# A General State-Space Formulation for Online Scheduling

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Background

#### 2.1. Chemical Production Scheduling

#### 2.1.1. General Problem Statement

- (i)
- Production facility data (e.g., unit capacities and connectivity),
- (ii)
- Production recipes (e.g., processing times and mixing rules),
- (iii)
- Production costs (e.g., material holding costs),
- (iv)
- Material availability (e.g., raw materials delivery amounts and dates),
- (v)
- Resource availability (e.g., maintenance schedule and utility levels), and
- (vi)
- Production targets or orders with due-times;

- (i)
- Number and the associated processing-sizes of the needed tasks,
- (ii)
- Assignment of these tasks to processing units, and
- (iii)
- Timing (or just the sequence) of these tasks on the assigned units;

#### 2.1.2. Problem Representation

#### 2.1.3. Model Classification

#### 2.1.4. Solution Methods

#### 2.2. Scheduling MILP Model

#### 2.3. Standard form of State-Space Models

#### 2.4. Scheduling State-Space Model

## 3. Modeling Generalizations

#### 3.1. New Basic Formulation

#### 3.2. Fractional Delays and Unit Downtimes

^{th}delay, the difference, $\varphi =\lceil ({\sum}_{r=1}^{\psi}{\pi}_{delay}^{r})/\delta \rceil -\lceil ({\sum}_{r=1}^{\psi -1}{\pi}_{delay}^{r})/\delta \rceil $, dictates how much additional, integer $\varphi $, delay is applied in the update steps. Figure 7A shows a numerical example for fractional delays.

#### 3.3. Variable Batch-Sizes

#### 3.4. Robust Scheduling: Batch-Sizes

#### 3.5. Robust Scheduling: Processing Times

#### 3.6. Feedback on Yield Estimates

#### 3.7. Task Termination

#### 3.8. Post-Production Storage in Unit

#### 3.9. Unit Capacity Degradation and Maintenance

## 4. Integrated Model

## 5. Case Study

## 6. Conclusions

## Acknowledgments

## Author Contributions

## Conflicts of Interest

## Abbreviations

MILP | mixed integer linear program |

MPC | model predictive control |

RTN | resource task network |

STN | state task network |

HOLD | hold (storage) task |

MT | maintenance (cleaning) task |

## Nomenclature

Indices/sets | |

$i\in \mathbf{I}$ | tasks |

$j\in \mathbf{J}$ | units (equipment) |

$k\in \mathbf{K}$ | materials |

$t\in \mathbf{T}$ | time-points/periods |

$\mathbf{I}\supseteq {\mathbf{I}}_{j}$ | tasks that can be carried out in unit j |

$\mathbf{I}\supseteq {\mathbf{I}}_{k}^{+}$ | tasks producing material k |

$\mathbf{I}\supseteq {\mathbf{I}}_{k}^{-}$ | tasks consuming material k |

$\mathbf{I}\supseteq {\mathbf{I}}_{\mathrm{HOLD}}$ | hold (storage) tasks |

$\mathbf{I}\supseteq {\mathbf{I}}_{\mathrm{MT}}$ | maintenance tasks |

$\mathbf{J}\supseteq {\mathbf{J}}_{i}$ | units suitable for carrying out task i |

$\mathbf{J}\supseteq {\mathbf{J}}_{\mathrm{MT}}$ | units which can degrade, and consequently, need a corresponding maintenance task |

$\mathbf{K}\supseteq {\mathbf{K}}^{F}$ | feed (raw) materials |

$\mathbf{K}\supseteq {\mathbf{K}}^{I}$ | intermediates |

$\mathbf{K}\supseteq {\mathbf{K}}^{P}$ | final products |

Parameters | |

${\alpha}_{ij}^{F}$ | fixed cost of running task i on unit j |

${\alpha}_{ij}^{P}$ | proportional cost of running task i on unit j |

${\alpha}_{ij}^{T}$ | cost of terminating task i on unit j |

${\alpha}_{ij}^{M}$ | proportional cost of maintenance task i on unit j |

${\beta}_{ij}$ | fixed batch-size of task i executed on unit j |

${\beta}_{ij}^{min}/{\beta}_{ij}^{max}$ | min/max capacity on batch-size of task i executed on unit j |

${\widehat{\beta}}_{ijkt}^{P}/{\widehat{\beta}}_{ijkt}^{C}$ | material unloading/loading loss during production/consumption of material k |

${\gamma}_{k}$ | selling price of material k |

${\gamma}_{k}^{INV}$ | inventory cost of material k |

${\gamma}_{k}^{BO}$ | backlog cost of material k |

$\delta $ | discretization of time-grid; length of time-periods |

${\zeta}_{kt}$ | incoming shipment of material k at time t |

${\dot{Z}}_{ij}^{n}$ | disturbance parameter denoting unit breakdown |

${}^{B}{\dot{Z}}_{ij}^{n}$ | batch-size of task suspended due to unit breakdown |

${}^{L}{\dot{Z}}_{ij}^{n}$ | yield-loss size of task suspended due to unit breakdown |

${}^{M}{\dot{Z}}_{ij}^{n}$ | maintenance-size of maintenance task suspended due to unit breakdown |

