You are on page 1of 7

Journal of Simulation

ISSN: 1747-7778 (Print) 1747-7786 (Online) Journal homepage: https://www.tandfonline.com/loi/tjsm20

Simulation-based dispatching in a production


system

N Höppe, F Seeanner & S Spieckermann

To cite this article: N Höppe, F Seeanner & S Spieckermann (2016) Simulation-based dispatching
in a production system, Journal of Simulation, 10:2, 89-94, DOI: 10.1057/jos.2015.17

To link to this article: https://doi.org/10.1057/jos.2015.17

Published online: 19 Dec 2017.

Submit your article to this journal

Article views: 34

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tjsm20
Journal of Simulation (2016) 10, 89–94 © 2016 Operational Research Society Ltd. All rights reserved. 1747-7778/16
www.palgrave-journals.com/jos/

Simulation-based dispatching in a production system


N Höppe1, F Seeanner2 and S Spieckermann3*
1
Reis GmbH & Co. KG Maschinenfabrik, Obernburg, Germany; 2SimPlan Systems GmbH, Maintal, Germany; and
3
SimPlan AG, Maintal, Germany
This paper describes the continuous application of simulation from system design to system operation including the integration of
the simulation in a manufacturing execution system (MES). This is done using the example of a real-world manufacturing system,
a complex flexible laser welding facility. The results of a simulation study during system design are presented and it is derived how
these results strengthened the decision to implement a simulation-based MES. Some technical details of the connection between the
Enterprise Resource Planning (ERP) system, the MES, and the Programmable Logic Control (PLC) are explained, as well as the
usage of the MES and the simulation by the operations planners (dispatchers).
Journal of Simulation (2016) 10(2), 89–94. doi:10.1057/jos.2015.17; published online 25 September 2015

Keywords: discrete-event simulation; manufacturing execution system; welding facility; simulation-based dispatching

1. Introduction the developed software solution. The paper closes with a


summary and an overview of some lessons learned from the
The application of discrete-event simulation (DES) to support
practical application of the solution.
the daily operations of a production or logistics system is being
discussed for more than 20 years by now (cp. Noche and
Scholtissek, 1993 as well as Wenzel and Meyer, 1993). The
VDI guideline on simulation in production and logistics (VDI, 2. Problem description
2014) as well as Klingstam and Olson (2000) or Kosturiak and
The laser welding facility under consideration has four parallel
Gregor (1999) present several examples of simulation during
manual workstations, two welding booths, one automatic trans-
system operation. However, simulation in production and logis-
portation system, a carrier storage, holding up to eleven empty
tics is still mainly applied during system design to support
carriers, and a station to transfer carriers in and out of the facility.
manufacturing or logistics engineering. The second most com-
The carriers are moved by a cross transfer car called ‘shuttle’
mon reason for using simulation models is to support the ramp-up
between workstations and the welding booths. The shuttle can
of real-world production control systems (discussed under the
hold up to two carriers. On each carrier there is at least one
terms virtual ramp-up or emulation; cf. Johnstone et al, 2007).
fixture. At the workstations, metal sheets are set up on the fixtures
Only a small number of simulation applications is actually
before being moved to a welding booth, where the sheets are
supporting daily business. This is confirmed by, for example the
welded to form metal containers. Figure 1 depicts the facility
surveys given in Jahangirian et al (2010) and Smith (2003), each
layout.
of which are summarizing more than 200 publications.
Most of the carriers hold one single fixture. However, there are
More recently, Negahban and Smith (2014) noticed a shift
also carriers equipped with two or more (smaller) fixtures for
towards the use of simulation for manufacturing system opera-
parallel assembly of various small parts. Basically, the assign-
tion. The contribution of this paper can be seen in this context, as
ment of fixtures to carriers is fixed even though a changeover is
its focus is on a simulation application being part of the
possible. A changeover is not done within the facility; carriers are
operational control of a laser welding facility. The components
moved out via the outfeed transfer, the fixture is changed, and the
and the processes in this facility as well as the challenges related
carriers are moved backed into the facility. There are identical
to schedule the orders for the facility are explained in the next
fixtures so that several carriers may be set up to hold the same
chapter. Chapter 3 presents a short overview of the state-of-the-
product.
art in tackling similar problems. The idea and the need to use a
A production order, typically comprising a lot of identical
simulation-based application during daily operation of the facility
products (ie, lot size > 1), is always processed exclusively at
was born during a simulation study supporting system engineer-
one of the four workstations, as pre-manufactured metal sheets
ing. The fourth chapter is dedicated to this ‘classic’ simulation
have to be available at that workstation. These metal sheets are
activity. Chapter 5 describes functional and technical details of
provided by forklift trucks, that is, they cannot easily be shifted
from one workplace to another. An operator puts the metal sheets
*Correspondence: S Spieckermann, SimPlan AG, Edmund-Seng-Str. 3-5,
Maintal, Hessen D-63477, Germany. into the fixture (or in case of small parts into the fixtures) on the
E-mail: sven.spieckermann@SimPlan.de carrier. As soon as this is done, the operator signals to the control
90 Journal of Simulation Vol. 10, No. 2

