Distribution Planning Systems, Vehicle Routing

Distribution Planning Systems, Vehicle Routing Problems

Distribution Planning for Make to Order Manufacturers

The work of Chang and Makatsoris (nd) entitled: "Supply Chain Modeling Using Simulation" published in the International Journal of Simulation states that the key factor for success in business in today's global market is the management of the "entire supply chain." (Chang and Makatsoris, nd) it is held that it is acknowledged in today's "World-class organizations...that non-integrated manufacturing processes, non-integrated distribution processes and poor relationships with suppliers and customers" are not enough for the success of these organizations as they have come to "realize the impact of an organization's plan on the other areas of the supply chain." (Chang and Makatsoris, nd) the following illustration is that of the 'typical supply chain' which has been adapted from the work of Chang and Makatsoris (nd).

Figure

Typical Supply Chain

Source: (Chang and Makatsorsis, nd)

Supply chain functionalities include those stated as follows:

(1) demand planning;

(2) master planning

(3) procurement

(4) transportation; and (5) Manufacturing.

The following illustration is a diagram of an example supply chain simulation model which has been adapted from the work of Chang and Makatsoris, nd (nd).

Figure 2

Diagram of Supply Chain Simulation Model

Source: Chang and Makatsoris (nd)

Stated as example data requirements are those listed as follows:

(1) manufacturing process and time information;

(2) inventory control policies information

(3) Procurement and logistics information

(4) Demand information; and (5) Policies and strategies information. (Chang and Makatsoris, nd)

The following illustration lists the data required for each of these five areas.

Figure 3

Source: (Chang and Makatsoris, nd)

The report of Chang and Makatsoris (nd) concludes by stating that meeting the customer demand for "guaranteed delivery of high quality and low cost with minimal lead time" is the objective of supply chain management." If this objective is to be achieved, then companies must have a "better visibility into the entire supply chain of their own as well as those of their suppliers and customers." (Chang and Makatsoris, nd) This requires agility to adjust and reformulate plans and to do so in real time, to meet the unplanned moments in the supply chain production process. It is related that these needs are that which have driven the "application of discrete event simulation for analyzing entire supply chain process." (Chang and Makatsoris, nd) Supply chain management that is efficient can be realized through considering carefully information related to "capacity and material information." (Chang and Makatsoris, nd) the source of such inefficiencies includes: (1) the company; (2) the suppliers and (3) some are caused by both the company and its suppliers. (Chang and Makatsoris, nd)

II. Production Planning for Make-to-Order Manufacturers -- an Overview

The work of Chandra and Fisher (nd) entitled: "Coordination of Production and Distribution Planning" reports a computational study that investigated that value of coordinating production and distribution planning. Chandra and Fisher consider a plant with various product production over time and in which maintained is an inventory of finished goods at the plant in the study. Product distribution takes place through a fleet of trucks to various retail outlets at which the demand for each product "is known for every period of a planning horizon." (Chandra and Fisher, nd) Chandra and Fisher state that the majority of consumer products flow through "a pipeline that begins with production at a plan, followed by transportation to a retail outlet for consumer purchase, perhaps passing through a distribution center on the way." (Chandra and Fisher, nd)

It is stated by Chandra and Fisher to be increasingly clear "that companies will need to make the necessary organizational changes that will facilitate coordination of these operational functions and develop an ability to make more complex decisions within this structure." (nd) Chandra and Fisher state that Pyke (1987) and Cohen and Lee (1988) "studied integrated production/distribution systems under stochastic demand" and that Pyke "developed an analytical model of a simple three node system (factory, finished goods stockpile and single retailer) and examined the properties of the cost functions arising from this model for a single product case." (Chandra and Fisher, nd) However, it is related that this model does not make consideration for costs or issues that are related to the transportation of goods from the stockpile to the retailer. Chandra and Fisher (nd) state that their work makes comparison of two approaches to management of this operation "one in which the production scheduling and vehicle routing problems are solved separately and another in which they are coordinated within a single model." Chandra and Fisher state that the comparison is made "...for a number of different values of the basic model parameters, which include the length of the planning horizon, the number of products and retail outlets, and the cost of setups, inventory holding and vehicle travel. The reduction in total operating cost from coordination ranged from 3% to 20%. The value of coordination increases as the length of the planning horizon, the number of products and retail outlets, and vehicle capacity increases, as production capacity becomes less binding, and as distribution costs increase relative to production costs." (Chandra and Fisher, nd)

Chandra and Fisher state in conclusion that their study has shown that "under the right conditions, the value of coordination production and distribution can be extremely high. The analysis further provides the capacity to make the decisions that are more complex that are required under coordination." (nd) Presently, the majority of companies are characterized by both the organization and incentive structure design lacking the capacity to support coordination of production and distribution, therefore the efforts for improvement are focused on these two functions. These efforts are stated to be "reaching the point of diminishing returns" resulting in it being the right time to take under consideration making the necessary changes in the organization needed to achieve production and distribution coordination.

