Professional Documents
Culture Documents
The process of model transformation is based on Candidate models from the example are sch and gr.
comparing vis--vis if all the defined relations are On the other hand, a transformation can be invoked to
verified. For those relations who are not verified, the verify the consistency of two models as well as to
mentioned process tries to carry out by inserting, transform a model into another. A transformation
eliminating or modifying the final model [16]. example can be seen on code 1.
The reserved identifier transformation is used to carry
3.4.1. QVT (Query, View, Transformations). The out a transformation in a one-way direction, i.e.
transformations are classified as relation and mapping. considering that one of the models is chosen as the final
The first ones specify multi directional transformations. one. The selected model must be empty or contain the
They do not create or modify models, but they can check necessary elements to make the transformation, meaning
the consistency among two or more related models. the destiny model must be totally empty or it should have
While mapping, transforms elements from a domain into some previous objects that must be mapped into the
the elements of another domain [17]. transformation model.
QVT allows carrying out transformations among There are elements such as relations and domain
models whose languages are defined in MOF (Meta within a transformation. The relations define restrictions
Object Facility) using the following elements: that must be satisfied by the elements of the candidate
models. The domains are variables that can be matched
Query: is an expression tested on a given model that to a model of a given metamodel [18].
has as a result one or more instances of types. The domains that correspond to the models sch and
View: is a model that is completely derived from gr can be seen in the example code 2. Each domain
another model. They are more general than Query. They specifies a pattern, i.e. name and schema. In the example
tend to be read-only entities. above the variable name is the same in both properties
Transformation: generates a final model from an name, in consequence, they have the same value. A
original one. The relations specify the transformations transformation has two kinds of relation: top-level where
all the relationships must be complied with and no-top using QVT. The transformation process is explained in
level that are satisfied only when they are invoked more detail in the sections below.
directly or transitively.
4.1. Models Transformation. To begin with, a
top relation ResourcesToNode() metamodel design is needed to transform models. Model-
{ Oriented Objects (MOO) was chosen in the design
Name: String process and it is implemented using UML.
Domain sch r: Resources { MOO was selected as the modeling tool since
Nombre = Name structural representation of objects as well as dynamic
} and functional behavior of requirement is allowed.
Domain gr n: Node{ Besides, it can be used several abstraction levels and
Nombre = Name natural language is an important characteristic of this
Label = r+Name methodology [25].
} In this paper, a class diagram was modeled with two
} domains that represent the schedule and the
corresponding graph respectively [18]. Once design is
finished, the original and the transformed domain are
Code 2. Top Relation declaration example. determined and selected. This means that, if model A is
chosen as the original one, transformations are made in
4. Schedule Modeling using Graphs. such manner that it is equivalent to model B (transformed
Production planification and control activities can be model).
considered as means to reduce the uncertainty and deal
with disturbances that invalidate assumptions made.
The steps to create a task schedule, generally, are:
Project analyzing
Tasks and resources identification
A work plan creation
Assessment of results
1. Start
2. Schedule state query Figure 8. Model transformation summary table
3. Select state to modify
4. Select modify option The graph model for the schedule state is made from
5. If attributes states change:
the overview table in figure 8, which will contain the
5.1.Change state attribute
5.2. Save modification (original value, new value, abstraction level presented in foregoing sections.
place, date)
6. If attribute batch change:
6.1. Change batch attribute
6.2. Save modification (original value, new value,
place, date)
7. If attribute resources change:
7.1. Change resources attribute
7.2. Save modification (original value, new value,
place, date)
8. If attribute tasks change:
8.1. Change tasks attribute
8.2. Save modification (original value, new value,
place, date)
9. Save changes Figure 9. Schedule graph model example
10. Save changes log
11. End The graph model, as can be seen in figure 9, presents a
more informative representation than alternative tools to
Code 3. Schedule state modification pseudo-code observe schedule states, batches, resources and tasks
along with their name and precedence and execution
Information stored in the change logs and the resulting relationships. Resource groups, model relations and task
traces of schedule states obtained are used to elicitate executions are discriminated by using different colors.
expert knowledge on schedule problem resolutions that is Also, the first abstraction level information is observed in
lost in other simpler diagrams such as Gantt diagrams. the graph diagram as was explained in previous sections.
Heuristics and learning rules are created based on that The schedule graph specifies element data obtained by
stored knowledge and can be incorporated in a cognitive steering over each node and resorting to the information
system using these systematic procedures to calculate stored in the second abstraction level.
state schedule similarities and graph distances [26].
According to the model conversion functionality, the 6. Concluding Remarks
prototype uses a detailed table of the converted elements,
that is to say, using the stored information from a This paper proposes a new modeling methodology
schedule state that performs the element transformation using graphs that results in a prototype implementation
that automates the model generation and the elicitation of to improve formal and mental models, System
knowledge for rescheduling. This proposal emerges of Dynamic Review, vol. 14, n 4, pp. 309-340, 1998.
the need to satisfy the earlier limitations mentioned about [6] N. Shadbolt y P. Smart, Knowledge Elicitation:
current tools and representation methodologies to model Methods, Tools and Techniques, Florida: J. R.
production schedules. Be aware that current tools do not Wilson & S. Sharples, 2015.
allow capturing expert human knowledge to address re-
scheduling problems. [7] V. Franzoni, A. Milani y S. Pallottelli, Multi-path
In the first place, concepts and definitions required to Traces in Semantic Graphs For Latent Knowledge
understand graphs, model languages and transformations Elicitation, de International Conference on Natural
were studied. Computation (ICNC), Italia, 2015.