Figure 1 3D-Layout of the laser welding facility.

software that the carrier is ready to enter a welding booth by Operations planning tries to find a schedule for the selected
pressing a release button on the operating panel at his work- production orders with a minimum makespan (ie the time
station. The control software picks up the carrier and moves it difference between the start and finish of the scheduled orders).
into a free welding booth. Both welding booths are equipped In this case, this objective is complementary to maximizing the
identically and can process all different types of parts. However, utilization of the laser equipment, which is the most expensive
welding is only possible in one booth at a time, as there is only key resource of the facility. Main restrictions are earliest start and
one laser source available, which is directed via reflectors to latest end date of the orders.
either booth. While the welding process is active in one booth, the The number of possible solutions is quite high: there are 90 part
other booth is to some extent buffering a second carrier in order to numbers, 30 fixtures, and 15 carriers. Some fixtures are identical,
optimize the utilization of the expensive laser source, by reducing meaning that it is possible to have identical part numbers on
waiting or idle times of the laser. After the welding is completed, different carriers in the system. The same carrier may in principle
the shuttle picks up the carrier again and moves it back to the start be loaded at different workstations. However, as described above,
workstation. Depending on the details of the order, additional because of constraints in the part logistics, not all assignments of
metal sheets have to be added, or the operator has to change the part numbers to workstations are valid. Furthermore, orders for
position of part and fixture before sending the carrier into a small parts need to be combined appropriately, as SAP requires
welding booth again for another welding operation. The control one production order for each (small) part number whereas the
logic of when to pick up a carrier and where to move it is part of carriers hold fixtures for more than one small part.
the PLC (programmable logic control) of the facility. In addition, blocks for single carriers, for dedicated fixtures, for
The dispatching of production orders for the welding booth is one of the two welding booths or for selected workstations need to
done by a team of operations planners (dispatchers) at the be considered because of equipment breakdown, maintenance or
operations planning department. The orders are provided via the missing part supply. Blocked resources have to be excluded when
Enterprise Resource Planning system (ERP) from SAP. The PP searching for an order schedule. Finally, shift patterns and the
(production planning) module of the SAP system schedules a available workforce (workstation operators) are limiting factors.
basic start date and a basic end date based on throughput times.
When dispatching the orders, these dates need to be factored in as
earliest start and latest end date, respectively (cp. Dickersbach
3. State of the art
and Keller, 2011 for an introduction and overview on SAP ERP).
The operations planner in charge decides which pending orders There are plenty of examples in the literature for both the
shall be started next by taking into account the work content per simulation of welding facilities like the one considered in this
order and the current state of the facility. A constraint is that paper, and for advanced-planning applications for order sequen-
orders are only released if at least one carrier with a matching cing. Chan and Chan (2004) present a survey on more than
fixture is within the facility, thereby ensuring that there are no 120 studies, all dedicated to the control and planning of orders
waiting times because of fixture changeover. and resources in flexible manufacturing systems (FMSs) using
N Höppe et al—Tautomation 91

