Integrated simulation-based facility layout and complex production line design under uncertainty

Integrated simulation-based facility layout and complex production line design under uncertainty

G Model CIRP-1844; No. of Pages 4 CIRP Annals - Manufacturing Technology xxx (2018) xxx–xxx Contents lists available at ScienceDirect CIRP Annals -...

1MB Sizes 0 Downloads 14 Views

G Model

CIRP-1844; No. of Pages 4 CIRP Annals - Manufacturing Technology xxx (2018) xxx–xxx

Contents lists available at 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

Integrated simulation-based facility layout and complex production line design under uncertainty Nikolaos Papakostas (2)*, Joseph O’Connor Moneley, Vincent Hargaden Lab for Advanced Manufacturing Simulation, School of Mechanical and Materials Engineering, University College Dublin, Belfield, Dublin 4, Ireland

A R T I C L E I N F O

A B S T R A C T

Keywords: Production planning Multi-level modelling Manufacturing system Facility layout problem

When designing production lines, a series of complex, interdependent phases need to be considered. These phases include the facility layout design, the allocation of tasks to available resources, the workload balancing as well as the validation of the proposed design against demand. This paper addresses all these phases in an integrated way, with the aim of reducing the overall time required to finalise the layout and production line design, while improving the overall performance of the manufacturing system under process and demand uncertainty, utilising a sophisticated, simulationbased software framework. The proposed approach is demonstrated using a realistic case scenario. © 2018 Published by Elsevier Ltd on behalf of CIRP.

1. Introduction The increase in complexity of manufacturing systems together with the fact that small-lot production is more frequent than mass production [1] are some of the challenges engineers need to address nowadays. Typically, as soon as the product structure is finalised, the production / assembly line design team needs to transform it into a bill of processes (BoP) and then identify all possible resources required to support this BoP [2]. This is part of the rough planning stage, which also includes the determination of the resources to be used, the rough estimation of the cost and the preliminary layout design. This phase is followed by the detailed resource planning phase, where the cell and process definition, the layout design, which deals with the identification of the exact location of each machine or workcentre in a facility, the material flow, the capacity planning as well as the simulation and optimisation are carried out [2]. The design of a production line usually requires a number of iterations involving different stages and engineering groups. This leads to additional time and cost requirements. Previous research has attempted to tackle different stages of the production line design process. In particular, simulation-based methodologies have been presented for tackling the detailed resource planning problem. In Ref. [3], an assembly line design approach is proposed, where alternative line configurations are generated and simulated, without, however, considering the layout design problem and the associated constraints. Tasks are also allocated to resources beforehand.

* Corresponding author. E-mail address: [email protected] (N. Papakostas).

Assembly line balancing methods aim to equally divide the workload among different workstations. A number of authors have attempted to tackle more complicated problems that are encountered in practice. In Ref. [4], a multi-objective genetic algorithm is proposed, which aims to minimise not only the number of workstations and the workload variance in assembly lines with deterministic process times but also the number of skilled workers among workstations, the number of assembly equipment and assembly direction changes. In Ref. [5], the assembly line balancing problem is addressed, while assigning dispatch rules to each workstation and then determining the layout, considering predefined locations for the different workstations as well as deterministic process times. Walking-worker assembly (WWA) lines are considered as an effective method to reduce work-inprocess and to eliminate communication problems in conventional assembly lines and cells, where there are stationary workers in each station [6]. In such WWA systems, workers, following each other, travel all workstations needed for performing all tasks pertaining to their products [6]. In Ref. [7] a mixed integer linear programming approach is presented for simultaneously balancing assembly tasks and assigning workers to each task. There, the location of the stations is considered known and the process times are deterministic. The facility layout design problem, on the other hand, usually deals with the minimisation of the average distance or transportation costs among workcentres or machines. The dynamic facility layout problem, where the demand varies over a number of time periods, is studied in Ref. [8]: a single production layout is generated that may perform sufficiently well under different demand profiles; the tasks are preassigned to resources and the key objective is to minimise the material handling and relocations costs over all planning horizon periods. On the other hand, Ref. [9] proposes a mixed integer linear programming method for achieving flexible production layouts that may satisfy

https://doi.org/10.1016/j.cirp.2018.04.111 0007-8506/© 2018 Published by Elsevier Ltd on behalf of CIRP.

