Identifier: TDX:1603
Authors: Rodríguez Martínez, Antonio
Abstract:
Industrial processes require periodic evaluations to verify their normal operation, as much in economical as in technical terms. These evaluations are necessary due to changes in markets, safety and environmental legislation. It is necessary to satisfy these demands in order to investigate process alternatives that allow the optimal use of existing resources with the possible minimum investment. One way to make this task is by means of retrofit.The objective of this work is to develop a methodology for retrofit of chemical processes based on hierarchical representations. The concepts involved in the methodology are,- The application in inverse manner of hierarchical procedure for conceptual design of Douglas [Douglas, 1988]- The representation of processes by means of generic block diagrams from Turton [Turton et al., 1998].- The structure of knowledge based systems (knowledge base, inference machine and user interface) [Han et al., 1996].The proposed methodology has learning from retrofit, design, synthesis and analysis of processes, so that it was managed to design and implement support tools for retrofit of processes. Knowledge from computer science area in addition to knowledge from chemical engineering were applied in order to develop a knowledge-based system by using computational algorithms from communication, visualization and artificial intelligence techniques.The methodology consists of four general steps,1. Data extraction. Information of chemical process is extracted from an initial representation.2. Analysis and representation of information. Extracted information is abstracted at several levels based on a set of hierarchical functions, precedence rules and user information.3. Generation of alternatives. The alternatives are generated based on the new specification for the original process and a Case-Based Reasoning (CBR) system.4. Adaptation and evaluation. The generated alternatives are evaluated until their new specifications are satisfied.The methodology was implemented as Knowledge-Based System (KBS). The KBS has three prototypes and one ontology. The ontology is used as common language between the prototypes. The prototypes are,- HEAD (Hysys Extraction Data) was developed in MS Visual Basic . The goal is to extract information from a process flow diagram (PFD) which is simulated in Hysys.Plant .- AHA! (Automatic Hierarchical Abstraction tool) is a prototype developed in Java that generates different levels of abstraction from the initial PFD in order to identify sections where retrofit can be applied.- RETRO (Reverse Engineering Tool for Retrofit Options) includes a CBR system to retrieve from a library cases with equipment/devices (Units) or groups of them (Meta-units) which were similar to the one selected by the user.A set of cases (Units and/or Meta-units) is obtained as output from the CBR system. The set is ordered according to similarity of cases. The user should take the most similar alternative and make the adaptation in a simulator (in our case, we have used Hysys.Plant ).The KBS has been validated with 50 chemical processes. It has been applied in detail to the processes of ammonia and acetone production with satisfactory results in reduction of time and complexity of the analysis, representation and generation of alternatives of solution during retrofit tasks.The advances and results of the research have been published in 9 articles (3 in international journals, 1 in a book collection and 5 in refereed conference papers).Also they have been presented in 10 international congresses (7 communications and 3 posters) and 3 nationals (3 posters).
Repositori URV