In the past few years, governmental initiatives and advances in sampling systems have led to an increased deployment of analytical technologies to characterize industrial processes. The proliferation of process analytical technology (PAT) has led to rapid increases in both the number of measurements performed and the volume of data generated. Traditional data analysis approaches require the development of calibration and validation models for each type of measurement. This approach can lead to gaps in the data space which in turn leads to uncertainty and deficiencies in process understanding, as each model is used to characterize the process independently.Data fusion provides a solution to this problem by fusing all of the analytical measurements into one continuous block of data. The fused data block is then used to produce one calibration and one validation model. Data fusion allows complete coverage of the entire process space; significantly decreasing uncertainty and providing optimal process characterization.This paper will focus on the application of data fusion to a petrochemical process. Particular attention will be paid to the data fusion procedure and variable selection methods. It will be demonstrated that better models, and therefore improved process control, results from the effective implementation of data fusion.

Surrounding and solving analytical problems with data fusion

246th National Meeting (2013)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.