weyand L, El Maraghy HA, Kien TK, Koren Y, Bley H, Chryssolouris G, Nasr N, Shpitalni M (2011) Assembly System Design and Operations for Product Variety. CIRP Annals 60(2):715–733. [6] Daaboul J, Da Cunha C, Bernard A, Laroche F (2011) Design for Mass Customization: Product Variety vs. Process Variety. CIRP Annals – Manufacturing Technology 60(1):169–174. [7] Umeda Y, Takata S, Kimura F, Tomiyama T, Sutherland JW, Karaf S, Herrmann C, Duflou JR (2012) Toward Integrated Product and Process Life Cycle Planning— An Environmental Perspective. CIRP Annals – Manufacturing Technology 61(2):681–702. [8] Duffie NA, Roy D, Shi L (2008) Dynamic Modeling of Production Networks of Autonomous Work Systems with Local Capacity Control. CIRP Annals – Manufacturing Technology 57(1):463–466. [9] ElMaraghy W, ElMaraghy H, Tomiyama T, Monostori L (2012) Complexity in Engineering Design and Manufacturing. CIRP Annals – Manufacturing Technology 61(2):793–814. [10] Perona M, Miragliotta G (2004) Complexity Management and Supply Chain Performance Assessment. A Field Study and a Conceptual Framework. International Journal of Production Economics 90(1):103–115. [11] Chryssolouris G, Efthymiou K, Papakostas N, Mourtzis D, Pagoropoulos A (2013) Flexibility and Complexity: Is It a Trade-off? To be Published to the 50th Anniversary of the International Journal of Production Research. [12] European Commission, 2012, Keeping Europe Ahead in Machine Tool Manufacture Industries, http://ec.europa.eu/research/industrial_technologies. [13] Mourtzis D, Doukas M, Psarommatis F (2012) A Multi-criteria Evaluation of Centralized and Decentralized Production Networks in a Highly CustomerDriven Environment. CIRP Annals – Manufacturing Technology 61(1):427–430. [14] Chryssolouris G, Dicke K, Lee M (1992) On the Resources Allocation Problem. International Journal of Production Research 30(12):2773–2795. [15] Phadke MS (1989) Quality Engineering Using Robust Design, 1st ed. Englewood Cliffs, Prentice Hall, New Jersey, USA. [16] Papakostas N, Efthymiou K, Mourtzis D, Chryssolouris G (2009) Modeling the Complexity of Manufacturing Systems Using Non-linear Dynamics Approaches. CIRP Annals – Manufacturing Technology 58(1):437–440. [17] Liao CJ, Shyu CH (1991) An Analytical Determination of Lead Time with Normal Demand. International Journal of Operations & Production Management 11(9):72–78. [18] EPA, 2010, URL: www.epa.gov [19] Chryssolouris G (2006) Manufacturing Systems: Theory and Practice, 2nd ed. Springer-Verlag, New York. [20] Evans JR (1993) Applied Production and Operations Management, 4th Ed. West Publishing Company, St. Paul, MN, USA. [21] Kolmogorov AN (1965) Three Approaches to the Quantitative Definition of Information. Problems of Information and Transmission 1(1):1–7. [22] Kaspar F, Schuster HG (1987) Easily Calculable Measure for the Complexity of Spatiotemporal Patterns. Physical Review A 36(2):842–848.
Contents lists available at SciVerse ScienceDirect
CIRP Annals - Manufacturing Technology jou rnal homep age : ht t p: // ees .e lse vi er . com /ci r p/ def a ult . asp
Design and operation of manufacturing networks for mass customisation Dimitris Mourtzis (2)*, Michalis Doukas, Foivos Psarommatis Laboratory for Manufacturing Systems and Automation, Department of Mechanical Engineering and Aeronautics, University of Patras, Patras 26500, Greece
A R T I C L E I N F O
A B S T R A C T
Keywords: Manufacturing system Complexity Mass customisation
The mass customisation paradigm, in combination with the volatility of globalised heterogeneous markets, directly affects industries towards realising efficient manufacturing network configurations. This research work aims to support the design and operation of manufacturing networks based on a multiobjective decision-making and simulation approach. The alternative network designs are evaluated through a set of multiple conflicting criteria including dynamic complexity, reliability, cost, time, quality and environmental footprint. Moreover, the impact of demand volatility to the operational performance of these networks is investigated through simulation. The proposed approach is validated through a real life case acquired from the CNC machine building industry. ß 2013 CIRP.
