(179c) Big Data Analytics for Batch Performance Improvement
AIChE Spring Meeting and Global Congress on Process Safety
Wednesday, March 29, 2017 - 2:30pm to 3:00pm
In the first case study, we examine a batch system with multiple operating units. The general approach consists of: (1) identify general step sequences, (2) assess the correlation between steps and other common process interruptions, and (3) identify key variables in the steps of interest for improvement. The methodology is helpful to understand the sources of the highest variability that affect the batch quality variable of interest. The proposed methodology is validated on an industrial system where three batch units (with over 100 variables and many recipe steps) are operating in series. After performing the data analysis and identifying critical steps and process variables, the production rate of this batch system has shown consistent improvement.
In the second case study, we combine data-driven models constructed from batch process data with a controller to reduce process variation of a batch reaction system. After analyzing the batch dataset from this process, key recipe step was identified to be the source of variability. However, conventional control strategies cannot be applied easily due to the unique process equipment. As a result, we propose an alternative control strategy based on the model from the batch data analysis, where both normal operating data and design of experiment data were utilized. In addition to minimizing variability, the safety and equipment constraints are also considered in the control action, leading to robust and stable control. The effectiveness of the proposed batch control strategy was demonstrated to show considerable improvement through both simulated case study and plant implementation.
1. Qin SJ. 2014. Process data analytics in the era of big data. AIChE J. 60(9):3092â3100
2. Colegrove LF, Seasholtz MB, Khare C. 2016. Big data: getting started on the journey. Chem. Eng. Prog., March, pp. 41â45
3. Kadlec P, Gabrys B, Strandt S. 2009. Data-driven soft sensors in the process industry. Comput. Chem. Eng. 33(4):795â814