III. Time Windows for Delivery and Production Planning for Make-to-Order Manufacturers

IV. Simulation Techniques in Logistics

The work of Umeda and Zhang (2006) entitled: "Supply Chain Simulation: Generic Models and Application Examples" proposes a simulation system that is designed for the purpose of supply chain management of operations and that the "most remarkable characteristic of the simulation" is the representation of business process activities in supply chain management. Included in these activities are such as: (1) demands predictions; (2) manufacturing planning; (3) material purchasing; (4) manufacturing and transportation ordering; and (4) products shipping. (Umeda and Zhang, 2006) Finally, Umeda and Zhang demonstrate typical forms of simulation and supply chain systems including those which are: (1) centre-controlled ordering systems; (2) vendor reorder-point systems; and (3) pull-operational systems. These simulations systems are then compared in terms of the performance of the supply chain. Stated as two goals in the supply chain operation are those of: (1) synchronization; and (2) either inside the enterprise or beyond the enterprise's boundaries on production and logistics issues. It is stated that the robust 'chain' makes a requirement of data exchange on a frequent basis and "among inner-company processes and outside supplier's processes according to the above business process behaviors." (Umeda and Zhang, 2006)

The scope of the supply chain oriented simulation is stated in the work of Umeda and Zhang to require: (1) material management activities in individual processes (manufacturing, receipt, shipping, storage, inspection; and transportation); (2) order processing activities in individual processes, such as purchasing, operations, shipment and transportation. The process of sending and receiving between suppliers must be explicitly expressed; (3) information processing activities which combine material management activities with ordering activities. (Umeda and Zhang, 2006)

It is necessary to understand that the activities of order processing and information processing "are tightly coupled with the material management activities in suppliers, main product factories and distributors." (Umeda and Zhang, 2006)

There are stated to be three problems on simulation modeling that the supply chain must overcome and those are stated as follows: (1) How business processes in each chain member company should be represented in simulation models; (2) How process synchronization and data exchange transactions among the chain member companies should be represented in simulation models; and (3) How the communication mechanisms between information-flow and material-flow should be represented in simulation models. (Umeda and Zhang, nd) the following is an illustration of the Supply Chain -- Operational Models of the Supply Chain System.

Figure 4

Supply Chain -- Operational Model of Supply Chain System

Source: (Umeda and Zhang, nd)

Umeda and Zhang conclude by stating that their work demonstrates that simulation models may be specified for discrete manufacturing supply chain systems and demonstrates a proposed model composed of supplier member models. Stated is that through combining these modeling elements realization of center-controlled (push) system, buffer-drive (pull) system, and their hybrid combined system (such as re-order point model) all models representing business process operations may be implemented through the use of simulation software.

The work of Carol C. Bienstock (1996) entitled: 'Sample Size Determination in Logistics Simulations" states that logistics research contains numerous applications of a computer simulation modeling of logistics/distribution systems." (Bienstock, 1996) Logistics/distribution systems are stated by Bienstock (1996) to:

(2) involve networks of 'fixed facilities and connecting linkages';

(3) are characterized by complex and stochastic interrelationships among system components; and (4) generate data that are relatively quantifiable. (Bienstock, 1996)

These are stated to be the reason that these systems "lend themselves to investigation using simulation methodology." (Bienstock, 1996) Simulation is stated to offer an alternative "for understanding these systems, since experimenting with the actual systems would be too costly." (Bienstock, 1996) Furthermore, simulation is stated to facilitate the "examination of dynamic processes or systems over time by allowing the compression of real time." (Bienstock, 1996) Bienstock states that the methods for adjusting the sample size 'n' in simulation studies are as follows:

(1) simulation runs for each experimental condition (each cell) may be replicated 'n' times;