1. Introduction Original Equipment Manufacturers (OEMs) operate in highly competitive, volatile markets, with fluctuating demand, increasing labour costs in developing countries, and new environmental regulations [1]. Driven by the ever increasing need to reduce cost and delivery times, OEMs are called to efficiently overcome these issues by designing and operating sustainable and efficient manufacturing networks. The selection of optimum manufacturing network configurations that satisfy these challenging objectives however, is a proven data-intensive, NP-hard problem [2]. Therefore, strategic level decision-making cannot be accurately performed solely based on the experience and past knowledge of a supply chain manager [3]. To support the decision process, this research work proposes a method for the design and operation of highly efficient modern manufacturing networks operating under demand fluctuations, economic and environmental constraints. 2. State of the art The manufacturing landscape is nowadays more complex and dynamic than ever, due to the consequences of globalisation and recent economic recession [4] among other. The design of the manufacturing network of cooperating companies and its operation are key decisions for companies in order to endure competition. The push-pull model followed by many modern industries in order to address the product personalisation requirements [1] has driven the development of postponement strategies [5]. Product and process variety has also been identified as an enabler towards improving the customer-perceived value [6]. Likewise, recent environmental directives consist of additional constraints when designing and managing supply chains [7]. Moreover, a large number of recent publications deals with the emerging aspects of increasing complexity of manufacturing activities and the dynamic nature of supply chains [8]. The * Corresponding author. 0007-8506/$ – see front matter ß 2013 CIRP. http://dx.doi.org/10.1016/j.cirp.2013.03.126
importance of managing the complexity in supply chains is indicated in [9] as the study depicted that lower manufacturing network complexity is associated with reduced costs and overall network performance [10]. Thus, complexity should be considered as a cost criterion that has to be minimised. Complexity has been studied through approaches based on information theory, time series analysis, axiomatic theory, coding systems for machines and products, and through methods inspired by fluid dynamics [11]. In this landscape, the industrial sector of CNC machine building is facing challenges. As far as Europe is concerned, the CNC sector accounts for 158,000 jobs spread over 1474 companies with a worth of 17,512 billion s. Thus, it makes a key contribution to the economy and balance of payments [12]. For producing highly customised products, flexible and configurable machines are required for providing the capability to perform diversified manufacturing tasks in the concept of mass customisation. The proposed method aims to the timely identification of optimum or near optimum manufacturing network configurations for the production of heavily customised CNC machines. Both static and dynamic characteristics of the system under diversified market demands are captured, when examining the performance of different manufacturing network configurations. A complexity measure is incorporated in the decisionmaking process. Moreover, the network reliability is considered during the manufacturing network design, along with criteria of cost, time, quality and environmental footprint. Finally, the applicability of the approach is validated via a real life case study with data acquired from a CNC machine building industry. 3. Design and operation of manufacturing network method 3.1. Modelling of the manufacturing network A modern manufacturing network is composed of cooperating OEM plants, suppliers and dealers that produce and deliver final products to the market. In this research work, Decentralised Manufacturing Network (DMN) and Centralised (CMN)
D. Mourtzis et al. / CIRP Annals - Manufacturing Technology 62 (2013) 467–470
3.2. Description of the design and operation method
Weeks
Volatile 6 Normal ( =2.5, =1.5) 5 4 3 2 1 0 Weeks
Orders
The decision-making procedure used for the identification of optimum manufacturing network configurations is based on resource-task assignment decisions. The process is based on 6 steps: (i) formation of alternatives, (ii) determination of criteria to satisfy objectives, (iii) definition of criteria weights, (iv) calculation and normalisation of criteria values, (v) calculation of utility value, and (vi) selection of alternative with the highest utility value [19]. A manufacturing network configuration alternative is defined as a set of partner-task assignments, capable to manufacture a customised product within a manufacturing network structure. Two methodologies are used in the decision-making process for the generation and evaluation of manufacturing network alternatives, namely the Exhaustive Search Algorithm (EXS) and the Intelligent Search Algorithm (ISA). The EXS is an enumerative method, whereas ISA generates a subset of the Total Number of Alternative (TNA) manufacturing network configurations through 3 adjustable control parameters (Fig. 1A) [13,14]. The control parameters are: the Selected Number of Alternatives (SNA) that defines the breadth of the search, the Decision Horizon (DH) that controls the depth of the search and the Sampling Rate (SR) that guides the search to high quality branches in the tree of alternatives. The required number of experiments (obtained through a suitable orthogonal array), the optimum values for these three factors, as well as their influence on the utility value, are obtained through a Statistical Design of Experiments (SDoE) [15] and were identified at SNA = 100, DH = 3 and SR = 10. The workflow of the method is depicted in Fig. 1.
