(26a) Unlocking Value through Better Quantification of Uncertainty in Asset Utilization Loss Using a Data-Driven Stochastic Approach

Motivated by the extensive labor and capital costs and potential loss of revenue at stake, optimization tools have been successfully developed to support turnaround planning for maximizing the value of manufacturing asset operations. For instance, Amaran et al. [1-2] consider medium and long-term turnaround planning in integrated sites using mathematical programming approaches to balance different trade-offs.

Some of the major drivers for a turnaround include asset performance renewal, regulatory inspections, and capital projects. Of these drivers, asset renewal presents a more flexible window of opportunity to unlock operational value through optimal turnaround interval planning. However, evaluating this opportunity requires characterization of Asset Utilization (AU) loss over time to subsequently incorporate into the existing planning framework.

Asset-level modeling based on individual equipment reliability analysis can be overly complex and expensive to maintain from a data collection and modeling standpoint. Resorting to higher level models which are less complex, may result, at times, in multiple individual effects getting aggregated. This can lead to a set of categories with observations lacking a discernible trend with respect to the actual root-cause factors. In such cases, Hidden Markov Models (HMM) [3] can be useful in modeling the uncertainty around the observations by assuming that the system transitions through non-observable or hidden states.

In this work, we consider a data-driven stochastic approach to model AU loss for an asset based on conceptualized production loss historian data. The conceptual historian tracks asset utilization loss over a set of equipment or process categories over time. We build HMM models for a subset of the categories without a discernable trend using the historical data as observations. Further, this approach combined with Monte Carlo simulations allows us not to be trapped by overly conservative predictions in AU losses. We illustrate the application of this approach within a turnaround planning model.

References:

[1] Amaran, S.; Sahinidis, N.V.; Sharda, B.; Morrison, M.; Bury, S. J.; Miller, S.; Wassick, J. M. Long-term turnaround planning for integrated chemical sites, Computers & Chemical Engineering, 2015, 72, pp. 145-158.

[2] Amaran, S.; Zhang, T; Sahinidis, N. V.; Sharda, B.; Bury, S. J. Medium-term maintenance turnaround planning under uncertainty for integrated chemical sites, Computers & Chemical Engineering, 2016, 84, pp. 422-433.

[3] Chen, Z.; Li, Y.; Xia, T.; Pan, E Hidden Markov model with auto-correlated observations for remaining useful life prediction and optimal maintenance policy, Reliability Engineering & System Safety, 2019, 184, pp. 123-136.