${\theta}_{ijt}^{\tau}$ | dummy parameter as defined in Equation (46) |

${\theta}_{ikt}^{\overline{\rho}}$ | dummy parameter as defined in Equation (37) |

${\theta}_{ijt}^{T}$ | dummy parameter as defined in Equation (63) |

${\dot{\lambda}}_{ij}^{n}$ | yield loss in task i running on unit j, with run status n |

${\widehat{\Lambda}}_{jt}$ | binary parameter, when 1, denotes unit j unavailable during time $[t,t+1)$ |

${\dot{\mu}}_{ij}$ | when negative, represents extent of sudden partial loss in unit capacity |

${\xi}_{kt}$ | demand for material k at time t |

${\widehat{\xi}}_{kt}$ | demand disturbance for material k at time t |

${\pi}_{break}$ | actual (fractional) time at which a unit breaks down |

${\pi}_{down}$ | fractional downtime in a unit |

${\pi}_{delay}$ | fractional delay in a task |

${\pi}_{delay}^{r}$ | r^{th} delay in a task |

${\rho}_{ik}$ | mass-conversion coefficient (material consumption) |

${\overline{\rho}}_{ik}$ | mass-conversion coefficient (material production) |

${\overline{\rho}}_{ik}^{r}$ | conservative mass-conversion coefficient for production (${\overline{\rho}}_{ik}^{r}<{\overline{\rho}}_{ik}$) |

${\rho}_{i{i}^{\prime}j}^{C}$ | deterioration in unit capacity to perform task i, due to performing task ${i}^{\prime}$ on that unit j. |

$\sigma $ | online iteration number |

${\tau}_{ij}$ | processing time of task i on unit j |

${\tau}_{ij}^{r}$ | conservative processing time of task i (${\tau}_{ij}^{r}<{\tau}_{ij}$) on unit j |

${\tau}_{j}^{T}$ | task independent unit j downtime after terminating a task |

${\tau}_{ij}^{T}$ | task dependent unit j downtime after terminating task i |

${\dot{Y}}_{ij}^{n}$,${\widehat{Y}}_{ijt}^{n}$ | single-/multi-period disturbance parameters denoting delay |

${}^{B}{\dot{Y}}_{ij}^{n}$,${}^{B}{\widehat{Y}}_{ijt}^{n}$ | single-/multi-period disturbance parameters denoting batch-size of a delayed task |

${}^{L}{\dot{Y}}_{ij}^{n}$,${}^{L}{\widehat{Y}}_{ijt}^{n}$ | single-/multi-period disturbance parameters denoting yield-loss size of a delayed task |

${}^{M}{\dot{Y}}_{ij}^{n}$,${}^{M}{\widehat{Y}}_{ijt}^{n}$ | single-/multi-period disturbance parameters denoting maintenance-size of delayed maintenance task |

$\varphi $ | duration of delay or breakdown, in multiples of $\delta $ |

$\psi $ | recurrence count of delay for a task |

Variables | |

${B}_{ijt}$ | batch-size of task i on unit j |

${\overline{B}}_{ijt}^{n}$ | lifted batch-size |

$B{O}_{kt}$ | backlog level of material k during period $(t-1,t]$ |

${C}_{ijt}$ | capacity of unit j to perform task i during period $(t-1,t]$ |

${\overline{L}}_{ijt}^{n}$ | lifted yield-loss variables |

${M}_{ijt}$ | maintenance-size of the maintenance task |

${\overline{M}}_{ijt}^{n}$ | lifted maintenance-size |

${S}_{kt}$ | inventory level of material k during period $(t-1,t]$ |

${T}_{ij}^{n}$ | binary variable, when 1, denotes termination of task i, with run-status n, on unit j |

${V}_{kt}$ | outgoing shipment to meet demand for material k at time t |

${W}_{ijt}$ | binary variable, when 1, denotes task i starts on unit j at time-point t |

${\overline{W}}_{ijt}^{n}$ | lifted task-start variables |

${X}_{ijt}$ | when 1, captures the information about delays in a task with progress status $n=0$ |

${}^{B}{X}_{ijt}$ | the batch-size of delayed task with progress status $n=0$ |

${}^{L}{X}_{ijt}$ | yield-loss of delayed task with progress status $n=0$ |

${}^{M}{X}_{ijt}$ | maintenance-size of delayed maintenance task with progress status $n=0$ |