simulation. They also present some definitions of FMSs and, design, simulation to support daily-system operation, scheduling
following Chan and Chan, the welding facility presented in approaches, and an implementation in an MES-like context.
Chapter 2 would fall into this category. According to the
definitions, FMSs are facilities where machine centres (such as
the welding booths including the laser), manual workstations and
4. Simulation supporting system design
a material handling system (the shuttle and the carriers including
the carrier storage) work jointly under computer control. During the design of the welding facility, a simulation model was
Smith (2003) presents a number of examples for the use of implemented in order to assess the complex dependencies of
simulation during system and manufacturing engineering of FMSs, carriers with different fixture set-ups, variant-depended proces-
that is during ‘flexible manufacturing system design’. He also sing times for the different part numbers, operators potentially
presents literature were simulation was used to support the daily serving multiple workstations, and stochastic breakdowns of the
operation of FMSs and classifies it as simulation for ‘manufactur- welding equipment. Main performance indicator for the simula-
ing systems operation’. In the latter case, however, the number of tion experiments carried out with the simulation model was the
examples is significantly smaller than in the first-mentioned case. utilization of the welding booths in relation to the number of
PP applications to support the operational control of FMSs operators and in relation to the order sequence.
might be considered to be part of scheduling or even part of Figure 2 shows the correlation of number of operators and the
advanced planning and scheduling (APS) from a management welding booth utilization. Not only was the operator number
science standpoint. From a system controls point of view, subject to modifications but also the way the operators are assigned
however, PP applications might be considered to be part of to workstations. The alternative ‘variable WS assignment’ allows
MES (Manufacturing Execution Systems) located between the each operator to work on any of the workstations where a carrier
company-wide ERP level and the shop floor control via PLC. For is waiting. With the alternative ‘fix WS assignment’ each operator
both areas, APS and MES, there is a huge amount of literature, is bound to a limited number of allotted workstations.
starting with text books (Pinedo, 2012), and including several The impact of the order sequence was assessed based on a
recent surveys (Ouelhadj and Petrovic, 2009, Framinan and Ruiz, given pool of production orders. Using the orders in this pool, 500
2010, Saenz de Ugarte et al, 2010). Several of the examples order sequences (500 order sequence permutations) were gener-
therein have characteristics similar to the application described in ated randomly. Each of the permutations was then assessed via the
detail in Chapter 5. simulation model. Figure 3 shows the results of the simulation
This brief context of simulation of FMSs, APS, and MES runs (using three operators with variable workstation assignment).
illustrates that the application presented in this paper is not new
from a research perspective and the authors do not claim it to be.
However, the joint use of DES and scheduling algorithms within
an operational planning system or within a MES in a real-world
context is still a ‘rare’ exception. Sauer (2010) sees a future trend
in supporting MES by means of simulation. Framinan and Ruiz
(2010) point out that the discussion about the ‘gap’ between theory
and practice of scheduling has been going on for more than
20 years by now and that only 3-5% of quantitative studies in
scheduling consider realistic production settings. In this sense, the
case study discussed here addresses exactly this gap by document-
ing a real-world application, combing simulation for system Figure 2 Number of operators and welding booth utilization.

Figure 3 Welding booth utilization for different order sequences.


92 Journal of Simulation Vol. 10, No. 2

Figure 4 Architecture of the planning and scheduling application.

Five replications were performed for each order sequence permu- planning application. The communication with SAP is based
tation; the graph shows the average welding booth utilization of on Remote Functions Calls (RFC), a SAP standard interface.
these replications. The resulting deviations in the utilization are More precisely, there is an RFC function block used to read
quite large or at least too large to accept them in the real-world the orders from SAP. This function block returns a list of all
context. If the best and the worst sequence are eliminated, there is released and not yet acknowledged production orders within
still a difference of about 10 percentage points. Evaluations for SAP ERP. This list comprises order number, process number,
one, two, and four operators are showing very similar results. part number, lot size, priority, earliest start date, and latest end
Even if it is fair to expect that a manual operations planning date. A second RFC function block is used to send acknowl-
will achieve better results than random sequence generation, edgements back to SAP once an order is finished. Here, order
these experiments were well-suited to demonstrate the sensitivity number, part number, order status (‘OKAY’, ‘NOT OKAY’),
of the welding booth utilization to the order scheduling. and a timestamp are passed back to the ERP system. Within
It resulted in additional encouragement and commitment to all imported, released, and open orders, the operations planner
implement an operative planning and scheduling system in the selects a subset for scheduling. The manual selection of an
real-world facility, able to automatically create order sequences, order subset was an explicit requirement, giving the
and to assess them using the simulation model of the facility. operations planner the opportunity (and responsibility) to
fully control the scheduling process. If the planner selects
only one order at a time, there is explicit control over when
and where this order is processed. If all open orders are
5. Simulation to support daily-system operation
selected, the optimization and simulation component has
The software tool to support daily-system operations is a role- the greatest flexibility in supporting the decision making.
based JAVA web application, which is basically usable via all However, even then the final selection of a schedule still is
browser-enabled PCs in the intranet of the company running the a manual decision of the operations planner, as will
welding facility. The application comprises several components become clearer in step 3.
depicted in Figure 4. There are interfaces to SAP ERP as well as 2. In the second step, the operations planner sets the number of
to the OPC server of the welding facility. At the heart of the workstation operators and the current shift patterns. In addi-
application, there are an optimization component and a simula- tion, selected resources, such as carriers or fixtures, might be
tion component. The simulation component is using the simula- blocked. Varying these parameters, the planner can create a
tion model which was implemented during system planning. The couple of different scenarios. The basic information on the
explanation of the planning process will provide some additional availability of workstations, welding booths, carriers, and
details on each component. fixtures comes from the shop floor control IT, that is, from
In principle, the software-based planning operations can be the PLC. However, the operations planner can block addi-
divided in four steps: tional equipment components in order to assess alternative
scenarios. Components which are not available according to
1. The operations planner in charge imports production orders, the information coming from the PLC cannot be made
including related status information from SAP, into the available for planning scenarios.
N Höppe et al—Tautomation 93

