material flow analysis at basf – a huge variety in ... · material flow analysis at basf - a huge...
TRANSCRIPT
Simulation in Production and Logistics 2015 Markus Rabe & Uwe Clausen (eds.) Fraunhofer IRB Verlag, Stuttgart 2015
Material Flow Analysis at BASF – A Huge Variety in Complexity
Materialflussanalyse bei BASF – Eine große Bandbreite an Komplexität
Markus Feist, Ana Maria Fernández Alcalde, BASF SE, Ludwigshafen (Germany), [email protected], [email protected]
Abstract: Material flow analysis (MFA) is a common method for the investigation of production and logistic processes. Its objective is finding, quantifying and (if possible) avoiding bottlenecks and saving potentials. Other applications of MFA are the designing of capacities of production parts or logistic devices such as loading stations or storages. Usually, the complexity of the process and the questions to be answered determine the required level of detail of the MFA. The huge variety in complexity starts from simple estimations and calculations, going up to building detailed and customised simulation models. One common simulation method is the so called discrete event simulation. In this contribution, three examples are presented for using MFA in the BASF SE, a company in the field of chemical industry. An overview is given on how established methods are used in an industrial setting to solve different tasks with various data basis and purposes.
1 Introduction In industry, material flow analysis (MFA) is often used to gain knowledge about existing production and logistic processes, or to determine capacities for plants or logistic devices for investment projects, Jahangirian et al. (2013) and references cited therein. In either case, it must be determined whether the benefit is worth the effort of setting up a complex and detailed simulation or if a less detailed calculation is sufficient. The required level of detail is determined by several boundary conditions. For example, the complexity of the underling processes and the respective tasks which should be answered are as important as the available data basis. In that way it could happen that a MFA covers only a data analysis of production data, followed by a straight-forward and not very complex calculation (notation: static MFA). But it can also cover a complex dynamic model for discrete event simulations (notation: dynamic MFA).
Independent of the level of detail, MFA takes into account the plants and logistic behaviors as well as the product output and delivery reliability with respect to capacity, product recipes and logistic setup. In case of static calculations, many
330 Feist, Markus; Fernández Alcalde, Ana Maria
assumptions, generalisations and simplifications usually have to be made. This process of decision is accompanied by a deep discussion on the expected impact on the calculation results. For dynamic modeling a detailed process model including fewer simplifications is built, aiming to approximate the real process behavior more closely. Using the method of discrete event simulations (Jahangirian et al. 2013), even unexpected breakdowns or planned events as well as simultaneous usage of equipment and resources can easily be considered.
Independent on the level of complexity, it is possible to investigate the influence of production order, delays in transport, lack of raw materials or buffer capacity. Upon the completion of the model, the verification (usually accomplished by comparison with historical data) is a milestone in the MFA process. Based on the historical data, results are evaluated to check whether the simulation captures the influence of the key parameters on the underling process qualitatively or even quantitatively. Afterwards, future scenarios can be investigated, for instance, variations in the production amount or changes of the product portfolio. Finally, with the help of these results, proposals for an increase of production efficiency such as the elimination of bottlenecks, new production strategies as product order or better resource utilisation can be developed and evaluated. Following a short general introduction, three representative examples of MFA at BASF SE illustrating typical applications are discussed. Since the applied methods are of common use, the presentation focusses on the underling questions, the data basis and boundary conditions as well as the effort to gain robust results. In the order of increasing complexity and level of model detail, the MFA examples focus on the following topics:
Example 1: Interwoven production of two products – static MFA Example 2: Raw material delivery and its influence on the production – dynamic
MFA Example 3: Simulation of an entire batch production plant – static and dynamic
MFA
2 Examples
2.1 Example 1: Interwoven Production of two Products – Static MFA
For MFA of this interwoven production three main questions should be answered:
Which production order is required to be reliable for the customers demand? What is the maximal annual production of products? Is it possible to make the calculations with a model/tool that is as complex as
needed yet easy to operate?
In addition, it was requested to make the calculation model adaptable for certain changes in boundary conditions and for further and repeated use during planning. Keeping static and time discrete calculation to be as easy as possible, MS Excel was chosen as an established tool.
Material
Figure 1
Figure 1differentcapacitiemaintain
Necessarthe analyfrom thebehaviorparameteThe out productishutdowcustomerare takemanually
2.2
For MFA
Whic What Whic
maxi
Flow Analys
1: Outline – In
1 shows the it capacities ofes and the paran reliability for
ry assumptionysis process. Oe minimal batr or stochasticers and their coming resu
on plan depewns of the plan
r demands anden into accouy.