Seasonal 4+3sin(0.5x+ /3)
10 9 8 7 6 5 4 3 2 1 0
Orders
Orders
configurations are modelled and their performance is compared. As opposed to traditional CMNs, a DMN is structured upon more flexible relations between these manufacturing entities. Assembly tasks in DMN can be performed by suppliers and dealers, in proximity to the final customer, thus leading to decreased transportation and environmental costs, as depicted in [13].
Orders
468
30 25 20 15 10 5 0
7 6 5 4 3 2 1 0
Rapidincrease 0.01x2
Weeks
Benchmarking Uniform (a=0, b=5)
Weeks
Fig. 2. Demand profiles used in the simulation experiments.
be noted here that the number of orders controlled by these statistical distributions respects the capabilities of the system. The ratio p = demand requirements/system manufacturing capacity is maintained below 1, in order not to lead the system to unstable behaviour [16]. 3.3. Criteria of the multi-objective decision-making process The criteria for determining the performance of an alternative manufacturing network configuration are the following: Production and Transportation Cost (PTC): the sum of the production cost (PC) for manufacturing network partner i to perform task k and of the transportation cost (TC) from partner i to partner j, where i, j, k 2 N, i = 0,1,. . .,I, j = 0,1,. . .,J and k = 0,1,. . .,K [13]. PTC ¼
J I X K I X X X PCik þ TCi j i
i
k
ðsÞ
j
Lead Time (LT): the sum of processing and setup time (PT) for partner i to perform task k and of the transportation time (TT) from partner i to partner j [17,20]. LT ¼
J I X K I X X X PTik þ TTi j ðdaysÞ i
i
k
j
Energy Consumption (EC): sum of energy consumption (EP) for partner i to perform task k and of the transportation energy (ET) required from partner i to partner j [18].
EC ¼
J I X K I X X X EPik þ ETi j ðJÞ i
i
k
j
CO2 emissions (CO): the emitted tonnes of CO2 for the transportation (CE) required from partner i to partner j [18]. CO ¼
J I X X CEi j ðtonnes CO2 Þ i
Fig. 1. Workflow of the proposed design and operation method.
The above procedure is followed for both the DMN and the CMN configurations. The 10 best identified alternatives from the EXS (based on their utility value Ui) (Fig. 1B) for these two manufacturing network types are automatically modelled and simulated against a set of demand profiles. Totally, 80 simulation experiments were conducted (20 alternative network configurations 4 demand profiles) (Fig. 1C). The simulation period was one year and the profile used for the generation of the number of orders per week are: (i) seasonal demand (sinusoidal), (ii) rapid demand increase (parabolic), (iii) volatile demand (Gaussian) and (iv) benchmarking (uniform) (Fig. 2). The results of simulation are used for the calculation of dynamic complexity (Fig. 1D). It should
j
Quality (Q): the mean quality of the partners of an alternative manufacturing network configuration [19]. PI QLi Q¼ i I Reliability (R): total reliability, where s represents a serial and p a parallel resource [20], s, p 2 N, s = 0,1,. . .,S and p = 0,1,. . .,P. Rstot ¼
S Y
Rs ; for serial resources
s
Rptot ¼ 1
R Y
ð1 R p Þ; for serial resources; for parallel resources
r
Dynamic Complexity (CLZ): expressed as the unpredictability of the flowtime timeseries (Fig. 3A). CLZ is calculated through a
D. Mourtzis et al. / CIRP Annals - Manufacturing Technology 62 (2013) 467–470
469
assembled and transferred between different partners, eventually delivering the customised CNC machine to the customer. The machine consists of 7 components (4 standard and 3 customisable), each one consisting of a set of subcomponents. The TNA for this customised product configuration is calculated at 83,980,800 for the DMN and at 3,110,400 for the CMN. 6. Results and discussion
4. Software tool implementation The proposed methodology is implemented into a web-based software tool for the Design and Planning of Manufacturing Networks, namely the DPMNß, where a digital simulator is also integrated. The algorithms for the generation and evaluation of the alternative manufacturing network configurations are programmed in the JAVATM framework, following the Software as a Service (SaaS) architectural pattern. Moreover, the integrated simulator exchanges data through customised XML files, generated on the fly. The data model is implemented in the Oracle 11i relational database. The experiments were performed on a typical IntelTM i7 3.4 GHz powered workstation, with 8 GB of RAM. 5. Industrial case from the CNC machine building sector The network of the industrial case study comprises a number of partner facilities (OEM, Supplier and Distribution Centre plants) that cooperate, in order to manufacture and deliver the customised CNC machine to the customer. The modelling of the networks follows the CMN and DMN approach, as described in [13]. The total number of cooperating partners is 19, and the total number of individual resources is 76. The customised Bill of Materials (BoM) of the CNC machine is depicted in Fig. 4, together with the Bill of Processes (BoP) of the case study. The BoP represents the sequence of operations for the production of a CNC machine. Components are