Table 1 Example for data transfer between MES and PLC schedule are possible, for example, because of missing metal
sheets at one of the workstations. In this case, parts of a
Address Data Item
production order might be skipped or postponed causing the
PCStation.WinLC RTX F EC. Order[1].OrderNo SAP_001 system to send a ‘NOT OKAY’ message back to SAP. If that
PCStation.WinLC RTX F EC. Order[1].CarrierID WT_05 happens, the order needs to be re-opened (in parts or in total) at a
PCStation.WinLC RTX F EC. Order[1].Fixture01.FixtureID VT_07
later point in time with another schedule. All operations are
PCStation.WinLC RTX F EC. Order[1]. Fixture01.PartNo MAT_013
PCStation.WinLC RTX F EC. Order[1]. Fixture01.Count 1 logged into a database, enabling all kinds of analyses, compar-
isons of target and actual indicators, or tracking of actual
production processes.
3. The third step covers the generation and assessment of
production schedules for the planning scenarios defined in
the second step. By experience, the operations planner 6. Experiences with the planning application
generates four to eight different production schedules. Typi-
First of all, a software application between SAP ERP and the
cally, there are four schedules with active morning shift plus
PLC level which transfers the order data, supports order schedul-
active evening shift, and the planner modifies the number of
ing and transforms an order schedule into PLC-conform move-
operators. If in that case the orders cannot be completed
ment instructions is simply a technological necessity, at least in a
with the maximum of four workstation operators in two
situation where SAP and shop floor are required to work together
shifts per day, additional schedules using a three-shift-pattern
by exchanging order data. This is one of the straightforward
(including night shifts) will be generated. Of course, the
reasons for the popularity of MES applications, let alone all
operations planner might also change additional parameters
planning and scheduling features.
in order to generate and evaluate even more alternative
With respect to the scheduling and evaluation step itself,
schedules. To compile the schedules, priority-based heuristics
however, it turned out that even with today’s computational
are applied. Each generated schedule is fed into the simulation
resources and the available software, using optimization and
model of the welding facility. The model is used to compute a
simulation, it is still a challenging job, even for a rather small
detailed forecast on key performance indicators. The indica-
facility like the one considered in this case study. The generation
tors are workstation and welding booth utilization, distances
of schedules followed by the simulation-based forecast on how
travelled by each worker and workload per worker, order
the facility will perform under this schedule within the next
throughput times, and so on. The simulated throughput times
48 hours requires a computational time of several minutes
are used as target times and are compared with real-world times.
measured for a typical schedule, covering about 130 production
4. After the generation of schedules using heuristics and the
orders leading up to 10 000 movement instructions for the PLC.
simulation-based assessment of these schedules, the results are
The feedback of the operations planners as the main users of
presented to the operations planner using table sheets and
the planning application is that the software significantly supports
different kinds of charts. The planner uses this information to
and simplifies their daily business. Compared with a purely
select a schedule and to forward it via interface to the PLC
manual planning process, the tool saves them a considerable
level. The PLC interface is implemented using the so-called
amount of working time. The utilization of the welding booths is
Distributed Component Object Model (DCOM) together with
in the upper level of the bandwidth presented in Chapter 4.
the Siemens OPC-Server, running on Siemens Simatic S7
Altogether, the concept of simulation-based support of the
PLCs (cp. Mahnke et al, 2009). Using DCOM, the data are
welding facility was a success. As one result, the system
sent and received in tuples of < address, data > , where
integrator and supplier of the facility included the application
‘address’ stands for a unique string and ‘data’ stands for a
concept into its IT product portfolio for other, similar, projects.
simple data type (integer, floating etc). In order to control the
Another more recent result is that the operator of the overall
shuttle, the PLC needs simple movement instructions.
facility is currently working on extending the simulation-based
A movement instruction comprises a carrier ID, a motion from
operation support concept to the plant logistics.
welding booth to workstation or vice versa, and storage and
retrieval movements. Each fixture needs to know the attached
part number in order to start the correct welding programme in
the welding booths. Table 1 shows an example for the data 7. Summary and outlook
transfer between planning application and PLC. Each row of
The presented case study demonstrates that the integrated and
the table represents a tuple of the form < address, data > .
continuous use of simulation from system design to daily-system
As soon as the first movement instructions are sent to the PLC,
operation is well-understood and that a successful application
the production itself will start (or continue, respectively).
in a real-world manufacturing environment is feasible. The
The planning application permanently monitors the production technology is available to integrate simulation models with ERP
progress. For each welded part, an acknowledgement is sent to systems such as SAP, with the PLC level, and with optimization
the SAP ERP. However, deviations during the execution of a algorithms.
94 Journal of Simulation Vol. 10, No. 2