Please cite this article in press as: Papakostas N, et al. Integrated simulation-based facility layout and complex production line design under uncertainty. CIRP Annals - Manufacturing Technology (2018), https://doi.org/10.1016/j.cirp.2018.04.111

G Model

CIRP-1844; No. of Pages 4 N. Papakostas et al. / CIRP Annals - Manufacturing Technology xxx (2018) xxx–xxx

2

different demand profiles over multiple time periods, without, however, addressing task allocation and balancing problems. In Ref. [10], a cellular layout design technique is proposed that can determine the positions of manufacturing components and simultaneously schedule the pending tasks. The positions of these components and of the parts are considered static and the processing times of the tasks cannot be evaluated accurately, since the operation of the cell is not simulated. Designing human–robot workplaces is a complex task and in Ref. [11] a method is proposed where tasks are allocated to human operators or robots, without, however, considering the layout design problem and other production constraints. It is apparent, therefore, that despite recent advances, facility layout design, line balancing, task allocation and validation still take place in different design stages. 2. Methodology Fig. 2. Core HS-based algorithm structure.

In the initial stage of the process of designing new production lines, a lot of information is still uncertain. Sources of uncertainty include multiple potential demand profiles as well as uncertain process and part transportation times, especially in cases where little or no knowledge is available regarding the assembly or manufacturing tasks to be completed. Once the product design is finalised, the potential demand profiles and the matching manufacturing technologies are specified. Then, the capacity and process requirements are detailed. The next steps include the design of the facility layout, the allocation of tasks to resources and the validation of the line configuration. In real production environments with uncertain process times and complex process plans, where human operators may be responsible for executing a multitude of tasks that require their movement across a series of workcentres and machines, it is often hard to evaluate the performance of a specific facility layout configuration in one step. A series of iterations is typically required, allowing for the fine adjustment of the layout as well as of the line design. In this paper, a novel, multi-stage simulation-based integrated approach is presented where alternative layout and production configurations are generated and validated via simulation. The proposed approach may be further extended to also include parts of the product design, resources selection, detailed process configuration and commissioning (Fig. 1).

Fig. 1. Production line design phases.

In particular, a new algorithm, based on the Harmony Search (HS) optimisation metaheuristic [12], is designed and implemented as part of a simulation-based production line design software platform. HS was inspired by how Jazz musicians would rapidly refine their individual improvisation through variation, resulting in an aesthetic harmony [13]. The pseudocode of the HS-based algorithm is presented in Fig. 2 (adapted from Ref. [13]). Each alternative contains the position and orientation of each workcentre w and the specification of which human operator is in charge of the operations taking place in each workcentre:

Fig. 3. An example of a layout and task allocation alternative.

where xw, yw are the coordinates of the centroid of the rectangle corresponding to workcentre w, ’w its orientation and hw the operator in charge of workcentre w tasks. In the example of Fig. 3, each workcentre has an entry and an exit point and requires a gap Ds from the facility walls or the other workcentres’ boundaries. The position and task allocation of workcentre W3 is adjusted in an iteration. The algorithm utilises good solutions already discovered (Alternatives) to influence the creation of new alternatives (newAlt) towards improving the overall solution quality [13]. This is achieved by stochastically developing alternative solutions iteratively, where every component (newAltw wrkc ) of each alternative is either selected randomly from the pool of existing alternatives or configured randomly, while respecting the production and layout constraints. In each iteration the worst alternative is replaced whenever a better one is found. The parameters of the algorithm include:  The maximum number of alternatives to be generated (MNA) and maintained during all iterations (maxIterations).  The Samplesmax number of times each alternative is simulated.  The parameter Consolidationrate 2 [0,1], which affects the rate of the algorithm convergence, controls the use of information from existing MNA alternatives or the generation of a new random position / orientation and operator assignment for a workcentre. A value of 0.7 would indicate that there is a 70% chance the configuration of a workcentre will be selected from existing alternatives and a 30% chance that a new configuration of a workcentre will be randomly generated.  The parameter Adjustrate 2 [0,1], controlling the frequency of adjustments of workcentres’ positions / orientations and operator assignments selected from any of the other MNA alternatives. Avalue of 0.6, for instance, indicates that there is a 60% chance a workcentre configuration that has been selected from existing alternatives will be modified, leading to a change of the workcentre location.  The maximum adjustment distance of the workcentre position (Xmax, Ymax), with:

