Computers and Chemical Engineering 34 (2010) 668–682

Contents lists available at ScienceDirect

Computers and Chemical Engineering

journal homepage: www.elsevier.com/locate/compchemeng

Towards an ontological infrastructure for chemical batch process management

E. Muñoz, A. Espuña, L. Puigjaner ∗
Chemical Engineering Department – CEPIMA, Universitat Politecnica de Catalunya, Av. Diagonal 647, Pab. G-2, 08028 Barcelona, Spain

a r t i c l e i n f o a b s t r a c t

Article history: A crucial step for batch process improvement and optimization is to develop information structures
Received 19 August 2009 that streamline data gathering and, above all, are capable of integrating transactional data into a system
Received in revised form using the analytical tools that are developed. Current trends in electronics, computer science, artificial
16 December 2009
intelligence and control system technology are providing technical capability that greatly facilitates the
Accepted 19 December 2009
development of multilevel decision-making support. In this paper, we present the batch process ontol-
Available online 4 January 2010
ogy (BaPrOn), wherein different concepts regarding batch processes are categorized and the relationships
between them are examined and structured in accordance with ANSI/ISA-88 standards, which provide
Keywords:
Ontology
a solid and transparent framework for integrating batch-related information. This paper also focuses
Knowledge representation on systematic integration of different actors within the control process. The proposed approach bases
Knowledge sharing the conceptualization through the ANSI/ISA-88 representation, providing the advantage of establish-
Batch process ing a more general conceptualization of the batch process domain. The capabilities of the envisaged
Decision-support systems ontological framework were assessed in a test bed PROCEL pilot plant: scheduling-monitoring and
control-rescheduling was closed, information quality was accessed by knowledge description, and an
optimum decision-making task was performed. The ontological structure can be extended in the future
to incorporate other hierarchical levels and their respective modeling knowledge.
© 2009 Elsevier Ltd. All rights reserved.

1. Introduction sic DSS design is comprised of components for sophisticated

database management capabilities with access to internal and
The chemical process industry (CPI) must face an ever-changing external data, information, and knowledge; modeling functions
environment to meet current market needs. The management of accessed by a model management system; simple user interface
chemical batch plants involves collecting and processing huge designs that enable interactive queries, reporting, and graphing
amounts of data, which are subsequently exploited. These data can functions; and optimization by mathematic algorithms and or intu-
be viewed as a valuable source of information for decision-making, ition/knowledge. Much research and practical design effort has
regardless of the use of its analysis (such as in preformulation and been conducted in each of these domains.
new process development, supply chain management, scheduling, Fig. 1 describes what probably came to be the most commonly
process control, fault analysis, etc.). used model of the decision-making process in a DSS environment.
For 40 years, companies have been developing management Typically, the phases overlap and blend together, with frequent
information systems to help end users to exploit data and models, looping back to earlier stages as more is learned about the prob-
with the final objective of use in discussions and decision-making. lem, as solutions fail, and so forth. A first step is the recognition
Current global competition has made some of these decisions of the problem or an opportunity. Once the problem has been rec-
(related to aspects of manufacturing such as economic efficiency, ognized, it is defined as a term that facilitates the creation of the
product quality, flexibility and reliability) essential for the viabil- model. Some authors consider that the emphasis is on the next
ity of the enterprise. Decision-support systems (DSS) are computer two steps: model development and alternatives analysis. Then, the
technology solutions that can be used to support complex decision- choice is made and implemented. As a final step and if necessary,
making and problem solving (Shim et al., 2002). a new recognition is performed. Obviously, no decision process is
DSS are defined as computer-aided systems at the company this clear-cut in an ill-structured situation (Shim et al., 2002).
management level that combine data and sophisticated analytic There has also been a huge effort in the DSS field to build a
models to support decision-making (Simon & Murray, 2007). Clas- group support system (GSS) or collaboration support systems to
enhance the communication-related activities of team members
who are engaged in computer-supported cooperative work. The
∗ Corresponding author. communication and coordination activities of team members are
E-mail address: [email protected] (L. Puigjaner). facilitated by technologies that can be characterized along the three

0098-1354/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compchemeng.2009.12.009
 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682 669

