Presentation on theme: "Basic Optimization Training"— Presentation transcript:
1Basic Optimization Training LLamasoft, IncOctober 2008
2Optimization Course Overview Course Goal: Understand the basics of supply chain optimization through lecture, computer exercises, and coachingCourse Objective: Students will be able to use the software, with minimal assistance, to correctly build a supply chain model; add the basic components of optimization to the model; perform an optimization; interpret the results through outputs; and perform an infeasibility analysis.
3Overview of Model Components Optimization Basic Components AgendaOverview of Model ComponentsOptimization Basic ComponentsStructureCostConstraintsInfeasibility AnalysisOptimization ResultsReview
6Understanding Optimization What is the optimal network structure?Thousands of possibilities…Evaluate millions of alternatives, find the global optimal structure: lowest cost structure that meets the constraintsDetermine the optimal supply chain network structure using MIP/LP programming…One Optimal Answer!
8Learning Objective- Structure Explain the basic components to modeling in the softwareUse the components to create a supply chain networkPerform a network optimizationGoal: Successfully design a new optimal supply chain network using the software
15Saving Models and Projects Save Model in Training Folder:C://Llamasoft/Training/OptimizationSave Model As:Opt_Training_BasicSave Project As:Guru_Training
16Site Types Customer Site Only customer sites can have demand Customer sites have sourcing policies but NO inventory policiesThey do not ship to other sites or produce any productsFlow into customer sites = $$$ (revenue!)Existing FacilityIndicates this site presently exists in the supply chainPotential FacilityIndicates this site presently does not exist in the supply chainType, and Choice combine to determine which costs to applyOpen Potential Facility = Add Startup CostClosing Existing Facility = Add Closing CostAlways incur Fixed Operating Costs if flow exists
17Creating the Opt_Training_Basic Model 4 SitesIPMFGDC 1CZDC 2IPSPSPTPTPIPSPSPTPTP2 ProductsProduct AProduct B
28Shipments Vs. DemandShipments are typically modeled through the Transportation Policies tableShipments table allows you to model shipments outside of the Transportation Policies tableThere are no cost fields in the Shipments data table- the costs in the Transportation Policies table are used to calculate the transportation costsAllows for accurate modeling of a “Push” system, instead of a demand driven supply chain
29Opt_Training_Basic Model: Sourcing Policy MFGDC 1CZDC 2Product_A, Product_BIdentifies which sites to send replenishment and customer orders, and whether product is ordered from an outside source or made at that site
30Types of Sourcing Policies Single SourceSingle Source (Select Closest)Multiple Sources (Most Inventory)Multiple Sources (Order of Preference)Multiple Sources (Probability)Source by TransferMakeMake by ScheduleMake (Single Process)Make (Order of Preference)Make (Probability)Hint: Use the Quick Reference Card for descriptions of these policies!
31Opt_Training_Basic Model: Adding Sourcing Policies Open the Sourcing Policies TableAdd a total of (10) Sourcing Policies4 Multiple Sources (Most Inventory)CZ can source Product_A from DC_1 and DC_2CZ can source Product_B from DC_1 and DC_24 Single SourceEach DC can only source from the MFG for Both Products2 Make1 for each product at the Manufacturer
32Opt_Training_Basic Model: Layout Map Display the Sourcing Policies on the Map
33Opt_Training_Basic Model: Transportation Policy MFGDC 1CZDC 2There must be at least one Transportation Policy that applies to source and destination which is not “Make”Each Transportation Policy defines a “Flow”Source SiteDestination SiteProduct (all applicable products if not explicitly entered)Mode (1 if not explicitly entered)
34Types of Transportation Policies ParcelLTL (Less than Truckload)Full TL (Full Truckload)Air, Rail and ShipDaily or Weekly ShipmentPeriodic ShipmentPooled Outbound/ Pooled InboundPooled Periodic Outbound/ Pooled Periodic InboundFlow (Optimization Only)Link To LaneAggregate ContainerDisaggregate Container
35Opt_Training_Basic Model: Add Transportation Policies Open the Transportation Policies TableAdd (4) Transportation PoliciesOne for each Source-Destination combination defined in the Sourcing PoliciesSource Sites: MFG, DC_1, DC_2Destination Sites: DC_1, DC_2, CZLeave all other fields at default value
36Opt_Training_Basic Model: Inventory Policy DC 1DC 2MFGProduct_AProduct_BDefinesInitial Inventory LevelsSafety/ Cycle Stock LevelsAssociated CostsOne Inventory Policy is optional for each product at the facility sites
37Opt_Training_Basic Model: Add Inventory Policies Open the Inventory Policies TableAdd (6) Inventory PoliciesOne for each Facility (non-Customer Site) and Product combinationSites: MFG, DC_1, DC_2Product Name: Product_A, Product_BLeave all other fields at default value
38Opt_Training_Basic Model: Add Costs Open the Sourcing Policies TableAdd a sourcing cost to one lane (2 policies)Use the filter bar to view only DC_1Update the Average Unit Cost to 1 and clear the filterHow do you think this will affect the optimization results?
