Level: Advanced Job Function: Analyst / Scientist / Engineer Mattia Vallerio, Data Scientist, BASF Antwerpen Salvador Santiago, Team Lead for Lean Six Sigma, BASF Antwerpen
Design of experiments (DOE) is a powerful tool to obtain the maximum amount of information within a limited amount of time or resources. Most of the available DOE techniques focus mainly on designing experiments for setups that can be largely controlled, i.e., in labs or pilot environments. What about a world-scale manufacturing process where the signal-to-noise ratio is very low, where some input parameters cannot be fully controlled, where interaction and correlation between parameters is the norm and where experimental cost is prohibitive? Is DOE still an option? How do we know if it makes sense to continue experimenting?
This talk explores the use of JMP to define DOE campaigns for world-scale manufacturing processes with the help of actual use cases. The authors will first highlight the challenges faced in DOE for such processes and will then present and demonstrate a methodology in JMP to tackle this problem. To do this, we exploit the one advantage such a manufacturing process typically has: the presence of historical data.
The main purpose of this contribution is to start a constructive discussion within the JMP community around the presented problem and to trigger the development of further improvement of DOE techniques.