Industry Pharmaceutical and Life Sciences, Manufacturing
Specialization Or Business Function Supply Chain and Logistics (Manufacturing Process Optimization), Chemistry
Technical Function Data Visualization (Time Series), Analytics (Data Mining, Machine Learning, Time Series Analysis, Descriptive Analysis)
Technology & Tools
This project is for Merck KGaA (known as Millipore Sigma in the US and Canada), one of the largest pharmaceutical and chemical companies in the world.
During the manufacturing of a particular product, a reaction is unexpectedly stalling and additional catalyst must be added. This is disruptive to the manufacturing process, and the team would like to understand what is causing it in order to prevent this from occurring in future batches. This product is manufactured infrequently so only about 30 batches of data are available from the past 3 years. There are two pieces of equipment the reaction may occur on; G144P420 or G144P422. For each batch, there are time series data recorded during the reaction. This time series data includes pressure, temperature, etc. This time series data is recorded roughly every second; however, there are gaps. Ideally, we would like to use feature extraction to identify differences between good and bad batches of product.
Attached is sample data for one good batch of product. Each batch has one file corresponding to it (of the same format as the one attached). Note that the tag name column refers to the piece of equipment and the metric being collected in German. For instance, "G144P420.Innentemp" means equipment G144P420 and the metric recorded was the Inner Temperature.
As a final deliverable, we would like to understand what key features may be driving differences in product quality.
We will not consider proposals over the rate that was specified.
Matching Providers