Fig. 4. BoM and BoP of the CNC machine building industrial case study.
4,435,531
EXS DMN
184
0.856
0.853
290
0.868
166,515
1E+7 1E+6 1E+5 1E+4 1E+3 1E+2 1E+1 1E+0
Computation Time (ms)
0,88 0,88 0,87 0,87 0,86 0,86 0,85 0,85 0,84
0.874
Lempel-Ziv analysis of the timeseries obtained through simulation (Fig. 3B). Firstly, the average value of the flowtime values of the timeseries obtained through simulation, is calculated. The values of the flowtime that are equal or exceed the average value are encoded with the value 1, and with 0 otherwise (Fig. 3C). The string of 0 s and 1 s is used for the analysis of the complexity (Fig. 3D). LZ has been proven an effective means for the measurement of the Kolmogorov complexity. According to Kolmogorov [21], the complexity of a given string of 0 s and 1 s is given by the number of bits of the shortest computer program which can generate this string. The Lempel-Ziv algorithm identifies the patterns in the 0–1 string, in order to estimate the complexity [22]. The dynamic complexity CLZ of the manufacturing network is affected by the following parameters: number and size of buffers, resource cycle time, Mean Time Between Failures (MTBF), Mean Time To Repair (MTTR) and from the demand to manufacturing capacity ratio (p) described in Section 3.2. Therefore, the dynamic complexity CLZ is a 5-tuple: CLZ = f(buffers, cycle time, MTTR, MTBF, p ratio).
Utility Value
Fig. 3. Dynamic complexity (CLZ) assessment methodology.
Based on the utility values obtained from the computer experiments for the EXS and ISA, DMN depicts a clear advantage over the CMN (Fig. 5). The calculated criteria values for the DMN and CMN are depicted in Table 1.
ISA EXS ISA DMN CMN CMN
Fig. 5. Comparison of the utility value and computation time of EXS vs. ISA for DMN and CMN. Table 1 Criteria values for DMN and CMN obtained through EXS and ISA. Criteria
Units
DMN EXS
DMN ISA
CMN EXS
CMN ISA
PTC LT EC CO R Q
(s) (days) (MJ) (tonne CO2) – –
168,238 53.79 3.14080 942,979 0.8909 76.73
168,326 54.38 3.07968 953,512 0.8890 74.38
168,215 53.71 3.13800 941,712 0.8815 77.6
167,505 54.04 3.13800 947,338 0.8875 76.8
The ISA results are 2.4% worse that the EXS results for the DMN case requiring however, 15,295 times less computation time. The ISA results are of high quality when compared to the EXS results, belonging to the top 1% of the highest utility value solutions (Fig. 6). Thus, ISA has an evident advantage over the EXS. Analogous is the trend for individual criteria values (Fig. 7). DMN values are better than CMN, in most cases, because DMN allows the formation of alternatives where suppliers can perform assembly tasks, thus, reducing the total transportation distance, which leads to lower costs, CO2 emissions and energy needs. Furthermore, the operational performance of these networks is investigated through their simulation over the demand profiles presented in section 3.2. The simulation is performed for 52 weeks, which is 6240 working hours (three 8 h shifts/day, 5 days per week). In Fig. 8, the complexity values calculated from processing the simulation results are illustrated. The experiments were performed for the 10 best CMN and DMN manufacturing network configurations identified by the EXS and for 10 runs of the ISA using the parameters indicated by SDoE. The complexity values of the DMN are lower than those of CMN, the latter being 0.062 lower at handling predefined market demand (seasonal). In a DMN, the number of transportation tasks, which act as buffers with processing time, is reduced due to the fact that assembly tasks may be performed at supplier or dealer sites. As the buffer characteristics is a variable affecting complexity, this leads to reduced values of CLZ. The ISA results moreover, are of high quality when compared to the EXS results. Furthermore, the same trend is observed in the complexity values for all three cases. Complexity increases from 0.2 to 0.4 for deterministic demands (seasonal and rapid increase) to 1.0–1.2 for the volatile demands (Gauss and uniform distributions).
