- Email: [email protected]
A Simulation Framework for the Evaluation of Production Planning and Order Management Strategies in the Sawmilling Industry
Proceedigs of the 15th IFAC Symposium on Proceedigs of the 15th IFAC Symposium on Proceedigs the IFAC on Information of Control Problems in Manufacturing Proceedigs the 15th 15th IFAC Symposium Symposium on Information of Control Problems in Manufacturing Available online at www.sciencedirect.com Information Control Problems in Manufacturing Manufacturing Proceedigs of theOttawa, 15th IFAC Symposium on May 11-13, 2015. Canada Information Control Problems in May 11-13, 2015. Ottawa, Canada May 11-13, 2015. Ottawa, Canada Information Control Problems in Manufacturing May 11-13, 2015. Ottawa, Canada May 11-13, 2015. Ottawa, Canada
ScienceDirect
IFAC-PapersOnLine 48-3 (2015) 622–627
A Simulation Framework for the Evaluation of Production Planning A Framework for the Evaluation of Production Planning A Simulation Simulation Framework for the Evaluation of Production Planning and Order Management Strategies in the Sawmilling Industry A Simulation Framework for the Evaluation of Production Planning and Order Management Strategies in the Sawmilling Industry and Order Management Strategies in the Sawmilling Industry and Order Management Strategies in the11 Sawmilling Industry 1 2 1 2 Ludwig DUMETZ , Jonathan GAUDREAULT , André THOMAS2,, Ludwig DUMETZ GAUDREAULT 1, Jonathan 1, André THOMAS 1 1 2, 1 1 2 Ludwig DUMETZ , Jonathan Jonathan GAUDREAULT , André THOMAS 1, NadiaGAUDREAULT 1, Hind 2 Ludwig DUMETZ , THOMAS Philippe MARIER LEHOUX EL-HAOUZI 1 1, André 2, Philippe MARIER , Nadia LEHOUX , Hind EL-HAOUZI 1 1 2 1 1 2 Ludwig DUMETZ , Jonathan GAUDREAULT , André THOMAS , Philippe MARIER MARIER ,, Nadia Nadia LEHOUX LEHOUX ,, Hind Hind EL-HAOUZI EL-HAOUZI Philippe 1 1 2 1 Philippe MARIER , Nadia LEHOUX , Hind EL-HAOUZI 1 FORAC Research Consortium, Université Laval, Québec, Canada
2 2 22 2
11 FORAC Research Consortium, Université Laval, Québec, Canada FORAC Consortium, Laval, ([email protected]; [email protected]; [email protected]; FORAC Research Research Consortium, Université Université Laval, Québec, Québec, Canada [email protected]) ([email protected]; [email protected]; [email protected]; [email protected]) 1 FORAC Research Consortium, Université Laval, Québec, [email protected]) ([email protected]; [email protected]; [email protected]; [email protected]) CRAN, Centre de Recherche en Automatique de Nancy, France ([email protected]; ([email protected]; [email protected]; [email protected]; CRAN, Centre de Recherche en Automatique de Nancy, France ([email protected]; [email protected]) [email protected]) ([email protected]; [email protected]; [email protected]; [email protected]) CRAN, Centre de [email protected]) CRAN, Centre de de Recherche Recherche en en Automatique Automatique de Nancy, Nancy, France France ([email protected]; ([email protected]; [email protected]) CRAN, Centre deRaw Recherche en Automatique de Nancy, France ([email protected]; [email protected]) Abstract: material heterogeneity, complex transformation processes, and divergent product flows
Abstract: Raw material heterogeneity, complex transformation processes, and divergent product flows Abstract: Raw heterogeneity, transformation processes, and divergent product flows Abstract: Raw material material heterogeneity, complex transformation processes, andsawmills divergent product flows make sawmilling operations difficult to to complex manage. Most Most north-American lumber sawmills apply a make-tomake-tomake sawmilling operations difficult manage. north-American lumber apply a Abstract: Raw material heterogeneity, complex transformation processes, and divergent product flows make sawmilling operations difficult to manage. Most north-American lumber sawmills apply a make-tomake sawmilling operations difficult to manage. Most north-American lumber sawmills apply a make-tostock production strategy, some accepting/refusing orders according to available-to-promise (ATP) stock production strategy, some accepting/refusing orders according to (ATP) make operations difficult to manage. Most north-American lumber sawmills apply framework a make-tostock sawmilling production strategy, some accepting/refusing orders according to available-to-promise available-to-promise (ATP) stock production strategy, some accepting/refusing orders according to available-to-promise (ATP) quantities, while a few uses more advanced approaches. This article introduces a simulation quantities, while few uses more approaches. This article introduces aa simulation framework stock production some accepting/refusing to as available-to-promise (ATP) quantities, while aaa strategy, few uses more advanced advanced approaches. This according article introduces simulation framework quantities, while few more advanced This article introduces a simulation framework allowing comparing anduses evaluating different approaches. productionorders planning strategies well as order order management management allowing comparing and evaluating different production planning strategies as well as quantities, while a few uses more advanced approaches. This article introduces a simulation framework allowing comparing and evaluating evaluating different production planning strategies as well well as as order management management allowing and different production strategies as order strategies.comparing A basic basic ERP system is is also integrated into planning the framework (inventory management, lumber strategies. A system integrated into the (inventory management, lumber allowing comparing and evaluating different production strategies as well as orderthemanagement strategies. Aplanning basic ERP ERP system ATP is also also integrated into planning the framework framework (inventory management, lumber strategies. A basic ERP system is also integrated into the framework (inventory management, lumber production algorithms, and CTP calculation, etc). The user can configure production production algorithms, ATP CTP calculation, etc). The user can configure the production strategies. Aplanning basic management ERP system process, is alsoand integrated intohow the they framework (inventory management, lumber production planning algorithms, ATP and CTP calculation, etc). The user can the production planning algorithms, ATP and CTP calculation, etc).will Theperform user canin configure the production production planning and and order and evaluate will perform inconfigure various market contexts planning order management process, and evaluate how they various market contexts production planning algorithms, ATP and CTP calculation, etc). The user can configure the production planning and order management process, and evaluate how they will perform in various market contexts planning orderevent management process, using the the and discrete event simulation model.and evaluate how they will perform in various market contexts using discrete simulation model. planning and orderevent management process, and evaluate how they will perform in various market contexts using the discrete simulation model. using the discrete event simulation model. Keywords: Production planning strategies, lumber, order order management, simulation © 2015, (International Federation of Automatic Control) Hosting by simulation Elsevier Ltd. All rights reserved. using theIFAC discrete eventplanning simulation model. Keywords: Production strategies, lumber, management, Keywords: Keywords: Production Production planning planning strategies, strategies, lumber, lumber, order order management, management, simulation simulation Keywords: Production planning strategies, lumber, order management, simulation strategies are used in order to verify the model. In Section 5, strategies are used order to 1. INTRODUCTION strategies are used in inand order to verify verify the the model. model. In In Section Section 5, 5, 1. INTRODUCTION strategies are used in order to verify the model. In Section 5, results are presented analysed. 1. INTRODUCTION results are presented and analysed. 1. INTRODUCTION strategies are used inand order to verify the model. In Section 5, results are presented analysed. results are presented and analysed. Sawmilling is a process difficult to manage. Raw material 1. INTRODUCTION Sawmilling is aa process difficult to manage. Raw material results are presented and analysed. Sawmilling is difficult to Raw material Sawmilling isfrom a process process difficult to manage. manage. Raw material (log) comes the forests and shows aa great diversity in (log) comes from the forests and shows great diversity in Sawmilling is a process difficult to manage. Raw material (log) comes from the forests and shows a great diversity in (log) comes fromquality, the forests and shows a etc. greatThe diversity in terms of wood diameter, length, sawmill 2. BACKGROUND terms of wood diameter, length, The sawmill 2. BACKGROUND (log) comes fromquality, the forests and shows a etc. great diversity in terms of wood quality, diameter, length, etc. The sawmill 2. BACKGROUND terms of wood quality, diameter, length, etc. The sawmill 2. BACKGROUND must take into account this heterogeneity while trying to must take into account this heterogeneity while trying to Lumber production is a three phase manufacturing process. terms of wood quality, diameter, length, etc. The sawmill 2. BACKGROUND must take into account this heterogeneity while trying to Lumber production is a three manufacturing process. must take into account this heterogeneity while trying to maximize produced value and/or meet customer expectations. Lumber production is aa three threeetphase phase manufacturing process. maximize produced value and/or meet customer expectations. Lumber production is phase manufacturing process. As described by Gaudreault al. (2010), it involves three must take into account this heterogeneity while trying to maximize produced produced value and/or meet customerreasons. expectations. As described by Gaudreault et al. (2010), it involves three maximize value and/or meet customer expectations. Satisfying demand is difficult for the following First, Lumber production is a three phase manufacturing process. As described by Gaudreault et al. (2010), it involves three Satisfying demand is difficult for the following reasons. First, As described by Gaudreault et al. (2010), it involves three facilities. First, the sawing unit