1 / 59

The Illusion of FAB Capacity in Central Planning

Learn how to incorporate the complexity of FAB capacity, tool deployment, and operating curve into central planning for demand-supply networks in semiconductor production. Discover the challenges and strategies for finding CAPAVAIL (available capacity) in FAB routes and deployment.

fmcnair
Download Presentation

The Illusion of FAB Capacity in Central Planning

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. The Ongoing Challenge - TutorialThe Illusion Of CapacityIncorporating the Complexity Of FAB Capacity (tool deployment & operating curve) into Central Planning for Demand-Supply Networks for the production of semiconductor based packaged goods with substantial non-FAB complexityTraditional CPE capacity with resource entitiesand its source in FAB Routes & Deploymentin steady-state start patterns”(“never mind transitions” – ramp ups or downs)part 3 of 4 Dr. Ken Fordyce & John Fournier, IBMProf. John Milne, Clarkson UniversityDr. Harpal Singh, CEO Arkieva Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  2. Hunt for CAPAVAIL(& CAPREQ) in FAB Routes and Deployment Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  3. Outline • Overview of the Demand Supply Network for the production of semiconductor based package goods • Warring factions • Decision Tiers • Aggregate FAB Planning • Central Planning • Two major challenges • Planned lack of tool uniformity • Inherent variability • Basics of Aggregate Factory Planning • Can this wafer start profile be supported • Near Term Deployment • WIP Projection • Basics of Central Planning • Basic Functions • Historical emphasis on non-FAB complexity • Alternate BOM for example • Handle FAB Capacity with limits stated as wafer starts • Wafer start equivalents evolved to nested wafer starts • Second look at capacity (CAPREQ and CAPAVAIL) • Linear methods in central planning engines • FAB complexity creates miss match • Operating Curve and Cycle time Tax • Creating CPE type capacity from routes and consumptions of tools • The complexity of deployment & routes Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  4. definitions • CAPREQ - establishing a consumption rate for each unit of production by that manufacturing activity for the selected resource • CAPAVAIL - providing the total available capacity for the resource. connecting manufacturing releases (starts) to resource consumption with a linear relationship • Route – sequence of manufacturing actions • Deployment – (alternative Machines); PSO – partially shared overlap between tools and operations Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  5. Deployment / Alternative Machines PSO – partially shared overlap between tools and operations 1 – oper/tool link active 0 – not allowed Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  6. Exampleto highlight search for CAPAVAIL & CAPREQ from resource entity to resource operations & tools Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  7. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Six options • Capacity Allocation Variable set and Dynamic CAPAVAIL • Quick look at Challenges for Heuristic Option • Complexity of Interactions Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  8. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Capacity Allocation Variable set and Dynamic CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  9. Traditional CPE Capacity Information – resource entity level no operations or tools Fixed Consumption Rate into Fixed Capacity Available Traditional CPE Model Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  10. Wafer Start Decision Variables • XA = number of wafers of Antelope • XG = number of wafers of Gazelle • XL = number of wafers of Lion Capacity Constraint Equations – one for each resource entity Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  11. Traditional CPE BCD (best can do) Model Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  12. Where do we get the CAPREQ & CAPAVAIL values? How does this relate to consumption of tools along the route Created from the complexity of FAB Routes Where do these values come from? FAB Routes Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  13. Where do we get the CAPREQ & CAPAVAIL values? How does this relate to consumption of tools along the route FAB Routes Transform To simpler Capacity Statements Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  14. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Capacity Allocation Variable set and Dynamic CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  15. Abbreviated Route for Antelope Focus MUV, DUV, ION, and ETCH toolsets FAB Routes Tool set, rpt iden tools sequence Each operation has an id An operation can be repeated within a route for the same part; operation can be used in multiple routes (parts) 2nd MUV operation Each Operation ID has set of specific tools within the tool set that are deployed to this operation. Therefore the lot links to the tool options via the operation id This near term “steady state” deployment. At dispatch, the tool options for the lot may be different then the deployment for manufacturing engineering reasons (temporary restriction) or business decision for flow control and allocation imposed SHARED Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  16. Yes We should Use Raw process time Pass count Is CAPREQ for Antelope CAPAVAIL is number of passes available each time unit Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  17. Antelope lot Tool Connection to find CAPAVAIL ? Lot Connects to Operation Operation Connects To Tools (deployment) Route Operations 010 & 095 Operation ID muvop01 Steady state deployment MUV Tool 03 MUV Tool 04 MUV Tool 05 MUV Tool 01 MUV Tool 02 Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  18. Antelope lot Hunt for CAPAVAIL no easy answer? Route Operations 010 & 095 Short term adjustment Operation ID muvop01 Steady state deployment MUV Tool 04 MUV Tool 03 MUV Tool 05 MUV Tool 01 MUV Tool 02 Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  19. RPT is HARD to Estimate Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  20. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Capacity Allocation Variable set and Dynamic CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  21. Focus on MUV Operations in Route Antelope & MUV op01 -> op02 -> op03 -> op01 -> op05 Convert from Sequence to Count (passes) Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  22. Focus on MUV Operations in Route Each Antelope Lot passes Through MUVop01 twice If we assume all MUV operations are the same Then the column sum is the CAPREQ for MUV Resource Entity Under “assumption every operation in 24 hours” this is CAPREQ by MUV operation If the start mix is constant over an extended period of time Then the average workload per operation is the same each day independent of cycle time (measured in CTM – cycle time multiplier) --- ignoring the need for “idle without WIP” capacity to be allocated to meet the cycle time objective; where the shorter the cycle time the more “spare” capacity has to be reserved Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  23. Extending Capacity Required to All Unique MUV Operations Original CAPREQ Lost visibility To which MUV operation 5 MUV passes are split across 7 MUV operations 2 – 1 – 1 – 0 – 1 - 0 - 0 From one MUV constraints to 7 – one for each unique MUV operation How do we determine cap0X? Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  24. The Path From Route to CAPREQ for each MUV Operation Sequence Count CAPREQ each unique MUV operation Where CAPREQ is pass count at each MUV operation Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  25. Extending MUV Resource Entity Capacity Equation To one equation for each unique MUV Operation Eq (1-1) original model is replaced by Equation Set 1 Link tools to operations How do We Determine cap0X? Hunt for CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  26. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Capacity Allocation Variable set and Dynamic CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  27. Linking MUV Operations to MUV Tools “?” is 0 if tool can not service operation, 1 if it can more advanced version value is between 0 and 1 inclusive Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  28. Linking MUV Operations to MUV Tools “???” raw capacity available for tool after accounting various factors Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  29. Raw “Effective” Capacity Available non-trivial to estimate Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  30. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Capacity Allocation Variable set and Dynamic CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  31. Linking MUV Operations to MUV Tools – case 1 Simplest Case – All Tools Can Handle All operations Assume CAPAVAIL each tool is 20 Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  32. Equation Set 1 1 equation each MUV Operation Equation Set 1 Can be replace with this equation When all tools handle all operations When All tools can handle all Operations Capacity Consumption for MUV Can be represented with a single equation Original MUV Capacity Equation Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  33. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Capacity Allocation Variable set and Dynamic CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  34. Linking MUV Operations to MUV Tools – Case 2 Case – Two Independent Groups MUV can be divided into two independent Resource Entities • MUV Resource Entity 1 (MUVRE1) • tools 1, 2, and 3 • servicing operations 1, 2, and 3 • MUV Resource Entity 2 (MUVRE2) • tools 4 and 5 • servicing operations 4, 5, 6, and 7 Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  35. Equation Set MUV-RE2 Equation Set MUV-RE1 Equation Set MUV – the 7 MUV Operations split into two Equation Sets (MUV-RE1 and MUV-RE2) Divide Equation Set into two groups Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  36. Equation Set 1 Is split into Equation Set 2 and Set 3 Equation Set 2 and 3 each can be replaced with single equation When all tools handle all operations within A specific group of tools and operations Equation Set 2 Operation 1, 2, 3 and tools 1,2, 3 Replaces Equation Set 2 Equation Set 3 Operation 4, 5, 6, 7 and tools 4, 5 Replaces Equation Set 3 When MUV can be split into two independent components Capacity Consumption for MUV can be represented with two equations Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  37. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Six options • Capacity Allocation Variable set and Dynamic CAPAVAIL Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  38. Linking MUV Operations to MUV Tools Real world complexity non-uniform deployment MUVRE1 all tools do not handle all operations op01 serviced by TL01 & TL02 op02 serviced by TL01 & TL03 op03 serviced by TL02 Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  39. The Critical Question How does non-uniform deployment Impact our ability to estimate cap01, cap02, and cap03 Equation Set MUV-RE1 It creates a situation that requires a careful balance between solution accuracy, model complexity, model performance, and stressing the social order Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  40. Six Options • Maximize Capacity Flexibility • Minimize Capacity Flexibility • projected wafer start profile • modify traditional method for capacity to handle or conditions • Capacity Allocation Decision Variable • Combination of options using heuristics to create resource entity Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  41. Six Options • Maximize Capacity Flexibility - make the assumption all tools can handle all operations and continue to use a single equation for MUV-RE1. The risk is we overstate capacity flexibility potentially committing the FAB to produce more than it is able to produce. If the “deviation” from uniformity is low, this is a reasonable option. • Minimize Capacity Flexibility - Establish a fixed allocation of tool capacity for tools 01, 02, and 03 to each operational constraint enabling us to estimate cap01, cap02, and cap03 and use a constraint equation for each operation and accept the risk of turning away business. When demand patterns have limited deviation over time, this is a reasonable option. Easiest But risk Pretty easy Lowest initial risk Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  42. Option 2: Fixed Allocation Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  43. Option 2: Fixed Allocation CAPREQ CAP need Allocation percentage Actual CAPAVAIL Need vs Available SIX SHORT Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  44. Six Options • Create a projected wafer start profile that would include demand priorities translated to the starts and use optimization to simultaneously make start decisions and allocate tools between operations to maximize supply against prioritized demand. (extension of CAPS / EPOS). Downside is we do not know the wafer start pattern until after the CPE runs to determine the best can do (BCD) start profile to meet prioritized demand. Therefore we would need to partition the CPE into an explode run and an implode run and loosely couple the FAB model with the rest of the CPE – probably in an iterative approach. • Modify the traditional method of stating capacity restrictions in CPE models to include the “or” conditions. This similar to, but not identical with, handling alternative operations in CPEs. Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  45. Six Options • Introduce a capacity allocation decision set to the model which is described in the next section. There is a rich history of this type of approach in FAB tool planning dating back to at least the early 1980s in practice and literature. This decision set is close to the “M” variable in Hung and Cheng (2002). • Institute a combination of these options guided by some heuristics to combine tools where reasonable into a single resource for the CPE and establish other structures to handle “problem tool sets.” More details Next slides Dynamic option 2 Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  46. Example Steps • Traditional CPE - capacity constraints at resource entity level • No details on operations and tools • FAB Routes – sequence, pass count • Using pass counts to create CAPREQ for each resource entity • Preliminary search for CAPAVAIL • Focus on MUV resource entity incorporating tools and operations • Creating capacity constraints for each MUV operation instead of one constraint for the MUV resource entity • Determining CAPREQ with pass count at each unique MUV operation • Hunt for CAPAVAIL • Cases / Options to find CAPAVAIL • Case 1: simplest, all tools can service all operations • Case 2: two independent groups • Case 3: asymmetric deployment – life gets complicated • Six options • Capacity Allocation Variable set and Dynamic CAPAVAIL Fixed or Input to Model Becomes a model decision Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  47. Capacity Allocation Decision Variable • The values for cap01?, cap02?, and cap03? (CAPAVAIL) in equation set MUV-RE1 are limited to being some combination of capacity allocated to each operational constraint from tools 01, 02, and 03 • that does not violate deployment restrictions and does not allocate more than 100% of each tool. • The decision on what percentage of each tool to allocate to each constraint (cap0N?) determines the CAPAVAIL for each capacity constraint equation which directly influences the wafer start profile the FAB can support. Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  48. Capacity Allocation Decision Variable Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  49. Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

  50. CAPREQ CAP need Allocation Decision Now made By model Not an input Allocation percentage Actual CAPAVAIL Need vs Available Fordyce, Fournier, Milne, Singh Illusion of FAB Capacity in Central Planning – hunt for CAPAVAIL

More Related