39Opt_Training_Basic Model: Model Options Go to Tools Model Options (F3)Review Optimization PeriodStart Date / TimeEnd Date / Time
40Opt_Training_Basic Model: Run the Optimization Optimize the Supply ChainSave the Project and the Model
42Check on Learning- Structure You should be able to:Open Supply Chain GuruTMAdd a new modelSave a projectSave a modelOpen a tableUnderstand differences between sites and customersUse the filter barMove from field to fieldOpen the layout mapChange settings on the layout mapView sites and policies on the layout mapUnderstand the types of Sourcing policiesUnderstand the types of Transportation PoliciesUnderstand when to use shipments vs demandUnderstand the types of Inventory PoliciesEnter data into tablesAccess the help systemSet Optimization OptionsRun a simple optimizationRun the error check on a modelAccess Optimization Outputs
44Learning Objective- Cost Distinguish between the different costs used in Supply Chain Optimization with the softwareApply costs to the network built in Exercise 1Goal: Successfully optimize the network with the new costs included
45Three Basic Components of Optimization StructureCostsConstraints
46Basic Costs for Optimization Site CostsFixed operatingFixed startupClosingTransportation CostsAverage CostDuty RateDiscount RateReturn Trip CostTransportation Asset CostsUnit Fixed CostInventory CostsSite inventoryIn-transit inventoryInbound and Outbound WarehousingProduction CostsWork Center CostsFixed OperatingFixed StartupClosingWork Resource Cost
49Site Costs: Fixed Operating Costs associated with the day to day operations of the facilityEnable use of a step- function to associate the operating cost based upon operating capacity
50Site Costs: Fixed Operating Facility is closed/not used if the throughput is zero.Facility is open at Level 1 if the throughput is between 0 and 5,000 pounds.Facility is open at level 2 if throughput greater than 5,000 pounds
51Site Costs: Fixed Start-Up Costs to open and begin operating a new facilityOnly applies to Potential FacilitiesNo thoughput constraintsAbility to use step- function to associate start up cost with operating capacity
52Site Costs: Closing Cost to end operations at an Existing Facility Does not apply to Potential Facilities or Customers
54Exercise 2a: Create New Cost Model From the Project Explorer, right click on the Opt_Training_Basic ModelSelect Copy ModelRight Click on the copied modelSelect Save Model AsSave model as: Opt_Training_Cost
55Exercise 2a: Add Fixed Operating Costs Open the Sites TableOpen the Field GuruEnter the following Fixed Costs:DC_1Capacity CostDC_2Capacity Cost0 50
56Exercise 2a: Complete Fixed Operating Costs Now run the optimization and view the results!
58Transportation Costs: The Concept of “Flow” In Optimization, there are no individual shipmentsInstead it is the total amount shipped, as determined by the optimizerThis total amount is the “Flow”.Site ASite B10,000 Units
59Transportation Cost “Average Unit Transportation Cost” Average Cost Cost BasisShipment WeightDistance
60Transportation Cost: Average Cost Related to Cost BasisTransportation Cost per Cost Basis UnitAssociated Field Guru
61Transportation Cost: Cost Basis Weight = Avg Cost * Weight of FlowQty = Avg Cost * Number of Units of FlowCubic = Avg Cost * Volume of FlowDistance = Cost per MileFixed = Fixed Cost per ShipmentWeight-Distance = Cost per Pound per MileQty-Distance = Cost per Unit per MileCubic-Distance = Cost per Volume per Mile
62Transportation Cost: Distance and Fixed Cost Basis In order to cost these correctly the optimizer needs to approximate the number of shipments madeSite ASite B10,000 UnitsTotal Flow
63Transportation Cost: Shipment Weight Since the optimizer only knows the flow (sum of all shipments), the only way to cost at the “shipment” level is to approximate the shipment size.If the flow is 10,000 pounds, and the average shipment weight is 1000 pounds that corresponds to 10 shipments.