is responsbile for sawing logs maximize produced value and/or meet customer expectations. Satisfying demand is is difficult for the theatfollowing following reasons. First, facilities. First, the sawing unit is responsbile for sawing logs Satisfying demand difficult for reasons. First, sawing generates many products the same time (i.e., As described by Gaudreault et al. (2010), it involves three facilities. First, the sawing unit is responsbile for sawing logs sawing generates many products at the same time (i.e., facilities. First, the sawing unit is responsbile for sawing logs into green rough lumber according to different cutting Satisfying demand is difficult for the following reasons. First, sawing generates generates many products at at the the which same cannot time (i.e., (i.e., into green rough lumber according to different cutting sawing many products same time divergent process with co-production), be facilities. First, the sawing unit is responsbile for sawing logs into green rough lumber according to different cutting divergent process with co-production), which cannot be into green rough lumber according to different cutting patterns. At this step, produced lumber vary in quality sawing generates many products at the same time (i.e., divergent process with co-production), which cannot be At this step, produced lumber vary in quality divergent process with co-production), which cannot be patterns. avoided (Wery et al. 2012). Many researchers have proposed into green rough lumber according to different cutting patterns. At this step, produced lumber vary in quality avoided (Wery et al. 2012). Many researchers have proposed At thisand step, producedThen, lumber vary in must quality (grade), length, dimension. the lumber be divergent process co-production), which cannot be patterns. avoided (Wery et 2012). Many have proposed (grade), length, dimension. Then, the lumber be avoided to (Wery et al. al.with 2012). Many researchers researchers havecompanies proposed models optimize lumber production. However, patterns. Ata kiln thisand step, produced lumber vary in must quality (grade), length, and dimension. Then, the moisture lumber must be models to optimize lumber production. However, companies and dimension. Then, lumber must be (grade), length, dried using unit in order to reduce the content. avoided (Wery et al. 2012). Many researchers have proposed models to optimize lumber production. However, companies dried using a kiln unit in order to reduce the moisture content. models to optimize lumber production. However, companies do not necessarily know the best way to integrate these (grade), length, and dimension. Then, the lumber must be dried using a kiln unit in order to reduce moisture content. do not necessarily know the best way to integrate these dried using a kiln unit in order to reduce the moisture content. This step is necessary to use the lumber in construction models optimize lumber production. However, companies do not necessarily know the best way to these step is necessary to useto the lumber in construction do not to necessarily know the best way to integrate integrate these This optimization models within their management processes. dried unit in order reduce theYan moisture content. This step is necessary to the lumber in construction optimization models within their management processes. This using step(Wery isa kiln necessary to use use the lumber in et construction industry et al. 2014). According to al. (2001), do not necessarily know the best way to integrate these optimization models within their management processes. industry (Wery et al. 2014). According to Yan et al. (2001), optimization models within their management processes. This step is necessary to use the lumber in construction industry (Wery et al. 2014). According to Yan et al. (2001), industry (Wery et al. 2014). According to Yan et al. (2001), This paper describes a simulation framework developed to drying operation is crucial to ensure quality (by reducing optimization models within their management processes. This paper describes a simulation framework developed to drying operation is crucial to ensure quality (by reducing industry (Wery et is al. 2014). According to Yanstability) et al.reducing (2001), This paper describes aa simulation framework developed to drying operation crucial to ensure quality (by drying operation is crucial to ensure quality (by reducing This paper describes simulation framework developed to compare and evaluate different planning and order biological damage, by increasing dimensional while compare evaluate different planning and order damage, increasing dimensional stability) while This paperand describes a simulation framework developed to biological drying operation isby crucial toThe ensure quality (by reducing compare and evaluate different planning and order biological damage, by increasing dimensional stability) while biological damage, by increasing dimensional stability) while compare and evaluate different planning and order management strategies. Each strategy is defined by: the reducing transportation cost. final step is conducted by reducing transportation cost. The final step is conducted by management strategies. Each strategy is defined by: the compare and evaluate different planning and order biological damage, by increasing dimensional stability) while management strategies. Each strategy is defined by: the reducing transportation cost. The final step is conducted by reducing transportation cost. The final step is conducted by management strategies. Each strategy is defined by: the production planning models used, the size of the planning the planing unit to obtain the desired surface and thickness. the planing unit to obtain the desired surface and thickness. production planning models used, the size of the planning management strategies. Eachused, strategy is order defined by: the the reducing transportation The final step and is conducted production planning models the of planning unit the surface thickness. the planing planing unit to to obtain obtaincost. the desired desired surface and thickness. by production planning models used, the size size of the the planning horizon, the re-planning frequency, and the acceptation horizon, the re-planning frequency, and the acceptation Many optimization models been developed to production models used, the size of the planning the planing unit to obtain thehave desired surface and thickness. horizon, theplanning re-planning frequency, and the order order acceptation Many optimization models have been developed to support support horizon, the re-planning frequency, and the order acceptation criteria (which can be based on stock levels, ATP, CTP, etc). Many optimization models have been developed to support criteria (which can be based on stock levels, ATP, CTP, etc). Many optimization models have been developed to support decision making process in the lumber industry. They lead horizon, the re-planning frequency, and the order acceptation criteria (which (which can can be be based based on on stock stock levels, levels, ATP, ATP, CTP, CTP, etc). etc). decision making process in the lumber industry. They lead to to criteria Many optimization models have been developed to type support decision making process in the lumber industry. They lead to decision making process in the lumber industry. They lead to These strategies can be evaluated for different market optimal or near-optimal solutions. The aim of this of criteria (which can be based on stock levels, ATP, CTP, etc). optimal or near-optimal solutions. The aim of this type of These strategies can be evaluated for different market decision making processmaximize in the lumber lead of to These can be for different market or near-optimal solutions. The aim of this type optimal or near-optimal solutions. Theindustry. aim of They thiscosts. type of These strategies strategies can be evaluated evaluated for different market optimal conditions in order to answer questions such as: What control optimization is value or minimize optimization is often often to to maximize value oraim minimize costs. conditions in order to answer questions such as: What control These strategies can be evaluated for different market optimal or near-optimal solutions. The of this type of conditions in order to answer questions such as: What control optimization is often to maximize value or minimize costs. conditions in order to answer questions such as:should What control strategy should be used for this market? What be the optimization is often to maximize value or minimize costs. strategy be used for this market? What should be and et (2014) tactical conditions in order to answer questions such as: What control optimization is often to maximize minimize aacosts. strategy should should beand used forplanning this market? What should be the the Marier Marier (2011) (2011) and Marier Marier et al. al. value (2014)orproposed proposed tactical strategy should be used for this market? What should be the planning horizon the interval to improve Marier (2011) and Marier et al. (2014) proposed aa tactical planning horizon and the planning interval to improve the Marier (2011) and Marier et al. (2014) proposed tactical MIP model integrating production (sawing, drying, strategy should be used for this market? What should be the planning horizon and and of thethe planning interval to improve improve MIP model integrating production (sawing, drying, planning horizon the planning interval to the financial performance company? If reducing the leadMarier (2011) and Marier etproduction al. (2014) proposed a tactical MIP model integrating (sawing, drying, financial performance of the company? If reducing the leadMIP model integrating production (sawing, drying, finishing), sales, and distribution. A Sales and Operation planning horizon and the planning interval to improve the financial performance of would the company? company? If reducing reducing the leadleadfinishing), sales, and distribution. A Sales and Operation financial performance of the If the time was possible, what be the rate of acceptance for MIP model production (sawing, drying, finishing), sales, and distribution. A and time was possible, what would be the rate of acceptance for finishing), sales,integrating and distribution. A Sales Sales and Operation Operation Planning approach is to sales, financial performance ofwe the re-schedule company? Ifmore