in a conceptual manner. The use of multiple models to represent

detailed and abstract knowledge of chemical processes has been
taken into account recently. In particular, this knowledge repre-
sentation enables us to identify process sections together with their
function, objectives and relations within the process. This enables
the automatic generation of alternative views of the process, orga-
nized in a hierarchy of different levels of abstraction. The approach
is similar to reverse engineering.
Other requirements that a knowledge-representation system
should meet are natural representation of the physical real-
ity/model (items of equipment/devices), automatic generation of
abstraction levels to identify which sections of the process can be
potentially improved, and the implementation of computer refer-
ence model sets (Rodríguez-Martínez et al., 2004). The computer
reference model is a large-scale, hierarchically arranged com-
puter system that integrates plant management, plant production
scheduling, inventory management, individual process optimiza-
tion, and unit process control for all of the plant’s operating units
as a whole. The reference model defines requirements that are com-
Fig. 1. Decision-making process. mon to all implementations, but it is independent of the specified
requirements of any particular implementation that are applicable
to existing processes.
continua of time, space, and level of group support (Alavi & Keen, The way to address these problems is to reduce or eliminate
1989; Shim et al., 2002). Teams can communicate synchronously conceptual and terminological confusion and come to a shared
or asynchronously; they may be located together or remotely; and understanding. Such an understanding can function as a unifying
the technology can provide task support primarily for the individ- framework for the different viewpoints of an ontology that has been
ual team member or for the group’s activities. These technologies adopted to develop an integrated framework, through the defini-
are utilized to overcome the space and time constraints that burden tion and semantic description of data and information. This is the
face-to-face meetings, to increase the range and depth of informa- basis for modeling the different forms of knowledge that are to be
tion access, and to improve group task performance effectiveness, organized or the contextualized information that can be used to
especially by overcoming “process losses”. produce new meanings and generate new information.
From the previous discussion, it is clear that the need for infras- Ontologies constitute a means of specifying the structure of
tructures that continuously and coherently support fast and reliable a domain of knowledge in a generic way that can be read by
decision-making activities related to the production process is now a computer (formal specification) and presented in a human-
of paramount importance (Venkatasubramanian et al., 2006). This readable form (informal specification). Moreover, ontologies are
need is more evident when we consider recent activity in the fields emerging as a key solution to knowledge sharing in a coopera-
of data warehousing, online analytical processing (OLAP), data min- tive business environment (Missikoff & Taglino, 2002). Since they
ing and Web-based DSS, followed by the treatment of collaborative can express knowledge (and the relationships in the knowledge)
support systems and optimization-based decision support (Shim with clear semantics, they are expected to play an important role
et al., 2002). It is quite common for batch process activities to in forthcoming information-management solutions to improve the
have large databases. Hence, an enormous amount of information is information search process (Gruber, 1993; Obrst, 2003). Recently,
created, stored and shared and it may be hard to find the right infor- several ontologies have been developed in the field of chemical
mation when it is required. Furthermore, because of the possible processes. They include OntoCAPE, which is a formal, heavyweight
use of different computer languages and differences in concep- ontology for the domain of (chemical) process engineering. In the
tualization, the interoperability between information in different chemical domain, the design, construction, and operation of chem-
systems is one of the most critical aspects in the daily operation of ical plants are considered the major engineering activities (Bayer
many organizations. & Marquardt, 2004; Morbach, Yang, & Marquardt, 2007; Morbach,
Indeed, shrinking profits have made it essential to exploit large Wiesner, & Marquardt, 2009). Another well-known ontology is
databases (as companies need to manufacture many non-cyclical ontological informatics infrastructure for pharmaceutical product
products with complex recipes, etc.) using non-generic/blind meth- development and manufacturing (POPE), which is an informatics
ods. One key aspect is information extraction, which should result framework in the pharmaceutical product development domain,
in the extraction of information quality. Information quality can to support decision-making in the entire process, including drug
be defined as precise information in terms of time, content, and formulation design, process simulation and process safety analysis
clarity (Eppler, 2006). A common problem is that this data extrac- for the active pharmaceutical ingredient process, as well as drug
tion process may be performed using blind methods in many cases. product development (Venkatasubramanian et al., 2008; Zhao et al.,
However, the performance of such methods can be drastically 2006). BioPortal is an open repository of biomedical ontologies. In
improved by combining them with knowledge or expertise of the this portal, researchers in biomedical informatics can submit their
process. ontologies and access them using the BioPortal user interface or
Information is data that is processed to be useful. Knowledge is Web services (Board et al., 2009). This allows searches for terms and
the application of data and information through the development of biomedical resources, and also provides comments within ontolo-
a system which models, in a structured way, the experience gained gies or the option of adding ontology mappings.
in some domain. Knowledge exists as soon as human interaction In general, the aforementioned ontologies offer alternatives
is or has been made available in any step of the product/process for sharing knowledge in the process and engineering domains.
development (Gebus & Leiviskä, 2009). In recent years, there has However, many of the ontologies that have been developed to
been an effort to create knowledge with a minimum human inter- date are focused on existing product development processes. In
face, either in a straight and formal way (e.g. expert systems) or other words, these approaches create an ontology that is suitable
670 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682

for on-going processes. However, when some standard struc- • In a batch plant production environment, the occurrence of
tures are taken into account, the opportunity to improve data, unpredictable events is usually unavoidable. Such events may be
information and knowledge is enormous. This work presents the related to external market factors or to the intrinsic plant opera-
development of a generic ontology following ANSI/ISA-88 stan- tion, and include equipment breakdowns and variable operation
dards (International Society for Measurement and Control, 1995, times. Despite the uncertainty in the environment, the scheduler
2006, 2007; Shirasuna, 2007), which allows infrastructure to be has to make some decisions in order to start production and to
created that should be general enough to be applied to any batch face uncertainty when an abnormal event occurs.
system. These standards were developed on the basis of the Pur- • In addition, the integration of a control and monitoring sys-
due reference model, which is a detailed collection of the functional tem into process management helps to provide the process
requirements of the generic information management and control state information opportunely at different levels in the decision-
to run a batch manufacturing process (Williams, 1989). Addition- making hierarchical structure, thus reducing the risk of incidents
ally, the proposed ontology may be used as a straightforward and improving the efficiency of the reactive scheduling by updat-
guideline for standardizing batch process management and con- ing the schedules in the most effective way, which improves the
trol. Moreover, it offers a robust structure for sharing knowledge, process yield.
which may help to make more appropriate decisions that improve
business performance. Unexpected events or disruptions can change the system status
and affect its performance. Deviations from the original schedule
and information about equipment breakdowns that is provided
2. Analysis requirements
by the control and monitoring system will eventually trigger
rescheduling. However, the schedule that is generated will be
The problem of batch process management has been studied
assessed according to the new plant situation. Thus, if some modifi-
in many works (Mendez, Cerda, Grossmann, Harjunkoski, & Fahl,
cations are made, the newly created schedule will be translated into
2006; Rippin, 1983; Shah, 1998; Subrahmanyam, Bassett, Pekny, &
some control recipes for the actual process. Consequently, ontology
Reklaitis, 1995). These works have made a rich and vast descrip-
can also be used to ensure the robustness of the running plan in the
tion of such problems and have pointed out open issues where
system.
research efforts may provide significant improvements, in which
The rescheduling system allows different dispatching rules,
the in-depth description aids understanding.
optimizers and objective functions to be selected, according to the
This section aims to describe the particular elements involved
process knowledge. Alternative rescheduling techniques (recalcu-
in current ontology development, which affect its subsequent per-
late a new robust schedule, update operation times, reassignment,
formance assessment.
etc.) are evaluated and a system should select the most suitable
ones, according to the objective function that is adopted. Optimiza-
2.1. Plant requirements tion algorithms may be included, depending on the interest of the
decision maker and the required reaction time.
Effective production is very important in today’s global com-
petitive environment. In the case of the batch process industry, 2.2. Recipe requirements
multi-product and multipurpose plants, as well as continuous
or semi-continuous processes, manufacture a variety of products This integration approach follows the ANSI/ISA-88 batch con-
through a sequence of operations that share available resources, trol standard, which differentiates between four types of recipes:
intermediate products, and raw materials. The efficient use of such general, site, master and control. The general and site recipes are
resources can be analyzed at different levels: general recipes that are outside the scope of the control system. At
control level, the information recipes are the master and control
• If we focus on the planning area and control level, a process recipes.
system involves multiple and interrelated activities that are per- Master recipes are derived from site recipes and are targeted
formed at single or multiple sites, with different durations and at the process cell. A master recipe is a required recipe level;
amounts of information. Generally, information flows from the without it, control recipes cannot be created and batches cannot
marketing department to the manufacturing department, which be produced. Master recipes take into consideration the equip-
determines the production schedule that is needed to meet the ment requirements within a given process cell. They include
sales strategies. In its most general form, the scheduling problem the following information categories: header, formula, equipment
requires information that is related to the configuration of the requirements and procedure. Control recipes are batches that are
plant (the available equipment units and resources), the prod- created from master recipes. They contain the product-specific
uct recipes (the set of processing tasks and resources required to process information that is required to manufacture a partic-
manufacture a given product), precedence relationships between ular batch of product. They also provide the detail needed to
materials and final product requirements (demands and related initiate and monitor equipment procedural entities in a process
due dates) (International Society for Measurement and Control, cell.
1995).
At this level, a closed-loop framework is proposed for the 2.3. Control activity requirements
decision-making task of batch chemical plants. This framework
integrates both scheduling and control and is based on ANSI/ISA- The control activity model from International Society for
88 standards (BaPrOn). Integration enables future events to be Measurement and Control (1995) (Fig. 2) provides an overall
taken into account at scheduling level, in order to minimize the perspective of batch control and shows the main relationships
occurrence of critical situations during the execution of the pro- between the various control activities. The control activities define
cess. It also provides any required information from the process how equipment in the batch manufacturing plant will be con-
measurements that are made by the control and fault diagnosis trolled.
system. On the other hand, the opportunity for reactive schedul- Using this approach, recipes can be modified without chang-
ing allows the process to respond under unexpected schedule ing the code of the PLCS and DCS that run the basic regulatory
deviations or abnormal events. control. Moreover, recipes can run on different sets of equipment.
 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682 671

