GoConqr has a new look! Let us know what you think.

Sign up free

Flashcards

Production Analysis

Marie Scholten logistik German 198 cards

The 4 P's of Operations Management

Cards in this deck (198)

  1. 1The 4 P's of Operations ManagementPolicies: Aims, Objectives & Strategies for the organization and the operations function
    Practices: Systems, People, Procedures & Technological capabilities
    Performance: strongly linked to practices (e.g. JIT). Important factors: Time, Cost, Quality, Flexibility
    Processes: Bring together policies, practices and performance. Analysis of processes very important. Operations Management also includes academic study of different operations practices.    The 4 P's of Operations Management
  2. 2American System of Manufactures (ASM) ( what is it and which aspects are still used"sequential series of operations carried out on successive special purpose machines that produce interchangeable parts"
    (Sequential series of operations -> successive specialized machines -> interchangeable parts)
    Key historical figures: Eli Whitney, Samuel Colt, Isaac Singer, Cyrus McCormick
    Aspects still present in modern manufacturing:
    Organized factory structure, specialized machines, Precision manufacture, interchangeable parts, co-ordinated work sequences, material flows & Quality techniques
  3. 3TaylorismFrederick.W.Taylor (1856-1915)
    (aka Scientific Management)
    - Setting Rates for piecework
    - Analysis of elements of any task
    - Development of planning departments for scheduling work
  4. 4Assembly Line Manufacturing at Ford- Development of moving assembly lines at Ford in 1914
    - Techniques for bicycle industry
    - Effective for manufacturing a single product in high volumes, on a continuous basis and to rigid (unflexibel) standards
    - Production in large batches
    - System for manufacturing Ford Model T set new standards until 1930s
    - Later, Ford focused on production of varied products
  5. 5Mass Production (Term, underlying thought & Drawbacks)Term evolved from an article about Ford.
    Grds Denkweise: Max profit, min production cost & Max production volume
    Drawbacks of mass production in the interwar period: low quality & flexibility
  6. 6Beyond Mass Production (since 19080 and since 1990)Economic expansion in Japan during 1980s
    Competitive advantage through automobile and electronics sectors
    Japanese products known for high quality and reliability (became known as "Total Quality Management")
    Key factors: re-organization of work, reduced stock-keeping, part supply as needed (-> JIT)
    Since 1990:
    - Lean Manufacturing, Agile Manufacturing, World Class Manufacturing
    - Alignment of internal and external processes
    - Supply Chain Management
  7. 7Planning HorizonsLongterm (several years):
    - Strategic decisions: Plant location, distribution network, supplier selection
    Mid-term (6-24 months):
    - Outline for regular operations
    - Rough quantities and times for material flows and resources
    Short term (up to 3 months):
    - detailed instructions for immediate execution
    - Depends on decisions on structure and quantities from upper levels
    The last two levels are called operational and some authors refer to the second level as "tactical"
  8. 8Enterprise Resource Planning (ERP)ERP = comprehensive software approach to support decisions concurrent with planning and controlling the business.
    Integrates application programs in: Finance, Manufacturing, Logistics, Sales & Marketing, HR,... Realization via common database
    Today: Focus on moving to coordinated planning and execution across companies. Requires consistent definitions of measures cross functional areas. Ability to link customers and suppliers into a complete supply chain
  9. 9Transaction Processing vs. Decision SupportTransaction Processing:
    - posting and tracking activities that document the business
    - efficient handling of transactions across all steps of production process
    Decision support:
    - System should help the user to make intelligent judgements
    - Support for better decision making
    - People are still the decision makers
    - Improved decision making through APS or APS-modules
  10. 10ERP Software - Top 4 & TrendSAP, Infor Global Solutions, Oracle and Microsoft
    Trend: Cloud based ERP
    Advantage: No need to host ERP system on company's servers.
    Mix of both worlds: Two-tier ERP systems
  11. 11Advanced Planning Systems (APS)3 main characterics:
    1. Intergral planning of whole supply chain: of single enterprise or network of enterprises
    2. True optimization:
    - Definition of constraints, alternatives and objectives for planning problems
    - Application of exact or heuristic optimization methods
    3. Hierarchical planning:
    - SC planning not possible in a "monolithic" manner
    - Successive solution of planning problems would miss optimality too
    - HPS try to find a compromise
  12. 12Links between APS and OR (Advanced Planning Systems and Operations Research)APS heavily rely on OR methods and concepts, such as: LP, MIP, Constraint Programming, Queueing theory & (Meta-)heuristics
    Strong model-based approaches: Abstraction, Relaxation, Omittance of unnecessary details and focus on relevant features
  13. 13Evolution of Operation PracticeCraft Production (19th century) -> Batch Manufacturing (ASM, around 1900) -> Mass Production (Taylorism, Ford, Beginning of 20th century )
    Since 1990: Lean,- Agile,- World Class,- Manufacturing
  14. 14What was the main process advancement during the transition from craft production to batch manufacturing?Craft work= individuals (or small firms) develop and deliver goods and services -> Low volumes, high cost per item.
    ASM = „the sequential series of operations carried out on successive special purpose machines that produce interchangable parts.“
    Development of methods for analysing and improving the organization of work. (E.g. Specializing machines, (re-)organizing factory structure, co-ordinating work sequences,…)
  15. 15What are the pros and cons of the mass production paradigm as implemented at Ford after 1930?Pros: Reduce the attendency of unskilled workers, workers need minimum level of knowledge, labour became a variable cost, labour component can be modified easily, we now had a big pool of unskilled workers who could be „used“, minimizes unit costs; very fast
    COns: work method was absolutely not flexible, only mass production possible, quality was a big problem as a large batch was producted, if a defect occured, it was not discovered until a big batch was produced the same way; a lot of time was used to fix mistakes that could have been avoided in the first place;
    - lack of flexibility
  16. 16What are (usually) the limits of ERP systems in terms of decision support? What are APS systems able to achieve in this context?ERP Systems only support decision making processes and data management but as people still make the decision, no real optimization.
    APS can do real (exact or heuristic) optimization by using constraints and objectives. Hierarchical planning is possible.
  17. 17Why are decision support and optimization (or at least quantitative approaches) often so closely linked? Give a few examples of planning problems from production and logistics to which this kind of perspective may apply! Explain why the definition of an objective function is reasonable (and advisable) in these cases.Inventory-> it is important to find a balance between being able to satisfy all customers demand and minizing the inventory holding costs.
    Production -> an objective function is needed to find the optimum of minimizing the variable costs in concern of the fix costs. For example, if a production plant is opend (therefore fixed costs occur) one should maximize production by holding certain constraints.
  18. 18How can one usually find the minimum (or the maximum) of a function F(t)? Hint: It is an approach that you may still know from school.Find the ‚extrem point‘ by computing the deviation for the function F(t). (Extremwertaufgaben)
  19. 19Sketch the principle of steepest ascent or hill-climbing methods in the three-dimensional space.Hill climbing: mathematical optimization technique. Local search, iterative algorithm, starting iteration: arbitrary solution ; step by step finding better solutions.
    Convex function (optimum can be found); konkarv function (only local optimum can be found)
    Simple hill climbing vs. steepest ascent hill climbing.
    Simple: choose first improvement
    Steepest ascent hill climbing: compare all successors and choose “best” improvment
  20. 20Why is it sufficient to consider only the corner points of the feasible region when solving a linear program?Optimum is always one of the corner point as there, the feasible area is maximized (by holding the constraints..
  21. 21Describe in your own words what the convexity property of a function means, based on the single variable case.convex functions only have one optimum. non-convex functions have several optimums.
    Therefore the deviation (finding the optimum) finds several extreme points (-> more complex).
    Examples: on page 5 & 6    Describe in your own words what the convexity property of a function means, based on the single variable case.