## References

- Harjunkoski, I.; Maravelias, C.T.; Bongers, P.; Castro, P.M.; Engell, S.; Grossmann, I.E.; Hooker, J.; Méndez, C.A.; Sand, G.; Wassick, J.M. Scope for industrial applications of production scheduling models and solution methods. Comput. Chem. Eng.
**2014**, 62, 161–193. [Google Scholar] [CrossRef] - Kelly, J.D.; Mann, J. Crude oil blend scheduling optimization: An application with multimillion dollar benefits. Hydrocarb. Process.
**2003**, 82, 47–54. [Google Scholar] - Méndez, C.A.; Cerdá, J.; Grossmann, I.E.; Harjunkoski, I.; Fahl, M. State-of-the-art review of optimization methods for short-term scheduling of batch processes. Comput. Chem. Eng.
**2006**, 30, 913–946. [Google Scholar] [CrossRef] - Maravelias, C.T. General framework and modeling approach classification for chemical production scheduling. AIChE J.
**2012**, 58, 1812–1828. [Google Scholar] [CrossRef] - Velez, S.; Maravelias, C.T. Reformulations and branching methods for mixed-integer programming chemical production scheduling models. Ind. Eng. Chem. Res.
**2013**, 52, 3832–3841. [Google Scholar] [CrossRef] - Wassick, J.M.; Ferrio, J. Extending the resource task network for industrial applications. Comput. Chem. Eng.
**2011**, 35, 2124–2140. [Google Scholar] [CrossRef] - Nie, Y.; Biegler, L.T.; Villa, C.M.; Wassick, J.M. Discrete Time Formulation for the Integration of Scheduling and Dynamic Optimization. Ind. Eng. Chem. Res.
**2015**, 54, 4303–4315. [Google Scholar] [CrossRef] - Gupta, D.; Maravelias, C.T.; Wassick, J.M. From rescheduling to online scheduling. Chem. Eng. Res. Des.
**2016**, 116, 83–97. [Google Scholar] [CrossRef] - Cott, B.J.; Macchietto, S. Minimizing the effects of batch process variability using online schedule modification. Comput. Chem. Eng.
**1989**, 13, 105–113. [Google Scholar] [CrossRef] - Kanakamedala, K.B.; Reklaitis, G.V.; Venkatasubramanian, V. Reactive schedule modification in multipurpose batch chemical plants. Ind. Eng. Chem. Res.
**1994**, 33, 77–90. [Google Scholar] [CrossRef] - Huercio, A.; Espuña, A.; Puigjaner, L. Incorporating on-line scheduling strategies in integrated batch productioncontrol. Comput. Chem. Eng.
**1995**, 19, 609–614. [Google Scholar] [CrossRef] - Kim, M.; Lee, I.B. Rule-based reactive rescheduling system for multi-purpose batch processes. Comput. Chem. Eng.
**1997**, 21, S1197–S1202. [Google Scholar] [CrossRef] - Ko, D.; Na, S.; Moon, I.; Oh, M.; Dong-Gu Samsung, T.S. Development of a Rescheduling System for the Optimal Operation of Pipeless Plants. Comput. Chem. Eng.
**1999**, 23, S523–S526. [Google Scholar] [CrossRef] - Huang, W.; Chung, P.W.H. A constraint approach for rescheduling batch processing plants including pipeless plants. Comput. Aided Chem. Eng.
**2003**, 14, 161–166. [Google Scholar] - Henning, G.P.; Cerdá, J. Knowledge-based predictive and reactive scheduling in industrial environments. Comput. Chem. Eng.
**2000**, 24, 2315–2338. [Google Scholar] [CrossRef] - Palombarini, J.; Martínez, E. SmartGantt—An interactive system for generating and updating rescheduling knowledge using relational abstractions. Comput. Chem. Eng.
**2012**, 47, 202–216. [Google Scholar] [CrossRef] - Elkamel, A.; Mohindra, A. A rolling horizon heuristic for reactive scheduling of batch process operations. Eng. Optim.
**1999**, 31, 763–792. [Google Scholar] [CrossRef] - Vin, J.; Ierapetritou, M.G. A new approach for efficient rescheduling of multiproduct batch plants. Ind. Eng. Chem. Res.
**2000**, 39, 4228–4238. [Google Scholar] [CrossRef] - Méndez, C.A.; Cerdá, J. Dynamic scheduling in multiproduct batch plants. Comput. Chem. Eng.
**2003**, 27, 1247–1259. [Google Scholar] [CrossRef] - Ferrer-Nadal, S.; Méndez, C.A.; Graells, M.; Puigjaner, L. Optimal reactive scheduling of manufacturing plants with flexible batch recipes. Ind. Eng. Chem. Res.
**2007**, 46, 6273–6283. [Google Scholar] [CrossRef] - Janak, S.L.; Floudas, C.A.; Kallrath, J.; Vormbrock, N. Production scheduling of a large-scale industrial batch plant. II. Reactive scheduling. Ind. Eng. Chem. Res.
**2006**, 45, 8253–8269. [Google Scholar] [CrossRef] - Novas, J.M.; Henning, G.P. Reactive scheduling framework based on domain knowledge and constraint programming. Comput. Chem. Eng.
**2010**, 34, 2129–2148. [Google Scholar] [CrossRef] - Honkomp, S.; Mockus, L.; Reklaitis, G.V. A framework for schedule evaluation with processing uncertainty. Comput. Chem. Eng.
**1999**, 23, 595–609. [Google Scholar] [CrossRef] - Subramanian, K.; Maravelias, C.T.; Rawlings, J.B. A state-space model for chemical production scheduling. Comput. Chem. Eng.
**2012**, 47, 97–110. [Google Scholar] [CrossRef] - Gupta, D.; Maravelias, C.T. On deterministic online scheduling: Major considerations, paradoxes and remedies. Comput. Chem. Eng.
**2016**, 94, 312–330. [Google Scholar] [CrossRef] - Velez, S.; Maravelias, C.T. Advances in Mixed-Integer Programming Methods for Chemical Production Scheduling. Annu. Rev. Chem. Biomol. Eng.
**2014**, 5, 97–121. [Google Scholar] [CrossRef] [PubMed] - Subramanian, K.; Rawlings, J.B.; Maravelias, C.T.; Flores-Cerrillo, J.; Megan, L. Integration of control theory and scheduling methods for supply chain management. Comput. Chem. Eng.
**2013**, 51, 4–20. [Google Scholar] [CrossRef] - Subramanian, K.; Rawlings, J.B.; Maravelias, C.T. Economic model predictive control for inventory management in supply chains. Comput. Chem. Eng.
**2014**, 64, 71–80. [Google Scholar] [CrossRef] - Kondili, E.; Pantelides, C.C.; Sargent, R.W.H. A general algorithm for short-term scheduling of batch operations-I. MILP formulation. Comput. Chem. Eng.
**1993**, 17, 211–227. [Google Scholar] [CrossRef] - Pantelides, C.C. Unified frameworks for optimal process planning and scheduling. In Proceedings of the Second Conference on Foundations of Computer Aided Operations; Cache: New York, NY, USA, 1994; pp. 253–274. [Google Scholar]
- Sundaramoorthy, A.; Maravelias, C.T. Computational Study of Network-Based Mixed-Integer Programming Approaches for Chemical Production Scheduling. Ind. Eng. Chem. Res.