Ontologies were created to help in knowledge reuse and

sharing: reuse means building new applications by assembling
components that have already been built, while sharing occurs
when different applications use the same resources. Reuse and
sharing have the following advantages: they are cost, time and
resources effective (Fensel, 2003). In this work the batch process
ontology (BaPrOn) is presented. In BaPrOn a conceptualization
through the ANSI/ISA-88 representation provides an advantage of
establishing a more general conceptualization in the batch pro-
cess domain. Such a generalization is behind years of joint work
by recognized batch manufacturing experts who met to define a
perceptive view of batch plants organization and its correspond-
ing hierarchy of control functions. As a consequence, following the
ANSI/ISA-88, virtually all activities concerning batch processes can
be properly represented.
BaPrOn is a procedural oriented ontology that supports the
management of different concepts (physical models, procedures,
functions and processes) in accordance with ANSI/ISA-88 batch
process standards, categorizing them and examining the relation-
ships between them. This will allow the association between the
Fig. 2. Control activity model. elements mentioned before, and the further identification of any
information resource if it would be required.

The use of the ANSI/ISA-88 standard significantly reduces costs and 3.1. Language
implementation time through:
Different ontology languages provide diverse facilities. Any lan-
• Effective utilization and optimization of plant equipment, which guage used to codify ontology-underpinned knowledge should be
maximizes total plant production capabilities. expressive, declarative, portable, domain independent and seman-
• Reductions in total validation costs and production down-time tically well defined. The language used in an ontology is essential
via separate validation of recipe procedures and equipment. for its future implementation and sharing. We adopted web ontol-
ogy language (OWL), as it has good characteristics for ontologies
(Bechhofer et al., 2004). OWL has been designed for use by appli-
2.4. Final problem statement
cations that need to process the content of information, instead of
just presenting the information to humans. OWL facilitates greater
In summary, the lack of integration among the control sys-
machine interpretability of Web content than that supported by
tem’s hierarchical levels does not allow complete optimization of
extensible markup language (XML), resource description frame-
scheduling. Information is needed from different hierarchical lev-
work (RDF), and resource description framework schema (RDF-S),
els when an important change is required. However, the desired
as it provides additional vocabulary along with formal semantics.
change cannot be made unless the system be enough robust. In
The ontology formally describes the meaning of the terminology
general, at planning level, the schedule is optimized with the static
used in documents. If machines are expected to perform useful rea-
information contained in the recipes, and a semi-improvement is
soning tasks on these documents, the language must go beyond the
made. In addition, the need to integrate the different modeling
basic semantics of RDF schema.
approaches in a hierarchical decision-support system means that
OWL has been designed to meet this need for a web ontology
consistent terminology and concepts must be used to improve the
language, and is part of the growing stack of W3C recommen-
communication and collaboration tasks over the entire system.
dations that are related to the semantic web.

3. Proposed approach: ontology-based infrastructure • XML provides a surface syntax for structured documents, but
imposes no semantic constraints on the meaning of these doc-
An ontology defines the basic terms and relations that comprise uments (XML-Core-Working-Group, 2009).
the vocabulary of a topic area as well as the rules for combining • XML Schema is a language that restricts the structure of XML doc-
terms and relations to define extensions to the vocabulary (Neches uments and extends XML with data types (McQueen & Thompson,
et al., 1991). It is understood as an explicit specification of a con- 2000).
ceptualization (Gruber, 1993). However, since Gruber’s definition, • RDF is a data model for objects (“resources”) and relations
various improvements have been made to the definition of “an between them. It provides simple semantics for data models,
ontology”. For example, an ontology has been described as a hier- which can be represented in XML syntax (Klyne & Carroll, 2002).
archically structured set of terms for describing a domain that can • Schema is a vocabulary for describing properties and classes of
be used as a skeletal foundation for a knowledge base (Swartout, RDF resources. It includes semantics for generalization hierar-
Neches, & Patil, 1993). chies of these properties and classes (Brickley & Guha, 2002).
Ontologies provide the shared and common domain struc-
tures that are required for the semantic integration of information OWL adds more vocabulary for describing properties and
sources. Though it is still difficult to find consensus among ontol- classes, including the relations between classes (e.g. disjointed-
ogy developers and users, there is some agreement about protocols, ness), cardinality (e.g. “exactly one”), equality, richer typing of
languages and frameworks. Ontologies are hierarchical domain properties, characteristics of properties (e.g. symmetry), and enu-
structures that provide a domain theory, have a syntactically and merated classes.
semantically rich language, and a shared and consensual terminol- The semantics in the ontology build on XML’s ability to define
ogy (Klein, Fensel, Kiryakov, & Ognyanov, 2002). customized tagging schemes and RDF’s flexible approach to rep-
672 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682

Fig. 3. XML schema for a recipe.

