Simulation and Modelling Collaboration with PLM Kenneth J Rasche, P.E. Senior Engineering Manager Whirlpool Corp.
Executive Summary PROJECT PROCESS Team rarely is able to reuse knowledge This part of the process is usually not consistent Rarely have the right data and info to create reusable knowledge Projects Create Data Sometimes SMEs Create Info Sometimes SMEs Create Knowledge Don t wait and see if we can and/or do create knowledge : Drive Reusable Knowledge with the design process! COLLABORATIVE PROCESS Project Needs viewed through knowledge Organizational Knowledge Framework BEFORE Reuse Update and Create SMEs pre-define Project Deliverables to update and create reusable knowledge chunks Organizational Knowledge Framework AFTER Knowledge Gaps for this specific project
Executive Summary Organizational Collaboration of simulation and modeling offers huge competitive advantages by allowing an organization to move closer to knowledge Nirvana where: No knowledge has to be learned more than once The organization only learns what it needs to learn PLM systems offer capabilities to enable an organizational collaboration process, but some lower level processes have to be in place to maximize the benefit Product Knowledge Structure Reusable Knowledge Surrogate Models Design for Population
Content Context - Whirlpool Product Knowledge Structure Reusable Knowledge Surrogate Models Design for Population Organizational Collaboration in PLM 4
Context - Whirlpool A very successful 100+ years Sales $2.3M $1.1B $3.4B $8.2B $14B (1942) (1970) (1985) (1995) (2005) ~$20B (Current)
Context - Whirlpool Main Product Lines Laundry Refrigeration Cooking Cleaning Small Appliances Vertical Axis Horizontal Axis Dryers Pedestals Accessories Fabric refreshers Side X Side Top Mount Bot Mount BIR SxS BIR Bot Mount CD Side X Side CD Bot Mount Refrig. Drawers Cook Top Range Built In Slide in Microwaves Freestanding Warming drawers Tall Tub Stainless Tub Plastic tub Water products Trash compactors Disposers Stand Mixer Toasters Blenders Food processor Coffee maker Immersion blenders Coffee grinders
NORTH AMERICA REGION ~30,000 Employees 7 Technical Centers 12 Factories LATIN AMERICA REGION ~21,000 Employees 3 Technical Centers 3 Factories Context - Whirlpool Global Footprint GLOBALLY 9+ Brands ~30 factories 20 Technical Centers 10,000+ SKU s 120+ Product platforms 100,000+ parts 5,000 Technical People EUROPE REGION ~14,000 Employees 5 Technical Centers 10 Factories ASIA REGION ~5,000 Employees 5 Technical Centers 5 Factories Numerous organizational borders across businesses, regions, functions, product categories, and locations
Product Knowledge Structure Like Dewey Decimal System, product knowledge structure shows us where the knowledge belongs Product Knowledge Structure needs to be defined to prevent gaps and overlaps Good search tool is not the answer A place for everything, and everything in its place Provides map to where knowledge is and where it should go Structure has 2 axes: Product Reusable pieces of the product Attributes System models of product performance Provides all teams with a live place for collaboration 8
Attributes, sub-attributes and functions Geometric Interfaces Product Knowledge Structure Engineering Sub-assemblies (sub-systems) Hinge - Bottom Attribute Interfaces Reliability Door Alignment Knowledge Chunk Soft sub-attribute requirements and collaboration allow for greater innovation and system optimization There is tremendous power in having one version of the truth 9
Application Input 1 Fail Pass Reusable Product Knowledge Chunk Attribute Output Force CFM Temperature Delta Pressure Displacement Wattage etc. Design B Design A Point data is typically not very reusable, we need to connect the dots Level 1 reusability is evaluating the design over a range of foreseeable system inputs Level 2 reusability is evaluating the design concept over a range of foreseeable system inputs and foreseeable dimensional changes Application Input Force CFM Temperature Delta Pressure Displacement Wattage etc. 10
Application Input 2 Application Input 1 Reusable Product Knowledge Chunk Knowledge Chunk: Initial version of the knowledge chunk shows clearly that Design A passes and by what margin knowledge is reusable and provides format for additional design development New Application: New application is added and clearly shows Design A will not work for this new application we need a new design Pass Fail Concept B2 Concept B1 Design A Concept B1: First concept is added and performance is clearly not enough improvement Concept B2: The concept is iterated and B2 clearly shows there is significant margin to the requirement 11
Vertical Deflection, mm Old Design Load New Design Load Reusable Product Knowledge Chunk 2.5 2.0 Pros Automate simulations upfront so additional work is mostly compute time, not human time 1.5 1.0 Old Design Target New Design Target If new designs are in inference space, the model does not have to be run again the knowledge chunk already has the answer AK3X_69mm Better development of understanding how the design performs helps AK3X - Given food Load ensure models are representative of physical parts and application AK3X_64mm Correlation and Calibration can be built up over multiple applications avoid chasing the latest set of test results 0.5 Cons 0.0 0.00 200.00 400.00 600.00 800.00 1000.00 Vertical Load Pressure to just do the minimum for my project Standardization can impede innovation 12
