1. A case study of configuration management methods in a major automotive OEM Keith Phelan ktphela@clemson.edu Mechanical Engineering Clemson University Clemson, SC Crystal Wilson crysta3@clemson.edu Industrial Engineering Clemson University Clemson, SC Joshua Summers jsummer@clemson.edu Mechanical Engineering Clemson University Clemson, SC Mary Beth Kurz mkurz@clemson.edu Industrial Engineering Clemson University Clemson, SC

2. 2/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Background Trend towards mass customization – Increased need to meet the demands of specific customers – Maintain the ability to produce items efficiently in a cost-effective manner – Blending of artisan craftsmanship with mass production The need for configuration management – Increased use of product families or configurable products – As the number of variants increases, the complexity of effectively managing all of the possible options increases significantly – Assists in executing change management on highly interrelated products

3. 3/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Existing Techniques Rule-based reasoning – Relies on a set of rules (typically if-then conditions) to manage configurations Model-based reasoning – The product is modeled as a system that can be broken down into components and relationships between components Description logics, constraint-based and resource-based models are examples Case-based reasoning – Uses preexisting product variants to assist in developing additional configurations

4. 4/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Case Study Justification Major automotive OEM in Spartanburg, SC – with high configuration complexity Four distinct vehicle models are assembled at the facility, each with numerous variants For a given model, approximately 800 configurator rules govern greater than 500 vehicle options, that affect the assembly of approximately 1500 parts Over 300,000 vehicles manufactured yearly, with no two vehicles being the same – with regularly changing configurations Three planned launches per year, one of which is a major launch Additional minor changes occur on a regular basis as necessary (approximately 350 changes to a single model over 5 years)

5. 5/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Case Study Procedure Interviews – Conducted with Launch and Change Control personnel – 24 hours of targeted interviews conducted over a 3-month period – Primarily conducted on-site at the manufacturing facility (2 interviews conducted over teleconference with personnel in Germany) PositionSection Time (hours) Topics of Discussion Section managerLaunch and Change Control 6Configuration management and change processes Launch and change coordinator Launch and Change Control 10Configuration management and change processes Launch planning coordinator Launch and Change Control 2Launch planning and configuration change process Release quality assurance specialist Launch and Change Control 2Vehicle ordering and configuration management systems Launch and change coordinator Launch and Change Control 1Launch planning and parts release Electronics specialistElectrical/Electronics Validation 1Configuration verification process Product data managerProduct Data1Rule database and configuration change process Product data managerSpecial vehicle projects1Configuration change process

6. 6/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Case Study Procedure Document analysis – Extensive analysis of the documents/systems used to manage configurations and configuration changes Rule database Configuration change request/implementation form Ethnographic research – Numerous visits to the manufacturing facility to observe Launch and Change Control personnel How changes to the configuration are verified How changes to the system are requested and implemented How vehicle assemblies are verified prior to a launch How vehicles are ordered by the consumer for production on the line

7. 7/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Current Configuration Process Based on a master rule database governing how vehicle options interact and which model codes are applicable for each rule Vehicles are ordered and configured external to the manufacturer (either by a dealership or a customer) Based on the vehicle options ordered by the consumer, a separate system specifies the vehicles for assembly at the plant Clear example of rule-based reasoning Limitations – With a large number of rules (>800), it is nearly impossible to ensure the accuracy of the rule database – Verification is largely conducted based on experiential knowledge – Often, errors in the ruleset are only identified during assembly – Difficult to understand complete implications of changes to the system

8. 8/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Current Management Process Much of the change verification process is conducted based on experiential knowledge Weekly approval meeting Implementation into rule set Approval Configuration change proposed Report issues back for consideration Change to affected parties for review Initial review Issue/feedback review Identify issues Distribution

9. 9/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Proposed Changes Implementation of a rule management system using design structure matrices (DSMs) – Models the rules that govern the option relationships as interactions in the matrix – Allows for simplified querying and organization of the data found in the master ruleset Use of data visualization techniques to better understand relationships – Using a modified connectivity graph to view the relationships between options and parts – Simplifies the process of identifying how the different components relate to each other – Makes it easier to understand changes to the system by direct comparison of graphs

10. 10/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Graph-Based Visualization Green (thick) lines indicate inclusions Red (thin) lines indicate exclusions Blue (medium) lines indicate either/or

11. 11/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Ongoing Work User study to determine the impact of different visualization methods Variables: – Graph layout – Inclusion of part information – Color-coding of the graph – Specific color-coding for part information Details – Full-factorial method – 74 participants Procedure – Participants given a version of the data visualization graph and asked to answer a series of questions about the configuration system described in the graph Findings – Layout and coloring of the graph did not have a significant impact – Removal of part information significantly increased the accuracy of responses for purely option-related questions

12. 12/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Ongoing Work Use of a visualization support method to identify potential problems in ongoing configuration changes – Windshield change – Indian language model change – Emission standards change Initial results: – Use of graph-based method allows easy discovery of why the change is written as is – Inclusion of parts (for a small subset of the VRM) makes it simple to determine if any parts are no longer valid – Use of graph-based method allows for the easy comparison between multiple model codes – Use of graph-based method assists in understanding the consequences of a complicated and confusing change

13. 13/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Future work Implementation of automated error detection – Queries were identified during user to understand relationships between components – From the queries, algorithms are being developed to use the DSM-based software to automatically check for errors in the system Increase user interface capabilities – Rule authoring using graph-based visualization tool – Sub-graph matching for creating visualization graph layouts Additional testing and implementation for in-progress configuration changes

14. 14/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Conclusions The Automotive OEM primarily uses rule-based reasoning for its configuration management processes – Limitations due to difficulty in verifying current configurations and changes to the configuration set Implementation of DSM-based rule management system – Increases the user’s ability to query and manipulate rules – Allows for future implementation of automated querying and rule validation Implementation of graph-based visualization techniques – Simplifies the process of identifying interactions between options – Allows for future implementation of rule authoring using a graphical interface

15. 15/15 2014.08.19 ktphela@clemson.edu http://www.clemson.edu/ces/cedar Questions?