Future research may very well tackle the challenge of setting Kosturiak J and Gregor M (1999). Simulation in production system
standards with respect to interfaces between the software compo- life cycle. Computers in Industry 382: 159–172.
nents ERP, MES, and PLC. A common standard comprising, for Mahnke W, Leitner SH and Damm M (2009). OPC Unified Architecture.
Springer: Berlin, Heidelberg.
example, communication protocols, would potentially simplify Negahban A and Smith JS (2014). Simulation for manufacturing system
the implementation of simulation-based MES significantly. design and operation: Literature review and analysis. Journal of
Manufacturing Systems 33(2): 241–261.
Noche B and Scholtissek P (1993). Anwendungen der Simulation in der
Note —This paper is a revised and expanded version of a paper entitled Unternehmensplanung. In: Kuhn A, Reinhardt A and Wiendahl HP
Simulationsgestützte Feinplanung im Produktionsumfeld – vom SAP zur (eds). Handbuch der Simulationsanwendungen in Produktion und
SPS presented at 15th ASIM Dedicated Conference Simulation in Produc- Logistik. Vieweg: Braunschweig, pp 267–307.
tion and Logistics, Paderborn, Germany, 9–11 October 2013. Ouelhadj D and Petrovic S (2009). Survey of dynamic scheduling in
manufacturing systems. Journal of Scheduling 12(4): 417–431.
References Pinedo M (2012). Scheduling—Theory, Algorithms, and Systems. 4.
Auflage. Springer: New York, Dordrecht, Heidelberg, London.
Chan FTS and Chan HK (2004). A comprehensive survey and future Saenz de Ugarte B, Artiba A and Pellerin R (2010). Manufacturing
trend of simulation study on FMS scheduling. Journal of Intelligent execution systems—A literature review. Production Planning &
Manufacturing 15(1): 87–102. Control 20(6): 525–539.
Dickersbach JT and Keller G (2011). Production Planning and Control Sauer O (2010). Trends in manufacturing execution systems.
with SAP ERP. 2nd edn, Galileo Press: Bonn, Boston. In: Huang GQ, Mak KL, Maropoulos PG (Hrsg.) (eds). Proceedings
Framinan JM and Ruiz R (2010). Architecture of manufacturing of the 6th CIRP-Sponsored International Conference on Digital
scheduling systems: Literature review and an integrated proposal. Enterprise Technology, Berlin, Heidelberg: Springer, pp 685–693.
European Journal of Operational Research 205(2): 237–246. Smith JS (2003). Survey on the use of simulation for manufacturing
Jahangirian M, Eldabi T, Naseer A, Stergioulas LK and Young T (2010). system design and operation. Journal of Manufacturing Systems
Simulation in manufacturing and business: A review. European 22(2): 157–161.
Journal of Operational Research 203(1): 1–13. VDI (2014). Guideline 3633 Part 1 ‘Simulation of Systems in Materials
Johnstone M, Creighton D and Nahavandi S (2007). Enabling industrial Handling, Logistics and Production’. Beuth: Berlin.
scale simulation/emulation models. In: Henderson SG, Biller B, Wenzel S and Meyer R (1993). Kopplung der Simulation mit Methoden
Hsieh MH, Shortle J, Tew JD, Barton RR (eds). Proceedings des Datenmanagements. In: Kuhn A, Reinhardt A and Wiendahl HP
of the 2007 Winter Simulation Conference. IEEE, Piscataway, (Hrsg.) (eds). Handbuch der Simulationsanwendungen in Produktion
pp 1028–1034. und Logistik. Vieweg: Braunschweig, pp 347–368.
Klingstam P and Olsson B-G (2000). Using simulation techniques for
continuous process verification in industrial system development.
In: Joines JA, Barton RR, Kong K, Fishwick PA (eds). Proceedings
of the 2000 Winter Simulation Conference. IEEE, Piscataway, Received 16 October 2014;
pp 1315–1321. accepted 5 August 2015 after one revision

You might also like