resenting data. This unifying aspect makes it easier to establish, from scratch. Methontology identifies the following activities in
through collaboration and consensus, the utilitarian vocabularies the development of an ontology: specification, knowledge acqui-
(between ontologies) needed for far-flung cooperative and integra- sition, conceptualization, integration, implementation, evaluation,
tive applications using the Word Wide Web and internal servers. and documentation. The life cycle of the ontology is based on the
The uses of these languages are helpful for the first task of the ontol- refinement of a prototype and ends with a maintenance state. The
ogy, which is to become a standard tool for vocabulary, format, and most distinctive aspect of Methontology is the focus on this main-
definitions. Restrictions and reasoning make communication pos- tenance stage.
sible between the different system elements. Fig. 3 shows a master In contrast, On-To-Knowledge methodology includes the
recipe standard imported from the World Batch Forum (Brandl & identification of goals that should be achieved by knowledge man-
Emerson, 2003). From left to right, a XML schema which may be agement tools and is based on an analysis of usage scenarios. The
more comprehensible to humans is shown, as well as XML lines steps proposed by On-To-Knowledge are: (i) kick-off, in which
that both machines and devices are capable of reading. some competency questions are identified, potentially reusable
ontologies are studied and a first draft of the ontology is built; (ii)
3.2. Methodology refinement, in which a mature and application-oriented ontology
is produced; (iii) evaluation, in which requirements and compe-
Currently, numerous ontologies are being developed and used tency questions are checked and the ontology is tested in the
in various research areas. Each development project usually follows application environment; and finally (iv) ontology maintenance.
its own set of principles, in order to design criteria and phases in the On-To-Knowledge stresses that the ontology modeling process
ontology development process. The absence of standard guidelines should start with a definition of the abstraction level, which is
and methods hinders the following: the development of shared and strongly dependent on the usage of the ontology.
concentrated ontologies within and between projects; the exten- All the aforementioned methodologies have been inserted
sion of a given ontology by others; and its reuse in other ontologies into the Plan, Do, Check and Act Cycle (PDCA), (http://www.
and final applications. It is widely recognized that constructing hci.com.au/hcisite3/toolkit/pdcacycl.htm), which results in an
ontologies, or domain models, is an important step in the devel- ordered sequence of steps, that are easy to understand and track
opment of knowledge-based systems (KBSs). However, there is a (Fig. 4).
lack of consensus on a uniform approach to designing and main-
taining these ontologies. Various methodologies exist to guide the 3.3. Ontology layering and architecture
theoretical approach that is taken, and numerous ontology build-
ing tools are available. The problem is that these procedures have The architecture of BaPrOn could be considered as nominal
not coalesced into popular development styles or protocols, and ontology and high ontology, taking into account the factors that
the tools have not yet matured to the degree one would expect in influence the complexity, such as concepts, taxonomy, patterns,
other software instances. constraints and instances, as show in Table 1.
The methodology used in this paper is based in two ontology The bases for the construction of this particular ontology include
development methodologies “Methontology” (López, Gómez- concepts and relations taken from the ANS/ISA-88 standard. The
Pérez, Sierra, & Sierra, 1999) and “On-To-Knowledge” (Sure & coherence between the concepts in the conceptualization ensures
Studer, 2002). that there is compatibility between the control elements at differ-
Methontology provides support for the entire life cycle of ontol- ent control levels. This provides more opportunities for applications
ogy development. It enables experts and ontology makers who are in which information is updated at the same rate as data are
unfamiliar with implementation environments to build ontologies received.
 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682 673

The layers represent the following:

Concepts: A concept defines a basic and abstract idea that is com-

monly used in an ontology domain. It is represented as a word or
phrase.
Relations: A relation describes the way in which two or more con-
cepts are interrelated. It is usually described by a verb or verb
phrase (basic properties).
Basic fact types: A basic fact type is a kind of primitive sentence or
fact. It is composed of concepts and relations. If the basic fact type
is always true in the ontology that contains it, it can play a role as
an axiom in the logic-based ontology (detailed properties).
Constraints: A constraint is the restriction that is applied to a fact
type (binary or numerical restrictions for properties).
Derivation rules: These are rules, functions or operators (including
mathematical calculations or logical inference) that are used to
derive new facts from existing ones.
Instances: Instances have the particularities of processing and are
Fig. 4. Methodology for developing BaPrOn. specifications in the applications of the upper layers.

3.3.1. Concepts
Table 1
Rating for the conceptualization complexity of ontologies.
The concepts were taken from ANSI/ISA-88. The first part of this
standard, entitled models and terminology, defines most of the
Rating Conceptual model criteria main concepts that the standard has contributed to manufactur-
Very low Only concepts ing automation. We extracted 83 concepts with their respective
Low Taxonomy, high number of patterns, no constrains descriptions, as shown in Table 2, in which the columns Name,
Nominal Taxonomy with properties general patterns available, some
Description, Synonyms, Acronyms and Type give the developer
constrains
High Taxonomy with properties and axioms, few modeling patterns, more information.
considerable number of constrains
Very high Taxonomy with properties, axioms and instances, no patterns, 3.3.2. Relations
considerable number of constrains The basic relations were taken from the assertions of the models
that ANSI/ISA-88 describes. These assertions show the basic rela-
tions between physical model, recipe model and procedural model
Logics are added to concepts, which express the relationships and are described by entity–relationship diagrams (E–R diagrams).
between the concepts and the relations. The BaPrOn ontology is For ANSI/ISA-88, recipes are needed because of the set of infor-
logic-based, includes the logics and has six main layers. Fig. 5 intro- mation that uniquely identifies the production requirements for
duces the six layers that are seen as the main axis of the ontology a specific product. Fig. 6(a) shows the relations between recipes
architecture. types. Some models are also described, such as the procedural con-

Table 2
Some concepts from ANSI/ISA-88 part 1: models and terminology.

Name Description Type

Batch process A process that leads to the production of finite quantities of material by subjecting quantities of input materials to an ordered set Concept
of processing . . .
Batch The material that is being produced or that has been produced by a single execution of a batch process. An entity that represents Concept
the production . . .
Recipe The necessary set of information that uniquely defines the production requirements for a specific product. There are four types of Concept
recipes defined . . .
Header Information about the purpose, source and version of the recipe such as recipe and product identification, creator, and issue date. Concept
Formula A category of recipe information that includes process inputs, process parameters, and process outputs. Concept
Process A sequence of chemical, physical, or biological activities for the conversion, transport, or storage of material or energy. A process Concept
consists of . . .
Process stage A part of a process that usually operates independently from other process stages and that usually results in a planned sequence Concept
of chemical or . . .
Process operation A major processing activity that usually results in a chemical or physical change in the material being processed and that is Concept
defined without consideration of the
Process action Minor processing activities that are combined to make up a process operation. Concept
Enterprise An organization that coordinates the operation of one or more sites. Must contain a site. Concept
Site A component of a batch manufacturing enterprise that is identified by physical, geographical, or logical segmentation within the Concept
enterprise. A site may . . .
Area A component of a batch manufacturing site that is identified by physical, geographical, or logical segmentation within the site. An Concept
area may . . .
Process cell A logical grouping of equipment that includes the equipment required for production of one or more batches. It defines the span Concept
of logical control of . . .
Unit A collection of associated control modules and/or equipment modules and other process equipment in which one or more major Concept
processing activities can
Equipment module A functional group of equipment that can carry out a finite number of specific minor processing activities. An equipment module Concept
is typically centered . . .
Control module A collection of sensors, actuators and associate processing equipment that acts as single entity from a control stand point, that can Concept
carry out basic control.
674 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682

Fig. 5. Construct ontology layering.