reducing the when leadtime was possible, what would be of for Planning (S&OP) (S&OP) approach is used used to correlate correlate sales, time was possible, what would be the the rate rate of acceptance acceptance for Planning new orders? Should often finishing), sales, and distribution. A Sales and Operation (S&OP) approach is used to correlate sales, new orders? Should we re-schedule more often when Planning (S&OP) approach is used to correlate sales, marketing, procurement, production, and finance, so as time was possible, what would be the rate of acceptance for new orders? Should we re-schedule more often when marketing,(S&OP) procurement, production, andtofinance, so sales, as to to new orders? Should we re-schedule more often when business activities are increased? Planning approach is used correlate marketing, procurement, production, and finance, so as to business activities are increased? marketing, procurement, production, and finance, so as to create an annual plan that takes into consideration different new orders? Should we re-schedule more often when business activities are increased? create an annual plan thatproduction, takes into consideration different business activities are increased? marketing, procurement, and finance, so as to create annual that into different create an anfamilies. annual plan plan that takes takes into consideration consideration different This paper is organized as follows: Section 22 presents a product A tactical planning was business activities are increased? product families. A similar similar tactical planning model model was This paper is as follows: Section create anfamilies. annual plan that takes into consideration different This paper is organized organized as to follows: Section 2 presents presents aaa proposed product A similar tactical planning model was product families. A similar tactical planning model was This paper is organized as follows: Section 2 presents review of existing tools used support the decision-making by Singer et al. (2007) for the Chilean sawmilling proposed by Singer et al. (2007) for the Chilean sawmilling review of existing tools used to support the decision-making This is organized as ofto follows: Section 2 presents a industry. product families. Aet similar tactical planning model was reviewpaper offorexisting existing tools used to support the decision-making proposed by Singer al. (2007) for the Chilean sawmilling proposed by Singer et al. (2007) for the Chilean sawmilling review of tools used support the decision-making process different stages a lumber production system. industry. by Singer et al. (2007) for the Chilean sawmilling process different stages of a lumber production system. review offor existing tools used the decision-making proposed process for different stages of lumber production system. industry. industry. process for different stages oftoaasupport lumber production system. Section 3 introduces the simulation framework. Section 4 Section 3 introduces the simulation framework. Section 4 At process for different stages of a lumber production system. industry. Section 3 introduces the simulation framework. Section 4 At the the operational operational level, level, Gaudreault Gaudreault et et al. al. (2010) (2010) proposed proposed Section 3 introduces the simulation framework. Section 4 presents a case study used to demonstrate how the framework At the operational level, Gaudreault et al. (2010) presents a case study used to demonstrate how the framework At the operational level, Gaudreault et al. (2010) proposed proposed three MIP models that can be used to plan/schedule sawing, Section 3 introduces the simulation framework. Section 4 presents a case study used to demonstrate how the framework three MIP models that can be used to plan/schedule sawing, presents aused case to study used todifferent demonstrate how theVery framework can be compare strategies. basic At theMIP operational level, Gaudreault etplan/schedule al. (2010) proposed three models that can be used to sawing, can be used to compare different strategies. Very basic three MIP models that can be used to plan/schedule sawing, presents a case study used to demonstrate how the framework can be used to compare different strategies. Very basic can be used to compare different strategies. Very basic three MIP models that can be used to plan/schedule sawing, can be used to compare different strategies. Very basic
2405-8963 © IFAC (International Federation of Automatic Control) Copyright © 2015, 2015 IFAC 654Hosting by Elsevier Ltd. All rights reserved. Copyright 2015 responsibility IFAC 654Control. Peer review© of International Federation of Automatic Copyright ©under 2015 IFAC IFAC 654 Copyright © 2015 654 10.1016/j.ifacol.2015.06.151 Copyright © 2015 IFAC 654
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
drying, and wood finishing (planing) operations. The objective function allows maximizing production value and/or minimizing orders lateness. A basic coordination mechanism (heuristic) is provided to synchronize those plans. Improved coordination mechanisms are proposed in Gaudreault et al. (2009) and Gaudreault et al. (2012). A stochastic version of the sawing operations planning was developed by Kazemi-Zanjani et al. (2013). An improved version of the drying model was also proposed in Gaudreault et al. (2011).
623
time between the order arrival and the delivery date D (Tony Arnold et al. 2010). Each order can be either accepted or rejected (2) according to a given policy. If the order is rejected, it leaves the system. If it is accepted, it waits until delivery date and material availability (3). The order is then shipped (4).
Even though the previous optimization models show many benefits, they still involve many challenging issues such as how they should best be used by a specific company evolving in a specific market context. Each company/production unit should put in place an operation management system integrating (1) optimization models and algorithms; (2) business processes and policies. To deal with this issue, discrete-event simulation can be used to test different scenarios and show how the different changes in the operating environment will impact the performance of the organization. Discrete-event simulation can be used in such context. For example, El Haouzi et al. (2008) used discrete-event simulation to compare different manufacturing system in a company implementing Demand Flow Technology (Costanza J. 1996). In Abdel-Malek et al (2005), the authors compared different supply chain outsourcing strategies. The key performance indicators used were the inventory levels and the total cost.
3.
Figure 1: Conceptual representation of the framework The ERP system is in charge of the planning production (a) using a model from Marier et al. (2014). The ERP also offers services for computing volumes that are available to promise (ATP) (b) and capable to promise (CTP) (c), while managing a list of accepted orders (d) and inventories (e).
SIMULATION FRAMEWORK
The simulation model “calls” the ERP each time a planning is needed, a new order is accepted, or when ATP, CTP or inventory information is needed.
The framework presented here allows comparing and evaluating different planning and orders management strategies. Each strategy is defined by: the production planning models used, the size of the planning horizon, the re-planning frequency, and the order acceptation criteria (which can be based on stock levels, ATP, CTP, etc.).
Parameters of the model specify the simulation horizon, the planning horizon, and the re-planning frequency. The user also needs to specify which policy should be used to accept/refuse an order. The order can be accepted based on current stocks, ATP, or CTP.
These strategies can be evaluated for different market conditions (order arrival rate per product, order size, demand lead time, etc.)
3.2- Order acceptation policies
A discrete event simulation model is developed using SIMIO. The user can therefore define scenarios visually (i.e. configure its operations management framework and market conditions). The simulation model is also connected to a basic ERP system (inventory management, lumber production, planning algorithms, ATP and CTP calculation, etc) we developed.
Stock: a tentative order of size Q is accepted if current inventory I minus the sum of commitments (accepted orders not delivered yet) is greater than or equal to Q. ATP: an order is accepted if Q ≤ Minimum foreseen stock after order due date t
D−1
Q ≤ I + ∑ (Pt − Et ) − max {∑(Ek − Pk )}
3.1- Simulation framework description
t=now
A conceptual representation of the framework is provided in Figure 1.
D≤t≤T
k=D
Where D is the order due date, T is the simulation horizon and I is the current inventory, Pt the production at period t and Et the commitment at period t.
For each product, orders are generated in (1) according to a given arrival rate. Following Ben Ali et al. (2014), orders in the lumber industry typically follow a Poisson distribution. Other distributions are provided to model the size of the order and the demand lead-time. This parameter corresponds to the
CTP: When processing an order, a tentative production plan is computed in order to check if we can satisfy the new order without compromising the previously accepted orders.
655
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
624
AcceptAll: For study/comparison purpose, the model can also be configured to accept all orders. 4.
A total of 450 scenarios are defined. We needed 50 replications to obtain a significant confidence interval (95%). The time needed to run one scenario considering the confidence interval was around 20 seconds, for a total of 150 hours of computation time.
EXPERIMENTS / MODEL VERIFICATION
The following experiment was carried out in order to perform model verification. We tested different scenarios (combination of order acceptation policies, market conditions, and planning parameters) for a case that was small enough for us to anticipate the results.
Although CTP is supported by the framework, it is not part of the experiment/results as it was too computing intensive to provide results on time. When using CTP, one replication needs more than 30 minutes of computation time. That would have increase simulation time by approximately 187 days. However, we have access to a super computer (8000 processors) that will allow us to provide the results in the future.