**2011**, 50, 5023–5040. [Google Scholar] [CrossRef] - Pinto, J.M.; Grossmann, I.E. A Continuous Time Mixed Integer Linear Programming Model for Short Term Scheduling of Multistage Batch Plants. Ind. Eng. Chem. Res.
**1995**, 34, 3037–3051. [Google Scholar] [CrossRef] - Blomer, F.; Gunther, H.O. LP-based heuristics for scheduling chemical batch processes. Ind. Eng. Chem. Res.
**2000**, 38, 1029–1051. [Google Scholar] [CrossRef] - Velez, S.; Maravelias, C.T. Mixed-integer programming model and tightening methods for scheduling in general chemical production environments. Ind. Eng. Chem. Res.
**2013**, 52, 3407–3423. [Google Scholar] [CrossRef] - Merchan, A.F.; Maravelias, C.T. Reformulations of Mixed-Integer Programming Continuous-Time Models for Chemical Production Scheduling. Ind. Eng. Chem. Res.
**2014**, 53, 10155–10165. [Google Scholar] [CrossRef] - Burkard, R.; Hatzl, J. Review, extensions and computational comparison of MILP formulations for scheduling of batch processes. Comput. Chem. Eng.
**2005**, 29, 1752–1769. [Google Scholar] [CrossRef] - Janak, S.L.; Floudas, C.A. Improving unit-specific event based continuous-time approaches for batch processes: Integrality gap and task splitting. Comput. Chem. Eng.
**2008**, 32, 913–955. [Google Scholar] [CrossRef] - Lee, H.; Maravelias, C.T. Discrete-time mixed-integer programming models for short-term scheduling in multipurpose environments. Comput. Chem. Eng.
**2017**, 107, 171–183. [Google Scholar] [CrossRef] - Sahinidis, N.; Grossmann, I. Reformulation of multiperiod MILP models for planning and scheduling of chemical processes. Comput. Chem. Eng.
**1991**, 15, 255–272. [Google Scholar] [CrossRef] - Yee, K.; Shah, N. Improving the efficiency of discrete time scheduling formulation. Comput. Chem. Eng.
**1998**, 22, S403–S410. [Google Scholar] [CrossRef] - Lee, H.; Maravelias, C.T. Mixed-integer programming models for simultaneous batching and scheduling in multipurpose batch plants. Comput. Chem. Eng.
**2017**, 106, 621–644. [Google Scholar] [CrossRef] - Papageorgiou, L.G.; Pantelides, C.C. Optimal campaign planning/scheduling of multipurpose batch/semicontinuous plants. 2. A mathematical decomposition approach. Ind. Eng. Chem. Res.
**1996**, 35, 510–529. [Google Scholar] [CrossRef] - Bassett, M.H.; Pekny, J.F.; Reklaitis, G.V. Decomposition techniques for the solution of large-scale scheduling problems. AIChE J.
**1996**, 42, 3373–3387. [Google Scholar] [CrossRef] - Kelly, J.D.; Zyngier, D. Hierarchical decomposition heuristic for scheduling: Coordinated reasoning for decentralized and distributed decision-making problems. Comput. Chem. Eng.
**2008**, 32, 2684–2705. [Google Scholar] [CrossRef] - Wu, D.; Ierapetritou, M.G. Decomposition approaches for the efficient solution of short-term scheduling problems. Comput. Chem. Eng.
**2003**, 27, 1261–1276. [Google Scholar] [CrossRef] - Calfa, B.A.; Agarwal, A.; Grossmann, I.E.; Wassick, J.M. Hybrid Bilevel-Lagrangean Decomposition Scheme for the Integration of Planning and Scheduling of a Network of Batch Plants. Ind. Eng. Chem. Res.
**2013**, 52, 2152–2167. [Google Scholar] [CrossRef] - Castro, P.M.; Harjunkoski, I.; Grossmann, I.E. Greedy algorithm for scheduling batch plants with sequence-dependent changeovers. AIChE J.
**2011**, 57, 373–387. [Google Scholar] [CrossRef] - Roslöf, J.; Harjunkoski, I.; Björkqvist, J.; Karlsson, S.; Westerlund, T. An MILP-based reordering algorithm for complex industrial scheduling and rescheduling. Comput. Chem. Eng.
**2001**, 25, 821–828. [Google Scholar] [CrossRef] - Kopanos, G.M.; Méndez, C.A.; Puigjaner, L. MIP-based decomposition strategies for large-scale scheduling problems in multiproduct multistage batch plants: A benchmark scheduling problem of the pharmaceutical industry. Eur. J. Oper. Res.
**2010**, 207, 644–655. [Google Scholar] [CrossRef] - Relvas, S.; Barbosa-Póvoa, A.P.F.; Matos, H.A. Heuristic batch sequencing on a multiproduct oil distribution system. Comput. Chem. Eng.
**2009**, 33, 712–730. [Google Scholar] [CrossRef] - Jain, V.; Grossmann, I.E. Algorithms for Hybrid MILP/CP Models for a Class of Optimization Problems. INFORMS J. Comput.
**2001**, 13, 258–276. [Google Scholar] [CrossRef] - Harjunkoski, I.; Grossmann, I.E. Decomposition techniques for multistage scheduling problems using mixed-integer and constraint programming methods. Comput. Chem. Eng.
**2002**, 26, 1533–1552. [Google Scholar] [CrossRef] - Maravelias, C.T.; Grossmann, I.E. A hybrid MILP/CP decomposition approach for the continuous time scheduling of multipurpose batch plants. Comput. Chem. Eng.
**2004**, 28, 1921–1949. [Google Scholar] [CrossRef] - Roe, B.; Papageorgiou, L.G.; Shah, N. A hybrid MILP/CLP algorithm for multipurpose batch process scheduling. Comput. Chem. Eng.
**2005**, 29, 1277–1291. [Google Scholar] [CrossRef] - Maravelias, C.T. A decomposition framework for the scheduling of single- and multi-stage processes. Comput. Chem. Eng.