  22. 22Give a definition of a convex set, preferably in your own words. Give examples (drawings) of both, convex and non-convex sets! Why is it not possible to describe a non-convex set by a feasible linear program?A convex set is a set in which a line is always completely part of the set. (illustration of convex set: page 7; illustration of non-convex set: page 8)Give a definition of a convex set, preferably in your own words. Give examples (drawings) of both, convex and non-convex sets! Why is it ...
  23. 23Manufacturing vs. ServicesIn manufacturing, the focus is on:
    - production rates
    - inventory
    In services, the focus is on:
    - staffing or
    - other labor-related factors
  24. 24Aggregate Planning - Classification (Planning Horizons, Unit of measure & subjects)Medium (6-18 months)
    Unit of measures: Product families (eg all Ford F-series trucks)
    Subjects: employment (hiring/firing), overtime, inventories, facility modifications, utilities, material-supply contracts
  25. 25Production Planning and Control (PPC) PerspectiveForecasting (of aggregate demand for t period planning horizon)
    ⬇︎
    Aggregate Production Planning (determine aggregate production and workforce levels for t period planning horizon)
    ⬇︎
    Master Production Schedule (production levels by item by time period)
    ⬇︎
    Material requirements planning system (detailed timetable for production and assembly of components and subassemblies)
  26. 26Hierarchical Production Planning (HPP)- developed by Hax & Meal (1975)
    - HPP is concerned with workforce sizes and production rates at different levels
    Aggregation hierarchy according to Hax and Meal:
    1. Items (correspondence to stock-keeping unit, most detailed level)
    2. Families (Group of items with common setup cost)
    3. Types (Groups of families)
  27. 27Purpose of AggregationGeneral:
    - Focus on course of action, consistent with the company's strategic goals
    - Avoid too much detail (and thus excessive effort to update plans)
    Products (Create product families):
    - common processing (-> setup times!), labor, materials and also marketing grouping
    Labor: Aggregation to single group along product family lines
    Time:
    - Aggregation to months, quarters, seasons
    - Balance between (a) limited number of decision points (-> complexity) and (b) flexibility to be able to adapt to demand changes
  28. 28Aggregation Planning Objective / GoalObjective: Determine...
    - work force levels [W(t)],
    - production levels [P(t)] and
    - storage quantities (inventory)
    to minimize total costs over T periods of the planning horizon
  29. 29Smoothing (Aggregation Planning)Refers to costs resulting from changing production and workforce levels.
    Key components: Costs caused by hiring and firing of workers
  30. 30Bottleneck Planning (Aggregation Planning)Bottleneck = inability of system to react to sudden changes in demand
    Occurs due to capacity restrictions
  31. 31Threat when it comes to deciding on the planning horizon of aggregation planningToo short: production levels may be insufficient to meet future demand
    Too long: inaccuracies may arise
    Potential danger: End-of-horizon effect
    Common principle: Rolling schedules (A rolling schedule is a type of scheduling system that plans activities or tasks over a fixed time horizon, such as a week, a month, or a quarter, but also includes a rolling element that updates the schedule at regular intervals to account for changes and new information.)
  32. 32Treatement of demand (Aggregation Planning)Important assumption: Demand is deterministic
    Weakness AND strength
    Possibility of forecast errors is simply ignored BUT systematic changes in demand (seasonality, business cycles) can be covered (random models require average demand to be constant)
  33. 33Relevant costs (Aggregation Planning)Smoothing costs: Result from changing production levels; Key factors: size of workforce
    Holding costs: Cost accrued by having capital tied up in inventory; Almost always of linear type; Assumption: Charge against ending inventory
    Shortage costs: Danger of lost sales, typically linear form. Possible reasons: (a) Demand is higher than expected or (b) demand exceeds production capacity. Assumption: Shortage results in back orders which have to be filled.
    Regular time costs: Cost of producing one unit of output during regular working time. Direct & indirect costs of materials; Manufacturing expenses
    Overtime & subcontracting costs: Production by regular-time employees beyond normal workday or by an external supplier
    Idle time costs: may result from underutilization of workforce. Normally zero, because covered by regular time costs
  34. 34Basic (Aggregation) Strategies Methodology- Chase (zero inventory) strategy
    - Level (constant workforce) strategy
    -> production per period= MaxPeriode(commulatedDemand aufsummiert bis zur jetzigen periode/jetzige periode)
    -> # of workers per period = maxPeriode(commulatedDemand aufsummiert bis zur jetzigen periode/ (units aufsummiert bis zur jetzigen periode / workers)
    - "mixed" strategy
  35. 35Backorderingmeans that production in t + 1 can be used to satisfy demand in t (at the additional per unit backorder costs)
    (net inventory at end of t) = (units produced in t) - (units demanded in t) + (net inventory at end of t − 1)
    Net inventory is positive → units on stock Net inventory is negative → units backordered
  36. 36Management Coefficient ModelMain Idea (Aggreation): Construct model for controlling production levels based on prior decisions made by management
    Decision rule for optimal production level similar to linear decision rule.
    Developed by Bowman (1963)
  37. 37Dealing with fractional solution obtained by LP(1) Rounding approach:
    - Rounding the variables values to nearest integer
    - Usually sufficient but optimality no more guaranteed
    (2) (Mixed) Integer Programming:
    - Yields only integer values for variables marked as such
    - Drawback: Solution algorithms more complex and time consuming (branch& bound, branch & cut)
  38. 38Main Characteristics of the Disaggregation ProblemIn general:
    - Important activity in top-down production planning
    - Always required when increasing detail level of planning
    - Should follow organizational lines (-> decision making levels)
    Aggregation Planning/Master Production Planning (MPS) context:
    - Decisions on product and unit basis required
    - Conversion of aggregate production plans into detailed MPS plans
    - Aggregate decisions constrain disaggregated actions (i.e. the sum of disaggregated quantities must equal the aggregated ones)
    - Dimensions: products, time, location
  39. 39Grafik Zusammenspiel Aggregate, disaggregate & Master Production ScheduleGrafik Zusammenspiel Aggregate, disaggregate & Master Production Schedule
  40. 40Disaggregation Methods(1) Straightforward "simple" methods:
    - Product-based: Planning hierarchies and proportional factors (% vom Parent Produkt/Step)
    - Time-based: Equally distribute quantities
    (2) Runout time-based approaches:
    - From shortest to longest runout time
    - Equalize runout times
    (3) Mathematical Programming:
    - Model stated in Sipper & Bulfin (1997)
    - Model by Chung an Krajewski
    - Hierarchical Production Planning (HPP) based models
  41. 41Disaggregation: Runout time-based approaches: From shortest to longest runout timeRunout time R(i)= how long current inventory of product i will last
    1. Compute runout time R(i) for every product i as R(i)=I(i)/D(i)
    2. Sort products according to R(i) in increasing order
    3. Begin by producing product with the smallest R(i) and then next smallest R(i)...
    Applicability conditions:
    - Setup times are small and
    - capacity is larger than demand
  42. 42Disaggregation: Runout time-based approaches: Equalize Runout TimesGenerally useful in:
    - capacity constrained operations
    - when batches of multiple products are produced on the same line
    Basic idea/Goal: Inventory for each product should run out at same time (if production stops)
    - use total available production capacity
    - produce just enough of each product such that above goal is reached
    Advantage: Recognizes that products share production caoacity
    Disadvantage: Does not account for economic lot sizes
    Computation Principle: Divide projected available inventory at end of period being planned by demand per week:
    Aggregate runout time=
    [(Total inventory on hand in prod.hours)
    + (Total production hours available per period)
    -(Demand forecast for next period in prod. hours)]
    /(Demand forecast for next period in prod hours)
  43. 43Model According to Sipper and Bulfin (1997)Disaggregation!
