(122f) Strategies for Systematically Optimizing the Operational Management of Lube-Oil Manufacturing and Packaging Facilities

In lube-oil industries, multiple lube-oil products are consecutively transported in batches through a single pipeline network. During a changeover operation from one product type to another, the lines must be cleaned to ensure the integrity of the new product batch. Lube-oil industries pose many restrictions and the use of any foreign aqueous/non-aqueous based solvents for cleaning/flushing the pipelines is considered a source of contamination to the highly sensitive finished lube-oil products. Hence, the industries use a finished product from the current batch to flush the pipelines from the residues of a previous batch. This leads to the generation of commingled (mixed) oil, which does not match the desired specifications of either of the two batches. The existing flushing operations are controlled by a flush timer. A certain flush time is chosen by an operator based on his previous experience with a particular product. At the end of the chosen flush time, the samples are collected and tested for various physical properties to ensure product quality. If the specifications do not meet the desired range, an additional flushing is conducted. The current procedure mostly relies on trial and error and results in long hold time/ downtime for the lab testing. Moreover, in most scenarios it results in excessive flushing, generating large volumes of commingled oil and leads to high economic losses to these industries. To this end, our work aims to address the existing drawbacks and explore alternative operational procedures aiming to minimize commingled oil volumes and improve the economic and environmental footprint of the flushing operations in lube-oil industries.

In recent years, optimization techniques have gained growing interest in the petroleum industry. MINLP models have been widely used for developing systematic scheduling plans and calculating commingled oil volumes for multiproduct pipelines (Chen et al. 2021; Abdellaoui et al. 2018). Major oil companies are developing models based on empirical correlations that are only applicable to their respective pipeline networks. There is no widely accepted correlation that can be used in all actual scenarios (He et al. 2018). Hence, based on these fundamentals and existing literature, we developed an idea of employing optimization techniques and formulating the flushing operation as an optimal control problem.

2. Mathematical Models

In this work, we investigate the lube-oil processing pipelines of our partnered industry, one of the world leaders in manufacturing lubricants. At this facility, the effectiveness of a flush is tested by collecting samples at the end of a designated flush time and analyzing them for their composition through various laboratory tests. These include viscosity, density, color, water content, and spectrophotometry. Viscosity is considered the preliminary and most crucial differentiable property for ensuring the product quality of individual batches. However, the conventional methods of testing viscosity require a long-running time leading to extended operational downtime. Thus, this work addresses the existing drawback and enables better in-line controllability of the flushing operation by developing models for predicting the viscosity of the lube-oil blends in real-time. The viscosity blending correlations recommended by API-TDB were combined with the component balance equations for the pipeline networks at our partnered lube-oil facility (Riazi 2005; Roegiers and Zhmud 2011). Our developed models were validated against experimental data that was collected through systematically designed and performed plant experiments. The agreement within ±5% error between the experimental and simulated data confirmed that our models hold value.

3. Solution Strategy

The flushing involves controlling a dynamic system, i.e., the system that evolves. Hence this work uses optimal control principles for optimizing the existing operations (Diwekar 2008). The developed problem was solved using two solution techniques viz. maximum principle and non-linear programming (Yenkie and Diwekar 2014b; 2014a). Our solution algorithm was developed using the programming language of MATLAB and the method involved the analysis of time-dependent first-order and second-order differential equations for generating an optimum flushing time. The results of the two solution methods were compared and the pros and cons are discussed.

4. Conclusion and Future Work

Our work highlights the advantages of using optimization techniques and API recommended property prediction correlations, for eliminating the drawbacks of the existing flushing operations in the lube-oil industries. We present methods that can enable industrial stakeholders to make informed decisions and thereby reduce the operational costs and downtime of these processes. This work will thus serve as an excellent starting point in improving the environmental and economic footprints of these operations.

5. References

Abdellaoui, Wassila, Asma Berrichi, Djamel Bennacer, Fouad Maliki, and Latéfa Ghomri. 2018. “Optimal Scheduling of Multiproduct Pipeline System Using MILP Continuous Approach.” In Computational Intelligence and Its Applications, edited by Abdelmalek Amine, Malek Mouhoub, Otmane Ait Mohamed, and Bachir Djebbar, 411–20. IFIP Advances in Information and Communication Technology. Cham: Springer International Publishing. https://doi.org/10.1007/978-3-319-89743-1_36.

Chen, Lei, Ziyun Yuan, JianXin Xu, Jingyang Gao, Yuhan Zhang, and Gang Liu. 2021. “A Novel Predictive Model of Mixed Oil Length of Products Pipeline Driven by Traditional Model and Data.” Journal of Petroleum Science and Engineering 205 (October): 108787. https://doi.org/10.1016/j.petrol.2021.108787.

Diwekar, Urmila. 2008. Introduction to Applied Optimization. Springer.

He, Guoxi, Mohan Lin, Baoying Wang, Yongtu Liang, and Qiyu Huang. 2018. “Experimental and Numerical Research on the Axial and Radial Concentration Distribution Feature of Miscible Fluid Interfacial Mixing Process in Products Pipeline for Industrial Applications.” International Journal of Heat and Mass Transfer 127 (December): 728–45. https://doi.org/10.1016/j.ijheatmasstransfer.2018.08.080.

Riazi, M. R. 2005. Characterization and Properties of Petroleum Fractions. W. Conshohocken, PA: ASTM International.

Roegiers, Michel, and Boris Zhmud. 2011. “Property Blending Relationships for Binary Mixtures of Mineral Oil and Elektrionised Vegetable Oil: Viscosity, Solvent Power, and Seal Compatibility Index.” Lubrication Science 23 (6): 263–78. https://doi.org/10.1002/ls.154.

Yenkie, Kirti M., and Urmila Diwekar. 2014a. “Optimal Control for Predicting Customized Drug Dosage for Superovulation Stage of in Vitro Fertilization.” Journal of Theoretical Biology 355 (August): 219–28. https://doi.org/10.1016/j.jtbi.2014.04.013.

Yenkie, Kirti M., and Urmila M. Diwekar. 2014b. “Comparison of Different Methods for Predicting Customized Drug Dosage in Superovulation Stage of In-Vitro Fertilization.” Computers & Chemical Engineering 71 (December): 708–14. https://doi.org/10.1016/j.compchemeng.2014.07.021.