Table 3 examples are shown in Table 3, which describes five categories of

Recipe contents.
information in a recipe.
Concept Description

Header Administrative Information and process summary 3.3.4. Constraints

Equipment requirements Information about the specific equipment Some information about basic fact types is required. This asser-
necessary to make a batch or a specific part of the tion leads to the use of constraints. One example could be the
batch
Procedure Defines the strategy for carrying out a process
following constraints, which are obtained from Table 3.
Formula Describes recipe process inputs, process
parameters, and process outputs • The recipe must contain one header.
Other Information Product safety, regulatory, and other information • The recipe must contain at least one procedure.
that does not fir in the other categories
• The recipe must content at least one formula.

trol model shown in Fig. 6(b), which describes the relation that 3.3.5. Derivation rules
occurs in a process cell. The process model shown in Fig. 6(c), which This class of rules is used to link parts within the classes and
is very conceptual and describes the functionality needed to create properties. One example is the suggestion in the standard that “a
a batch, describes the steps that must occur to make a product. batch management system and equipment control system should
All these relations must involve the control model, process exist separately”. In other words, for the control recipe procedure
model and the physical model to accomplish process function- and equipment control to occur, the control recipe procedure and
ality. Fig. 7 shows some basic relations of the aforementioned an equipment phase must always exist.
models.
3.3.5.1. Instances. The instances represent the particular reality of
3.3.3. Basic fact types the domain. One example is a unit used in the process cell of
Within ANSI/ISA-88, many facts are found that provide the CEPIMA’s group Reactor 1 from PROCEL, which is described in
necessary information for related classes in more detail. Some Table 4.

Fig. 6. ANSI/ISA-88 recipe, procedure and process models.

 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682 675

Fig. 7. ANSI/ISA-88 basic relations between models.

3.4. Implementation with a friendly user interface. Furthermore, we used Collaborative

Protégé, a Protégé extension that enables users who develop an
We selected Protégé, a description logic reasoning system as a ontology collaboratively to hold discussions, chat, annotate ontol-
tool for ontology editing and knowledge acquisition (Horridge et ogy components and changes, all as an integral part of the ontology
al., 2007). This system acts as an inference engine to check data development process. From Protégé, it is possible to export ontolo-
consistency and validity (http://protege.stanford.edu/). Protégé is gies to other knowledge-representation systems, such as:
a widely used open-source ontology and knowledge base editor
• Resource description framework (RDF) is mainly intended for use
Table 4 in the semantic web, but it has also been described as a content
Instance description (PROCEL unit).
management technology, a knowledge management technology,
Concept Description a portal technology, and as one of the pillars of e-commerce
Type Glass reactor
(Klyne & Carroll, 2002).
Label EQ1 • Ontology inference layer (OIL) is intended to solve the findability
Capacity 10 L problem, support e-commerce, and enable knowledge manage-
Process Heating and holding ment (Horrocks et al., 2000).

Fig. 8. Ontology modeling approach.

676 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682

Fig. 9. System active part.

• The DARPA agent markup language (DAML) focuses on support- of the overall process, new modeling structures have been devel-
ing the semantic web, though one would assume that it also has oped within BaPrOn. These structures can recognize the existing
other uses (Pagels, 2006). trade-offs and impacts of the available alternatives at the different
information aggregation levels, and discard non-significant effects,
The proposed ontology is intended to promote transversal through retuning the decision-making/optimization model accord-
process-oriented management, to enable crossover among the dif- ing to the current process status. Thus, effective decisions can
ferent functionality silos in which businesses have typically been be made by avoiding both a greedy/myopic hierarchical decision-
structured. In order to obtain (and manage) a comprehensive view making structure and a monolithic optimization model (which is

Fig. 10. PROCEL scheme of the flow shop plant of the case study.
 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682 677

Fig. 11. Information required by the recipe at different levels.

Table 5 interface and implementation for the W3C web ontology language
Batches processing times (min).
(OWL).
Product i1 i2 i3 The user interface interacts with a model of the ontology (batch
processes domain) on the client side via a listener pattern. When
Stage Unit (j) Time Unit (j) Time Unit (j) Time
the model needs to be filled with new information, the remote pro-
l1 j1 30 j2 24 j3 50
cedure call (RPC) module on the client side will invoke a request
l2 j2 60 j3 12 j2 12
l3 j3 59 j1 54 – – to the RPC module of the server, which interacts with the ontol-
ogy and the collaboration APIs to provide the requested data. The
aim is to make it easier for knowledge engineers and experts to
impossible to solve in industrial-size scenarios). The management manage knowledge. The working process of the ontology modeling
process structures incorporated into this framework were designed approach is shown in Fig. 8, which illustrates the relation between
following ANSI/ISA-88 standard recommendations (International different actors. In addition, it enables the user to reuse existing
Society for Measurement and Control, 1995, 2001, 2003, 2006, information from the modeling databases or to create and add new
2007). However, it is envisaged that this transversal ontology can design information.
be further extended to incorporate higher decision-making hier- For the application, we activated the parts of the ontology
archies (supply chain) and to include the entire life cycle of the that concern the management of recipe information between the
organization, from the design stages to the delivery of final prod- schedule and control, as showed in Section 3. One of the main goals
ucts. of the ontology is to act like the ANSI/ISA-88. XML recipe standards
have been defined to provide an automated standardization for the
generation and management of master recipes and control recipes.
4. Proposed ontology exploitation
Fig. 9 shows the parts of the ontology that are activated in the
control process module, the schedule module and the recipe struc-
A first application of this ontology has been implemented
tures, which were mentioned in Section 3 and control the behavior
to “close” the typical scheduling-fault analysis-rescheduling loop
of the interaction between real classes and properties, ruled by the
(control levels 0 to 3 of the Purdue Reference Model) (International
axioms.
Society for Measurement and Control, 1995; Williams, 1989).
The system is coordinated by an internal server acting as an
Table 7
information administrator that is consistent with the ontol-
Cleaning time (min).
ogy structure. This can be achieved through the ontology web
language-application program interface OWL-API, which is a Java Product Unit (j) Time

i1 j1 20
i2 j2 10
Table 6 i3 j3 20
Pipes transfer times (min). i1 j1 40
i2 j2 20
From-to Unit (j1) Unit (j2) Unit (j3)
i3 j3 40
Unit (j1) 0 5.8 9.6 i1 j1 10
Unit (j2) 10.4 0 5.6 i2 j2 5
Unit (j3) 10.3 11.8 0 i3 j3 10
678 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682

• Regarding the sequencing decisions, X pis is a binary variable that

pis
establishes the general precedence relationship between a pair
of tasks (p, i, s) and (p, i, s) executed at the same processing unit
pis pis
(otherwise Xpis is meaningless). If Xpis is equal to 1, task (p, i, s)
is a direct or no direct predecessor of task (p, i, s) on the waiting
line for allocated unit.
• Alternatively, in the case where task (p, i, s) is processed before
pis
task (p, i, s) in the same unit, Xpis takes the value zero.