    Assumptions:
    - n products on L lines
    - AP yields quarterly plan on monthly basis
    - Aggregation of both, products and facilities
    - Goal: Conversion to weekly production plan (individual products and lines)
    - Keep inventory at level suggested by aggregate plan
    - Starting point: first month of AP
  44. 44Model according to chung and KrajewskiGoals:
    - Disaggregate AP to Master Production Schedule (MPS)
    - Specify lot sizes and timing for individual end products
    - Adhere to targets (inventory and family production levels) from AP
  45. 45Disaggregation in Hierarchical Production PlanningOverall Goal: Optimize lot-sizes at the base level
    Top Level (G): Product Groups - Families possessing the same or similar cost parameters, consumption rates and demand patterns (seasonality)
    - Exogenous Information: Group demand of period t aggregated over all families
    - Split Objective: Minimize capacity, inventory and production costs
    - Output: Medium-term production program for product groups -> Communicate plan to next level
    - Main point of critique: Similarity assumptions concerning families that belong to the same group:
    *Aggregation is trivial (take parameter that is common for all families in that group)
    *Problem becomes difficult if parameters are allowed to be different
    Medial-level (F): Product Families - Items which share the same setup cost
    - Exogenous information: Family demand of period t as an aggregation over all items in the family
    - Split Objective: Minimize stationary setup costs
    Base Level (I): Product Items
    - Exogenous information: Forecast of item demand of period t
    - Split Objective: Equalization of runout times (guarantee mutual setup)
  46. 46Manufacturing SystemA manufacturing system is a set of machines,
    transportation elements, computers, storage buffers, people, and other
    items that are used together for manufacturing. These items are
    resources.
  47. 47Production rateHow much is made in a given time
    or
    the average number of parts produced in a time unit (also called throughput).
    If nothing interesting ever happens (no failures, etc.):
    Production rate = 1/ operation time
  48. 48Lead TimeHow much time before delivery
  49. 49Cycle timeHow much time a part spends in the factory
  50. 50Delivery reliabilityHow often a factory delivers on time
  51. 51Capital pay-back periodThe time before the company get its investment back
  52. 52Linear ProgrammingA special case of nonlinear
    programming in which the objective and the
    constraints are all linear.
  53. 53Were to find basic feasible solutions? (Geometrically)basic feasible solutions are corners of
    the constraint set.
  54. 54Aggregate Plans: Statement of...- Production rates
    - workforce levels
    - inventory holdings
    ... based on demand and capacity limits, on a macro Level.
  55. 55Aggregate UnitsUnits of measure:
    - weight
    - volume
    - time (worker hours)
    - money (dollars of sale)
    - fictitious quantities
  56. 56Linear Decision RuleMain idea (Aggregation): All relevant costs represented by quadratic functions.
    Function build on constant variables which depend on cost parameters.
    It is a better approximation of non-linear functions (Taylor series).
  57. 57Master Production Schedule (MPS) Overview ActivitiesInterface between
    (1) Sales and operations planning (incl. aggr planning)
    (2) Material requirement Planning (MRP)
    Input: Disaggregated sales and operations plan
    Output: Manufacturing schedule in terms of (end-)products
    [- Statement of specific products that represent output
    - Specify mix and volume of output
    - Determine quantities and timing for producible products]
    Midterm planning task (weeks/months)
  58. 58Sales & Operations Planning (SOP)- Aggregate statement of required manufacturing output to meet
    company objectives
    - Consolidation and incorporation of demand forecasts into sales and
    operations plan
    - Balance supply and demand volume
  59. 59MPS Responsibilities (Day2day & Operational)Day-to-day:
    (1) Coordination between sales and manufacturing
    (2) Shows WHEN products will be available in the future
    (3) Basis for promising delivery to customer (-> ATP)
    Operational: Construction of MPS record + periodic review and update (Rolling horizon planning)
    (1) Processing MPS transactions
    (2) Responding to exception conditions
    (3) Measuring MPS effectiveness
  60. 60MPS as a Statement of Future Output(1) Specifies WHICH products to produce WHEN and HOW MANY
    (2) Provides anticipated build schedule
    (3) Differs from forecasts in various different ways (EG it takes into account capacity limits/resource consideration, inventory & costs of production inkl setup)
    (4) MUST be stated in terms of producible products
    (5) However: Products = end-item must not hold!
  61. 61Business Environments/Strategies for MPSMake-to- stock (Production in batches)
    -> MPS units: end items or model grouping (eg only diff colour)
    Make-to-order
    -> MPS units: Particular end-item or sets of items composing customer order
    -> Difficulties: Production often starts before BOM is fully known
    Assembly to order (Production of components and subassemblies IN ADVANCE)
    -> MPS unit: Planning bill of materials
    -> MPS statement in terms of "average" product
    Siehe MIRO!
  62. 62Input Parameters for MPSMain:
    - Products
    - Lot sizes (standard, minimum)
    - Setup/changeover times
    Rest:
    - Planning horizon
    - Gross requirements (forecasts, customer orders)
    - On-hand inventory
    - Safety stock
    - Production rate or per-unit capacity consumption
    - Resource capacities (measured in hours or days)
    - Overtime
    - Subcontracted quantities
    - Backlog (max. quantities)
  63. 63Available to promise Logic (ATP)Part of MPS!
    Computation of "Available" Row = previous available + MPS- max(forecast, orders)
    Computation of "ATP" Row:
    First period= onHandInventory + first period MPS - sum of orders next MPS
    Later periods= MPS - sum of orders until next MPS
    Adaptions necessary in case of subsequent period ATP deficiencies
  64. 64Rough-cut Capacity Planning (Purpose & Techniques)Purpose:
    - Estimate capacity requirements of master schedule
    - Modify resource levels or material plan accordingly
    Techniques:
    - Capacity planning using overall factors
    - Capacity bills
    - Resource Profile
  65. 65Capacity Bills (Rough cut capacity planning technique)- More direct linkage between end products in MPS and capacity
    requirements on work centers
    - Degree of detail may vary (e.g., critical work centers only)
    - Basen on "bill of capacity" concept: Product structure data
  66. 66Resource Profiles (Rough cut capacity planning technique)- Takes into account timing of projected workloads at individual work centers
    - Provides time-phased projections of capacity requirements
    - Requirements:
    (1) Bill of Material
    (2) Standard times
    (3) Routing information
    (4) Production lead times
    -> More accurate and sophisticated approach to RCCP
  67. 67Capacity Planning Overall Factors (Rough cut capacity planning technique)Overall Factors (CPOF):
    Two-step procedure (given standard hours per unit):
    1. Compute overall capacity requirements per time period
    2. Use historical ratios to allocate total capacities to work centers
    Advantages: Easy to compute, minimal data requirements
    Disadvantages: Inherent inaccuracy
  68. 68Finite Capacity SchedulingCapacity planning approach so far: Plan materials and then check capacity requirements for established plan.