**2006**, 30, 407–420. [Google Scholar] [CrossRef] - Subrahmanyam, S.; Kudva, G.K.; Bassett, M.H.; Pekny, J.F. Application of distributed computing to batch plant design and scheduling. AIChE J.
**1996**, 42, 1648–1661. [Google Scholar] [CrossRef] - Ferris, M.C.; Maravelias, C.T.; Sundaramoorthy, A. Simultaneous Batching and Scheduling Using Dynamic Decomposition on a Grid. INFORMS J. Comput.
**2009**, 21, 398–410. [Google Scholar] [CrossRef] - Velez, S.; Maravelias, C.T. A branch-and-bound algorithm for the solution of chemical production scheduling MIP models using parallel computing. Comput. Chem. Eng.
**2013**, 55, 28–39. [Google Scholar] [CrossRef] - Shah, N.; Pantelides, C.C.; Sargent, R.W.H. A general algorithm for short-term scheduling of batch operations-II. Computational issues. Comput. Chem. Eng.
**1993**, 17, 229–244. [Google Scholar] [CrossRef] - Stephanopoulos, G. Chemical Process Control: An Introduction to Theory and Practice; Prentice-Hall: Englewood Cliffs, NJ, USA, 1984; p. 696. [Google Scholar]
- Ogunnaike, B.A.; Ray, W.H. Process Dynamics, Modeling, and Control; Oxford University Press: New York, NY, USA, 1994; p. 1260. [Google Scholar]
- Bequette, B.W. Process Control: Modeling, Design, and Simulation; Prentice Hall PTR: Upper Saddle River, NJ, USA, 2003; p. 769. [Google Scholar]
- Rawlings, J.B.; Mayne, D. Model Predictive Control: Theory and Design; Nob Hill Pub: Madison, WI, USA, 2009; p. 669. [Google Scholar]
- Seborg, D.E.; Edgar, T.F.; Duncan, M.A.; Doyle, F.J., III. Process Dynamics and Control; Wiley: Hoboken, NJ, USA, 2016; p. 502. [Google Scholar]
- Amrit, R.; Rawlings, J.B.; Biegler, L.T. Optimizing process economics online using model predictive control. Comput. Chem. Eng.
**2013**, 58, 334–343. [Google Scholar] [CrossRef] - Ellis, M.; Durand, H.; Christofides, P.D. A tutorial review of economic model predictive control methods. J. Process Control
**2014**, 24, 1156–1178. [Google Scholar] [CrossRef] - Rawlings, J.B.; Risbeck, M.J. Model predictive control with discrete actuators: Theory and application. Automatica
**2017**, 78, 258–265. [Google Scholar] [CrossRef] - Baldea, M.; Harjunkoski, I. Integrated production scheduling and process control: A systematic review. Comput. Chem. Eng.
**2014**, 71, 377–390. [Google Scholar] [CrossRef] - Li, Z.; Ierapetritou, M.G. Process scheduling under uncertainty: Review and challenges. Comput. Chem. Eng.
**2008**, 32, 715–727. [Google Scholar] [CrossRef] - Janak, S.L.; Lin, X.; Floudas, C.A. A new robust optimization approach for scheduling under uncertainty. II. Uncertainty with known probability distribution. Comput. Chem. Eng.
**2007**, 31, 171–195. [Google Scholar] [CrossRef] - Sand, G.; Engell, S. Modeling and solving real-time scheduling problems by stochastic integer programming. Comput. Chem. Eng.
**2004**, 28, 1087–1103. [Google Scholar] [CrossRef] - Sabuncuoglu, I.; Karabuk, S. Rescheduling frequency in an fms with uncertain processing times and unreliable machines. J. Manuf. Syst.
**1999**, 18, 268–283. [Google Scholar] [CrossRef] - Chaari, T.; Chaabane, S.; Aissani, N.; Trentesaux, D. Scheduling under uncertainty: Survey and research directions. In Proceedings of the 2014 International Conference on Advanced Logistics and Transport, Hammamet, Tunisia, 1–3 May 2014; pp. 229–234. [Google Scholar]
- Martagan, T.; Krishnamurthy, A. Control and Optimization of Bioprocesses Using Markov Decision Process. In Proceedings of the 2012 Industrial and Systems Engineering Research Conference, Orlando, FL, USA, 19–23 May 2012; pp. 1–8. [Google Scholar]
- Martagan, T.; Krishnamurthy, A.; Maravelias, C.T. Optimal condition-based harvesting policies for biomanufacturing operations with failure risks. IIE Trans.
**2016**, 48, 440–461. [Google Scholar] [CrossRef] - Dedopoulos, I.T.; Shah, N. Optimal Short-Term Scheduling of Maintenance and Production for Multipurpose Plants. Ind. Eng. Chem. Res.
**1995**, 34, 192–201. [Google Scholar] [CrossRef] - Sanmartí, E.; Espuña, A.; Puigjaner, L. Batch production and preventive maintenance scheduling under equipment failure uncertainty. Comput. Chem. Eng.
**1997**, 21, 1157–1168. [Google Scholar] [CrossRef] - Vassiliadis, C.; Pistikopoulos, E. Maintenance scheduling and process optimization under uncertainty. Comput. Chem. Eng.
**2001**, 25, 217–236. [Google Scholar] [CrossRef] - Kopanos, G.M.; Xenos, D.P.; Cicciotti, M.; Pistikopoulos, E.N.; Thornhill, N.F. Optimization of a network of compressors in parallel: Operational and maintenance planning – The air separation plant case. Appl. Energy
**2015**, 146, 453–470. [Google Scholar] [CrossRef] - Xenos, D.P.; Kopanos, G.M.; Cicciotti, M.; Thornhill, N.F. Operational optimization of networks of compressors considering condition-based maintenance. Comput. Chem. Eng.
**2016**, 84, 117–131. [Google Scholar] [CrossRef] - Biondi, M.; Sand, G.; Harjunkoski, I. Optimization of multipurpose process plant operations: A multi-time-scale maintenance and production scheduling approach. Comput. Chem. Eng.
**2017**, 99, 325–339. [Google Scholar] [CrossRef] - Zhang, Y.H.P.; Sun, J.; Ma, Y. Biomanufacturing: History and perspective. J. Ind. Microbiol. Biotechnol.
**2017**, 44, 773–784. [Google Scholar] [CrossRef] [PubMed] - Clomburg, J.M.; Crumbley, A.M.; Gonzalez, R. Industrial biomanufacturing: The future of chemical production. Science
**2017**, 355. [Google Scholar] [CrossRef] [PubMed]