Surrogate Models Pro s Interactions Fast Optimize system configuration Develop control logic System Models (Model of system level performance) Con s +/- 5% to 10% can t evaluate many design changes miss cost or performance opportunities Get the best of both worlds! Pro s Can evaluate all design changes Optimize performance vs cost Detailed Models (FEA, CFD, captures geometry) Con s Slow Sub-optimization risk 13
Surrogate Models Clearly define interfaces between all models Characterize on the variables that drive the system design Cascade inputs and outputs up and down the system of models Avoid skipping levels
Surrogate Models Characterize sub-assembly models to create surrogate models Characterize around the right variables Make the surrogate model as reusable as possible Don t skip levels System of Models (initial rev) Basic equations characterizations System of Models (next rev) Characterizations replace basic equations Improved characterizations Time Building Knowledge 15
Surrogate Models Multi-dimensional Data sets to capture the complete inference space 16
Occurrences Design for Population Variation of system level performance from sample test to sample test makes it difficult to correlate and calibrate Test 1 Sample Test 2 Sample Nominal Structuring the problem around the knowledge chunks allows us to understand the variation of each piece of the product and how it contributes to the overall performance X% Population Value Spec. limit System Population Performance 17
Design for Population Door Alignment = f(cab Square, Door Stiff, Hinge Stiff, Hinge Loc, Cab Stiff, etc.) Cabinet Square Door Stiffness Hinge Stiffness System Performance Evaluating the knowledge chunks individually allows you to remove much of the variation Hinge Location Monte Carlo simulations are required since very few of the distributions are normal Cabinet Stiffness 18
Collaboration with PLM Structure Team Moldflow Team CFD Team Project Manufacturin g Str001 Thm001 Perf001 Mat003 Str002 Str003 Thm002 Perf003 Perf004 System Model Str002 Mat003 PLM enables sharing Shares/reuses models with multiple deliverables/project and product teams Creates and clones their own models Revisions to models are reflected immediately everywhere they are shared Enables driving product knowledge chunk development through projects Perf004 Thm001 System Model Str001 Mat003 Perf004 Thm002 1
Soft sub-system Requirements Collaboration with PLM System Requirement $10 $13 $22 Rigid sub-system Requirements $11 $13 $18 Collaboration = $$ $45 $42 = $3 savings Cost functions are not linear Achieving the last 2% of a sub-system requirement may cost $4 Exceeding a different sub-system requirement may only cost $1 Sub-system designs migrate towards cost inflection points in optimum designs 20
Collaboration with PLM PROJECT PROCESS (not collaborative) Team rarely is able to reuse knowledge This part of the process is usually not consistent Rarely have the right data and info to create reusable knowledge Projects Create Data Sometimes SMEs Create Info Sometimes SMEs Create Knowledge Typical project design process generates lots of data. Teams are often off to the next project before much/any of the learnings are captured May or may not be able to distill or format into reusable information and/or knowledge New project teams often don t know where to find existing knowledge New project teams don t trust the existing knowledge
Collaboration with PLM PLM database capabilities allow functional teams and project teams to share information real time Information and knowledge sharing does not have to wait until it is published Definition of knowledge chunks allows functional teams to share information with several teams because it is reusable Information/knowledge is not system application specific Information/knowledge can be reused in the future Knowledge structure definition allows predefining project deliverables to be reusable knowledge Don t wait and see if we can create reusable knowledge after the project is complete Allows multiple teams to share the knowledge as it is created Imagine the power of your organization never having to learn the same thing more than once! COLLABORATIVE PROCESS Project Needs viewed through knowledge Organizational Knowledge Framework BEFORE Reuse Update and Create SMEs pre-define Project Deliverables to update and create reusable knowledge chunks Organizational Knowledge Framework AFTER Knowledge Gap for this specific project
Attributes, subattributes and functions Conclusions Organizational Collaboration of simulation and modeling offers huge competitive advantages by allowing an organization to move closer to knowledge Nirvana where: No knowledge has to be learned more than once The organization only learns what it needs to learn PLM systems offer capabilities to enable an organizational collaboration process, but some lower level processes have to be in place to maximize the benefit Product Knowledge Structure Reusable Knowledge Surrogate Models Design for Population Engineering Sub-assemblies (sub-systems) Hinge - Bottom Reliability Door Alignment 23
Thank You! 24