    Underlying assumption: Adjustments to capacity possible
    Otherwise: Simultaneous perspective required
    Implication of finite capacity scheduling:
    (1) Increasing level of detail
    (2) Sequencing decisions necessary (eg. by priority)
    (3) Requires more sophisticated solution approaches: MIP or advanced heuristic
  69. 69Material Requirement Planning (MRP) OverviewInterface between Master Production Scheduling (MPS) and detailed Scheduling
    Input:
    Output of MPS (ie. production plan for end-items or planning BOMs
    Product structure given by (detailed) bill of materials (BOM)
    Output: Detailed production plan for ALL involved subassemblies and components
    Production quantities are determined for each level of the system
    Consideration of production (and procurement) lead times -> Time-phased net requirements
    Planning horizon: weeks
  70. 70Lot sizing in MRPStronger focus on lot-sizing decisions (than in MPS)
    - Lot-4-lot vs. cost oriented perspective
    - Setup costs vs. inventory holding cost
    - Decisions have impact on lower levels
    Goal: Determine the production quantities such that the total holding and setup costs are minimized over the planning horizon
  71. 71Capacity Consideration in MRPResource feasibility important:
    - Capacitated lot-sizing models & algorithms
    - Capacity Requirement Planning
  72. 72Push philosophy(1) Strict top-down planning approach
    (2) Planning is done for all (BOM) levels in advance
    (3) MPS quantities (end-item net demands) serve as basis (“forecasts”)
    (4) Items are pushed to the next higher level when finished
    (moved downstream)
    (5) Opposite to JIT
  73. 73Explosion CalculusSet of rules for translation of gross requirements (MPS) into:
    (a) production schedule at the same level (net) and
    (b) requirements at lower level (gross)
    of the "product structure"
    Required calculations:
    (1) Offset time for beginning of production based on lead times -> time-phased requirements
    (2) Multiplication of higher-level requirements by gozinto factor    Explosion Calculus
  74. 74Product StructureRelationship between components at adjacent levels of the system
    Required Infos:
    (1) Parent-child relationship of components and end items at each level
    (2) lead time of each component (production time)
    (3) # of components required at the child level to produce one unit at the parent level (gozinto factor)
    Ways of representation:
    - Textual: BOM
    - Visual: Product structure diagram, Gozinto graph    Product Structure
  75. 75Production Lead Time (Components)- Machine processing time (material count*standard proc. time)
    - Setup (+ teardown) time
    - Queue time (up to 80%)
    - Wait time
    - Move time (stageing delays)
  76. 76Approaches for Lot-sizing in MRP (Single-level, single-item approaches/selections)- Economic Order Quantity
    - Silver-Meal Heuristic
    - Last Unit Cost
    - Part Period Balancing
    - Wagner-Whitin Algorithm
    Common drawback: No consideration of capacity constraints -> Adjustments to available resource capacities and / or material plan
    Remedy: Capacitated lot-sizing models
  77. 77Capacity Requirements Planning (CRP) Difference to RCCPDifferences to rough-cut capacity planning:
    (1) Uses time-phased material plan as provided by MRP, including:
    actual lot sizes and
    lead times
    for BOTH open shop orders (scheduled receipts) AND future releases (planned orders)
    (2)Incorporates gross-to-net calculations for end products and components
    (3) Accounts for status of work-in-progress on shop floor: Only remaining work on open shop orders is considered
    (4) Incorporation of demand for service parts, additional capacity required to react to scrap, item record errors, etc.
  78. 78MRP & UncertaintyKey factors of uncertainty:
    - Forecasts of future demands
    - Estimation of production lead times from one level to the other
    Impact: ALL lot-sizing decisions that were determined in the last run of the system could be incorrect
    Countermeasures:
    - Safety stocks
    - Safety lead times
  79. 79MRP System NervousnessBasic issue: Changes in the schedule when the horizon is moved
    forward one period
    Possible reasons: unanticipated changes in the MPS because of
    updated forecasts, late deliveries of raw material, failure of key
    equipment etc.
    Possible solution approach:
    Penalize setup in such periods by artificially increasing setup costs
  80. 80Best known JIT examplesFirms with high-volume repetitive manufacturing methods (e.g., Toyota)
    JIT requires physical changes and programs to make the changes
  81. 81Reduction of setup times and lot sizes (JIT)- industrial engineering techniques
    - thorough analysis of setup process(es)
    -> eg. Single-minute exchange of dies (SMED)
  82. 82Quality improvements (JIT)"no defects" goal, 2 critical aspects:
    (1) Total preventive/productive maintenance (TPM)
    (2) Poka-yoke: Integration of checking operations into processes
    Continual improvements (-> Kaizen): Also worker involvement + participation
  83. 83JIT Benefits-Lead time reductions
    - reduced WIP inventory
    -Less material movement in/out of storage
    -Reduced transactions
    - reduce need for shop floor tracking
    -reduction of complexity of detailed material planning
    -Simplified MPC systems
    -Reduced changeover times
    -Inventory reductions
    -Quality cost reductions
    -Quality improvements
  84. 84JIT Influences on Front end (ERP)- Level based capacity loading (smoothing production)
    - creation of rate-based MPS
    - Lead time reduction may increase responsiveness (from MTS to MTO or ATO)
  85. 85JIT Influences on Engine (ERP)Decrease of:
    - # of BOM levels
    - # of planned part numbers
    -> Complexity reduction in detailed planning
  86. 86JIT Influences on Backend (ERP)Shop floor:
    Fast order movement:
    - Progress tracking by MPC systems is not necessary
    - Backflushing: WIP accounting is done via a post-process BOM explosion rather than detailed shop order transactions
    Purchasing:
    Purchased items are rapidly converted into finished goods:
    - no transfer to stockrooms
    - no picking
  87. 87Pull SystemUpstream work center is allowed to produce only when corresponding
    signal from downstream work center has been received -> no pushing of parts allowed & no production just in order to keep workers busy
    Realisation: Kanban is just one way, others include:
    - empty containers
    - painted areas on the shop floor
    -e-kanban
  88. 88Types of Kanbans- Production ordering Kanbans: describes quantity that preceding process must produce
    - Withdrawal kanbans: describes quantity that subsequent process must withdraw
    - Express kanban: in case of shortages
    - Emergency kanban: replacement of defective parts, machine failures,...
    - Through Kanban: for directly adjacent work centers
    - Supplier Kanban: same as withdrawal, except for external supplier
  89. 89Kanban systemInformation system for JIT-based production, developed at Toyota
    (1) When a machine is available, and it has a production kanban, a part and a withdrawal kanban move from upstream buffer to machine.
    (2) After the operation, the withdrawal kanban returns to the upstream buffer, and a part (material) and a production kanban move to the downstream buffer
    (3) when a buffer has a free withdrawal kanban, a part moves to it from the upstream buffer, and a production kanban moves to the machine
    (4) the # of each kind of kanban is fixed at each stage
    (5) if a machine fails, the next machine can keep operating
  90. 90Kanban rules:1. Each process withdraws the NECESSARY products from the previous station in NEC. quantities at the NEC. time (Kanban must be attached to the product)
    2. Each process produces only what is withdrawn by subsequent process
    3. defective products are never moved to next process
    4. The # of Kanbans should be minimized
    5. The kanban system should be used to adapt to small fluctuations in demand
    Kanban is designed to absorb fluctuations; MRP must recalculate MPS weekly
  91. 91Kanban Card Set Size[Expected demand per time unit (D)]
    *
    [Lead time(L)]
    +
    [policy variable specifying the level of buffer stock (usually ca 10% of D*L)]
    /
    [Container capacity (usually <=10% of daily demand)]
  92. 92MRP & JIT Seperation(a) JIT MPS is rate-based
    (b) MPS is passed to JIT planning and execution subsystem relying on:
    - simplified BOMs
    - cellular manufacturing
    (c) detailed planning is also passed to JIT suppliers
  93. 93JIT Planning & Execution- Driven by daily build schedule
    - Uses inventory management subsystem:
    (1) for MRP-planned components
    (2) keeps track of finished goods, hub inventories, and deduct points
    Deduct point = stage in manufacturing where inventories are backflushed
    (a) Accounting is often done in stages
    (b) Helps in migrating from MRP-based planning to JIT
    (c) Goal: As few deduct points as possible
  94. 94JIT - Small WIP Inventory Pros & ConsPros:
    - Decrease inventory costs
    - Improves production efficiency
    - Points out quality problems quickly
    Cons:
    - May result in increased worker idle time
    - May decrease the production rate
  95. 95JIT - Kanban Information Flow Pros & ConsPros:
    - Efficient lot tracking
    - Inexpensive means of implementing JIT
    - Predetermined level of WIP inventory (based on the # of kanban tickets)
    Cons:
    - Slow to react to changes in demand
    - Ignores information about future demand patterns
  96. 96JIT - Coordinated Inventory & Purchasing Pros & ConsPros:
    - Inventory reduction
    - Improved relationships with vendors
    Cons:
    - Decreased opportunity for multiple sourcing
    - Suppliers must react more quickly
    - Improved reliability required of suppliers
  97. 97JIT vs. MRP(1) If problems develope and a line shuts down, JIT reacts immediately,
    MRP does not.