Finally, in order to achieve the aforementioned process, we cre-

ated a recipe schema that is ontology aligned. The recipe schema
was fitted for information concerns to the master and control
recipes and was built in XML language. The recipes are filled auto-
matically by a java application, which takes the required data from
the different database modules (control and schedule). In gen-
eral for recipes, the unit procedure was mapped to unit instances,
and operation and phase levels were mapped to equipment level
instances.
Since a production order exists, data for the master recipe are
built from the information found within that order. The production
order also contains information that is used to create and optimize
the schedule. However, control recipes are built from the infor-
mation resulting from the optimization process (schedule), which
converts each piece of schedule into a control recipe for a spe-
cific product, unit and batch. These recipes are sorted and sent to
the control module, using the respective data and information to
complete the process.

5. Case study

A batch pilot plant (PROCEL), which is a basic environment

for open simulation and optimization in a real time environment
package scenario, is located in the laboratory facilities at the UPC
Chemical Engineering Department. PROCEL provides an appro-
priate scenario for evaluating the ontology performance and for
studying and developing new process strategies. This case study
deals with the production of three chemical products, each with dif-
ferent production requirements (Table 5). The production system is
comprised of three principal processing units, two reactors and one
tank (Fig. 10). We considered one recipe for each product. Nomi-
nal processing times, transfer times (Table 6) and cleaning times
(Table 7), are considered fixed for initial scheduling. However, they
will be subject to changes, if required, to react to unexpected events.
The changes follow the corresponding models by which the process
Fig. 12. Temperature parameter following the ANSI/ISA-88 standard. is updated according to the information received from the plant. In
Table 8 some control set points are taking in to account, which are
The process control module is responsible for controlling the the maximum values for maintaining the plant under control.
phases, the actions and the specific values of the process variables BaPrOn consists of 76 classes, 115 axioms, and 86 sub-object
of each unit as well as the transition logics between each phase properties. These components make the ontology reasoning and
inside an operation or procedure. its use possible. The instances have been introduced into another
In contrast, the schedule module applies typical schedul-
ing techniques and generates a priori production schedules that
assume known and stationary operating conditions and demands Table 8
Control set points.
along the entire time horizon. In the schedule, the underlying
structure of the proposed mathematical model relies both on a Control parameter Value
continuous-time representation and the notion of general prece- Discharge EQl LTlmin 2.4 L
dence. The generalized precedence notion extends the immediate Charge QE1 LTlmax 4.4 L
predecessor concept to consider all the batches that belong to the Hold tank time 60 s
SP clean EQ1 time 90 s
same processing sequence. Broadly, this model handles allocation
SP clean EQ3 time 120 s
and sequencing decisions through different sets of binary variables, Heat EQl SP Tl 50 ◦ C
which are defined as follows: Heat EQ3 SP T3 50 ◦ C
Heat EQl SP Rl 100%
• Ypisu is a binary variable equal to one whenever task (p, i, s), that Heat EQ3 SP R3 50%
is the sth stage for manufacturing the ith batch of product p and Heat EQl SPAG1 100%
is allocated to equipment unit u. Heat EQ3 SP AG2 20%
 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682 679

Fig. 13. Values shown by protégé from the process database.

ontology (instance layering; see Fig. 5), which is called from the The control task in the units was applied to bring the processing
main ontology as an extension. Thus, we take into account one of time from the control level, through the ontology, to the sched-
the main ontology principles, which is to be universal (subjective) ule level. In this way variable values were updated when planning
in a particular domain, in this case the main ontology. The recipes level asked for using them as input for scheduling. In order to
need information founded inside the process. This information is succeed in this activity a link between data source from the exter-
shown in Fig. 11 and arrives to the schedule module and the control nal software variable and the real values captured by the control
module, covering the required information. system in the process database has been created. Using the Pro-
Fig. 12 shows in a summarized way, how from the master recipe tégé Database Manager facilitated this task. An example is show in
(unique file) some temperature parameter values (found from the Fig. 13 where structured query language (SQL) process databases
5 to 19 code string) are carried out to some control recipe (1 of 8 have been loaded for future information access.
files) temperature parameters (found from 299 to 311 string line), The main BaPrOn classes and subclasses hierarchical taxonomy
by the application of the model standardized at these two control (Fig. 14) was built in accordance with ANSI/ISA-88 standards. Con-
levels. sequently, a better understanding of this standard is attained for

Fig. 14. ANSI/ISA-88 classes and subclasses hierarchical taxonomy.

680 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682

Table 9
Comparison of process models at scheduling and control levels.

No Present steps Time (min) No Proposed steps Time (min)

Scheduling level 1 Input of external information 2 1 XML information quality input 3

source
2 Selection of specific and necessary 20 2 Capturing of information to the 4
information for the particular used particular language
process
3 Translation and capturing of 30 3 Scheduling optimization process 4
information to the used particular
language, in the scheduling process
4 Scheduling optimization process 4 4 XML quality information output 1
5 Output as information source 2

Control level 1 Input of external information 2 1 XML information quality input 3

source
2 Selection of specific and necessary 20 2 Capturing of information to the 4
information for the particular used particular language
process
3 Translation and capturing of 30 3 Control parameter process 6
information to the used particular
language, in the control process
4 Control parameter process 6 4 XML quality information output 1
5 Output as information source 2