64Transportation Cost: Distance Calculate using Straight LineBased on latitude and longitude of source and destination sitesAdds a circuity factor (17%- in Model Options)Calculate Using Mappoint RoutingInterfaces with Microsoft Mappoint to determine actual road distanceMust have Map Point installed on the same computer
65Transportation Cost: Transportation Assets Total cost of owning or using each unit of this asset This is a fixed cost, not used to calculate profits or expenses in the network operationIncluded on the summary report, can be used to compare various scenarios
66Transportation Cost: Other Costs DutyDiscount RateReturn Trip Cost
67Exercise 2b: Add Transportation Costs to Cost Model
68Exercise 2b: Add Transportation Costs Copy Opt_Training_Cost ModelSave as Opt_Training_Cost_TranspoAdd the following costs:MFG facility always costs 2.00 per unit shipped to any locationDC_1 costs per unit, per mile to ship to the customerDC_2 costs per unit, per mile to ship to the customer
69Exercise 2b: ResultsNow run the optimization and view results!
72Inventory Costs: Facility Inventory Holding Inv Holding Cost = Avg Inv * Product Value * (i/365) * TAvg Inv = Average InventoryProduct Value = Value in Products Tablei = Annual inventory carrying cost %T = Optimization period in days
73Inventory Costs: Average Inventory Calculation Method of CalculationInventory TurnsORConstituent PartsSafety Stock InventoryCycle Stock InventoryPre-Build Inventory
74Facility Inventory Level Determination Factors which affect inventory levelsVolume/ Qty of product throughput (Tput)Number of facilities in the networkAs the number of facilities decreases, the average inventory in the remaining facilities increases due to increased throughput, but at a decreasing rate.Total Facility Inventory
75Average Inventory Calculation 1: Inventory Turns Ratio of inventory throughput to average inventoryIncreasing Inventory Turns reduces Facility Holding CostsMust balance turnover with safety stock to avoid stockoutAlso called Stock turns, turns, stock turnover
76Inventory Turns: Linear Approximation Inv Turns = 8 Avg Inv = m * Tputm = Inverse of Inventory TurnsTput = Volume of Product Throughput
77Inventory Turns: Pooling Effect Piecewise linear approximationUsed in locations with considerable amounts of product, typically called a distribution centerAverage Inventory can defined over multiple ranges of throughput.Format for the relationship is a series of pairs <lower range value, turns value>Piecewise linear functions are usually only used at DCs and other locations where considerable product is held. Other locations which hold less inventory tend to exhibit liner holding costs for most products- Modeling the Supply Chain, 2nd Ed. Shapiro, Jeremy. Page 417
78Average Inventory Calculation 2: Constituent Parts Pre-buildSafety StockCycle Stock
79Constituent Parts: Pre-Build Inventory Results from demand exceeding production capability in one period, but excess production completed in the previous periodExampleIn a 2 period model MFG has a production capacity of 50 unitsThe demand is 20 and 80 units in Periods 1 and 2 respectivelyThe MFG site produces 50 units each periodIn the first period the 30 excess units produced are stored as Pre-Build InventoryViewed in the Optimization Output- Inventory Table
80Constituent Parts: Safety Stock Held excess productAlso called a bufferThe model may tap into the safety stock when necessary
81Constituent Parts: Cycle Stock Portion of inventory allocated to meet anticipated demandIn a simple model where demand is constant, cycle stock equals half the order sizeThe blue line refers to actual cycle inventoryThe red line refers to the average cycle stockThe order size is 2 units and occurs once per unit time
82Average Inventory Calculation: Constituent Parts Sum of pre-build inventory, safety stock and the cycle stock
83Inventory Costs: Facility Inventory Holding Why determine Average Inventory?Avg Inv is used to calculate Facility Inventory Holding Costs in the OptimizationInv Holding Cost = Avg Inv * Product Value * (i/365) * TAvg Inv = Average InventoryProduct Value = The Product’s Value in the Products Tablei = Annual inventory holding cost %T = Optimization period in days
84Inventory Costs: In-transit Inventory Total Cost due to value of products being transported and transport timeIn-transit Inventory Cost = Q * Product Value * (i/365) * TQ = Quantity of Products in-transitProduct Value = Value in Products Tablei = Capacity Cost % in Model OptionsT = Transport Time in Days