    (2) JIT does not incorporate knowledge about fluctuating demand while
    MRP does (therefore, in this case JIT will not work well).
    (3) For JIT, suppliers have to be located close to the production site,
    setup times and setup costs in general have to be low. MRP does not
    require these things.
    (4) Researchers did large scale simulations to compare JIT and MRP. The
    result was that JIT works best in a favorable environment (low
    demand fluctuations, reliable vendors, etc) and very poorly in an
    unfavorable environment. MRP also performs well in a favorable
    environment (comparable to JIT) and better than JIT in unfavorable
    ones.
  98. 98Alternatives /Derivatives of JIT- CONWIP (Constant Work-in-Process)
    - POLCA (Paired-cell overlapping loops of cards with authorization)
    - Just-in-sequence
  99. 99From JIT to Lean ProductionTerm according to Womack (1990)
    JIT + additional concepts, such as:
    - Six-sigma quality programs
    - Focused factory
    - Cellular manufacturing systems
    - Total productive maintenance
  100. 100Production Planning TasksProduction program planning
    Production requirements planning
    Planning in-house production
    Planning external procurements
  101. 101Manufacturing Control TasksOrder generation
    Order release
    Capacity control
    Sequencing
  102. 102Order ReleaseOrder release = point in time after which the shop floor is allowed to
    process an order
    Consequences:
    - material provision and allocation
    - influence on WIP level(s)
    - influence on order throughput times
    Important aspects:
    - Criteria
    - Degree of detail
    - Trigger logic
  103. 103Criteria (Order Release)(1) No criterion = immediate order release (after generation)
    (2) Due date-based order release (planned start date)
    (3) (constant) WIP on the shop floor or at specific work centers
    (4) Load-based (workstation-specific)
    Simplest form: Immediate order release
    Not well-suited for MTO enviornment
  104. 104Degree of Detail (Order release)Low degree of detail (order level) ⇒ centralized WIP control
    *Simpler to implement than decentr.
    * No blocked WIP, no or only few sequence interchanges
    * More WIP fluctuation, especially under complex material flow
    High degree of detail (individual operations) ⇒ decentralized WIP control
    *Regulation of WIP and throughput times on individual workstations
    *Better predictability of throughput times
    *Parametrization more difficult
    *Possibility of blocked WIP, deadlocks, sequencing interchanges
    Hybrid WIP control:
    Only difference to centralized / decentralized version: Processing of an order’s first operation
    * Suspend order when overall WIP or workstation WIP exceeds limit
  105. 105Trigger Logic (Order release)Periodic order release:
    *Typical release points: start of work day or shift
    *Causes WIP fluctuations (degree depends on length of release period)
    *Complicates determination of WIP limits
    Event-based order release:
    *Conceivable events: order generation, attainment of planned start date, WIP drops below limit, parameter change (e.g., WIP limit)
    *Tends to prevent WIP fluctuations
    *Requires greater implementation effort
  106. 106Due Date based Order ReleaseInability to regulate WIP in production -> requires higher mean WIP
    No load balancing / alignment on workstations
    No sequence interchanges on workstations caused by order release itself
    No backlog control
    Advantage: very simple
  107. 107Constant Work in Process Order Release (CONWIP)Release an order for a production as soon as (overall) WIP falls below a predefined WIP limit
    WIP Measures: # of parts, # of orders, planned hours
    Planned hours WIP:
    *Sum of planned hours of each order’s operations
    *Sum of planned hours of each order’s not yet completed operations
    *Planned hours on bottleneck workstation
  108. 108The role of schedulingScheduling
    * deals with the allocation of scarce resources to tasks over time
    * is a decision-making process with the goal of optimizing one or more
    objectives
    Typical Objectives:
    1. Max throughput, utilization, customer satisfaction (due date),...
    2. Min total setup cost, time,...
  109. 109Finite Scheduling/Finit loadingPerform scheduling not based on lead times but rather on true work
    contents (execution times)
  110. 110Sequence & Schedule DefinitionSequence: Just a permutation of task numbers (eg 1-4-3-5-2)
    Schedule:
    - Set of task starting times (for every resource)
    - Uniquely determined by a sequence, if
    preemption is not allowed and
    tasks start at earliest possible time instants
  111. 111Complexity of Sequencingassumption: N tasks and M machines
    On one machine: n! possible sequences
    On multiple machines: N!^M
  112. 112Types of schedulingMachine scheduling (single machine, job shop, flow shop, etc.)
    Personnel scheduling
    Facilities scheduling (e.g., operation rooms in hospitals)
    Vehicle scheduling (→ vehicle routing problems)
    Vendor scheduling
    Project scheduling
    Dynamic and / or stochastic scheduling problems
  113. 113Characteristics of Shop Scheduling Problems(1) Job arrival pattern: Static vs. dynamic
    (2) Process flow patterns:
    - Job shop
    - Flow shop
    - Open shop
    (3) Machine configuration at each station: Single vs. parallel machines
    (4) Additional constraints, for example
    - Precedence constraints
    - Worker / tooling constraints
    - No-wait constraints
    Common objectives:
    - Meet due dates
    Min WIP inventory
    - Min avg. flow time through the system
    - Max resource utilization
    - Reduce setup times
    - Min production and worker costs
    (!) Cost related obj are difficult quantify -> use "proxy" measures (time-related) instead
  114. 114Sequencing rules for single machine scheduling1. FCFS
    2. Shortest processing time (SPT)
    3. Earliest due date (EDD)
    4. Critical ration (CR):
    (dueDate-currentTime) / processing time
    More complex rule incoporate job weights:
    - Weighted shortest processing time (WSPT)
    - Weight-to-processing-time ration (WTPTR)
    - Rachamadugu and Morton's rule (R&M)
  115. 115Single Machine Scheduling: Which objectives are minimized by the shortest processing timeMean flow time, mean waiting time, mean lateness
  116. 116Minimize weighted flow time (Single Machine Scheduling)Basic Idea: Schedule jobs with small processing time and large weight toward the front & Schedule jobs with large processing time and small weight toward the back
    Compute processing-time-to-weight ratios p(i)/w(i) and sort jobs in
    nondecreasing order of ratios → Weighted shortest processing time
    (WSPT) schedule
  117. 117Minimize # of tardy jobs (Single Machine Scheduling)Moore's algorithm:
    1. Sequence the jobs according to EDD to obtain the initial solution
    2. Find the first tardy job in the current sequence, say job i
    3. Consider all jobs up to i and reject the job(s) with the largest processing time & return to step 2
    4. Form an optimal sequence by taking the current sequence and append the rejected job (in any order)
  118. 118Minimize total tardiness on a Single Machine (Scheduling)NP-hard! We need enumerative algorithms or heuristics
    Efficiently solvable cases:
    - If all jobs have a common due date, SPT minimizes total tardiness
    -If all jobs have a common proc. time, EDD yields the optimal solution
    -If all jobs are tardy, (W)SPT gives the optimal sequence
  119. 119Lawler's algorithm (1973) - Single Machine SchedulingIf the obj. function is of the form min max gi(Fi) with g(i) any nondecreasing function of the flow time and precedence constraints between jobs exists!
    (1) Determine set ob jobs V (all remaining jobs that are not required to precede any other)
    (2) Choose the job k that satisfies the objective function min gi(Fi) the best (NO MAX!)
