None Notebook

This notebook contains material from ND-Pyomo-Cookbook; content is available on Github. The text is released under the CC-BY-NC-ND-4.0 license, and code is released under the MIT license.

< 4.4 Maintenance Planning | Contents | Tag Index | 4.6 Unit Commitment >

Open in Colab

Download

4.5 Scheduling Multipurpose Batch Processes using State-Task Networks

Keywords: cbc usage, state-task networks, gdp, disjunctive programming, batch processes

The State-Task Network (STN) is an approach to modeling multipurpose batch process for the purpose of short term scheduling. It was first developed by Kondili, et al., in 1993, and subsequently developed and extended by others.

4.5.1 Imports

4.5.2 References

Floudas, C. A., & Lin, X. (2005). Mixed integer linear programming in process scheduling: Modeling, algorithms, and applications. Annals of Operations Research, 139(1), 131-162.

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 & Chemical Engineering, 62, 161-193.

Kondili, E., Pantelides, C. C., & Sargent, R. W. H. (1993). A general algorithm for short-term scheduling of batch operations—I. MILP formulation. Computers & Chemical Engineering, 17(2), 211-227.

Méndez, C. A., Cerdá, J., Grossmann, I. E., Harjunkoski, I., & Fahl, M. (2006). State-of-the-art review of optimization methods for short-term scheduling of batch processes. Computers & Chemical Engineering, 30(6), 913-946.

Shah, N., Pantelides, C. C., & Sargent, R. W. H. (1993). A general algorithm for short-term scheduling of batch operations—II. Computational issues. Computers & Chemical Engineering, 17(2), 229-244.

Wassick, J. M., & Ferrio, J. (2011). Extending the resource task network for industrial applications. Computers & chemical engineering, 35(10), 2124-2140.

4.5.3 Example (Kondili, et al., 1993)

A state-task network is a graphical representation of the activities in a multiproduct batch process. The representation includes the minimum details needed for short term scheduling of batch operations.

A well-studied example due to Kondili (1993) is shown below. Other examples are available in the references cited above.

Kondili_1993.png

Each circular node in the diagram designates material in a particular state. The materials are generally held in suitable vessels with a known capacity. The relevant information for each state is the initial inventory, storage capacity, and the unit price of the material in each state. The price of materials in intermediate states may be assigned penalities in order to minimize the amount of work in progress.

The rectangular nodes denote process tasks. When scheduled for execution, each task is assigned an appropriate piece of equipment, and assigned a batch of material according to the incoming arcs. Each incoming arc begins at a state where the associated label indicates the mass fraction of the batch coming from that particular state. Outgoing arcs indicate the disposition of the batch to product states. The outgoing are labels indicate the fraction of the batch assigned to each product state, and the time necessary to produce that product.

Not shown in the diagram is the process equipment used to execute the tasks. A separate list of process units is available, each characterized by a capacity and list of tasks which can be performed in that unit.

4.5.3.1 Exercise

Read this reciped for Hollandaise Sauce: http://www.foodnetwork.com/recipes/tyler-florence/hollandaise-sauce-recipe-1910043. Assume the available equipment consists of one sauce pan and a double-boiler on a stove. Draw a state-task network outlining the basic steps in the recipe.

4.5.4 Encoding the STN data

The basic data structure specifies the states, tasks, and units comprising a state-task network. The intention is for all relevant problem data to be contained in a single JSON-like structure.

4.5.4.1 Setting a time grid

The following computations can be done on any time grid, including real-valued time points. TIME is a list of time points commencing at 0.

4.5.5 Creating a Pyomo model

The following Pyomo model closely follows the development in Kondili, et al. (1993). In particular, the first step in the model is to process the STN data to create sets as given in Kondili.

One important difference from Kondili is the adoption of a more natural time scale that starts at $t = 0$ and extends to $t = H$ (rather than from 1 to H+1).

A second difference is the introduction of an additional decision variable denoted by $Q_{j,t}$ indicating the amount of material in unit $j$ at time $t$. A material balance then reads

\begin{align*} Q_{jt} & = Q_{j(t-1)} + \sum_{i\in I_j}B_{ijt} - \sum_{i\in I_j}\sum_{\substack{s \in \bar{S}_i\\s\ni t-P_{is} \geq 0}}\bar{\rho}_{is}B_{ij(t-P_{is})} \qquad \forall j,t \end{align*}

Following Kondili's notation, $I_j$ is the set of tasks that can be performed in unit $j$, and $\bar{S}_i$ is the set of states fed by task $j$. We assume the units are empty at the beginning and end of production period, i.e.,

\begin{align*} Q_{j(-1)} & = 0 \qquad \forall j \\ Q_{j,H} & = 0 \qquad \forall j \end{align*}

The unit allocation constraints are written the full backward aggregation method described by Shah (1993). The allocation constraint reads

\begin{align*} \sum_{i \in I_j} \sum_{t'=t}^{t-p_i+1} W_{ijt'} & \leq 1 \qquad \forall j,t \end{align*}

Each processing unit $j$ is tagged with a minimum and maximum capacity, $B_{ij}^{min}$ and $B_{ij}^{max}$, respectively, denoting the minimum and maximum batch sizes for each task $i$. A minimum capacity may be needed to cover heat exchange coils in a reactor or mixing blades in a blender, for example. The capacity may depend on the nature of the task being performed. These constraints are written

\begin{align*} B_{ij}^{min}W_{ijt} & \leq B_{ijt} \leq B_{ij}^{max}W_{ijt} \qquad \forall j, \forall i\in I_j, \forall t \end{align*}

4.5.5.1 Characterization of tasks

4.5.5.2 Characterization of states

4.5.5.3 Characterization of units

4.5.5.4 Pyomo model

4.5.6 Analysis

4.5.6.1 Profitability

4.5.6.2 Unit assignment

4.5.6.3 State inventories

4.5.6.4 Unit batch inventories

4.5.6.5 Gannt chart

4.5.7 Trace of events and states

< 4.4 Maintenance Planning | Contents | Tag Index | 4.6 Unit Commitment >

Open in Colab

Download