Secondly, the mentioned concepts were implemented [8] C. Zapata Jaramillo, A. Gelbukh y F. Arango Isaza,
to model, transform metamodels and develop a prototype Pre-conceptual schema: a conceptual-graph-like
that allows to automatize the schedule representation and kownledge representation for requierements
to elicitate expert domain knowledge gained by trial-and- elicitation, Berlin: Springer-Verlag, 2006.
error experience in dealing with abnormal events and [9] N. Cooke, Varieties of knowledge elicitation
unplanned disturbances. techniques de International Journal of Human-
Finally, a case study of a multi-product batch factory Computer StudiesUSA, 1994.
is presented to demonstrate the use and advantages of the
[10] G. Wagner de Garca y A. Barredo, Introduccin a
modeling and implementation of the methodology
la teora de grafos, Armenia: Ediciones Elizcom,
presented in this paper. 2010.
7. Future Work [11] N. Mart Oliet, Y. Ortega Malln y J. Verdejo
Lpez, Estructura de datos y mtodos algoritmicos:
Knowledge elicitation and production schedules ejercicios resueltos, Madrid: Pearson Educacin,
modeling, along with the developed prototype, serve as a 2004.
base of future works with Inverse Reinforcement [12] M. Fowller y K. Scott, UML gota a gota, Mjico:
Learning (IRL) [27]. Addison Wesley, 1999.
Future developments will use learning algorithms
[13] J. Rumbaugh, I. Jacobson y G. Booch, El lenguaje
where distances among schedules are need as input data.
de modelado unificado - Manual de referencia,
The proposal of this paper, using graph modeling is a
Madrid: Addison Wesley - Pearson, 2006.
fundamental advantage to quantify distances since exist
several options of algorithms to calculate them. Finally, [14] C. Ariste, J. Ponisio, N. Leopoldo y R. Giandini,
the proposed approach provides the foundations to future Diseando transformaciones de modelos CIM/PIM:
distance edition calculations and human expert desde un enfoque de negocios hacia un enforque de
knowledge modeling that will be used for automated sistemas, 2015.
learning in artificial cognitive systems. [15] M. Hebach, MDA with QVT, Presentation Boreland
Together, 2006.
8. References [16] L. Favre, Model driven architectures for reverse
engineering technologies: strategic directions and
[1] B. MacCarthy y J. Wilson, Human Performance in system evolution, de Engineering Sciece
Planning and Scheduling, London: Taylor & Referencia - IGI Global, 2010.
Francis, 2001.
[17] P. Barendrecht, Modeling transformations using
[2] Giudici y Bris Luch, Introduccin a la teora de QVT operational mappings, Eindhoven: University
grafos, Venezuela: Ediciones de la Universidad de of Technology, 2010.
Bolivar, 1997.
[18] M. Djedjigo, D. Maurad y P. Mokan, Aspect-
[3] F. Gonzlez Fernndez, Teora y prctica del Oriented security hardening of UML design models,
mantenimiento industrial avanzado, Madrid: Switzerland: Springer, 2015.
Editorial FC, 2003.
[19] M. Gonzlez Riesco, Gestin de la produccin:
[4] M. Gracia Ramos, M. Lopez-Jurado Gonzlez, M. cmo planificar y controlar la produccin
Yagez Insa y J. Mering Lindahl, Gua prctica de industrial, Vigo: Ideas Propias, 2016.
economa de la empresa I: empresa y entorno
[20] J. Verdoy y E. al., Manual de control estadistico de
(Teora y ejercicios), Barcelona: Universidad de
calidad: teoria y aplicaciones, Espaa: Universitat
Barcelona, 2008.
Jaume, 2006.
[5] D. Ford y J. Sterman, Expert knowledge elicitation
[21] M. Roln, M. Canavesio y E. Martinez, Agent
based modelling and simulation of intelligent
distributed scheduling systems, Computer aided
chemical engineering, n 26, pp. 985-990, 2009.
[22] M. Roln, M. Canavesio y E. Martinez, Agent
based generative simulation of an intelligent
distributed scheduling world with Netlogo, de
Proceedings of CICA Applied Computational
Intelligence Conference, Buenos Aires, 2009.
[23] J. Palombarini, Un enfoque cognitivo a la
reparacin automtica de planes y schedules
integrando aprendizaje por refuerzos con
abstracciones lgicas y relacionales, Santa Fe,
2014.
[24] R. Musier y L. Evans, An aproximate method for
the production scheduling of industrial batch
processes with parallel units, Computers and
chemical engineering, n 13, p. 229, 1989.
[25] D. Coleman, Object-oriented development - The
fusion method, Michigan: Prentice Hall, 1994.
[26] Van Otterlo, The Logic Of Adaptive Behavior:
knowledge representation and algorithms for
adaptative sequential decision making under
uncertainty in first-order and relational domains,
Amsterdam: IOS Press, 2009.
[27] S. Levine, Z. Popovi y V. Koltun, Nonlinear
inverse reinforcement learning with gaussian
processes, de Advances in neural information
processing systems 24 (NIPS), 2011.