    (3) schedule job k in the last position of the sequence and return to step 1
  120. 120Objectives and corresponding algorithms for single machine schedulingSPT (Shortest Processing time):
    Mean flow time, mean lateness
    W(eighted)SPT:
    Mean weighted flow time
    EDD:
    Maximum lateness/tardiness
    Moore:
    Number of tardy jobs
    Lawler
    Min-max with prec. constraints
  121. 121Cyclic vs. Acyclic scheduleCyclic vs. Acyclic schedule
  122. 122Underlying optimization problem of Scheduling with sequence dependent setup times (Cyclic & Acyclic) + ComplexityTraveling Salesman Problem
    The TSP(and thus the single machine scheduling problem (SMSP) with sequence-dependent setup times) is NP-HARD!
  123. 123Shortest Setup Time heuristicInitialization: Arbitrarily pick a job as the first job, denote this job i
    Step 1: Find the job j that is not scheduled yet and that has the
    shortest setup time if appended to i. Append the job j to the
    sequence. Now, job j becomes the new job i.
    Step 2: If all jobs are scheduled, Stop. Otherwise, go to Step 1.
    Lower Bound calculations:
    (a) Matrix based: Sum up n smallest elements in the matrix
    (b) Row based (or column based): Identify min value in each row (or col) and build the sum of those values
  124. 124A regret based algorithm (Scheduling with sequence dependent setup times)Characteristics:
    (1) Regret is a penalty of a decision NOT made
    (2) Provides a good heuristic for the TSP
    (3) LB computation is inherent part of the procedure
    Main steps:
    (1) Matrix reduction (row & col reduction)
    (1.1) The sum of all reduction coefficient is the LB
    (2) Regret calculations
    (2.1) For every zero in the matrix sum up the next smallest value in its row + second smalles value in its column (if we choose j to follow i -> some other task must preceed i and another one must follow j)
    (3) Identification of largest regret value
    (4) Assignment of corresponding job pair
    Per step: Do not forget to set matrix element that would lead to a cycle to ∞
  125. 125Branch and Bound conceptA general method for solving (mixed) integer optimization problems (with a minimization
    objective). It uses two concepts: branching and bounding.
    Root node solve relaxation of the optimization problem (e.g., LP of IP). Set LB to the
    objective function value of the relaxed problem z0 and set UB = ∞. Then identify on what to branch.
    Branch: Partition the problem P(k) at k into a given # of subproblems s (eg. which arc to choose next) and compute the LB at the child nodes
    Bound: the lowest bound across al nodes gives the current LB; a feasible solution to the original problem at a given node gives a new UB; All open nodes associated with LBs>=UBs can be pruned
  126. 126Little et al.(B&B algorithm for the Asymetric traveling salesman problem for single machine scheduling with setup times)
    1. LB = reduction of the cost matrix (Row&Col Reduction)
    2. Choose a node k with smalles LB for branching
    3. Branching scheme: At node k choose the arc with the larget regret value and generate childnodes: In the first forbid all tours containing this arc (LB: cost value of parent node plus regret value of forbidden arc), in the second node, include this arc (exclude according row and column and set arc that would lead to a cycle to M). If new reduction can be obtained after eliminating corresponding row and col from matrix, LB of second node= LBold + newReductionCoefficient
  127. 127Neighborhood Search for SMSPs (Iterative vs. one-pass heuristics)Quality of solutions delivered by one-pass heuristics may vary considerably. Remedy: Systematic search in the solution space.
    Relationship between solutions is based on a certain neighborhood principle
    Examine more than one, a potentially large number of solutions.
  128. 128Definition: Neighboring solution (Neighbor) vs. the neighborhood of a solutionNeighboring solution: Assume a given schedule S, then a neighboring schedule S′
    is a schedule that is “close” to the given schedule and obtained by applying a given
    operation to the given schedule.
    The neighborhood of a solution: consists of all neighboring solutions that can be obtained from the given solution (in our case schedule) by a applying a given operation to it.
  129. 129Neighborhood search (Definition)An initial schedule is created and neighbors or neighboring solutions are generated and evaluated, and a better neighbor is selected as the new current solution. This is repeated until no improving neighbor can be found. At this point the procedure terminates. The obtained schedule is a local optimum.
    Performance measures: any of the previously defined measures (average/total flow time, average/total tardiness, makespan, combined measures,...)
  130. 130Neighborhood operatorsAdjacent pairwise interchange (API): Exchange two adjacent jobs!
    API results in n-1 neighbors of the initial solution
    Pairwise interchange (PI):
    Exchange two jobs (not necessarily adjacent)!
    PI results in n(n-1)/2 neighbors of the initial solution
    Insertion (INS):
    Move a job to a different position! No exchange!
    INS results in (n-1)^2 neighbors of the initial solution
    Best improvement (evaluate the costs of all neighbors and take the one with the lowest
    cost) vs. first improvement (take the first neighbor of lower cost)
  131. 131Possible Machine Types (Parallel Machine Scheduling(1) Identical machines
    (2) Uniform machines (processing speed may vary)
    (3) Unrelated machines (individual processing time per job and machine
  132. 132Job routing /routeJob is split into multiple operations.
    The job routing gives to the processing order of operations within a job.
    Implied by the required technological sequence of processing steps.
  133. 133Types of (job) routings(1) Unspecified -> open shop
    (2) Predefined, but may vary from job to job -> job shop
    (3) Same for all jobs -> flow shop
  134. 134Common Performance Measures (Job shop scheduling)(1) Maximum completion time
    (2) Total weighted flow time (TWFT)
    Note: Min weighted flow time is equivalent to min weighted completion time (sum(w(i)*C(i))
    (3) Maximum lateness
    (4) Total weighted tardiness (TWT)
    (5) Min earliness and tardiness (ET):
    alpha(i) & beta(i): earliness and tardiness cost coefficient.
  135. 135Regular Performance Measures (Job Shop Scheduling)A performance measure is called regular if it is nondecreasing in C1, . . . , Cn
    Examples: - Max completion time
    - Total weighted completion time
    - Max lateness
    - Total weighted tardiness
    Example for nonregular: Total earliness (+ tardiness)
  136. 136Non-delay ScheduleA feasible schedule is called non-delay if no machine is ever kept idle if
    there is at least one operation which is ready for processing.
  137. 137Active ScheduleA feasible schedule is called active if no operation can be started earlier by
    changing the processing order on any of the machines without delaying
    some other operation.
  138. 138Semi-active ScheduleA feasible schedule is called semi-active if no operation can start earlier
    without changing the processing sequence on the respective machine.
  139. 139Optimality property of active schedulesThe set of feasible schedules which are optimal with respect to a regular
    optimization objective contains at least one active schedule.    Optimality property of active schedules
  140. 140Complexity of Job shop scheduling problem (JSSP)n!^m
    2 jobs, 2 machines = 4 schedules
    3 j,3m = 216 schedules
  141. 141Disjunctive Graph Representation (Job shop scheduling problem)A graph G for a C(max) job shop problem is defined by:
    (1) The node set V containing one node for each operation and 2 dummy nodes (source & sink)
    (2) The set of conjunctive arcs W representing the technological orders of operations
    (3) The set of disjunctive arcs Z (all possible operation pairs per machine)
    (4) The mapping p, giving the processing time for each node (node-weighted graph)    Disjunctive Graph Representation (Job shop scheduling problem)
  142. 142Disjunctive Graph Selection (Job shop scheduling problem)A feasible selection is an orientation of arcs from Z such that the resulting
    digraph is acylic.    Disjunctive Graph Selection (Job shop scheduling problem)
  143. 143Determine Operation starting time from an Disjunctive Graph (Job shop scheduling problem)The earliest starting time es(i) of an operation in the graph G can be
    computed as:
    max[es(predT)+p(predT), es(predM)+p(predM)]
    predT: technological predecessor
    predM: opereation wich is sequenced immediately before i on the same machine
    FORWARD SCHEDULING!