Example 2Productio
A of this proce
ch is the maximt is the order och improvemeimum?
sis at BASF - A
nterwoven pro
interwoven prf parts in the allel usage of r the customer
ns and simpliOne main assutch time of onc breakdownsinfluences on
ult is an MS ending on the nt. As a gened a limited capunt. Further,
2: Raw Matn – Dynam
ess three main
mal annual prof bottlenecksents in the p
A Huge Varie
oduction chain
roduction. Thproduction chcertain reactors demand ind
ifications resuumption is thene reactor. Ths in the procen the productiExcel-tool thforecast of t
eralised rule, pacity of the p
non-general
erial Delivemic MFA
n questions sh
roduction volus by raising throduction cha
ety in Complex
n
he main challhain, the limi
ors. The main dependent of i
ult from sevee time discrethat way, espeess are neglecion chain cou
hat automaticathe customersa certain prodproduction capised knowled
ery and its
ould be answe
ume? e output of theain are possib
xity
lenges result ited buffer angoal is to conits seasonality
eral discussionisation of 12 h
ecially a discrted. However
uld still be coally generatess demand andduction order,pacity and budge can be
Influence
ered:
e products? ble to raise t
331
from the nd storage ntinuously y.
ns during h derived
rete event r, the key onsidered. s a rough d planned , variable ffer tanks specified
on the
the actual
332
Figure 2
The startmaterial plant. Tmaterial.fixed timone and are proddownstreprocess,
At somefirst exaproductidelivery of the pldiscretisfine and supply otimes resby GPS informathence comaneuvedelays cblock thvariationdifferentsupply clow buff
To sum chain, o
2: Outline of t
ting point of is delivered
The loading s. Further a ma
me slots. The a half day ful
duced in cameam process another buffe
points, the coample. This on chain and reliability but
lant. Because ation would lhence would
of raw materiasulting from u
tracking of tion the time fovering behaver the rail carcan cause the he loading stns during the t loading timeircle. Finally,
fer capacities,
up: all unceron the storage
the production
this productioacross severa
station cannoaneuvering ofstation is con
ll production qmpaigns having
started againer tank and a f
omplexity of tis mainly atto the more d
t on the raw mof the low bulead to a relatd not be suitabals by rail carunexpected deall rail cars,
for the delivervior in the pars, and the un
missing of atation, until tunloading pro
es caused by ththe productiolead to a bloc
rtainties have e capacity an
Feist, Mar
n chain
on chain is a ral intermediat
ot be used ef the rail cars nnected to a rquantity in cag different ra
n with a bufffilling station,
this problem ittributed to tdetailed quest
material supplyuffer capacity tively high estble. The stochrs. One usuallelays. The dat, collected dury circle was mast. The loadinnloading procea maneuver sthe next freeocess result frhe outer temp
on contains unckage of the lo
a significant nd in the end
rkus; Fernánd
rail car supplyte stops to a xclusively fois possible on
raw material apacity. In the aw material ufer tank follosee figure 2.
is greater thanthe many untions. Here, thy and the maxand the presentimation error
hastic behavioly has to accota basis for thuring the lastmodeled usingng station hasess shows varlot. Consequee slot for marom few breaperature and thnforeseeable boading station
influence on d on the max
dez Alcalde, A
y circle. Hereloading stati
or supplying nly twice a datank of approreactors, two
utilisation facowed by a co
n the complexncertainties ache focus is noximal possiblent uncertaintier - or it has toor mainly occuount for variahis analysis is t half year. Wg the same diss defined timeriable durationently, the railaneuvering. T
akdowns and ahe waiting tim
breakdowns th.
the whole prximal outcom
Ana Maria
e, the raw on at the this raw
ay during oximately o products tors. The ontinuous
xity of the cross the ot laid on e outcome es, a time o be very urs at the
able cycle gathered With this stribution, e slots to ns. These l car will The time also from
mes in the hat, due to
roduction me of the
Material Flow Analysis at BASF - A Huge Variety in Complexity 333
products. Consequently, a detailed dynamic model was required. Such a model was implemented using the software INOSIM, (Hellenkamp and Balling 2012). It is used to investigate the influence of an additional loading station, a change in the campaign production, more buffer capacity etc. on the product chain. In a final step, the results of the scenarios were evaluated with respect to the benefit for the production and the best investment strategy.
