(397c) The Ontology System for Easy and Reusable Model Knowledge Representation
AIChE Annual Meeting
Tuesday, November 18, 2014 - 3:55pm to 4:15pm
The development and application of process models are usually not trivial due to the complexity of the system being studied and/or scales being considered. Therefore, it is useful to store the modelling results in some knowledge base with the possibility to update, access and extract information about the model, in order not to repeat the same model development steps, and also to reuse or improve existing know-how. Hence, there is a need for a simple but efficient system for knowledge representation and management of models and their connections to other tools for different model-based applications.
This work presents design and implementation of an ontology for knowledge representation related to computer-aided modelling. This ontology is built on the basis of the modelling language, which includes classes, layers, and blocks (these classification names are related to the graphical representation of the model in the modelling-tool software). Class is a pattern with specific parameters or characteristics that is used as template for creation of modelling objects. These objects are then called the instances of the class. Classes interrelate between each other; upper layers represent classes and contain sub-layers; sub-layers contain blocks. The collection of instances of the classes represents the model in the system.
The structure of the knowledge ontology is based on a model decomposition technique . Every model representation includes four upper layers, which are System information, Balance equations, Constitutive equations and Connection equations. Every process model could be decomposed to the equations and relations, which can be linked to one of the upper layers. Each upper layer includes a number of sub-layers, representing specific phenomena occurring in the system. In the upper layer there are predefined sub-layers (giving the basic and essential description of the system) as well as user-defined sub- layers, which might be unique for the specific case. The predefined sub-layers are, for example, number of balance-volumes, number of phases, presence of the reaction etc., given in the System information upper layer; similarly, mass balance, energy balance, momentum balance equations belong to the Balance equations upper layer; kinetic, thermodynamic, diffusion models belong to the Constitutive equations upper layer; relations between volumes inside the system and between system and outside world, closure equations, initial conditions belong to the Connection equations upper layer. Each sub-layer contains one or more blocks, representing various options or scenarios, which are possible in relation to the phenomena, linked to the corresponding sub-layer. Model equations are included in the corresponding blocks and this gives the model developer the possibility to quickly generate-create the required type of model by choosing one of the options available in the sub-layers. Also, the system allows updating of existing models, therefore, storing knowledge for further reuse.
On top of simple representation of the model structure this system is linked to a tool for model solution and analysis. Also, a model transfer via XML option for the created or existing model is available. Therefore, any tool, which accepts this type of XML file as an input, could be linked to the modelling tool-box.
This modelling ontology can be applied to a wide range of modelling applications in chemical engineering, as it is not associated to any specific modelling tools. The software implementation of the ontology makes it also easy to use even for the model developer with limited knowledge about modelling. The ontology also represents the knowledge in a way that allows model reuse and transfer to other software applications.
 I. Cameron, R. Gani, 2011, Product and Process Modelling. A Case Study Approach., Elsivier.