xw ¼ xw þ ex ; ex 2 ½X max ; X max ; yw ¼ yw þ ey ; ey 2 ½Y max ; Y max ; 8w 0

Alternative ¼

fðAltw wrkc Þg

¼ fðxw ; yw ; ’w ; hw Þg; w ¼ 1:::W

ð1Þ

0

ð2Þ

Please cite this article in press as: Papakostas N, et al. Integrated simulation-based facility layout and complex production line design under uncertainty. CIRP Annals - Manufacturing Technology (2018), https://doi.org/10.1016/j.cirp.2018.04.111

G Model

CIRP-1844; No. of Pages 4 N. Papakostas et al. / CIRP Annals - Manufacturing Technology xxx (2018) xxx–xxx

In each iteration, when a workcentre configuration is adjusted, the assignment of the operator is carried out randomly. A parametric discrete event simulation model is then adapted and associated to each alternative. The position / orientation of each workcentre and the operator in charge are reflected in the simulation model. Samplesmax experiments are executed and the average performance of each alternative in terms of time and cost criteria are evaluated. The alternative Z with the highest utility is identified (Fig. 4).

3

Fig. 6. Case scenario production line.

for the initial period, it has been concluded that up to two operators may be hired for setting up and running the production line. In order to validate the proposed approach, a parametric model was built for simulating the operation of the production line. A software library was developed on top of the Desmo-J framework [14]. The questions that need to be answered are:

Fig. 4. Integrated simulation-based approach.

The proposed methodology may be further extended from the current macro simulation-based approach to a more system-wide platform, where exact digital models of the manufacturing and control processes (micro-simulation and automation) are also integrated in alternative production configurations of wider scope (Fig. 5).

 How should each workcentre of stages 1–4 be placed, and where should their exact position and orientation be, given the available facility space as well as the production constraints?  How many operators will be needed, and which operator should be in charge of setting up, loading, unloading each workcentre as well as moving the parts among them? The software library is capable of representing the principle components of a typical production line in a generic way, thus allowing for the easy modelling of real production environments. For instance, when operator 1 is assigned to stage 2 in one alternative, the simulation model is adjusted accordingly, and all tasks related to that stage are assigned to that operator. Similarly, different positions and orientations of the workcentres will correspond to different distances and therefore transportation times for the operators to travel among workstations. The process times for the set-up, loading, processing, unloading and transportation tasks were approximated with normal distributions, based on former engineers’ experience and on the information available by the equipment vendors. The demand profiles were generated by considering the testing and validation requirements as well as available market information. Both candidate operators were assumed to have the same skills and their walking speed was approximately 1 m/s. 3.2. Experiments and discussion

Fig. 5. Extended, system-wide scope of the proposed approach.