85In-transit Inventory Costs: Example CustomerDemand = 1000 units10 days90 daysDC_B_OverseasDC_A_LocalProduct Value = $500In-transit Inventory Cost = Q * Product Value * (i/365) * TA to C 1000 * $500 * (15%/365) * 10 = $2,055B to C * $500 * (15%/365) * 90 = $18,493
86Inventory Costs: Inbound and Outbound Warehousing Inbound Warehousing Cost: activity cost of handling and moving one unit of product from receiving dock to inventory Outbound Warehousing Cost: activity cost of removing one unit of this product from inventory to the shipping dockIncludes such costs as paper tracking procedures, handling equipment, and personnelDoes not include Transportation Costs
94Learning Objective- Constraints The learner will be able to explain the different constraints involved in the software, identify potential constraints to a supply chain model, apply constraints to a practice model, and successfully perform an optimization on the model
95Constraints in Optimization Basics of ConstraintsAggregate ConstraintsFlowInventoryProductionSiteService ConstraintsMax Sourcing DistanceDue DateEnd to EndBundled Demand
96The Basics of Constraints: Definition Restrictions placed upon the modelAggregate Constraints: restriction defined for a sum over multiple objects, with at least one object having two or more valuesFlowInventoryProductionSiteService Constraints: restriction placed on the service to a customer
97The Basics of Constraints: Use in the Software Types of ConstraintsMinimumMaximumFixedConditional MinimumConstraint Variable InputsSpecific: Refers to one site/ product/ time period/ modeSet: Refers to a group of sites/ products/time periods/ modesAll: Refers to all sites/ products/ time periods/ modes
99Aggregate Constraints: Throughput Site is restricted by the amount of flow (basis)in the model during the specified periodStep function depicts capacity limit with INF
100Aggregate Constraints: Flow Flow requirement, flow requirement type, flow requirement basis, and time period that the restriction occursRestricted by 5 elements:Site, Destination, Mode, Product , or Time PeriodDC 1CZ 1
101Aggregate Constraints: Flow Places a restriction on the product flow over a set of time periods, between source and destination sites, for products or when using a specified mode
102Aggregate Constraints: Flow Count Sets up intricate constraints in the model linking the following 5 variables; Source, Destination, Product, Mode and PeriodBy aggregating the Destination Sites, Products and Modes it disregards the various possible flows that are due to these variables
103Aggregate Constraints: Inventory Restricted by 3 ElementsSiteProductTime Period
104Aggregate Constraints: Inventory Allows the specification of additional rules regarding inventoryDefines aggregated quantities over sites, products, and time periods
105Aggregate Constraints: Inventory Count Similar to aggregate flow countCan utilize the “Set” feature of the Groups Table
106Aggregate Constraints: Production Restricted by 4 elementsSiteProcessProductTime Period
107Aggregate Constraints: Production Defines aggregated productions that need to be restricted by a plant, or set of plants and by products, or set of products
108Aggregate Constraints: Production Count Similar to Aggregate Flow Count, but pertains to Productions
109Aggregate Constraints: Site Defines the minimum and maximum number of open sites allowed in a set of periods
110Aggregate Constraints at Sites Allows the user to customize the number of sites that can be used in a specific time period
111Service ConstraintsMaximum Sourcing DistanceDue DateEnd to EndBundled Demand
112Service Constraints: Maximum Sourcing Distance Consider a network with manufacturing, warehousing, and customer echelons. All flows between two successive echelons are permitted.DISTANCESM2M3M1CZ_1CZ_2CZ_3CZ_4WH2WH3WH1WH1WH2WH3M1500800120M26001000200M3300750CZ_1CZ_2CZ_3CZ_4WH1180720340600WH2700150280100WH320070640
113Maximum Sourcing Distance If the maximum sourcing distance is 200 miles for customers and 500 for the warehouses, the network is reduced to the following flow alternatives.DISTANCESCZ_1WH1WH2WH3M1500800120M26001000200M3300750M1WH1CZ_2M2WH2CZ_3CZ_1CZ_2CZ_3CZ_4WH1180720340600WH2700150280100WH320070640M3WH3CZ_4
114Maximum Sourcing Distance Between the end site and the source nodeDistance BasedCan be set in either the Sourcing Policy Table or the Service Requirements Table