D. Mourtzis et al. / CIRP Annals - Manufacturing Technology 62 (2013) 467–470
470
% of the utility value of ISA results compared to the EXS 0.75- 1 0- 0.45 1% 9% 0.650.75 0.45- 0.55 17% 31% 0.55 - 0.65 42%
Fig. 6. % of Utility value of the ISA results vs. EXS results.
DMNandCMN Leadtime (days) 51-55
4% 8% 38% 50%
55-59
results 11% 13%
24%
0.80-0.85
52%
59-62.5
Reliability 0.74-0.80
7%6%
CO2 emissions (tonnesCO 2) 1-6
35%
0.85-0.88
6-9
52%
9-13
0.89-1.00
62.5-68
the networks. Furthermore, DMN structures are more suitable than traditional CMNs for the needs of the demanding product personalisation environment. DMN configuration handles unexpected market requests more efficiently, as it functions on a more flexible manner, facilitating possible disturbances in excitations. In addition, the aggregated utility, which encapsulates the static and dynamic performance of the manufacturing network, verified that DMN is more suitable for today’s volatile environment. Finally, the ISA results were of high quality when compared to the EXS for all examined cases. Future work will focus on channelling order batches within the manufacturing network through the incorporation of production related constraints, effectively reducing the search space. Moreover, measures will be devised for estimating the structural manufacturing network complexity, with regards to the product characteristics. The relation between static and dynamic complexity will be thus investigated. Finally, a legacy Product Data Management system will be integrated for the automated extraction of the BoM of the product under investigation, effectively limiting the required human operator data entry task.
13-21
Acknowledgement Fig. 7. % of ISA vs. EXS results for individual criteria. Complexity vs demand proiles for DMN and CMN conigurations DMNEXS DMNISA CMNEXS DMNEXSmeanvalue DMNISAmeanvalue CMNEXSmeanvalue
The work reported in this paper has been partially supported by the EC funded project ‘‘e-CUSTOM-A web-based collaboration system for mass customization’’ (260067). References
1,2
Complexity
1,0 0,8 0,6 0,4 0,2 0,0
Seasonal
Rapi din crease Volatile Demand Profiles
Uniform
Fig. 8. Complexity for DMN and CMN for different demand excitations. Aggregated utility value vs. different demand proiles for CMN rapid seasonal uniform volatile
Utility Value (Uai)
1,0
Wu=0.5 Wc=0.5
0,8 0,6 0,4 0,2 0,0
Aggregated utility value vs. different demand proiles for DMN
Utility Value (Uai)
1,0 0,8 0,6 0,4
Wu=0.5 Wc=0.5
0,2 0,0 1 Best
2 3 4 5 6 7 8 10 bests alternatives from the EXS experiments
9
10
Fig. 9. Uai for different demand excitations, for CMN and DMN.