3. Case study 3.1. Description and modelling In order to test and validate the proposed methodology, a case study in a start-up company, planning to manufacture high-tech, high-value devices is examined. For illustration purposes, only a part of the company’s manufacturing system is considered. Aspects of the technical information regarding the products and processes are omitted for confidentiality reasons. The line studied in this section includes distinct production stages and a final inspection stage (Fig. 6). Each stage includes one workcentre. Stage 2 requires 4 parts as input, while stage 4 requires 40 parts. All other stages require one part as input. The position of the workcentres of the first and last stage is fixed since they have to be located close to the entrance and exit of the facility respectively. Due to the nature of the product and the stage in its lifecycle, it is not possible to introduce a higher degree of automation at the current point in time, so the transfer of the parts among stages is planned to be carried out manually. Production engineers have already specified the production processes and the suitable equipment. Having considered the potential demand profiles

A series of experiments was carried out to evaluate the efficiency of the proposed approach. The algorithm parameters are shown in Table 1. A conventional Facility Layout Problem solving approach was followed initially, where the proposed algorithm (Fig. 2) was used to determine the position of each workcentre in the facility space. In each iteration all alternatives were evaluated on the basis of estimating the average distance among workcentres with the aim to minimise it. The tasks were allocated to the available human operators manually by considering the process times of all tasks in the line. It has to be noted that the process times in Stage 3 are quite longer than the process times of all other tasks and this is why it was decided to allocate one of the two operators to this Stage only.

Table 1 Experiments parameters. Parameter

Value

Maximum number of alternatives (MNA) Maximum number of iterations (max Iterations) Simulation samples (Samplesmax) Frequency of use of other alternatives (Consolidationrate) Frequency of adjusting workcentres configuration (Adjustrate) Max. adjustment distance of workcentre position (Xmax, Ymax)

20 2000 10 70% 70% 1m

Please cite this article in press as: Papakostas N, et al. Integrated simulation-based facility layout and complex production line design under uncertainty. CIRP Annals - Manufacturing Technology (2018), https://doi.org/10.1016/j.cirp.2018.04.111

G Model

CIRP-1844; No. of Pages 4 N. Papakostas et al. / CIRP Annals - Manufacturing Technology xxx (2018) xxx–xxx

4

Fig. 7. Distance-based determination of layout and task allocation.

Fig. 8. Layout and task allocation with the proposed approach.

The simulation model was run 10 times (Samplesmax) for evaluating the average performance of the proposed configuration in terms of cost, order tardiness and flowtime against two different demand profiles. Then, the simulation-based approach was tested and in each iteration, all generated alternatives were evaluated based on the average cost, order tardiness and flowtime achieved via simulation against the same demand profiles, utilising the simple additive weighting multi-criteria evaluation method. Oneoperator configurations were also part of the solution space. The best alternatives obtained for these 2 groups of experiments are shown in Figs. 7 and 8 respectively. The average production cost, tardiness and flowtime per order were reduced by 0.8%, 3.6% and 2.3% respectively compared to the performance achieved with the distance-based approach. The main advantages, however, stem from the fact that the layout design and the configuration of the line may be tested, evaluated and validated accurately in one step, following the proposed, simulation-based approach. Following the results obtained in Ref. [15], calendar time savings may be expected, in terms of the start of production (SoP), ranging from one week for small production lines to up to two months for large lines. 4. Conclusions The proposed approach provides an integrated methodology that would allow engineers to better design production lines that are not easy to model with conventional facility layout design and line balancing modelling approaches. For instance, lines, allowing more than one moving operator to carry out the same task, or cases, where in each stage different numbers of parts are needed as input, necessitate the employment of more advanced approaches. At the same time, uncertainty regarding the process plans, especially in the early stages of the product development, when prototypes or batches of early product versions need to be produced for testing or validation purposes, cannot be easily

considered with standard production line design and balancing approaches. The uncertainty pertaining to the process times and the process plans themselves require the use of accurate simulation models that are well integrated within search and optimisation algorithms. The potential reduction in the SoP by up to two months would offer a significant competitive advantage to companies that plan to introduce new products to the market. At the same time, the accurate modelling of a production line would allow engineers not only to validate the line designs early enough but also to test different setups and configurations, involving diverse demand profiles and what-if scenarios. This capability would lead to the reduction of early design errors and would minimise the number of design iterations required for finalising the line design. Although the development of an accurate simulation model of the production line is not a simple task, the parametric simulation framework that was built and used in the context of this paper may consider all standard elements of today’s complex production systems. At the same time, modern discrete event simulation platforms provide rich Application Programming Interfaces that would allow the integration of off-the-shelf simulation platforms with optimisation algorithms and heuristics. On the other hand, the proper fine-tuning of the simulation parameters is of critical importance for reducing the total simulation runtime and for increasing the overall usability of the proposed methodology. In the future, the consideration of the next design stages, including the specification of the process configurations and the optimisation of the automation and control software and hardware, would open new opportunities for further reducing the overall time and effort needed for designing production lines.