The simulation horizon covers a full year, each day being divided into 2 production shifts (periods) of 7 hours of work. We consider that enough raw materials are available for the production of finished goods (i.e., infinite supply availability). Each order is for one single product and there are ten different products. The initial state of the model is as follow: the quantity available for each product is set between 50 and 200 MBFM. The starting quantity for each product was chosen to have a little inventory at the beginning of the simulation. Values are multiple of the order size and take into account the importance of each product (i.e., the number of sales of each product in one year). It is possible to have other starting values like previous commitments
5. RESULTS AND ANALYSIS To analyse the results, some key performance indicators have been considered: the simulation model then allows choosing the scenario that may maximize accepted orders, maximize orders delivered on time, minimize inventory, or simply highlight different parameters where an interaction between them occurs.
Table 1 below shows the full factorial design. It defines parameters values for orders acceptation policy, production planning policy, and market conditions.
5.1- Impact of the size of the planning horizon on the accepted volume of orders and inventory levels Figure 2 on the next page shows the impact of the planning horizon size on the total volume of orders accepted, as well as on the average inventory level. The parameters of the model are set for a demand intensity corresponding to 130% of the production capacity, a triangular demand lead-time distribution (1, 2, 3), and a re-planning frequency of 1 week. We show results for ATP, Stock, and AcceptAll orders acceptation policies.
Table 1: Full factorial design
Orders acceptation policy
Production planning policy
Market conditions
Parameters
Level
Value
Orders acceptation policy
3
Stock, ATP, AccepAll
Demand lead time
2
Random triangular (1,2,3)
AcceptAll is utopic because accepted volume exceeds the total capacity while generating backorders. On the other hand, it defines an upper bound for the total volume of accepted orders and a lower bound for the inventory level. As for the policy where we accept orders based on Stocks, it is our lower bound for the total volume of accepted orders and our upper bound for inventory levels.
Random triangular (0.5,1,2) Re-planning frequency
3
1, 2, 3 weeks
Planning horizon size
5
1, 1.5, 2, 3, 4 weeks
Demand Intensity1
5
90, 100, 110, 130, 150 %
Order Size
1
50 MBFM (capacity of a full truck load)
If we look at the volume of orders accepted for ATP, as expected, they are greater than for Stock. Volume of accepted orders for ATP increases with the size of the planning horizon (the smaller the horizon, the more we need to refuse some orders because our production plan and ATP do not reach that point). In our specific case, with a cumulative lead time of 3 weeks and a re-planning frequency of 1 week, there would be no purpose having a planning horizon superior to 4 weeks since no order can be received after the fourth week (although industry often use a longer planning horizon to have a better visibility, as mentioned by Tony Arnold et al., 2010). This result was expected (see Vollman et al. 1997) and contributed to establish the validity of the simulation model. Conversely, the inventory level associated to the ATP policy decreases when the size of the planning horizon increases, until we reach a planning horizon of four weeks. This result is also coherent.
1
Demand intensity is a parameter we defined to express the total number of orders received as a percentage of the production capacity. It is used to define the arrival rate. 656
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
Figure 2: Impact of the size of the planning horizon
Figure 3: Impact of the demand intensity
657
625
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
626
organization (business process reengineering). Some work to integrate a complete tactical planning, differentiate the operation planning (sawing, drying, planing) and have stochastic event in the production and supply, are underway. Therefore, the framework will allow simulating different coordination mechanisms between the tactical and operational planning level, as well as between the different departments (ex: raw material procurement, production and sales). The goal will be to recommend configurations adapted to different market conditions.
Finally, we note that for the ATP policy, even though the accepted volume is only slightly higher than the Stock policy (as the total production capacity remains the same), the reduction of the average inventory is significant (48,5% for a planning horizon of three weeks). 5.2- Impact of the demand intensity We first recall that demand intensity is the total demand expressed as a percentage of the total production capacity. Figure 3 on the previous page shows the total volume of accepted orders and the average inventory according to the demand intensity.
7. REFERENCES
As expected, the greater demand intensity is, the greater the total volume of accepted orders will be. This is true until we reach a point where all the production can be sold. This point is not represented in the figure; for the specific case reported, it was reached at around 170% (the volume of accepted orders is then equalled to the global production capacity). An intensity of 100% of the production capacity would thus not be enough (due to the stochastic environment, demand for some specific products would be less than their production volumes; some orders would have due date outside the simulation horizon; too many orders could have the same due date, forcing the reject for some of them).
Abdel-Malek, L. K. (2005). A framework for comparing outsourcing strategies in multi-layered supply chains. International Journal of Production Economics , 97, p. 318-328. Ben Ali, M., Gaudreault, J., D'Amours, S., & Carle, M.-A. (2014, Octobre). A Multi-Level Framework for Demand Fulfillment in a Make-to-Stock Environment - A Case Study in Canadian Softwood Lumber Industry. Costanza, J. Just-In-Time manufacturing excellence (1996) John Costanza Institute of Technology Inc.; 3 rd Edition. El Haouzi, H., Thomas, A., & Pétin, J.-F. (2008). Contribution to reusability and modularity of Manufacturing Systems Simulation Models: application to distributed control simulation within DFT context. International Journal of Production Economics , 112 (1), p.48-61. Gaudreault J, Frayret. JM. (2011). Combined planning and scheduling in a divergent production system with co-production. Computers and Operations Reasearch , 38(9), p. 1238-1250. Gaudreault J, Frayret. JM. (2009). Distributed search for supply chain coordination. Computers in Industry , 60(6), p. 441- 451. Gaudreault J, Pesant. G, Frayret JM, D’Amours S. (2012). Supply chain coordination using an adaptive distributed search strategy. IEEE Transactions on Systems Man and Cybernetics Part C , 42(6), p. 1424- 1438. Gaudreault, J., Forget, P., Frayet, J.-M., Rousseau, A., Lemieux, S., & D'Amours, S. (2010). Distributed operations planning in the lumber supply chain: models and coordination. International Journal of Industrial Engineering Theory, Applications & Pratice , 17(3): p168-189. Kazemi Zanjani, M., Ait-Kadi, D., & Nourelfath, M. (s.d.). A stochastic programming approach for sawmill production planning. International Journal Of Mathematics in Operational Research, Vol. 5, No. 1, 2013 p. 1-18. Marier, P. (2011). Gestion intégrée des ventes et des opérations dans l'industrie du sciage. ExpoConférence. Université Laval, Canada, Québec. Marier, P., Gaudreault, J., & Robichaud, B. (2014, Novembre 5-7). Implementing a MIP model to plane and schedue wood finishing operation in a sawmill:
Regarding the average inventory level, the greater the intensity of demand is, the smaller the average inventory has to be. This is true for any policy. However, the greater the intensity is, the bigger is the difference between ATP and Stock policies. We recall that AcceptAll policy may look attractive (less inventory and many orders accepted). However, there is a huge number of late deliveries and therefore the customer satisfaction is very poor. By comparison, on-time delivery reaches 39% for AcceptAll, against 100% for Stock strategy and ATP. 6. CONCLUSION This article proposed a simulation framework to compare and evaluate different planning and order management strategies. It also encompasses a basic ERP system that covers inventory management, lumber production, planning algorithms, ATP and CTP calculation. The user can configure the production planning and order management process directly on the framework and then evaluate how they will perform in various market contexts. This tool could be used in a company as a decision-making tool by allowing choosing the right production planning and ordering management strategies. Even though this simulation model is at its first stage, the results of the experiments refer to recognized practices in the literature (Vollman et al. 1997) and are verified. In future work, this framework will be used as the backbone of a more complex study. The goal is to propose guidelines for more agile operations management driven by demand. We need to propose an operation management framework describing how to combine algorithms, humans, and decision processes in order to maximize the overall performance of the 658
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
lessons learned. 10th International Conference of Modeling and Simuling- MOSIM'14 . Singer, M and Donoso, P. (2007). Internal supply chain management in the Chilean sawmill industry. International Journal of Operations & Production Management , 27(5), p. 524-541. Tony Arnold, J.R., Chapman, S., & Clive, L. (2012). Introduction to Materials Management. Pearson, Seventh Edition. Vollmann, T., Berry, W., & Whybark, D. (1997). Manufacturing planning and control for supply chain management. New-York: McGraw-Hill. Wery, J., Gaudreault, J., Thomas, A., & Marier, P. (2012). Improving sawmill agility through log classification. 4th International Conference on Information Systems, Logistics and Supply Chain Québec. Wery, J., Marier, P., Gaudreault, J., & Thomas, A. (2014). Decision-making framework for tactical planning taking into account market opportunities (new products and new suppliers) in a co-production context. MOSIM. Nancy. Yan, G. C. (2001). Experimental modelling and intelligent control of a wood-drying kiln. International journal of adaptive control and signal processing , 15(8), p. 787-814 .
