Level: Advanced
Michelle Brockway, Statistical Engineer, Corning Inc
In manufacturing, integration of data analytics can often be slow and under-utilized. However, the earlier the integration occurs the more rapid insights are generated. In this use-case, a predictive model was implemented alongside the startup of several production lines, showing the value in the early adoption of analytics. The optimized predictive model not only allowed our team to predict parts out of specification, but also gain understanding of key parameters to create process limits. Using incoming material attributes alone, we can adequately predict the amount of post-processed material that will be out of specification. This type of prediction allows for the validation of material attributes prior to shipping to our manufacturing plant, as well as insight on process changes, saving the business a significant amount of resources. The presentation will walk through the process in which the model was built, the background, implementation, and plan for maintaining the model in a manufacturing setting.
For proprietary purposes, the exact product and the raw data cannot be disclosed. The data will be transformed to show the methods used and implementation, but not disclose any proprietary information.