    Latest starting time: ls(i)=min(ls(succT),ls(succM))-p(i)
    BACKWARD SCHEDULING!
  144. 144Critical Path: Disjunctive Graph (Job shop scheduling problem)If es(i)=ls(i) -> i is part of critical path (no slack/buffer time)
    Multiple critical paths may occur in the same schedule    Critical Path: Disjunctive Graph (Job shop scheduling problem)
  145. 145Job shop schedule building heuristicsFact: Exact solution of problem in realistic dimension out of question
    Heuristic approaches:
    (1) Insertion algorithms
    (2) Bottleneck-based scheduling
    (3) List scheduling algorithms
  146. 146List scheduling algorithms (Job shop)Principles:
    (1) Chronological scheduling
    (2) Determine a lots of schedulable operations at each (relevant) time instant
    (3) Choose one of the available operations (priority rule)
    Simple schemes for creating non-delay schedules and active schedules
  147. 147Giffler-Thompson (GT) Algorithm (1960)Creates active schedule! and relies on priority dispatching rules for selecting operations! Is a List scheduling algo.
    Q: Set of currently schedulable operations
    se(o) = earliest possible starting time
    ce(o) = earlist completion time
    K = Conflict set
    Step1: Find operation with min completion time (c(min)) from Q
    Step 2: K= all schedulable operation which can start before c(min) on the same machine
    Step 4: Choose operation from conflict set & perform updates on es(o) and ce(o)    Giffler-Thompson (GT) Algorithm (1960)
  148. 148List Scheduling Algorithm to build non-delay schedulesOnly slight modifications of GT algorithm necessary:
    (1) Choose operation with minimum earliest starting time
    (2) Conflict set consists of operation which can start at the same time
  149. 149List Scheduling Algorithms (Job shop) Priority Rulesmin completion time -> Giffler-Thompson
    min earliest start date
    most remaining operations
    most work remaining
    minimum slack time
    FCFS
  150. 150Improvements for job shop scheduling heuristicsApproaches:
    (1) Try out (many) different Priority Rules
    (2) Create schedule and repeatedly apply SMALL improvement modifications
    Swap-based Improvements! Perform modifications ONLY on critical Path to ensure feasibility! And only those may lead to improvements. But it is unlikely to find optimal solution (only local optimum)
  151. 151Critical Blockscontiguous sequence of critical operations on a machine
    Important observation: Only swaps at the borders (first / last two
    operations) lead to an improvement
  152. 152Metaheuristic local search algorithms (Job shop scheduling improvements)In order to "escape" from local optimum!
    (a) Consist of problem-specific and problem-independent components
    (b) High-level components are responsible for leaving local optima
    (c) Often inspired by natural phenomena / processes
    (d) Prominent representatives: Tabu search, simulated annealing, VNS
  153. 153Extensions of the standard job shop model(1) Multiple simultaneously required resources (multi-constrained JSSPs)
    (2) Alternative (parallel) machines (flexible JSSPs)
    (3) Cumulative scheduling: resources have non-unary capacities
    (4) Sequence dependent setup times
    (5) No-wait restrictions / limited intermediate buffers
    (6) Stochastic processing times
    Note: Some of these extensions are covered by the more general notion of
    resource-constrained project scheduling (RCPSP)
  154. 154How to cover randomness (uncertainty) in job shop scheduling - stochastic schedulingFor example: Machine breakdowns
    Option 1: Modify processing time distribution to consider breakdowns
    Option 2: Model breakdowns explicitly by restricting machine availability dynamically (-> simulation)
  155. 155Stochastic Single Machine SchedulingAssumption:
    (1) Processing times are random variables with arbitrary distributions
    (2) Release times, due dates and weights are known in advance
    Shortest processing time scheduling: The weighted shortest expected processing times rule (WSEPT) minimizes
    expected total (average) weighted flow (completion) time criteria.
    Earliest Due Date Scheduling: The EDD (earliest due date) rule minimizes expected maximum lateness.
    Underlying observation: EDD minimizes L(max) under any realization of
    processing times
  156. 156Stochastic Multiple Machine SchedulingAssumption: Processing times follow an exponential distribution
    Expected makespan on two identical parallel machines: The expected value of the makespan is minimized by the longest expected
    processing time (LEPT) rule.
    Expected makespan in a two machine flow/job shop:
    Let a(i) and b(i) be the rates of the processing times distributions on the first and second machine. Sequencing jobs in decreasing order of a(i)-b(i) minimizes the expected value of the makespan.
    Note: Problems with more than 2 machines require more advanced approaches.
  157. 157Project - DefinitionSet of partially ordered, interrelated activities that must be completed to
    achieve a goal
  158. 158Project - Properties(1) Well-defined beginning and end
    (2) Happen only once or infrequently (exception: cyclic projects)
    (3) Goal: complete as quickly / as cheaply as possible
  159. 159Planning (establish project network)(1) Identify activities and precedences
    (2) Estimate time and resource requirements
  160. 160Project Scheduling(1) Determine critical activities and project duration
    (2) Reduce time for critical activities if necessary → subcontracting, overtime, time/cost trade-off
    (3) Reduce need for critical resources if necessary → subcontracting, more resources, shift start dates of activities
  161. 161Project ControlInvolves feedback from actual project execution (activity status, etc.)
  162. 162Project scheduling GoalDetermination of
    - activities’ earliest start / finish times
    - activities’ latest start / finish times
    - overall project completion time
    - critical activities & critical path
  163. 163Project Scheduling - Critical Path Method (CPM)(1) Forward pass: earliest start & finish times
    (2) Backward pass: latest start & finish times
    Result: activity slack, critical path
  164. 164Resource Types (PERT)- renewable resources (personnel, machines,...)
    - non-renewable resources (eg material)
  165. 165Project Evaluation and Review Technique (PERT)Activity duration is random variable as single estimate for activity duration is often unrealistic.
    Method characteristics: Critical Path Method with stocastic duration
    Results: Expected project completion date & Probability that project is completed by a certain date. All based on assumptions that do not always hold.
  166. 166Types of distribution (PERT)1. Uniform distribution
    2. Triangular distribution
    3. Beta distribution (historically used in PERT)
  167. 167Project completion time as a random variable (PERT)(1) Assumption: normally distributed (→ central limit theorem)
    (2) Requirements:
    (2.1) Independence of activity durations
    (2.2) Number of activities on the critical path is large enough
    (3) Advantage: Easy probabilistic computations, e.g., probability that project is completed within T weeks
  168. 168Limitation PERT(1) Multiple paths with lenths close to mean of the CP and Large variance
    (2) Tendency to underestimate project length
  169. 169Solution Approaches for Resource constrained project scheduling problem1. Graphic approach
    2. Finite capacity scheduling:
    2.1 Serial schedule generation scheme (SSGS) -> Active schedule
    (2.2) Parallel schedule generation schedule generation scheme (PSGS) -> non-delay schedule
    (2.3) Branch & Bound methods
    (2.4) Metaheuristics
  170. 170The Role of PlanningPlanning supports decision making:
    1. Identification of alternatives of future activities
    2. Selecting a good one or ideally the best
  171. 171Phases of Planning1. recognition and analysis of a decision problem
    2. definition of objectives
    3. forecasting of future developments
    4. identification and evaluation of feasible activities (solutions)
    5. selection of good solutions
  172. 172Model BuildingModel = simplified mapping of reality!