659
627
ScienceDirect
IFAC-PapersOnLine 48-3 (2015) 622–627
A Simulation Framework for the Evaluation of Production Planning A Framework for the Evaluation of Production Planning A Simulation Simulation Framework for the Evaluation of Production Planning and Order Management Strategies in the Sawmilling Industry A Simulation Framework for the Evaluation of Production Planning and Order Management Strategies in the Sawmilling Industry and Order Management Strategies in the Sawmilling Industry and Order Management Strategies in the11 Sawmilling Industry 1 2 1 2 Ludwig DUMETZ , Jonathan GAUDREAULT , André THOMAS2,, Ludwig DUMETZ GAUDREAULT 1, Jonathan 1, André THOMAS 1 1 2, 1 1 2 Ludwig DUMETZ , Jonathan Jonathan GAUDREAULT , André THOMAS 1, NadiaGAUDREAULT 1, Hind 2 Ludwig DUMETZ , THOMAS Philippe MARIER LEHOUX EL-HAOUZI 1 1, André 2, Philippe MARIER , Nadia LEHOUX , Hind EL-HAOUZI 1 1 2 1 1 2 Ludwig DUMETZ , Jonathan GAUDREAULT , André THOMAS , Philippe MARIER MARIER ,, Nadia Nadia LEHOUX LEHOUX ,, Hind Hind EL-HAOUZI EL-HAOUZI Philippe 1 1 2 1 Philippe MARIER , Nadia LEHOUX , Hind EL-HAOUZI 1 FORAC Research Consortium, Université Laval, Québec, Canada
2 2 22 2
11 FORAC Research Consortium, Université Laval, Québec, Canada FORAC Consortium, Laval, ([email protected]; [email protected]; [email protected]; FORAC Research Research Consortium, Université Université Laval, Québec, Québec, Canada [email protected]) ([email protected]; [email protected]; [email protected]; [email protected]) 1 FORAC Research Consortium, Université Laval, Québec, [email protected]) ([email protected]; [email protected]; [email protected]; [email protected]) CRAN, Centre de Recherche en Automatique de Nancy, France ([email protected]; ([email protected]; [email protected]; [email protected]; CRAN, Centre de Recherche en Automatique de Nancy, France ([email protected]; [email protected]) [email protected]) ([email protected]; [email protected]; [email protected]; [email protected]) CRAN, Centre de [email protected]) CRAN, Centre de de Recherche Recherche en en Automatique Automatique de Nancy, Nancy, France France ([email protected]; ([email protected]; [email protected]) CRAN, Centre deRaw Recherche en Automatique de Nancy, France ([email protected]; [email protected]) Abstract: material heterogeneity, complex transformation processes, and divergent product flows
Abstract: Raw material heterogeneity, complex transformation processes, and divergent product flows Abstract: Raw heterogeneity, transformation processes, and divergent product flows Abstract: Raw material material heterogeneity, complex transformation processes, andsawmills divergent product flows make sawmilling operations difficult to to complex manage. Most Most north-American lumber sawmills apply a make-tomake-tomake sawmilling operations difficult manage. north-American lumber apply a Abstract: Raw material heterogeneity, complex transformation processes, and divergent product flows make sawmilling operations difficult to manage. Most north-American lumber sawmills apply a make-tomake sawmilling operations difficult to manage. Most north-American lumber sawmills apply a make-tostock production strategy, some accepting/refusing orders according to available-to-promise (ATP) stock production strategy, some accepting/refusing orders according to (ATP) make operations difficult to manage. Most north-American lumber sawmills apply framework a make-tostock sawmilling production strategy, some accepting/refusing orders according to available-to-promise available-to-promise (ATP) stock production strategy, some accepting/refusing orders according to available-to-promise (ATP) quantities, while a few uses more advanced approaches. This article introduces a simulation quantities, while few uses more approaches. This article introduces aa simulation framework stock production some accepting/refusing to as available-to-promise (ATP) quantities, while aaa strategy, few uses more advanced advanced approaches. This according article introduces simulation framework quantities, while few more advanced This article introduces a simulation framework allowing comparing anduses evaluating different approaches. productionorders planning strategies well as order order management management allowing comparing and evaluating different production planning strategies as well as quantities, while a few uses more advanced approaches. This article introduces a simulation framework allowing comparing and evaluating evaluating different production planning strategies as well well as as order management management allowing and different production strategies as order strategies.comparing A basic basic ERP system is is also integrated into planning the framework (inventory management, lumber strategies. A system integrated into the (inventory management, lumber allowing comparing and evaluating different production strategies as well as orderthemanagement strategies. Aplanning basic ERP ERP system ATP is also also integrated into planning the framework framework (inventory management, lumber strategies. A basic ERP system is also integrated into the framework (inventory management, lumber production algorithms, and CTP calculation, etc). The user can configure production production algorithms, ATP CTP calculation, etc). The user can configure the production strategies. Aplanning basic management ERP system process, is alsoand integrated intohow the they framework (inventory management, lumber production planning algorithms, ATP and CTP calculation, etc). The user can the production planning algorithms, ATP and CTP calculation, etc).will Theperform user canin configure the production production planning and and order and evaluate will perform inconfigure various market contexts planning order management process, and evaluate how they various market contexts production planning algorithms, ATP and CTP calculation, etc). The user can configure the production planning and order management process, and evaluate how they will perform in various market contexts planning orderevent management process, using the the and discrete event simulation model.and evaluate how they will perform in various market contexts using discrete simulation model. planning and orderevent management process, and evaluate how they will perform in various market contexts using the discrete simulation model. using the discrete event simulation model. Keywords: Production planning strategies, lumber, order order management, simulation © 2015, (International Federation of Automatic Control) Hosting by simulation Elsevier Ltd. All rights reserved. using theIFAC discrete eventplanning simulation model. Keywords: Production strategies, lumber, management, Keywords: Keywords: Production Production planning planning strategies, strategies, lumber, lumber, order order management, management, simulation simulation Keywords: Production planning strategies, lumber, order management, simulation strategies are used in order to verify the model. In Section 5, strategies are used order to 1. INTRODUCTION strategies are used in inand order to verify verify the the model. model. In In Section Section 5, 5, 1. INTRODUCTION strategies are used in order to verify the model. In Section 5, results are presented analysed. 1. INTRODUCTION results are presented and analysed. 1. INTRODUCTION strategies are used inand order to verify the model. In Section 5, results are presented analysed. results are presented and analysed. Sawmilling is a process difficult to manage. Raw material 1. INTRODUCTION Sawmilling is aa process difficult to manage. Raw material results are presented and analysed. Sawmilling is difficult to Raw material Sawmilling isfrom a process process difficult to manage. manage. Raw material (log) comes the forests and shows aa great diversity in (log) comes from the forests and shows great diversity in Sawmilling is a process difficult to manage. Raw material (log) comes from the forests and shows a great diversity in (log) comes fromquality, the forests and shows a etc. greatThe diversity in terms of wood diameter, length, sawmill 2. BACKGROUND terms of wood diameter, length, The sawmill 2. BACKGROUND (log) comes fromquality, the forests and shows a etc. great diversity in terms of wood quality, diameter, length, etc. The sawmill 2. BACKGROUND terms of wood quality, diameter, length, etc. The sawmill 2. BACKGROUND must take into account this heterogeneity while trying to must take into account this heterogeneity while trying to Lumber production is a three phase manufacturing process. terms of wood quality, diameter, length, etc. The sawmill 2. BACKGROUND must take into account this heterogeneity while trying to Lumber production is a three manufacturing process. must take into account this heterogeneity while trying to maximize produced value and/or meet customer expectations. Lumber production is aa three threeetphase phase manufacturing process. maximize produced value and/or meet customer expectations. Lumber production is phase manufacturing process. As described by Gaudreault al. (2010), it involves three must take into account this heterogeneity while trying to maximize produced produced value and/or meet customerreasons. expectations. As described by Gaudreault et al. (2010), it involves three maximize value and/or meet customer expectations. Satisfying demand is difficult for the following First, Lumber production is a three phase manufacturing process. As described by Gaudreault et al. (2010), it involves three Satisfying demand is difficult for the following reasons. First, As described by Gaudreault et al. (2010), it involves three facilities. First, the sawing unit is responsbile for sawing logs maximize produced value and/or meet customer expectations. Satisfying demand is is difficult for the theatfollowing following reasons. First, facilities. First, the sawing unit is responsbile for sawing logs Satisfying demand difficult for reasons. First, sawing generates many products the same time (i.e., As described by Gaudreault et al. (2010), it involves three facilities. First, the sawing unit is responsbile for sawing logs sawing generates many products at the same time (i.e., facilities. First, the sawing unit is responsbile for sawing logs into green rough lumber according to different cutting Satisfying demand is difficult for the following reasons. First, sawing generates generates many products at at the the which same cannot time (i.e., (i.e., into green rough lumber according to different cutting sawing many products same time