(2) observation of 'n' subintervals of the simulation of an experimental condition may be increased by decreasing the length of the subintervals; or (3) the simulation of an experimental condition may be continued for a longer period of time, thereby increasing the number of subintervals (i.e. The sample size). (Bienstock, 1996)

The technique described by Bienstock (1996) is one that "...enables a logistics researcher to determine the number of replications necessary to achieve a relative degree of precision. A relative precision goal for a logistics/distribution experiment ensures a reasonable degree of precision within the context of the system being investigated. Use of this technique will provide a researcher with the number of replications which will yield the degree of precision necessary for drawing conclusions about the behavior of the system under the various experimental conditions." (Bienstock, 1996) This technique is stated as appropriate for "simulation modeling that employs successive independent replications of simulation runs; it is not appropriate for determination of achieved relative precision on subintervals of a single simulation run." (Bienstock, 1996) Furthermore, Bienstock states that this technique "...cannot be used in experimental designs that utilize VRT." (1996)

V. Appraisal of existing Simulation Models in Logistics

The work of Rosenfield, Copacino, Little and Payne (nd) entitled: "Logistics Planning and Evaluation Using 'What-if' Simulation" states that the planning for the configuration of "large, complex distribution systems for optimal balance of cost and service can be best accomplished through use of sophisticated computerized models. The use of such models of all types has gained attention in the solution of large scale logistics problems." (Rosenfield, Copacino, Little and Payne) One such model is that developed by Geoffrion and Graves which was a large-scale logistics planning and optimization model for a major large food company introduced in the early 1970s. Another was developed by Bender, Northrup and Shapiro and Klingman, Napier and Stutz which were mathematical approaches for logistics planning and optimization models. Other models have focused on simulation of the wide range of "costs and activities in the logistics system." (Rosenfield, Copacino, Little and Payne)

Rosenfield, Copacino, Little and Payne state that the various approaches of computerized and manual evaluations are inclusive of those as follows:

(1) Manual evaluation of alternatives;

(2) What-if simulation modeling;

(3) Optimization modeling; and (4) Heuristic modeling. (Rosenfield, Copacino, Little and Payne, nd)

It is related that in problems that are 'small-scale' or in which the number of shipment alternatives is limited, the analysis can be performed "manually and alternative scenarios can be explicitly evaluated." (Rosenfield, Copacino, Little and Payne, nd) the manual approach is stated to have been used "historically" although "the power and availability of computers have made other approaches more desirable." (Rosenfield, Copacino, Little and Payne, nd)

What-if simulation modeling is one of the two most widely used approaches and the other is optimization modeling. Rosenfield, Copacino, Little and Payne state that 'What-if' simulation modeling "generally connotes scenario evaluation, while optimization involves a determination of the optimal or best solution." (Rosenfield, Copacino, Little and Payne, nd) a third stated approach is that of "heuristic modeling" which is defined as a "trial and error process to reduce the multitude of possible problem solutions to a small, manageable number of feasible solutions." (Rosenfield, Copacino, Little and Payne, nd) the heuristic approach is one that is based most often upon "a criterion which managers seek to optimize. Hence heuristic modeling in this case is really a form of optimization modeling." (Rosenfield, Copacino, Little and Payne, nd) the What-if approach is stated to be feasible only when there are a limited number of alternatives for consideration. It is also related that the optimization approach is generally necessary when there are "significant (typically resource allocation) constraints in the logistics system activities." (Rosenfield, Copacino, Little and Payne, nd) What-if simulation is stated to be "cumbersome to apply" when "systems alternatives are restricted, the complexities of considering many policy variables" exists. (Rosenfield, Copacino, Little and Payne, nd) What-if simulation is preferred when the company is unable to "undertake a sweeping revision of their logistics system" since the cost "in terms of investment and organizational description is often too great." (Rosenfield, Copacino, Little and Payne, nd) Therefore, if the company desires to know the effects of a change to only one or just a few variables the 'What-if' simulation is very useful.