    Leaves out unnecessary detail, focus on relevant aspects & structural equality
    Examples of models: Forecasting models, simulation models, optimization models
  173. 173Planning Difficulties(1) Consideration of multiple cirteria or objectives: Multi-objective decision problem
    (2) Huge Number of alternatives: Combinatorial optimization problems
    (3) Uncertainty: Stochastic / random influences
  174. 174Approaches in APS for Multi objective problemsProblems: conflicting objectives, ambiquous preferences between alternatives
    Approaches:
    (1) Define minimum/maximum satisfaction levels for certain objectives
    (2) Try to price objectives as revenues or costs and maximize marginal profit
    (3) Use a weighted sum of scores or scale values for objectives
  175. 175Approaches in APS for huge number of alternativesAPS rely on:
    (1) mathematical methods of operations research (Linear Programming, network flow algorithms, dynamic programming,...)
    (2) heuristic optimization approaches (eg. local search)
  176. 176Approaches in APS for "uncertainity"APS use different concepts for dealing with uncertainty:
    (1) Buffer time and safety stocks to mitigate forecast errors
    (2) Rolling horizon planning to react to deviations from the plan
    (3) Event-driven planning (machine breakdown, arrival of urgent
    customer order, etc.)
    → Requires availability of data at any time
  177. 177Characteristics of advanced Planning System (APS)(1) Integral planning of whole supply chain (of single enterprise or network of enterprises)
    (2) True optimization:
    (2.1) Definition of constraints, alternatives and obj. for planning problems
    (2.2) Application of exact or heuristic optimization methods
    (3) Hierarchical planning:
    (3.1) SC planning not possible in a "monolithic" manner
    (3.2) Successive solution of planning problems would miss optimality too
    (3.3) HPS try to find a compromise
  178. 1785 main elements of Hierarchical Production Planning(1) Decomposition and hierarchical structure
    (2) Aggregation
    (3) Hierarchical coordination
    (4) Model building
    (5) Model solving
  179. 179Decomposition in Hierarchical Production PlanningOverall decision problem is split into two or more decision levels:
    1. Top level represents decisions with greatest impact
    2. One single upper level for each decision level
    3. Upper level defines the constraints for decisions
    4. Multiple decision units (e.g., production sites) possible at each level
  180. 180Hierarchical Coordination in Hierarchical Production PlanningAchieved by directives and feedback, for example
    (1) Target inventory levels for products
    (2) Cost coefficients
    (3) Information on achievement of objectives from lower level
  181. 181Aggregation in Hierarchical Production PlanningMain goals:
    (1) Reduce complexity
    (2) Mitigate uncertainty
    Time, products, resources, locations,...
  182. 182Model building in Hierarchical Production Planning(1) Model should adequately represent the decision situation
    (2) Involves disaggregation
    (3) Meet targets from upper level and consider effects on next lower level
  183. 183Model solving in Hierarchical Production PlanningHeuristics, optimization algorithms, manual procedures, …
  184. 184Structure of Advanced Planning SystemsStructure of Advanced Planning Systems
  185. 185Order Promising in APSGoal: Calculation of first promise date for customer orders → Affects order lead time and on time delivery
    Fast and reliable order promising is important but gets more and more difficult because of # of products, products life cycle get shorter, # of customer increases,...
    Standard MRP logic: Quote orders against scheduled receipts or standard
    lead times
    Common today in ERP and APS: Quote orders against master plan
  186. 186ATP in Assemble-to-Order EnvironmentsMain properties:
    - Components are produced based on forecast
    - Final assembly is order-driven
    - Actual configuration determined at order entry time (-> Configure- to- order (CTO))
    - MPS based on planning BOM
    => Multi-level ATP
  187. 187Multi level ATP (Available-to-promise)(1) Explode actual bill of material for configured product
    (2) Compute component request dates from customer requested date
    (3) Latest availability date among components crucial for order promising
    (4) Synchronize ATP consumptions according to final promised date
  188. 188Available-to-promise in Make-to-order environmentsSimilar to ATO, but earlier decoupling point.
    ATP represents:
    - Availability of raw material
    - Availability of capacity
    => Capable-to-promise
  189. 189Capable-to-promise (Dickersbach, 2004)Relies on production planning and scheduling module for order promising:
    *Forward customer order to PP/DS module
    *Insert it into current schedule
    *Return completion date as promised delivery date
    → Simulate execution of production order
  190. 190Production Planning & SchedulingGoal: Creating detailed production schedule for the shop floor
    Production schedule defines:
    (1) for each prod. order: start + completion time on required resources
    (2) sequence of orders on each resource
    Planning horizon: May vary between 1 day and several weeks (depending on largest throughput time of an order & availability of enough customer orders or reliable forecasts)
    Level of detail depending on production type!
    Job shop type: Scheduling of order on potential bottlenecks
    Cellular manufacturing/GT: Bucket-oriented capacity check
  191. 191Linkage between APS Scheduling to MPSDirectives / constraints received from upper level:
    (1) Available capacity (including overtime, additional shifts)
    (2) Availability of items from upstream units in the SC
    (3) Purchase agreements concerning materials from external suppliers
    (4) Target inventory levels (e.g., seasonal stock)
    (5) due dates for orders to be delivered to next downstream unit
  192. 192From a Model to a Production ScheduleFrom a Model to a Production Schedule
  193. 193Model Building in Detailed SchedulingStructural data: parts/component + BOM (single level only), resources, setup matrices, factory calendar,...
    Production process model (PPM):
    (1) Covers routings and processing instructions
    (2) Records both number of items per order + resource consumption per item
    (3) Linkage between different BOM levels via pegging
    Situation-dependent data: Initial inventory levels, WIP, Setup state of resources, set of orders to be processed withina given time interval, lot sizing rules
    Objectives: Throughput related, due date related, setup related... Multi-criteria optimization mostly via combined objectives (weighted sums)
  194. 194One of the goals of aggregation is to reduce the level of detail in planning. What is the main motivation for this ambition? a. We do not have the information in the required level of detail. b. It is impossible to create a production plan at a more detailed level. c. Limiting the computational effort, because otherwise it would take too long to solve AP problems. d. Reducing the effort needed to maintain and update detailed production plans (e.g., from week to week).d.
    Reducing the effort needed to maintain and update detailed production plans (e.g., from week to week).
  195. 195Choosing "sales dollars" as an aggregate unit for planning... a. has to be used with caution because capacity requirements for products may not be directly proportional to their selling price. b. is a natural aggregate unit for a company with large variety of products. c. is applicable independently of the ratios between selling prices and required working hours of products.a.
    has to be used with caution because capacity requirements for products may not be directly proportional to their selling price.
    b.
    is a natural aggregate unit for a company with large variety of products.
  196. 196In AP, we consider demand as deterministic, because... a. considering demand in a stochastic fashion does not allow for covering effects like trends or seasonality. b. we would not be able to solve an AP problem in a stochastic fashion (by stochastic optimization, for example). c. demand can usually be well anticipated when using appropriate forecasting techniques. d. it is easier to do so.a.
    considering demand in a stochastic fashion does not allow for covering effects like trends or seasonality.
  197. 197Capacity requirements planning (Prerequisites, Advantages & Disadvantages=Prerequisites:
    - BOM, routing/relationships between components, modules,..., time standards, lead time
    - Status of MRP-planned order & open shop orders
    Advantages: Provides more precise capacity requirements information than MPS resource profiles, especially for immediate time periods
    Disadvantages:
    - Large number of source data required
    - Computational effort
  198. 198Is the information flow from Sales & Operations Planning to Aggregate Planning and finally Master Production Scheduling is one-way?No. The information flow may in fact be bi-directional. Consider, for example, the result of ATP (Available-to-Promise) calculations. This is valuable information that has to be passed back to the demand management unit to help them better coordinate the customer order delivery dates.

Information

Description
logistik Flashcards on Production Analysis, created by Marie Scholten on 31/05/2023.
Created by
Marie Scholten
Views
0
History
  • Created by Marie Scholten · May 2, 2023
  • Copied by Marie Scholten · May 18, 2023
  • Copied by Marie Scholten · May 31, 2023
Rating

Feedback

What is it about?

What do you think of the new look?

How do you like it?