2.3 Example 3: Simulation of an Entire Batch Production Plant – Static and Dynamic MFA
For MFA of the whole plant three main questions should be answered:
Where are the bottlenecks of the production? Can additional buffer tanks reduce waiting times? What improvements could be done in the logistics to get a higher overall
capacity?
Figure 3: Illustration of the production with main boundary conditions
The entire production plant consists of 20 batch reactors of different sizes, clustered in 5 classes. In these, approximately 100 products are produced from 20 major raw materials. The products are stored in several dedicated or swing tanks and containers and are delivered to customers via filling stations as packed or via loading station as bulk ware, respectively. The production is not linear for all products, i. e., some products are produced from intermediate products that are also stored in product tanks. Each product can be produced in several reactors and even parallel, if needed. In addition, they can have different batch sizes and batch times.
The initial step of the MFA was a detailed data analysis. In order to gain a deep understanding of the plant and the production rules as well as of the planning process, regularly interviews with plant operators and planning personnel were conducted. The batch times and batch sizes for each product were analysed by using process data from a PCS system. Here, a huge variance in the batch times was dis-covered, related to waiting times due to full storage tanks or blocked loading
2 stations forliquid filling
~20 raw material
~100 products
5 reactor classes20 reactors
~20 dedicated tanks
~20 swing tanks
4 loading and unload. stations
Products: • 5-10 production ways• No dedicated reactors• Intermediate products• Variable batch time/size
Reactors: • Also used as buffer• Frequent cleaning
Load./unloading stations: • Packed or bulk• Variable filling capacities
Tanks and containers: • Swing tanks
Challenges
334 Feist, Markus; Fernández Alcalde, Ana Maria
stations. Therefore, one main task was to separate waiting times from production times to define the ”golden batch” for each product. Due to the high flexibility in the production and since the whole complexity of the production had to be covered, the generalisation of the production rules was another challenge. Finally, the product demands and seasonality were analysed by using historical selling data. Altogether lead to the basis of a high detailed discrete event model of the whole production and loading process.
The simulation model was able to manage over 100 products, up to 10 different production paths, variable batch times and sizes as well as dedicated and swing tanks. This high level of detail was necessary in order to account for the high flexibility and complexity of the real production process. By validation it has been proved that the model was as detailed as necessary to approximate the historical production sufficiently and to correctly describe the influence of the key parameters.
In a final step, the model was used to calculate scenarios to work out and quantify possible improvements. As a main result it turned out that a low filling capacity during the weekend leads to a product accumulation after approximately one day. Consequently, the reactors are used as buffers and have to wait. In addition, the production during the first days of the week is affected catching up these product amounts. Finally, a higher filling capacity during a weekend was assumed and its improvement was quantified with respect to the overall production amount.
3 Summary In summary, this presentation gives an overview on how different ‘Material Flow Analysis’ topics at BASF SE really are. Selected examples and results illustrate the variety of questions and constraints, like complexity, level of detail and data basis. Special attention was paid to stress the different approaches with respect to the questions and the final usage of the MFA results. The examples show that a static MFA is used for not very complex problems or as basis for validation of a complex and dynamic MFA. In the first example straightforward calculations lead to appropriate results. The main advantage is the relatively low effort to gain suitable statements as basis for best decisions. The use of a dynamic MFA is shown in the second example where unexpected events can cause big influence on the final results. There a static MFA would lead to an over- or underestimation of the capacity of the considered parts. The combination of static and dynamic MFA is used if the underlying conditions are very complex. A static MFA is needed to achieve a well analysed data basis. In addition static calculation results are used for validation of simulation results to be sure that the model is appropriate to cover the main influencing parameters.
Overall, the discussed case studies illustrate that the MFA provides a deep insight and understanding in the status quo of the production, and it embodies a comprehensive analysis basis for future decisions and investments.
Material Flow Analysis at BASF - A Huge Variety in Complexity 335
References Hellenkamp, T.; Balling, P.: Production Logistics Analysis with INOSIM
Professional In: Laroque, C.; Himmelspach, J.; Pasupathy, R.; Rose, O..; Uhrmacher, A.M. eds.: Proceedings of the 2012 Winter Simulation Conference (WSC), Berlin (Germany), 09.-12. December 2012.
Jahangirian, M; Eldabi, T.; Naseer, A. Stergioulas, A. N.; Young, T.: Simulation in manufacturing and business: A Review. European Journal of Operational Research 203 (2010) 1-13