Stated as advantages to the 'What-if' approach are: (1) this simulation model permits more exact representations of the process being modeled; (2) this simulation offers more flexibility; (3) this type of direct calculation can enhance management understanding. (Rosenfield, Copacino, Little and Payne, nd) Rosenfield, Copacino, Little and Payne state that the optimization approach and the 'What-if' simulation modeling are "not necessarily mutually exclusive, and indeed, often overlap." (Rosenfield, Copacino, Little and Payne, nd) Rosenfield, Copacino, Little and Payne report that three separate sets of computer programs used in the model reported in their work were those of:

(1) Data-base management programs;

(2) Freight and lead time analysis models; and (3) the What-if scenario evaluator. (Rosenfield, Copacino, Little and Payne, nd)

The data-base management programs are stated to have processed and converted the data into files. These programs then converted the supply/demand data into a set of demands by product class broken down by combination of origin node and destination node. (Rosenfield, Copacino, Little and Payne, nd) While the actual part-by-part modeling is stated to have not been performed due to limitations of a practical nature related to computer storage and time it is stated that the "aggregation of each suppliers parts into a part class provided a reasonable estimate for individual parts." (Rosenfield, Copacino, Little and Payne, nd)

A detailed account of the number of parts each supplier shipped and for each model year was contained in the corporate data base including information of the piece weight, dollar value, quantity of each part, assigned shipping class, and destination. Large amounts of freight cost and lead time data were processed by freight and lead time analysis models into functional regression relationships between costs and lead times distances, weights per shipment and other factors. (Rosenfield, Copacino, Little and Payne, nd) System mileages are stated to have been determined "on the basis of the latitude and longitude specifications for each of the nodes." (Rosenfield, Copacino, Little and Payne, nd)

Therefore, "any two nodes could be assigned locations and hence a mileage fro lead time and freight rate calculations." (Rosenfield, Copacino, Little and Payne, nd) Important to the model was the use of arbitrary locations with the model's key part of the software being the 'what-if' simulator based on the supply/demand data" which made determination of the overall system costs. The software implemented in this analysis had the capacity to trace a part from the supplier across the entire supply network and to the part's final destination.

The what-if scenario evaluator was designed to determine total system costs given:

(1) the supply/demand matrix as a function of the origin/destination pair and part class; and (2) the specific scenario. (Rosenfield, Copacino, Little and Payne, nd)

Rosenfield, Copacino, Little and Payne state that in broad terms "the scenario was a specification of a path for each element of demand." (nd)

Rosenfield, Copacino, Little and Payne state that they created a module that: (1) recalculated all system mileages; (2) recomputed all freight rates and lead times; (3) updated arc costs to reflect modifications; and Updated all paths to reflect new arcs. (Rosenfield, Copacino, Little and Payne, nd) Therefore, the arcs and paths were "identified in the same way, all paths and arcs were modified." (Rosenfield, Copacino, Little and Payne, nd) the model is reported to have been implemented "...in two separate sets of runs." (Rosenfield, Copacino, Little and Payne, nd)

The appropriate broad alternatives and specific candidate scenarios were identified but only following much analysis and scenario evaluation. The majority of the lead times underwent regression calculation and the majority of the freight rates on arcs that originated at the suppliers were based on formulas for calculation as well. The primary strategic question was related to the issue of the facility and the actual freight rates were utilized for "all of the paths in the system in the second set of runs." (Rosenfield, Copacino, Little and Payne, nd)

It is reported that in the runs, and where "specific freight rates were used to verify the final decisions, shipment sizes were calculated and used to determine the appropriate input freight rate over a selection of weight breaks." (Rosenfield, Copacino, Little and Payne, nd) the savings that were identified in this model were comprised by freight and transit inventory savings arising from reduced distances, consolidation of loads and for a portion of material, elimination of a transit leg." (Rosenfield, Copacino, Little and Payne, nd) it is stated that the model "was able to determine the complex effects of these changes." (Rosenfield, Copacino, Little and Payne, nd)

The final evaluation rendered implementation of an alternative system relating to the relocation of a major facility." The model's clarity and the flexibility in its capacity to report are stated as key factors to the model's acceptance in terms of its results for the corporation. Arising from this study are several conclusions concerning the use of logistics planning models in the area of corporate decision-making. Stated first is that one factor for successful implementation of any model type is a "flexible, understandable and clear model." (Rosenfield, Copacino, Little and Payne, nd)

Rosenfield, Copacino, Little and Payne state that the explicit calculation of transit and inventory costs, the clear reporting and tracing of all costs, and the ability of managers to sit at a terminal, specify an alternative and evaluate the impacts of that choice contributed to decisive decision-making and subsequent rapid implementation." (nd) it is implied that "lower-level management and model developers must work together closely in the design of the model." (Rosenfield, Copacino, Little and Payne, nd)