divergent process with co-production), be facilities. First, the sawing unit is responsbile for sawing logs into green rough lumber according to different cutting divergent process with co-production), which cannot be into green rough lumber according to different cutting patterns. At this step, produced lumber vary in quality sawing generates many products at the same time (i.e., divergent process with co-production), which cannot be At this step, produced lumber vary in quality divergent process with co-production), which cannot be patterns. avoided (Wery et al. 2012). Many researchers have proposed into green rough lumber according to different cutting patterns. At this step, produced lumber vary in quality avoided (Wery et al. 2012). Many researchers have proposed At thisand step, producedThen, lumber vary in must quality (grade), length, dimension. the lumber be divergent process co-production), which cannot be patterns. avoided (Wery et 2012). Many have proposed (grade), length, dimension. Then, the lumber be avoided to (Wery et al. al.with 2012). Many researchers researchers havecompanies proposed models optimize lumber production. However, patterns. Ata kiln thisand step, produced lumber vary in must quality (grade), length, and dimension. Then, the moisture lumber must be models to optimize lumber production. However, companies and dimension. Then, lumber must be (grade), length, dried using unit in order to reduce the content. avoided (Wery et al. 2012). Many researchers have proposed models to optimize lumber production. However, companies dried using a kiln unit in order to reduce the moisture content. models to optimize lumber production. However, companies do not necessarily know the best way to integrate these (grade), length, and dimension. Then, the lumber must be dried using a kiln unit in order to reduce moisture content. do not necessarily know the best way to integrate these dried using a kiln unit in order to reduce the moisture content. This step is necessary to use the lumber in construction models optimize lumber production. However, companies do not necessarily know the best way to these step is necessary to useto the lumber in construction do not to necessarily know the best way to integrate integrate these This optimization models within their management processes. dried unit in order reduce theYan moisture content. This step is necessary to the lumber in construction optimization models within their management processes. This using step(Wery isa kiln necessary to use use the lumber in et construction industry et al. 2014). According to al. (2001), do not necessarily know the best way to integrate these optimization models within their management processes. industry (Wery et al. 2014). According to Yan et al. (2001), optimization models within their management processes. This step is necessary to use the lumber in construction industry (Wery et al. 2014). According to Yan et al. (2001), industry (Wery et al. 2014). According to Yan et al. (2001), This paper describes a simulation framework developed to drying operation is crucial to ensure quality (by reducing optimization models within their management processes. This paper describes a simulation framework developed to drying operation is crucial to ensure quality (by reducing industry (Wery et is al. 2014). According to Yanstability) et al.reducing (2001), This paper describes aa simulation framework developed to drying operation crucial to ensure quality (by drying operation is crucial to ensure quality (by reducing This paper describes simulation framework developed to compare and evaluate different planning and order biological damage, by increasing dimensional while compare evaluate different planning and order damage, increasing dimensional stability) while This paperand describes a simulation framework developed to biological drying operation isby crucial toThe ensure quality (by reducing compare and evaluate different planning and order biological damage, by increasing dimensional stability) while biological damage, by increasing dimensional stability) while compare and evaluate different planning and order management strategies. Each strategy is defined by: the reducing transportation cost. final step is conducted by reducing transportation cost. The final step is conducted by management strategies. Each strategy is defined by: the compare and evaluate different planning and order biological damage, by increasing dimensional stability) while management strategies. Each strategy is defined by: the reducing transportation cost. The final step is conducted by reducing transportation cost. The final step is conducted by management strategies. Each strategy is defined by: the production planning models used, the size of the planning the planing unit to obtain the desired surface and thickness. the planing unit to obtain the desired surface and thickness. production planning models used, the size of the planning management strategies. Eachused, strategy is order defined by: the the reducing transportation The final step and is conducted production planning models the of planning unit the surface thickness. the planing planing unit to to obtain obtaincost. the desired desired surface and thickness. by production planning models used, the size size of the the planning horizon, the re-planning frequency, and the acceptation horizon, the re-planning frequency, and the acceptation Many optimization models been developed to production models used, the size of the planning the planing unit to obtain thehave desired surface and thickness. horizon, theplanning re-planning frequency, and the order order acceptation Many optimization models have been developed to support support horizon, the re-planning frequency, and the order acceptation criteria (which can be based on stock levels, ATP, CTP, etc). Many optimization models have been developed to support criteria (which can be based on stock levels, ATP, CTP, etc). Many optimization models have been developed to support decision making process in the lumber industry. They lead horizon, the re-planning frequency, and the order acceptation criteria (which (which can can be be based based on on stock stock levels, levels, ATP, ATP, CTP, CTP, etc). etc). decision making process in the lumber industry. They lead to to criteria Many optimization models have been developed to type support decision making process in the lumber industry. They lead to decision making process in the lumber industry. They lead to These strategies can be evaluated for different market optimal or near-optimal solutions. The aim of this of criteria (which can be based on stock levels, ATP, CTP, etc). optimal or near-optimal solutions. The aim of this type of These strategies can be evaluated for different market decision making processmaximize in the lumber lead of to These can be for different market or near-optimal solutions. The aim of this type optimal or near-optimal solutions. Theindustry. aim of They thiscosts. type of These strategies strategies can be evaluated evaluated for different market optimal conditions in order to answer questions such as: What control optimization is value or minimize optimization is often often to to maximize value oraim minimize costs. conditions in order to answer questions such as: What control These strategies can be evaluated for different market optimal or near-optimal solutions. The of this type of conditions in order to answer questions such as: What control optimization is often to maximize value or minimize costs. conditions in order to answer questions such as:should What control strategy should be used for this market? What be the optimization is often to maximize value or minimize costs. strategy be used for this market? What should be and et (2014) tactical conditions in order to answer questions such as: What control optimization is often to maximize minimize aacosts. strategy should should beand used forplanning this market? What should be the the Marier Marier (2011) (2011) and Marier Marier et al. al. value (2014)orproposed proposed tactical strategy should be used for this market? What should be the planning horizon the interval to improve Marier (2011) and Marier et al. (2014) proposed aa tactical planning horizon and the planning interval to improve the Marier (2011) and Marier et al. (2014) proposed tactical MIP model integrating production (sawing, drying, strategy should be used for this market? What should be the planning horizon and and of thethe planning interval to improve improve MIP model integrating production (sawing, drying, planning horizon the planning interval to the financial performance company? If reducing the leadMarier (2011) and Marier etproduction al. (2014) proposed a tactical MIP model integrating (sawing, drying, financial performance of the company? If reducing the leadMIP model integrating production (sawing, drying, finishing), sales, and distribution. A Sales and Operation planning horizon and the planning interval to improve the financial performance of would the company? company? If reducing reducing the leadleadfinishing), sales, and distribution. A Sales and Operation financial performance of the If the time was possible, what be the rate of acceptance for MIP model production (sawing, drying, finishing), sales, and distribution. A and time was possible, what would be the rate of acceptance for finishing), sales,integrating and distribution. A Sales Sales and Operation Operation Planning approach is to sales, financial performance ofwe the re-schedule company? Ifmore reducing the when leadtime was possible, what would be of for Planning (S&OP) (S&OP) approach is used used to correlate correlate sales, time was possible, what would be the the rate rate of acceptance acceptance for Planning new orders? Should often finishing), sales, and distribution. A Sales and Operation (S&OP) approach is used to correlate sales, new orders? Should we re-schedule more often when Planning (S&OP) approach is used to correlate sales, marketing, procurement, production, and finance, so as time was possible, what would be the rate of acceptance for new orders? Should we re-schedule more often when marketing,(S&OP) procurement, production, andtofinance, so sales, as to to new orders? Should we re-schedule more often when business activities are increased? Planning approach is used correlate marketing, procurement, production, and finance, so as to business activities are increased? marketing, procurement, production, and finance, so as to create an annual