115Service Constraints: Customer Due Date Customer due date-driven service constraints force the demand to be classified by customer lead times. Suppose P1 demand at each customer is 100 units.Classified demandCZ_1CZ_2CZ_3WH2WH1P1 in 7 days=75P1 in 3 days=25P1 in 5 days=50P1 in 1 days=50P1 in 6 days=40P1 in 5 days=604326571AirTruckRail
116Customer Due DateAll supply alternatives are feasible for the first demand classification, but only the following alternatives are feasible for the second classificationCZ_1CZ_2CZ_3WH2WH1P1 in 3 daysP1 in 1 daysP1 in 5 days42531AirTruckRail
117Customer Due Date Only from the last echelon site to the customer Time- basedSet in the Demand Table
118Service Constraints: End-to-End End-to-end service requirements are given from a make-node to a customer node.M2M1CZ_1CZ_2WH2WH3WH1Time from M1 to CZ_1 for Product1 <= 5 daysTime from M2 to CZ_1 for all products <= 7 daysDistance from M1 to CZ_2 for Product2 <= 250 miles
119End to End ConstraintsSource Site does not have to directly deliver to the customer; there may be other facilities in the network where the order will pass throughSpecified by maximum time for an order to leave the facility and reach the customer OR by maximum allowable distance between the facility and the customerSet in Service Requirements Table
120Service Constraints: Bundle Demand When choosing to bundle demand, demand for all products at one customer site will be sourced from one or multiple facilities at the same ratioWH145075Demand CZ_1(P1) = 600Demand CZ_1(P2) = 100CZ_115025WH2
121Bundled Demand Check this box to aggregate all the demand by customers When a customer demands multiple products, these are sourced in equal ratios from one or multiple sites (proportional to the demand quantities for these products)
122Exercise 3a: Constraining the Optimization Model
123Unconstrained Model Al Five DCs in Use Houston Processing Plant supplies only DC_KC
124With Aggregate Flow Constraints Max Flow Reqt Type means at most 500 units of flow can go through DC_Albany.Cond Min Flow Reqt Type means we either have at least 1000 units flow through DC_Portland or none at all.How does this change our optimized results?
125With Aggregate Flow Constraints DC_Portland not used, customers now served by DC_PhoenixFewer CZs in Northeast are served by DC_Albany, more by DC_Atlanta
126With Aggregate Production Constraints Max Flow Requirement Type means that at most 1000 units can be produced at Norfolk.Min Flow Requirement Type means at least 850 units must be produced at Reno.How does this change our optimized results?
127With Aggregate Production Constraints Fewer CZs in Midwest served by DC_Atlanta, more by DC_KC.
128With Aggregate Site Constraints Create Group that contains all five DCs.Constrain Optimizer to select between one and three sites from within that group.How does this change our optimized results?
129With Aggregate Site Constraints Portland and Phoenix DCs are unused, KC picks up the slack.
130With Aggregate Inventory Constraints Open the Optimization Output Inventory table and note the inventory costs at DC_KC.Set Minimum Inventory at DC_KC to 100 units.Optimize the model.How does this change inventory costs at DC_KC?
132Exercise 3b: Add Constraints Copy the Final Cost ModelSave as Opt_Training_ConstraintsAdd the following Constraint:DC_1 can only ship a maximum of 50 units of Product_A to CZ_1 for the entire model period (Horizon)Now run the model and view the results!How this affect the network design?
133Check on Learning- Constraints You should be able to:Define aggregate constraintsDefine service constraintsOpen the service requirements tableOpen the aggregate constraint tablesCreate service constraintsCreate aggregate constraintsDefine and distinguish between serve and aggregate constraintsExplain the constraint requirement typesApply aggregate constraints to a modelApply service constraints to a modelUnderstand aggregate constraints sum and objects
135Guru Infeasibility Analysis Sometimes the optimization solver returns with a “Problem Infeasible” error messageInfeasibility refers to a problem with input data- there is no solution that fulfills all the constraintsGuru provides the following tools to help the user identify the source of infeasibilityCheck for supply-demand imbalanceCheck for logic errors in defining the network structureRemove all or some hard constraints and solve again
136Infeasibility Analysis Select the hard constraints to impose