Total time 118 Total time 26

potential users, which facilitates the application and the imple- ried at the scheduling control level. The following are the most
mentation by the instantiation of the current batch process system. significant benefits found in the case study:
Another remarkable aspect is that when a system or a simple pro-
cess is instantiated within the ontology structure, the relations Attending to knowledge diversity the technical effort required to
ensure that every part or instance follows ANSI/ISA-88 standards deal with the process system and its representation along the
requirements. As shown in Fig. 15, the slot information is required knowledge found at different levels, ensured that all parts within
when a control recipe instantiation is performed. the ontology were easily accessible. This is a particular and explicit
way of representing the knowledge by the content format and the
5.1. Results content type attributes found in the ontology structure classes.
These improvements were brought about by making this informa-
BaPrOn was successfully applied to a complex scenario of closing tion visible and readily available to diverse entities (human and
the scheduling-monitoring and control-rescheduling loop tanks for computers) at different control levels. In addition, as a result of
agile and quick information sharing and exchange among diverse testing the ontological system with the real database structure,
control levels (0, 1, 2 and 3 levels from the Purdue CIM reference we could apply a strengths, weaknesses, opportunities and threats
model) within the loop. As a result of the knowledge description, (SWOT) analysis. This enabled us to identify potential modifica-
access to information quality was improved. The best decision- tions in the database structure, to improve it and make it more
making task was achieved in this particular case study when the efficient.
specified data from the control and fault diagnosis level were car- Information within the recipe steps and process information (real
values stored and referenced to the batch in a database), raw
materials, associated characteristics, and quantities consumed are
automatically logged to the batch journal. When this is integrated
with batch historian tools, process data can be assigned to batch
context and used to generate reports, comparisons and trends for
detailed batch analysis.
While the instantiation of the case study is made, the pro-
cess becomes standardized automatically, as established by
ANSI/ISA-88. This provides several automation benefits, includ-
ing a reduction in the implementation time and cycle times. It
also allows for improved (batch-to-batch) product consistency,
improved product and process quality management, and better
cost accounting capability.
The monitoring information is inside the ontological structure.
This means that we have the guidelines for better control, as the
monitoring information shows which elements are the sources or
are affected in a fault detection process.

Finally, we made a contribution to the separation of recipe

procedure and equipment control logic, and provided a clearer def-
inition of manufacturing processes and production requirements.
The use of common terminology and manufacturing models sim-
plifies communication between customers and suppliers in the
Fig. 15. Protégé required parts in order to instance the process. system.
 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682 681

5.2. Ontology performance • It creates a modular and internally consistent standard that can
help reduce engineering cost, and it can finally serve as a man-
The use of BaPrOn enabled the integration of the scheduling task agement tool.
and the fault diagnosis system (FDS) and supplied the process state • It constitutes an appropriate tool to help filling the existing gap of
information to different levels in the decision-making hierarchical the missing implementation technology of the ANSI/ISA-88 stan-
structure, which led to the following improvements: dard, that is to say, the utilization of an ontology as a form of
knowledge representation.
Behavior: The efficiency of the reactive scheduling was improved
by updating the schedules in the most effective way, which leads The Methodology proposed versus the actual state of the art:
to an increase in process yield. nowadays many improvements concerned to the ontologies study
Language: The information presented proper, common and stan- area and its application is in process. However, actually there is not
dardized language within the hierarchical levels, regardless of its a certain methodology for their development. This work tries to
origins. ensure an ordered quality manner to develop an ontology. This is
Documentation: Ontology provides a certain and easy descrip- made trough the inclusion of the two most cited methodologies
tion for formal (read by a computer) and informal (presented in within a well known continuous improvement tool PDCA cycle.
a human-readable form) specifications of the system elements’ From this point of view this work is an ongoing successful effort
content. to improve ontology process development in a friendly way.
Storage: Efficient access is provided to process databases using a In addition, this ontology opens the way for achieving success-
database management system and to establish the links and loca- ful flexible control in adapting and recognizing different elements
tions of these databases. found through the hierarchy models that are associated with man-
Navigation: The Web-Protégé portal provides the option of inter- ufacturing multilevel control systems.
action among potential users of the ontology, by interchanging Finally, this work represents a step forward to support the
some notes about the different parts. integration (not just “communication”) of different software tools
applicable to the management and exploitation of plant database
A comparison between the process of how different levels inter- information, resulting into an enhancement of the entire process
act in a traditional way and the proposed one, are shown in Table 9. management structure.
We can see an important improvement of 20% decrease in the num- The ontological structure can be extended in the future to
ber of steps for carrying out each process. The major improvement incorporate other hierarchical levels and their respective model-
appears when a comparison of implementation time required by ing knowledge. Concepts of functions (order processing, detailed
the two models is made showing a 78% of time saved. production scheduling, production control, quality assurance, etc.)
can thus be developed according to the ANSI/ISA-95 standards, and
the whole supply chain (demand, capacity expansion, inventory
6. Conclusions and future work
control, etc.) can be improved.

The fields of ontology, knowledge management and decision-

Acknowledgements
support systems have matured significantly in recent years.
Currently, the integration of various fields and technological
The financial support received from the Dirección General de
solutions facilitates knowledge creation, storage and sharing in
Educación Superior Tecnológica (DGEST), Academy Excellence Pro-
a specific domain, and improves the effectiveness of decision-
gram, reference 072007004—E.A. from México and the Spanish
support systems.
Ministry of Science and Innovation (MICINN) through the project
The creation of a bridge between the batch process and related
“Advanced Support System for Flexible Chemical and Petrochemi-
information enabled formal and informal models, knowledge and
cal Production Processes” (DPI2006-05673) are fully appreciated.
experience to exploit different optimization tools through the sup-
port of common methods, languages and tools that simplify the
generation and processing of conceptual models. References
The BaPrOn ontology, which was created using the ANSI/ISA-
Alavi, M., & Keen, P. G. W. (1989). Business teams in an information age. Information
88 standards as a guideline, performed well at capturing common Society, 6(4), 179–195.
understanding in conceptual design and helping to utilize the rela- Bayer, B., & Marquardt, W. (2004). Towards integrated information models for data
tions that are mined from the databases. This helps users to obtain and documents. Computers and Chemical Engineering, 28, 1249–1266.
Bechhofer, S., van Harmelen, F., Hendler, J., Horrocks, I., McGuinness, D. L., Patel-
data and information in a proper, fast and standard way through Schneider, P. F., et al. (2004). OWL web ontology language reference-website.
an internal web page portal. Available at http://www.w3.org/TR/2004/REC-owl-ref-20040210/
The creation of consensual methodology aided the development Board, N. S. A., et al. (2009). Bioportal reference-website. Available at
http://bioportal.bioontology.org/
of the ontology and made it easy and friendly to use. We considered
Brandl, D., & Emerson, D. (2003). Batch markup language batchml. World Batch
On-To-Knowledge and Methontology, which are two of the most Forum.
commonly used existing methodologies. Brickley, D., & Guha, R. (2002). Rdf vocabulary description language 1.0: RDF schema.
Available at http://www.w3.org/TR/2002/WD-rdf-schema-20021112/
With the application of the ontology, in a short part of batch
Eppler, M. J. (2006). Introducing the notion of information quality (2nd Ed.).
control process, the follow advantages can be mentioned: Berlin/Heidelberg: Springer., pp. 1–16.
Fensel, D. (2003). Ontologies: A silver bullet for knowledge management and electronic
commerce. Springer.
• The systematic integration of different actors within the con-
Gebus, S., & Leiviskä, K. (2009). Knowledge acquisition for decision support systems
trol process. In particular planning (scheduling) and plant floor on an electronic assembly line. Expert Systems Applications, 36(1), 93–101.
control (unit control). Gruber, T. R. (1993). A translation approach to portable ontology specifications.
• This integration in a major application could allow that access Knowledge Acquisition, 5(2), 199–220.
Horridge, M., Jupp, S., Moulton, G., Rector, A., Stevens, R., & Wroe, C., et al. (2007).
that the information be not a constrain that limits the plant man- A practical guide to building owl ontologies using protege 4 and co-ODE tools.
agement capacity. Technical report, The University Of Manchester.
• It provides a common language for better communication about Horrocks, F. D., Broekstra, J., Decker, S., Erdmann, M., Goble, C., van Harme-
len, F., et al. (2000). Ontology inference layer oil. Available at http://www.
automation opportunities and manufacturing requirements. ontoknowledge.org/oil/
682 E. Muñoz et al. / Computers and Chemical Engineering 34 (2010) 668–682