**Figure 1.**STN representation of a process network. This network, for a bio-manufacturing process, is described in detail in Section 5. It has three steps (tasks) for production of a pharmaceutical ingredient (material M3). T1 is the task of preparing the cell cultures in lab-size beakers. T2 denotes the task of having the cell culture grow and produce the pharmaceutical active ingredient, on a feed of sugars, in a bio-reactor. T3 is a purification task, which is carried out in chromatograph columns. Finally, T4 is a dummy task to model storage of material M2 inside unit U2. ${\rho}_{ik}/{\overline{\rho}}_{ik}$ values (not shown in the figure) are either $-1$ or $+1$ depending on whether the task is consuming the material or producing it.

**Figure 2.**Task-states are shown for two online iterations – numbered $\sigma -1$ and $\sigma $. Each iteration uses its own local time-grid which is reset to start from 0. Here, ${\tau}_{ij}$ for the tasks is assumed to be 3. Lifting of past inputs enables knowing the complete status of the plant by looking at the states (variables) only at that moment in time. In the absence of delays or breakdowns, the lifting equations effectively represent the relation: ${\overline{W}}_{ijt}^{n}={W}_{ij(t-n)}\phantom{\rule{4pt}{0ex}}\phantom{\rule{4pt}{0ex}}\forall j,i\in {\mathbf{I}}_{j},n$. Arrows show which variables are equal due to the lifting equations (Equations (11) and (12), with no delays or breakdowns). Variables in green or red have a value of 1, rest have value 0. Information is carried over from one iteration to the next through the update step (Equations (17)–(19)).

