(530h) Scheduling and Analytics – Towards Better Planning

There is a lot of hype ongoing on big data analytics (Qin, 2014) and machine learning. Among others, AIChE conference has also arranged topical conferences on the topic, and for a good reason. Most companies collect continuously data from sensors that is stored for a certain time but never actually used, unless there is a need for post analytics as part of trouble shooting (Yidan et al., 2016). One attempt to create true value from the data is to use it proactively to improve the quality and actuality of planning. Nevertheless, often a schedule that is based on statistical average data is outdated already by the time it gets sent to the plant floor and due to the hierarchical planning structures it is very difficult to quickly adapt a schedule to changing conditions. This is a challenge that has also been looked into in integration of scheduling and control studies (Touretzky et al., 2017). The presented project SINGPRO will merge Big Data platforms, machine learning and data analytics methods with process planning and scheduling optimization. The goal is to create online, reactive and anticipative tools for more sustainable and efficient operation. The currently employed classical mathematical optimization models (Harjunkoski et al., 2014) are often limited by fixed parameter sets, which are commonly updated off-line and represent only statistical averages. Such parameters could be estimated much more precisely in an on-line fashion using Big Data technologies. By creating collaboration interfaces between scheduling optimization, big data analytics and machine learning, the process related decision-making loop will become much more agile, self-aware and flexible.

With sophisticated data analytics methods, one can embed to the overall key performance indicators (KPI) also all information about the process, e.g., tracking abnormal situations (anomaly detection), individual process equipment performance degradations (predictive maintenance), anticipated process timings (prediction of process behavior) and scenario simulation (e.g., artificial intelligence AI planning). Such an approach will help to select the best production strategies in order to maintain, e.g., production and energy efficiency as well as sustainability in rapidly changing market situations through data-driven self-adaptive scheduling models. The topic of data-driven models have already been investigated in other domains (Van der Aalst et al., 2004) and tools become available for the process industry (Wilson and Sahinidis, 2017). It can be expected that Industrial Internet of Things (IIoT) provides the needed seamless connectivity, cloud computing infrastructure and service-based business models to realize this vision.

References

Qin, S. J. (2014), Process data analytics in the era of big data. AIChE J., 60, 3092-3100. doi:10.1002/aic.14523

Yidan Shu, Liang Ming, Feifan Cheng, Zhanpeng Zhang, Jinsong Zhao (2016), Abnormal situation management: Challenges and opportunities in the big data era. Computers & Chemical Engineering, 91m 104-113. doi:10.1016/j.compchemeng.2016.04.011

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. doi:10.1002/aic.15564

Harjunkoski, I., Maravelias, C. T., Bongers, P., Castro, P. M., Engell, S., Grossmann, I. E., . . . Wassick, J. (2014). Scope for industrial applications of production scheduling models and solution methods. Computers and Chemical Engineering, 62, 161-193. doi:10.1016/j.compchemeng.2013.12.001

Van der Aalst, W., T. Weijters, L. Maruster, (2004). Workflow mining: Discovering process models from event logs", Knowledge and Data Engineering IEEE Transactions on, vol. 16, no. 9, 1128-1142

Wilson, Z. T., & Sahinidis, N. V. (2017). The ALAMO approach to machine learning. Computers and Chemical Engineering, 106, 785-795. doi:10.1016/j.compchemeng.2017.02.010