The incorporation of management input early on in relation to features of "clarity and understandability" increased the chances for success. Benefits of using computer planning models are benefits that are due to savings realized from alternative configurations. It is stated however, that the catalyst to selecting such alternatives is the model itself as management concludes at times that a particular change in the logistics system is needed. A model that is clear and that is easily understood will assist in gaining the agreement of all in implementing the changes that are needed and that will be approved by all. The design should be comprehensive so as to enable the organizational parts to participate in its development and additionally this type of design "gains credence for model accuracy and importance." (Rosenfield, Copacino, Little and Payne, nd)

VI. Logistics Simulation as Decision Support Models

The work of Cohen and Lee (1987) entitled: "Strategic Analysis of Integrated Production-Distribution Systems: Models and Methods" reports a "comprehensive model framework for linking decisions and performance throughout the material-production-distribution supply chain." (Cohen and Lee, 1987) Cohen and Lee relate that the threat of foreign competition has resulted in many U.S. firms re-evaluating their strategies in basic manufacturing "in order to regain a position of competitive advantage." (Cohen and Lee, 1987) Reported by Cohen and Lee is the development of a "model structure that can be used to predict" the firm's performance in regards to:

(1) cost of its products;

(2) the level of service provided to its customers; and (3) the responsiveness and flexibility of the production/distribution system. (Cohen and Lee, 1987)

The analysis conducted by Cohen and Lee is one that "takes into account the nature of the products produced, the process technologies used to manufacture the products, the structure of the facility network used to manage the flow of materials and the competitive environment in which the firm operates." (Cohen and Lee, 1987) Stated as the problem of interest is the concern of measuring the tradeoffs of production/distribution systems cost, service, and flexibility for various management strategies for materials under alternative conditions relating to the environment and structure of the corporation. Specifically considered in the methodology of Cohen and Lee are the relationships that exist between "production and distribution control policies" and specifically those affecting inventory control, plant production mix and production scheduling." (Cohen and Lee, 1987)

Cohen and Lee state that material flow in the production/distribution system is managed "by a variety of mechanisms. The inputs to each factory consist of materials and intermediate products which can be sourced from different vendors or other manufacturing facilities belonging to the firm." (1987) it is stated that such input flows as these are managed "by the firm's material requirement inventory control system." (Cohen and Lee, 1987) the material flow in the factory is affected by the layout of the plant, routing of products, lot size production and schedules for manufacturing. The outputs of finished goods can be stockpiled and then can be shipped directly to appropriate locations within the distribution network." (Cohen and Lee, 1987) Cohen and Lee state a goal of introducing a "model framework and an analytic procedure for evaluating the performance attributes of the class of production/distribution systems as shown in the following illustration:

Figure 5

Model Framework and Analytical Procedure for Evaluation of Performance Attributes of the Class of Production/Distribution Systems

Source: (Cohen and Lee, 1987)

The expressed stated of Cohen and Lee (1987) is to accurately predict the affect on performance of manufacturing and material strategies in the form of alternative methods. Questions addressed by Cohen and Lee include those as follows:

(1) How can production and distribution control policies be coordinated to achieve synergies in performance?

(2) How do service level requirements for material input, work-in-process and finished goods availability affect costs, lead times and flexibility? (Cohen and Lee, 1987)

The demand requirements in finished goods in the material billing consideration involves a great deal of uncertainty in both the production and distribution areas of the manufacturing process since "material requirements are not deterministic." (Cohen and Lee, 1987 ) in additional lead-times on material re-supply are randomly fluctuating resulting in "safety stocks for input materials [becoming] necessary to minimize production delays due to material shortages." (Cohen and Lee, 1987) it is related that the 'material control submodel' models the randomness of: (1) the demand process for materials; and (2) the resupply times from vendors. (Cohen and Lee, 1987) This model further "takes into account the cost of material inventory (setup and holding) and the cost of delay in order to determine the delay impacts of material shortages on production processing." (Cohen and Lee, 1987) This is sued for the purpose of: (1) determining ordering policies for materials required at each plan which results in service/availability levels (fill rates, stockout frequencies) for each of the raw materials used in the production lead-times of the product since material shortage can lead to delays in production." (Cohen and Lee, 1987 )