Plant-wide utility disturbance management in the process industry
Highlights
► A generic method for utility disturbance management (UDM) is presented. ► The UDM method is presented by step-by-step instructions and is easy to use. ► The UDM method is applied on a real industrial case study. ► The UDM method gives ordering of utilities according to the revenue loss they cause. ► A matrix representation of a site is introduced, that simplifies use of the method.
Introduction
In the chemical process industry, companies must continuously improve their operational efficiency and profitability to remain competitive (Bakhrankova, 2010, Bansal et al., 2005). This means it is of great importance to minimize losses in revenues due to e.g. disturbances in operation. Plant-wide disturbances cause considerable revenue losses at industrial sites (Bauer et al., 2007, Thornhill et al., 2002). Some of these plant-wide disturbances are caused by utilities, such as steam or cooling water, which are used at most industrial large-scale production sites. At a disturbance in the supply of a utility, the production in all areas that use the utility is affected. Furthermore, areas are often connected by the product flow at the site, which makes the consequences of utility disturbances hard to predict. Two examples of utility disturbances at an industrial site are given in Fig. 1, Fig. 2. Both figures show a pressure drop in the middle-pressure steam net and the production in four areas that all require middle-pressure steam during the same time-period. The dashed line shows the ideal steam pressure of 14 bar.
In Fig. 1 it can be seen that the production of product 1 is not affected by the disturbance, whereas mainly the production of product 4 has to handle the variations. The reason that the production of product 4 is reduced already before the pressure drop is detected is that the steam boiler fails before the pressure drops, and thus the operators can start to react to the disturbance before it is visible in the measurement data of the pressure. In Fig. 2, the operators at the site handled the disturbance by starting to reduce the production of product 1 immediately as the pressure in the steam net drops. In this case, the production of products 3 and 4 is back to normal shortly after the disturbance, whereas the production of products 1 and 2 is reduced for a longer time. Consequently, the same type of disturbance is not handled using the same product flow control strategy each time. Furthermore, since utilities are shared by production areas, it is not evident which utility that causes the greatest economical loss. This motivates the need for a method that orders utilities according to the loss of revenue they cause, and suggests strategies for handling utility disturbances so that the loss of revenue is minimized. Economical loss is a useful measure, since it is a measure that is easily understandable by the managers at a site.
Section snippets
Related research areas
To the best of our knowledge, managing disturbances in the supply of utilities at the site level is an unexplored topic which does not quite fit into any current research area. To show this, and to distinguish the contributions of this paper, a few related research areas are discussed below.
Disturbances in utilities are often plant-wide disturbances. Detection and diagnosis of plant-wide disturbances is discussed by, among others, Bauer et al., 2007, Thornhill and Horch, 2007 and Thornhill et
Equipment hierarchy
The role based equipment hierarchy of an enterprise is defined in the standard ISA-95.00.01 (2009). For continuous production, there are five levels in the hierarchy; the enterprise, site, area, production unit, and unit levels. A plant is a concept that is not well defined with respect to the role based equipment hierarchy. In some contexts, a plant denotes a site, and in other contexts an area is intended. To avoid confusion regarding this term, the words site and area are used exclusively in
Utilities in the process industry
Utilities are support processes that are utilized in production, but are not part of the final product. Utilities could either be specific to an area or be shared between production areas. In Brennan (1998), some utilities are described. Here, devices for combustion of tail gas and the vacuum system utility have been added to expand the list of common utilities in the process industry. Examples of uses of these utilities are described below:
Steam
The steam net is commonly used for heating, for
A matrix representation of utilities at a site
To enable efficient computation of key performance indicators such as utility availability and area availability (described in 6 Utility availability, 7 Area availability), a matrix representation of a site and its utilities is introduced in this section. A simple site with four areas and five utilities is used as an example throughout the paper, to show how the matrix representation is utilized. A flowchart of the product flow at the example site is shown in Fig. 5.
The example site requires
Utility availability
A key performance indicator that could be used for determining how often a utility suffers a disturbance is the availability of the utility. The availability of a production unit is according to the standard ISO/WD-22400-2 (2011) the ratio between the actual production time and the planned allocation time, where the planned allocation time is the time in which the unit can be used (the operation time) minus the planned downtime. For utilities, the suggestion is to define availability as the
Area availability
A simple estimate of the availability of each area, with respect to utilities, is as the fraction of time all utilities needed at the area are available; i.e. the intersection of the operation of all concerned utilities. The measure of area availability should be interpreted as the fraction of time an area has a possibility of operating at maximum production rate, with respect to utilities. Area availability computed without considering the connection of areas at a site is denoted direct area
Production modeling approaches
Utilities are often shared between the production areas at a site. Three approaches for modeling the production at a site with respect to utilities are suggested in this paper, here listed according to level of detail of the obtained model. The model should describe how the production in all areas is affected at utility disturbances.
- 1.