References [1] Colledani M, Tolio T, Fischer A, Iung B, Lanza G, Schmitt R, Váncza J (2014) Design and Management of Manufacturing Systems for Production Quality. CIRP Annals — Manufacturing Technology 63(2):773–796. [2] Michalos G, Makris S, Mourtzis D (2012) An Intelligent Search Algorithm-based Method to Derive Assembly Line Design Alternatives. International Journal of Computer Integrated Manufacturing 25(3):211–229. [3] Michalos G, Fysikopoulos A, Makris S, Mourtzis D, Chryssolouris G (2015) Multi Criteria Assembly Line Design and Configuration — An Automotive Case Study. CIRP Journal of Manufacturing Science and Technology 9:69–87. [4] Dalle Mura M, Dini G (2017) A Multi-objective Software Tool for Manual Assembly Line Balancing Using a Genetic Algorithm. CIRP Journal of Manufacturing Science and Technology 19:72–83. [5] Baykasoglu A, Tasan SO, Tasan AS, Akyol SD (2017) Modeling and Solving Assembly Line Design Problems by Considering Human Factors with a Real-life Application. Human Factors and Ergonomics In Manufacturing 27(2):96–115. [6] Cevikcan E (2016) An Optimization Methodology for Multi Model Walkingworker Assembly Systems: An Application from Busbar Energy Distribution Systems. Assembly Automation 36(4):439–459. [7] Sikora CGS, Lopes TC, Magatão L (2017) Traveling Worker Assembly Line (re) balancing Problem: Model, Reduction Techniques, and Real Case Studies. European Journal of Operational Research 259(3):949–971. [8] Madhusudanan Pillai V, Hunagund IB, Krishnan KK (2011) Design of Robust Layout for Dynamic Plant Layout Problems. Computers and Industrial Engineering 61(3):813–823. [9] Xiao Y, Xie Y, Kulturel-Konak S, Konak A (2017) A Problem Evolution Algorithm with Linear Programming for the Dynamic Facility Layout Problem — A General Layout Formulation. Computers and Operations Research 88:187–207. [10] Suemitsu I, Izui K, Yamada T, Nishiwaki S, Noda A, Nagatani T (2016) Simultaneous Optimization of Layout and Task Schedule for Robotic Cellular Manufacturing Systems. Computers and Industrial Engineering 102:396–407. [11] Tsarouchi P, Michalos G, Makris S, Athanasatos T, Dimoulas K, Chryssolouris G (2017) On a Human–Robot Workplace Design and Task Allocation System. International Journal of Computer Integrated Manufacturing 30(12):1272–1277. [12] Geem ZW, Kim JH, Loganathan GV (2001) A New Heuristic Optimisation Algorithm: Harmony Search. Simulation 76:60–68. [13] Brownlee J (2012) Clever Algorithms: Nature-Inspired Programming Recipeslulu. com. [14] Page B, Kreutzer W (2005) The Java Simulation Handbook: Simulating Discrete Event Systems, Shaker Verlag. [15] Papakostas N, Pintzos G, Giannoulis C, Chryssolouris G (2016) An Agent-based Collaborative Platform for the Design of Assembly Lines. International Journal of Computer Integrated Manufacturing 4:374–385.

Please cite this article in press as: Papakostas N, et al. Integrated simulation-based facility layout and complex production line design under uncertainty. CIRP Annals - Manufacturing Technology (2018), https://doi.org/10.1016/j.cirp.2018.04.111