**Figure 3.**When a 2 h delay is observed, through the lifting equations, ${\overline{W}}_{ijt}^{n}$ evolve over the green trajectory, leading to the task correctly finishing 2 h late in iteration $\sigma $. Here, $\tau $ for the task is 3. Arrows show which variables are enforced as equal by the lifting equations (Equations (11) and (12), with delays present). Variables and parameters in green have a value of 1, rest have value 0. (

**A**) The task now finishes at $t=4$, instead of at $t=2$. (

**B**) The task now finishes at $t=2$, instead of at $t=0$. Through Equation (13), the unit is kept busy at $t=0$ and 1, by the inclusion of the terms ${\widehat{Y}}_{(t=0)}^{\tau =3}$ and ${\widehat{Y}}_{(t=1)}^{\tau =3}$, and hence a new task is prevented from starting at these times. In addition, these terms in Equation (14), prevent the task’s produce from erroneously contributing to inventory (${S}_{k,(t=1)}$ and ${S}_{k,(t=2)}$).

**Figure 4.**When a 2 h delay is observed, through the lifting equations, ${\overline{W}}_{ijt}^{n}$ evolve over the green trajectory, leading to the task correctly finishing 2 h late in iteration $\sigma $. Here, $\tau $ for the task is 3. Arrows show which variables are enforced as equal by the lifting equations (Equations (22)–(24), with delays present). Variables and parameters in green have a value of 1, rest have value 0. (

**A**) The task now finishes at $t=4$, instead of at $t=2$. Due to the update step, parameters $\dot{Y}$, ${\widehat{Y}}_{0}^{0}$ and variable ${X}_{0}$ are 1, hence, in iteration $\sigma $, due to the optimization model, ${\overline{W}}_{0}^{0}$, ${X}_{1}$, and ${\overline{W}}_{1}^{0}$ are also 1. (

**B**) The task now finishes at $t=2$, instead of at $t=0$. The update step ensures that the true progress, n, of the task is reflected in the task-states at $t=0$, i.e., $n=2$.

**Figure 5.**When a breakdown is observed, further evolution of the task-states for the task (the green trajectory), running on the unit that broke, stops. Here, $\tau $ for the task is 3 and the unit downtime (blue) is 2 h. Arrows show which variables are enforced as equal by the lifting equations (Equations (11) and (12), with breakdown present). Variables and parameters in green, blue, or red have a value of 1, rest have value 0. The green task is suspended at $t=0$. A new task (red) can only start at $t=2$, once the unit downtime is over. (

**A**) Through Equation (13), the unit is kept busy at $t=0$ and 1, due to the terms ${\overline{W}}_{t=0}^{1}$ and ${\widehat{Z}}_{\mathrm{IT},1}^{1}$, respectively. (

**B**) Through Equation (13), the unit is kept busy at $t=0$ and 1, due to the terms ${\overline{Z}}_{t=0}^{\tau =3}$ and ${\widehat{Z}}_{\mathrm{IT},1}^{\tau =1}$, respectively. Additionally, the term ${\widehat{Z}}_{(t=0)}^{\tau =3}$ in Equation (14), prevents the green task-state (${\overline{W}}_{t=0}^{{\tau}_{ij}=3}$) from erroneously contributing to the inventory.

**Figure 6.**When a breakdown is observed, further evolution of the task-states for the task (the green trajectory), running on the unit that broke, stops. Here, $\tau $ for the task is 3 and the unit downtime (blue) is 2 h. Arrows show which variables are enforced as equal by the lifting equations (Equations (22)–(24)). Variables and parameters in green, blue, or red have a value of 1, rest have value 0. The green task is suspended at $t=0$. A new task (red) can only start at $t=2$, once the unit downtime is over. Through Equation (25), the unit is kept busy at $t=0$ and 1, by the inclusion of the terms ${\widehat{\Lambda}}_{t=0}$ and ${\widehat{\Lambda}}_{t=1}$. (

**A**) The parameter ${\dot{Z}}^{0}$, through Equation (20), prevents the task-state from evolving from ${}_{(\sigma -1)}{\overline{W}}_{0}^{0}$ to ${}_{\sigma}{\overline{W}}_{0}^{1}$. (

**B**) The parameter ${\dot{Z}}^{2}$, through Equation (20), prevents the task-state from evolving from ${}_{(\sigma -1)}{\overline{W}}_{0}^{2}$ to ${}_{\sigma}{\overline{W}}_{0}^{3}$.

**Figure 7.**The task has a nominal processing time $\tau =3$. Time-grid spacing is $\delta =1$ h. Here, for ease of discussion, the time-grid is global, i.e., it is not reset for each iteration. (

**A**) A delay (shown as oblique green pattern) of 0.66 h in time-period (0, 1) is observed. Since, $\lceil 0.66\rceil =1$, a delay of 1 h is applied at $t=1$. This would ensure that the task finish is aligned with $t=4$, even if the task actually ends at $t=3.66$. The horizontal green pattern represents the fictitious extra task runtime to align with the discrete time-grid. Next, another delay of 0.2 h is observed in time-period (1, 2). Since, $\lceil 0.66+0.2\rceil -\lceil 0.66\rceil =0$, no additional delay is applied in the update steps. This makes sense, because, the task finishes at $t=3.86$ in reality. Since, the previous delay was applied as 1 h, the task is now still thought to finish at $t=4$, in alignment with the time-grid. Finally, when another delay of 0.66 h is observed in time-period (2, 3). Since, $\lceil 0.66+0.2+0.66\rceil -\lceil 0.66+0.2\rceil =1$, a 1 h delay is applied in the update step. This correctly ensures that the task is now thought to end at $t=5$, which is the round up of the true end time of $t=4.52$ h. (