Additionally, the fusion of complexity together with the utility value (Ui) (Fig. 1D) yielded the aggregated utility (Uai) diagram presented for the CMN and DMN (Fig. 9). The weight values assigned to the Ui (Wu) and complexity (Wc) are 0.5. The results for Uai further validate the superiority of DMN over CMN. 7. Conclusions and outlook The presented method and tool can support strategic level decisions related to the design of efficient manufacturing network configurations. The incorporation of criteria of cost, time, environmental impact and quality encapsulated some of the most significant objectives that manufacturing industries are striving to achieve nowadays. Moreover, the inclusion of complexity as a decision making criterion depicted the operational characteristics of
[1] Simchi-Levi D (2010) Operation Rules, MIT Press Ltd, Cambridge, Massachusetts, USA. [2] Zhou G, Min H, Gen M (2002) The Balanced Allocation of Customers to Multiple Distribution Centers in the Supply Chain Network: A Genetic Algorithm Approach. Computers & Industrial Engineering 43(1–2):251–261. [3] Shapiro J (2001) Modelling the Supply Chain, Thomson Learning Inc, Duxbury. [4] European Commission, 2010, Impact of the Economic Crisis on Key Sectors of the EU: The Case of Manufacturing and Construction Industries. [5] Hu SJ, Ko J, Weyand L, El Maraghy HA, Kien TK, Koren Y, Bley H, Chryssolouris G, Nasr N, Shpitalni M (2011) Assembly System Design and Operations for Product Variety. CIRP Annals 60(2):715–733. [6] Daaboul J, Da Cunha C, Bernard A, Laroche F (2011) Design for Mass Customization: Product Variety vs. Process Variety. CIRP Annals – Manufacturing Technology 60(1):169–174. [7] Umeda Y, Takata S, Kimura F, Tomiyama T, Sutherland JW, Karaf S, Herrmann C, Duflou JR (2012) Toward Integrated Product and Process Life Cycle Planning— An Environmental Perspective. CIRP Annals – Manufacturing Technology 61(2):681–702. [8] Duffie NA, Roy D, Shi L (2008) Dynamic Modeling of Production Networks of Autonomous Work Systems with Local Capacity Control. CIRP Annals – Manufacturing Technology 57(1):463–466. [9] ElMaraghy W, ElMaraghy H, Tomiyama T, Monostori L (2012) Complexity in Engineering Design and Manufacturing. CIRP Annals – Manufacturing Technology 61(2):793–814. [10] Perona M, Miragliotta G (2004) Complexity Management and Supply Chain Performance Assessment. A Field Study and a Conceptual Framework. International Journal of Production Economics 90(1):103–115. [11] Chryssolouris G, Efthymiou K, Papakostas N, Mourtzis D, Pagoropoulos A (2013) Flexibility and Complexity: Is It a Trade-off? To be Published to the 50th Anniversary of the International Journal of Production Research. [12] European Commission, 2012, Keeping Europe Ahead in Machine Tool Manufacture Industries, http://ec.europa.eu/research/industrial_technologies. [13] Mourtzis D, Doukas M, Psarommatis F (2012) A Multi-criteria Evaluation of Centralized and Decentralized Production Networks in a Highly CustomerDriven Environment. CIRP Annals – Manufacturing Technology 61(1):427–430. [14] Chryssolouris G, Dicke K, Lee M (1992) On the Resources Allocation Problem. International Journal of Production Research 30(12):2773–2795. [15] Phadke MS (1989) Quality Engineering Using Robust Design, 1st ed. Englewood Cliffs, Prentice Hall, New Jersey, USA. [16] Papakostas N, Efthymiou K, Mourtzis D, Chryssolouris G (2009) Modeling the Complexity of Manufacturing Systems Using Non-linear Dynamics Approaches. CIRP Annals – Manufacturing Technology 58(1):437–440. [17] Liao CJ, Shyu CH (1991) An Analytical Determination of Lead Time with Normal Demand. International Journal of Operations & Production Management 11(9):72–78. [18] EPA, 2010, URL: www.epa.gov [19] Chryssolouris G (2006) Manufacturing Systems: Theory and Practice, 2nd ed. Springer-Verlag, New York. [20] Evans JR (1993) Applied Production and Operations Management, 4th Ed. West Publishing Company, St. Paul, MN, USA. [21] Kolmogorov AN (1965) Three Approaches to the Quantitative Definition of Information. Problems of Information and Transmission 1(1):1–7. [22] Kaspar F, Schuster HG (1987) Easily Calculable Measure for the Complexity of Spatiotemporal Patterns. Physical Review A 36(2):842–848.