(708g) Combining Data Analytics and Scheduling – First Results and Open Challenges

Last year as part of many related AIChE presentations, we presented the SINGPRO project that aims at finding benefits from integrating data analytics, machine learning and optimization. The hype on big data analytics (Qin, 2014) and machine learning is converging to concrete projects and first results. A good perspective on this topic is given in Venkatasubmaranian (2019). It is clear that there is something new in the novel approaches for handling large amounts of data. It also cannot be disputed that new methodologies will be needed to handle large data sets and identify the best ways of deriving their value.

In this presentation, we want to revisit the SINGPRO project target and present the first findings. We aim at using the available data proactively to improve the quality and actuality of planning instead of relying on static data that do not reflect current operation conditions. This can be seen as a way of supporting the integration of scheduling and control (Touretzky et al., 2017). However, instead of defining workflows between the levels, we aim to provide more accurate estimates in advance in order to reduce the mismatch between planning and online operations. We have studied cases of merging Big Data platforms, machine learning and data analytics methods with process planning and scheduling optimization and will provide an overview of the first promising results. We will show results that can improve the estimation of processing times, leading to more robust schedules. We also show examples where using historical operational data allows us to exclude some decisions leading to smaller scheduling problems. Finally, the process data can also be used in estimating equipment conditions leading to better approaches that combine operational and maintenance scheduling optimization.

This is just a beginning, but the examples show that by creating collaboration interfaces between scheduling optimization, big data analytics and machine learning the process related decision-making loop can become more agile, self-aware and flexible.

References

Qin, S. J. (2014), Process data analytics in the era of big data. AIChE J., 60, 3092-3100.

Touretzky, C. R., Harjunkoski, I., & Baldea, M. (2017). Dynamic models and fault diagnosis-based triggers for closed-loop scheduling. AIChE Journal, 63(6), 1959-1973.

Venkatasubramanian, V. (2019). The promise of artificial intelligence in chemical engineering: Is it here, finally? AIChE Journal, 65(2), 466-478.