On/off production without buffer tanks
Utilities and areas are considered to be either operating or not operating, i.e. ‘on’ or ‘off’. An area operates at maximum
Estimation of revenue loss due to utility disturbances
The on/off production modeling approach without buffer tanks, described in Section 8, enables estimation of revenue losses due to disturbances in utilities using the matrix representation introduced in Section 5. The on/off model does not adequately reflect how the actual production is managed for most sites, but is because of its simplicity useful for obtaining quick estimates of production losses due to disturbances. The estimates may be used as decision support for proactive disturbance
Utility disturbance management method
The general method for reducing the revenue loss due to disturbances in utilities is denoted the utility disturbance management (UDM) method. The strategies for reducing the loss may be both proactive and reactive disturbance management strategies, depending on which production modeling approach that is selected. Some suggestions of modeling approaches are given in Section 8. The accuracy of the strategies for reducing the revenue loss depends on the level of detail of the model of the site. A
Case study: application of the UDM method
The case study is performed at Perstorp, at their site in Stenungsund, Sweden. Perstorp is a worldwide enterprise that is a world leader in several sectors of the specialty chemicals market. Their products can be found in for example automotive, food, packaging, and electronics applications. Site Stenungsund is located on the Swedish west coast, approximately 50 km north of Gothenburg. The main products of the site are aldehydes, organic acids, alcohols, and plasticizers (Perstorp, 2012).
Simplicity vs. accuracy
The level of detail of the model of the site determines the level of detail of the strategies for minimizing the revenue loss that are given by the UDM method. When choosing modeling approach for the site, there is a trade-off between simplicity and accuracy of the results. This makes different modeling approaches suitable for different situations. For cases when modeling effort and time is limited, the on/off approach without buffer tanks might be a good choice, whereas if a detailed reactive
Conclusions
A general method for reducing the revenue loss due to disturbances in utilities was presented, denoted the utility disturbance management (UDM) method. In the method, both direct effects on areas due to disturbances in utilities, and indirect effects because of the connections of production areas at the site was investigated. The UDM method is easy to apply to any site by following the step-by-step instructions. A model of the site with respect to utilities is required to complete all steps of
Acknowledgement
The research is performed within the framework of the Process Industrial Centre at Lund University (PIC-LU) supported by the Swedish Foundation for Strategic Research (SSF).
References (29)
- et al.
Control of systems integrating logic, dynamics, and constraints
Automatica
(1999) - et al.
An industrial and academic perspective on plantwide control
Annual Reviews in Control
(2011) - et al.
A plant-wide industrial process control problem
Computers & Chemical Engineering
(1993) - et al.
Synthesis and operational planning of utility systems for multiperiod operation
Computers & Chemical Engineering
(1998) - et al.
Synthesis of utility systems by simulated annealing
Computers & Chemical Engineering
(1995) - et al.
Synthesis of utility systems with variable demands using simulated annealing
Computers & Chemical Engineering
(1997) - et al.
Management of changes and disturbances in manufacturing systems
Annual Reviews in Control
(1998) - et al.
A structural optimization approach in process synthesis-I: Utility systems
Computers & Chemical Engineering
(1983) Control structure design for complete chemical plants
Computers & Chemical Engineering
(2004)- et al.
Advances and new directions in plant-wide disturbance detection and diagnosis
Control Engineering Practice
(2007)
Enterprise-wide modeling & optimization – An overview of emerging research challenges and opportunities
Computers & Chemical Engineering
Decision support system for continuous production
Industrial Management & Data Systems
Toward a resilient supply chain with supply disturbances
Cited by (11)
Data preprocessing for multiblock modelling – A systematization with new methods
2020, Chemometrics and Intelligent Laboratory SystemsCitation Excerpt :By applying a multiblock framework on these processes it is possible to decrease the computational burden significantly, as well as to simplify the reconfiguration and update operations of the distributed model thus adding flexibility to the analysis. In this context, several methods have been hitherto developed for plant-wide process monitoring, such as hierarchical and multiblock statistical based approaches [5], decentralized fault diagnosis using multiblock methods [6], distributed modelling methods [7], decentralized fault detection and diagnosis via sparse models [8], and multi-level monitoring methods [9], among others [10,11]. More recently in order to cope with the challenges of the big data era, distributed and parallel data-driven methods have also been developed based on the MapReduce framework for plant-wide process modelling and monitoring [12] and to solve the quality prediction issue under very large scale of process data [13,14].
Review on data-driven modeling and monitoring for plant-wide industrial processes
2017, Chemometrics and Intelligent Laboratory SystemsCitation Excerpt :Troup and Georgakis [100] provided process systems tools in the pharmaceutical industry, which are related to process monitoring and fault diagnosis. Lindholm and Johnsson [101] proposed a plant-wide utility disturbance management method for the process industry. Detroja et al. [102] developed a method for plant-wide fault detection and diagnosis by using correspondence analysis.
KPIs as the interface between scheduling and control
2016, IFAC-PapersOnLineA periodic review lot sizing problem with random yields, disruptions and inventory capacity
2014, International Journal of Production EconomicsImproved two-level monitoring system for plant-wide processes
2014, Chemometrics and Intelligent Laboratory SystemsEmpirical models for utility disturbances in the process industry
2014, IFAC Proceedings Volumes (IFAC-PapersOnline)