**B**) A break-down is observed at $t=0.2$. Thus, ${\dot{Z}}^{0}=1$ for the update step between the iterations starting at $t=0$ and $t=1$. If the downtime (blue) is 0.66 h, then the unit actually becomes available at $t=0.86$. Thus, ${\widehat{\Lambda}}_{(t=1)}$ is not activated. This is indeed the case from our mathematical procedure as well, since, $t\in (0.2,0.86]$ does not include any integer time-point. If ${\pi}_{down}=1.5$ h, then $t\in (0.2,1.7]$ does span $t=1$, and consequently ${\widehat{\Lambda}}_{(t=1)}=1$. Finally, if ${\pi}_{down}=2.25$ h, then $t\in (0.2,2.45]$ spans $t=1$ and $t=2$, which results in ${\widehat{\Lambda}}_{(t=1)}=1$ and ${\widehat{\Lambda}}_{(t=2)}=1$.

**Figure 8.**In iteration $\sigma -1$, the green task ($\tau =2$) with a “large batch” is finishing at ${t}^{\prime}=1$, but, due to the use of ${\overline{\rho}}_{k}^{r}$, is anticipated to produce only half of what the demand is. Thus, another identical green task is scheduled to start at ${t}^{\prime}=4$, to satisfy the demand. With the use of ${\overline{\rho}}_{k}^{r}$, if the demand could be still be satisfied with a single “large” batch, a second batch wouldn’t be scheduled. In iteration $\sigma $, the earlier green task yields a large amount of material, in line with its large batch-size due to the true value of ${\overline{\rho}}_{ik}$ used at $t=0$. (

**A**) Since, here, there was no material handling loss, thus the second green task need not run now, as the demand was satisfied by the first batch itself. (

**B**) Although the green task results in a large yield at $t=0$, a small material handling loss (${\beta}_{k(t=0)}^{P}<0$) at $t=0$, requires the second green task to still be scheduled in order to meet the demand, but now with a smaller batch-size. If there are no further material handling losses, there would be a small excess inventory of material produced by the second green task, since its batch-size was decided assuming the yield to be lower (${\overline{\rho}}_{k}^{r}$) but would in reality by higher (${\overline{\rho}}_{k}$).

**Figure 9.**For the green task ($\tau =2$), a delay buffer of 1 h is chosen, i.e., ${\tau}_{ij}^{r}={\tau}_{ij}+1$. (

**A**) In iteration $\sigma -1$, the model predicts that the green task will finish at ${t}^{\prime}=2$. In the next iteration $\sigma $, when there is no delay in the green task, having run for 2 h, it correctly finishes at $t=0$. No ad-hoc model changes are required in between the iterations to make this possible. Consequently, to take advantage of this nominal finish time of the green task, the red task, now starts 1 h earlier than what it was scheduled for in the previous iteration. (

**B**) The green task, with conservative processing time, is scheduled to finish at ${t}^{\u2033}=2$ in iteration $\sigma -2$. In iteration $\sigma -1$, a delay of 1 h is observed. The processing time delay buffer is maintained, and now the task is anticipated to finish at ${t}^{\prime}=2$. In the next iteration ($\sigma $), the task correctly finishes at $t=0$, accounting for the 1 h delay. The red task can now start on time at $t=0$, or equivalently ${t}^{\u2033}=2$. Thus, having a delay buffer was useful, since it predicted a realistic start time for the red task in iteration $\sigma -2$ itself.

**Figure 10.**Selected online scheduling iterations for the case study. The color of the tasks corresponds to their color in Figure 1. Executed schedule, with respect to each iteration, is shown in lighter (fainted) colors. The clock-time is the global time. Each iteration has its own local discrete time-grid.

${\gamma}_{k}^{INV}=0.15{\gamma}_{k}$ | ${\gamma}_{k}^{BO}=1.5{\gamma}_{k}$ | ${\alpha}_{ij}^{F}=1$ | ${\alpha}_{ij}^{P}=0$ | ${\alpha}_{ij}^{T}=2$ | ${\alpha}_{ij}^{M}=0$ |

${\tau}_{\mathrm{U}3}^{T}=2$ | ${\tau}_{\mathrm{T}1,\mathrm{U}1}^{r}=3$ | ${\tau}_{\mathrm{MT},\mathrm{U}3}=2$ | ${\overline{\rho}}_{\mathrm{M}2,\mathrm{T}2}^{r}=0.9$ | ${\rho}_{\mathrm{T}3,\mathrm{T}3,\mathrm{U}3}^{C}=-0.2$ | $\delta =1$ |

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Gupta, D.; Maravelias, C.T.
A General State-Space Formulation for Online Scheduling. *Processes* **2017**, *5*, 69.
https://doi.org/10.3390/pr5040069

**AMA Style**

Gupta D, Maravelias CT.
A General State-Space Formulation for Online Scheduling. *Processes*. 2017; 5(4):69.
https://doi.org/10.3390/pr5040069

**Chicago/Turabian Style**

Gupta, Dhruv, and Christos T. Maravelias.
2017. "A General State-Space Formulation for Online Scheduling" *Processes* 5, no. 4: 69.
https://doi.org/10.3390/pr5040069