plan that takes into consideration different new orders? Should we re-schedule more often when business activities are increased? create an annual plan thatproduction, takes into consideration different business activities are increased? marketing, procurement, and finance, so as to create annual that into different create an anfamilies. annual plan plan that takes takes into consideration consideration different This paper is organized as follows: Section 22 presents a product A tactical planning was business activities are increased? product families. A similar similar tactical planning model model was This paper is as follows: Section create anfamilies. annual plan that takes into consideration different This paper is organized organized as to follows: Section 2 presents presents aaa proposed product A similar tactical planning model was product families. A similar tactical planning model was This paper is organized as follows: Section 2 presents review of existing tools used support the decision-making by Singer et al. (2007) for the Chilean sawmilling proposed by Singer et al. (2007) for the Chilean sawmilling review of existing tools used to support the decision-making This is organized as ofto follows: Section 2 presents a industry. product families. Aet similar tactical planning model was reviewpaper offorexisting existing tools used to support the decision-making proposed by Singer al. (2007) for the Chilean sawmilling proposed by Singer et al. (2007) for the Chilean sawmilling review of tools used support the decision-making process different stages a lumber production system. industry. by Singer et al. (2007) for the Chilean sawmilling process different stages of a lumber production system. review offor existing tools used the decision-making proposed process for different stages of lumber production system. industry. industry. process for different stages oftoaasupport lumber production system. Section 3 introduces the simulation framework. Section 4 Section 3 introduces the simulation framework. Section 4 At process for different stages of a lumber production system. industry. Section 3 introduces the simulation framework. Section 4 At the the operational operational level, level, Gaudreault Gaudreault et et al. al. (2010) (2010) proposed proposed Section 3 introduces the simulation framework. Section 4 presents a case study used to demonstrate how the framework At the operational level, Gaudreault et al. (2010) presents a case study used to demonstrate how the framework At the operational level, Gaudreault et al. (2010) proposed proposed three MIP models that can be used to plan/schedule sawing, Section 3 introduces the simulation framework. Section 4 presents a case study used to demonstrate how the framework three MIP models that can be used to plan/schedule sawing, presents aused case to study used todifferent demonstrate how theVery framework can be compare strategies. basic At theMIP operational level, Gaudreault etplan/schedule al. (2010) proposed three models that can be used to sawing, can be used to compare different strategies. Very basic three MIP models that can be used to plan/schedule sawing, presents a case study used to demonstrate how the framework can be used to compare different strategies. Very basic can be used to compare different strategies. Very basic three MIP models that can be used to plan/schedule sawing, can be used to compare different strategies. Very basic
2405-8963 © IFAC (International Federation of Automatic Control) Copyright © 2015, 2015 IFAC 654Hosting by Elsevier Ltd. All rights reserved. Copyright 2015 responsibility IFAC 654Control. Peer review© of International Federation of Automatic Copyright ©under 2015 IFAC IFAC 654 Copyright © 2015 654 10.1016/j.ifacol.2015.06.151 Copyright © 2015 IFAC 654
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
drying, and wood finishing (planing) operations. The objective function allows maximizing production value and/or minimizing orders lateness. A basic coordination mechanism (heuristic) is provided to synchronize those plans. Improved coordination mechanisms are proposed in Gaudreault et al. (2009) and Gaudreault et al. (2012). A stochastic version of the sawing operations planning was developed by Kazemi-Zanjani et al. (2013). An improved version of the drying model was also proposed in Gaudreault et al. (2011).
623
time between the order arrival and the delivery date D (Tony Arnold et al. 2010). Each order can be either accepted or rejected (2) according to a given policy. If the order is rejected, it leaves the system. If it is accepted, it waits until delivery date and material availability (3). The order is then shipped (4).
Even though the previous optimization models show many benefits, they still involve many challenging issues such as how they should best be used by a specific company evolving in a specific market context. Each company/production unit should put in place an operation management system integrating (1) optimization models and algorithms; (2) business processes and policies. To deal with this issue, discrete-event simulation can be used to test different scenarios and show how the different changes in the operating environment will impact the performance of the organization. Discrete-event simulation can be used in such context. For example, El Haouzi et al. (2008) used discrete-event simulation to compare different manufacturing system in a company implementing Demand Flow Technology (Costanza J. 1996). In Abdel-Malek et al (2005), the authors compared different supply chain outsourcing strategies. The key performance indicators used were the inventory levels and the total cost.
3.
Figure 1: Conceptual representation of the framework The ERP system is in charge of the planning production (a) using a model from Marier et al. (2014). The ERP also offers services for computing volumes that are available to promise (ATP) (b) and capable to promise (CTP) (c), while managing a list of accepted orders (d) and inventories (e).
SIMULATION FRAMEWORK
The simulation model “calls” the ERP each time a planning is needed, a new order is accepted, or when ATP, CTP or inventory information is needed.
The framework presented here allows comparing and evaluating different planning and orders management strategies. Each strategy is defined by: the production planning models used, the size of the planning horizon, the re-planning frequency, and the order acceptation criteria (which can be based on stock levels, ATP, CTP, etc.).
Parameters of the model specify the simulation horizon, the planning horizon, and the re-planning frequency. The user also needs to specify which policy should be used to accept/refuse an order. The order can be accepted based on current stocks, ATP, or CTP.
These strategies can be evaluated for different market conditions (order arrival rate per product, order size, demand lead time, etc.)
3.2- Order acceptation policies
A discrete event simulation model is developed using SIMIO. The user can therefore define scenarios visually (i.e. configure its operations management framework and market conditions). The simulation model is also connected to a basic ERP system (inventory management, lumber production, planning algorithms, ATP and CTP calculation, etc) we developed.
Stock: a tentative order of size Q is accepted if current inventory I minus the sum of commitments (accepted orders not delivered yet) is greater than or equal to Q. ATP: an order is accepted if Q ≤ Minimum foreseen stock after order due date t
D−1
Q ≤ I + ∑ (Pt − Et ) − max {∑(Ek − Pk )}
3.1- Simulation framework description
t=now
A conceptual representation of the framework is provided in Figure 1.
D≤t≤T
k=D
Where D is the order due date, T is the simulation horizon and I is the current inventory, Pt the production at period t and Et the commitment at period t.
For each product, orders are generated in (1) according to a given arrival rate. Following Ben Ali et al. (2014), orders in the lumber industry typically follow a Poisson distribution. Other distributions are provided to model the size of the order and the demand lead-time. This parameter corresponds to the
CTP: When processing an order, a tentative production plan is computed in order to check if we can satisfy the new order without compromising the previously accepted orders.
655
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
624
AcceptAll: For study/comparison purpose, the model can also be configured to accept all orders. 4.
A total of 450 scenarios are defined. We needed 50 replications to obtain a significant confidence interval (95%). The time needed to run one scenario considering the confidence interval was around 20 seconds, for a total of 150 hours of computation time.
EXPERIMENTS / MODEL VERIFICATION
The following experiment was carried out in order to perform model verification. We tested different scenarios (combination of order acceptation policies, market conditions, and planning parameters) for a case that was small enough for us to anticipate the results.
Although CTP is supported by the framework, it is not part of the experiment/results as it was too computing intensive to provide results on time. When using CTP, one replication needs more than 30 minutes of computation time. That would have increase simulation time by approximately 187 days. However, we have access to a super computer (8000 processors) that will allow us to provide the results in the future.
The simulation horizon covers a full year, each day being divided into 2 production shifts (periods) of 7 hours of work. We consider that enough raw materials are available for the production of finished goods (i.e., infinite supply availability). Each order is for one single product and there are ten different products. The initial state of the model is as follow: the quantity available for each product is set between 50 and 200 MBFM. The starting quantity for each product was chosen to have a little inventory at the beginning of the simulation. Values are multiple of the order size and take into account the importance of each product (i.e., the number of sales of each product in one year). It is possible to have other starting values like previous commitments
5. RESULTS AND ANALYSIS To analyse the results, some key performance indicators have been considered: the simulation model then allows choosing the scenario that may maximize accepted orders, maximize orders delivered on time, minimize inventory, or simply highlight different parameters where an interaction between them occurs.