International Society for Measurement and Control. (1995). Batch control. Part 1. Pagels, M. (2006 January). The darpa agent markup language (DAML) reference website.
Models and terminology. International Society for Measurement and Control. Available at http://www.daml.org/
International Society for Measurement and Control. (2001). Data structures and Rippin, D. (1983). Batch process systems—Engineering: A retrospective and prospec-
guidelines for languages. International Society for Measurement and Control. tive review. Computers and Chemical Engineering, 17, S1–S13.
International Society for Measurement and Control. (2003). Batch control. Part 3. Rodríguez-Martínez, A., López, I., Arévalo, Bañares, R., Alcántara, Aldea, A., et al.
General and site recipe models and representation. International Society for Mea- (2004). Multi-model knowledge representation in the retrofit of processes. Com-
surement and Control. puters and Chemical Engineering, 28, 781–788.
International Society for Measurement and Control. (2006). Control batch parte 4 reg- Shah, N. (1998). Single-and multisite planning and scheduling: Current status and
istros de producción de lote. International Society for Measurement and Control. future challenges. In J. F. Pekny, & G. E. Blau (Eds.), Proceedings of the FOCAPO
International Society for Measurement and Control. (2007). Batch control part 5 1998, CACHE (pp. 75–90).
automated equipment control models & terminology. International Society for Shim, J. P., Warkentin, M., Courtney, J. F., Power, D. J., Sharda, R., & Carlsson, C.
Measurement and Control. (2002). Past, present, and future of decision support technology. Decision Support
Klein, M., Fensel, D., Kiryakov, A., & Ognyanov, D. (2002). Ontology versioning and Systems, 33(2), 111–126.
change detection on the web. In Proceedings of the 13th international conference Shirasuna, M. (2007). Optimización de la producción en una empresa de cervezas
on knowledge engineering and knowledge management EKAW02 (pp. 197–212). usando el estándar ISA s88 (caso polar). Rockwell Automatization.
Klyne, G., & Carroll, J. J. (2002). Resource description framework (RFD): Concepts and Simon, F., & Murray, T. (2007). Decision support systems. Communications of the ACM,
abstract syntax reference-website. Available at http://www.w3.org/TR/2002/WD- 50(3), 39–40.
rdf-concepts-20021108/ Subrahmanyam, S., Bassett, M., Pekny, J., & Reklaitis, G. (1995). Issues in solving
López, M. F., Gómez-Pérez, A., Sierra, J. P., & Sierra, A. P. (1999). Building a chem- large scale planning, design and scheduling problems in batch chemical plants.
ical ontology using methontology and the ontology design environment. IEEE Computers and Chemical Engineering, 19, 577–582.
Intelligent Systems, 14(1), 37–46. Sure, Y. & Studer, R., (2002). On-to-knowledge methodology. Technical report, Uni-
McQueen, S., & Thompson, H. (2000). Xml schema reference-website. Available at versity of Karlsruhe.
http://www.w3.org/XML/Schema Swartout, W., Neches, R., & Patil, R. (1993). Knowledge sharing: Prospects and chal-
Mendez, C. A., Cerda, J., Grossmann, I. E., Harjunkoski, I., & Fahl, M. (2006). State- lenges. In K. Fuchi, & T. Yokoi (Eds.), Proceedings of the international conference
of-the-art review of optimization methods for short-term scheduling of batch on building and sharing of very large-scale knowledge bases.
processes. Computers and Chemical Engineering, 30, 913–946. Venkatasubramanian, V., Zhao, C., Joglekar, G., Jain, A., Hailemariam, L., Suresh, P.,
Missikoff, M., & Taglino, F. (2002). Business and enterprise ontology manage- et al. (2006). Ontological informatics infrastructure for pharmaceutical prod-
ment with symontox. In S. B. Heidelberg (Ed.), The semantic web—ISWC 2002, uct development and manufacturing. Computers and Chemical Engineering, 30,
2342/2002 (pp. 442–447). 1482–1496.
Morbach, J., Yang, A., & Marquardt, W. (2007). Ontocape: A large-scale ontology for Venkatasubramanian, V., Reklaitis, G., Hsu, S. -H., Jain, A., Hailemariam, L., Suresh,
chemical process engineering. Engineering Applications of Artificial Intelligence, P., et al. (2008). Overview of the ontological informatics infrastructure for pharma-
20, 147–161. ceutical product development. Technical report, School of Chemical Engineering
Morbach, J., Wiesner, A., & Marquardt, W. (2009). Ontocape 2.0: A (re)usable Purdue University.
ontology for computer-aided process engineering. Computers and Chemical Engi- Williams, T. J. (1989). A reference model for computer integrated manufacturing (CIM).
neering, 33, 1546–1556. Research Triangle Park: ISA.
Neches, R., Fikes, R., Finin, T., Gruber, T., Patil, R., Senator, T., et al. (1991). Enabling XML-Core-Working-Group. (2009 April). Extensible markup language (xml) reference-
technology for knowledge sharing. AI Magazine, 12(3), 36–56. website. Available at http://www.w3.org/XML/
Obrst, L. (2003). Ontologies for semantically interoperable systems. In CIKM’03: Pro- Zhao, C., Jain, A., Hailemariam, L., Suresh, P., Akkisetty, P., Joglekar, G., et al. (2006).
ceedings of the twelfth international conference on information and knowledge Toward intelligent decision support for pharmaceutical product development.
management, ACM (pp. 366–369). Journal of Pharmaceutical Innovation, 1(1), 23–35.