Table 1 below shows the full factorial design. It defines parameters values for orders acceptation policy, production planning policy, and market conditions.
5.1- Impact of the size of the planning horizon on the accepted volume of orders and inventory levels Figure 2 on the next page shows the impact of the planning horizon size on the total volume of orders accepted, as well as on the average inventory level. The parameters of the model are set for a demand intensity corresponding to 130% of the production capacity, a triangular demand lead-time distribution (1, 2, 3), and a re-planning frequency of 1 week. We show results for ATP, Stock, and AcceptAll orders acceptation policies.
Table 1: Full factorial design
Orders acceptation policy
Production planning policy
Market conditions
Parameters
Level
Value
Orders acceptation policy
3
Stock, ATP, AccepAll
Demand lead time
2
Random triangular (1,2,3)
AcceptAll is utopic because accepted volume exceeds the total capacity while generating backorders. On the other hand, it defines an upper bound for the total volume of accepted orders and a lower bound for the inventory level. As for the policy where we accept orders based on Stocks, it is our lower bound for the total volume of accepted orders and our upper bound for inventory levels.
Random triangular (0.5,1,2) Re-planning frequency
3
1, 2, 3 weeks
Planning horizon size
5
1, 1.5, 2, 3, 4 weeks
Demand Intensity1
5
90, 100, 110, 130, 150 %
Order Size
1
50 MBFM (capacity of a full truck load)
If we look at the volume of orders accepted for ATP, as expected, they are greater than for Stock. Volume of accepted orders for ATP increases with the size of the planning horizon (the smaller the horizon, the more we need to refuse some orders because our production plan and ATP do not reach that point). In our specific case, with a cumulative lead time of 3 weeks and a re-planning frequency of 1 week, there would be no purpose having a planning horizon superior to 4 weeks since no order can be received after the fourth week (although industry often use a longer planning horizon to have a better visibility, as mentioned by Tony Arnold et al., 2010). This result was expected (see Vollman et al. 1997) and contributed to establish the validity of the simulation model. Conversely, the inventory level associated to the ATP policy decreases when the size of the planning horizon increases, until we reach a planning horizon of four weeks. This result is also coherent.
1
Demand intensity is a parameter we defined to express the total number of orders received as a percentage of the production capacity. It is used to define the arrival rate. 656
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
Figure 2: Impact of the size of the planning horizon
Figure 3: Impact of the demand intensity
657
625
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
626
organization (business process reengineering). Some work to integrate a complete tactical planning, differentiate the operation planning (sawing, drying, planing) and have stochastic event in the production and supply, are underway. Therefore, the framework will allow simulating different coordination mechanisms between the tactical and operational planning level, as well as between the different departments (ex: raw material procurement, production and sales). The goal will be to recommend configurations adapted to different market conditions.
Finally, we note that for the ATP policy, even though the accepted volume is only slightly higher than the Stock policy (as the total production capacity remains the same), the reduction of the average inventory is significant (48,5% for a planning horizon of three weeks). 5.2- Impact of the demand intensity We first recall that demand intensity is the total demand expressed as a percentage of the total production capacity. Figure 3 on the previous page shows the total volume of accepted orders and the average inventory according to the demand intensity.
7. REFERENCES
As expected, the greater demand intensity is, the greater the total volume of accepted orders will be. This is true until we reach a point where all the production can be sold. This point is not represented in the figure; for the specific case reported, it was reached at around 170% (the volume of accepted orders is then equalled to the global production capacity). An intensity of 100% of the production capacity would thus not be enough (due to the stochastic environment, demand for some specific products would be less than their production volumes; some orders would have due date outside the simulation horizon; too many orders could have the same due date, forcing the reject for some of them).
Abdel-Malek, L. K. (2005). A framework for comparing outsourcing strategies in multi-layered supply chains. International Journal of Production Economics , 97, p. 318-328. Ben Ali, M., Gaudreault, J., D'Amours, S., & Carle, M.-A. (2014, Octobre). A Multi-Level Framework for Demand Fulfillment in a Make-to-Stock Environment - A Case Study in Canadian Softwood Lumber Industry. Costanza, J. Just-In-Time manufacturing excellence (1996) John Costanza Institute of Technology Inc.; 3 rd Edition. El Haouzi, H., Thomas, A., & Pétin, J.-F. (2008). Contribution to reusability and modularity of Manufacturing Systems Simulation Models: application to distributed control simulation within DFT context. International Journal of Production Economics , 112 (1), p.48-61. Gaudreault J, Frayret. JM. (2011). Combined planning and scheduling in a divergent production system with co-production. Computers and Operations Reasearch , 38(9), p. 1238-1250. Gaudreault J, Frayret. JM. (2009). Distributed search for supply chain coordination. Computers in Industry , 60(6), p. 441- 451. Gaudreault J, Pesant. G, Frayret JM, D’Amours S. (2012). Supply chain coordination using an adaptive distributed search strategy. IEEE Transactions on Systems Man and Cybernetics Part C , 42(6), p. 1424- 1438. Gaudreault, J., Forget, P., Frayet, J.-M., Rousseau, A., Lemieux, S., & D'Amours, S. (2010). Distributed operations planning in the lumber supply chain: models and coordination. International Journal of Industrial Engineering Theory, Applications & Pratice , 17(3): p168-189. Kazemi Zanjani, M., Ait-Kadi, D., & Nourelfath, M. (s.d.). A stochastic programming approach for sawmill production planning. International Journal Of Mathematics in Operational Research, Vol. 5, No. 1, 2013 p. 1-18. Marier, P. (2011). Gestion intégrée des ventes et des opérations dans l'industrie du sciage. ExpoConférence. Université Laval, Canada, Québec. Marier, P., Gaudreault, J., & Robichaud, B. (2014, Novembre 5-7). Implementing a MIP model to plane and schedue wood finishing operation in a sawmill:
Regarding the average inventory level, the greater the intensity of demand is, the smaller the average inventory has to be. This is true for any policy. However, the greater the intensity is, the bigger is the difference between ATP and Stock policies. We recall that AcceptAll policy may look attractive (less inventory and many orders accepted). However, there is a huge number of late deliveries and therefore the customer satisfaction is very poor. By comparison, on-time delivery reaches 39% for AcceptAll, against 100% for Stock strategy and ATP. 6. CONCLUSION This article proposed a simulation framework to compare and evaluate different planning and order management strategies. It also encompasses a basic ERP system that covers inventory management, lumber production, planning algorithms, ATP and CTP calculation. The user can configure the production planning and order management process directly on the framework and then evaluate how they will perform in various market contexts. This tool could be used in a company as a decision-making tool by allowing choosing the right production planning and ordering management strategies. Even though this simulation model is at its first stage, the results of the experiments refer to recognized practices in the literature (Vollman et al. 1997) and are verified. In future work, this framework will be used as the backbone of a more complex study. The goal is to propose guidelines for more agile operations management driven by demand. We need to propose an operation management framework describing how to combine algorithms, humans, and decision processes in order to maximize the overall performance of the 658
INCOM 2015 Ludwig DUMETZ et al. / IFAC-PapersOnLine 48-3 (2015) 622–627 May 11-13, 2015. Ottawa, Canada
lessons learned. 10th International Conference of Modeling and Simuling- MOSIM'14 . Singer, M and Donoso, P. (2007). Internal supply chain management in the Chilean sawmill industry. International Journal of Operations & Production Management , 27(5), p. 524-541. Tony Arnold, J.R., Chapman, S., & Clive, L. (2012). Introduction to Materials Management. Pearson, Seventh Edition. Vollmann, T., Berry, W., & Whybark, D. (1997). Manufacturing planning and control for supply chain management. New-York: McGraw-Hill. Wery, J., Gaudreault, J., Thomas, A., & Marier, P. (2012). Improving sawmill agility through log classification. 4th International Conference on Information Systems, Logistics and Supply Chain Québec. Wery, J., Marier, P., Gaudreault, J., & Thomas, A. (2014). Decision-making framework for tactical planning taking into account market opportunities (new products and new suppliers) in a co-production context. MOSIM. Nancy. Yan, G. C. (2001). Experimental modelling and intelligent control of a wood-drying kiln. International journal of adaptive control and signal processing , 15(8), p. 787-814 .
659
627