(399g) Ontology Engineering: Support to Decision Making in Biorefining

Authors: 
Cecelja, F., University of Surrey
Koo, L., University of Surrey
Kalemi, E., University of Surrey
Trokanas, N., University of Surrey
Biorefining is a dynamic field with ever grooving number of processes operating in different environment, using different feedstocks and providing expanding range of products. Option scoping and process synthesis and process design are becoming increasingly complex [1]. While development and integration of process and feedstock models and datasets from laboratory experiments are crucial at design stage [2], process and feedstock discovery and valorisation are important at operational stage [3]. This in turn necessitates efficient decision support tool addressing stated complexity and assuring timely and accurate decisions made at increasingly shorter time span.

This paper proposes a knowledge based decision support tool for biorefining which:

  • Coordinates operation of biorefining repositories of process and feedstock models and datasets, as well as repositories of existing biorefining processes and available feedstocks, all with fast and precise discovery and audited updates;
  • Supports model and data integration used for process synthesis and design [4];
  • Enables screening opportunities and integration of concomitant networks based on technological capability and resource availability for targeted products [3, 5];
  • Assessment of economic and environmental potential and effects [6].

The proposed decision making tool employs ontologies to embed tacit knowledge in the domain of biorefining, knowledge gained from past experience but also from the latest research and otherwise advances in biorefining [1, 2]. More specifically, a set of integrated ontologies address i) functionality and otherwise properties of biorefining models and datasets from laboratory experimentation and biorefining processes, or simply from analytical assessments, ii) nature of feedstock, i.e. variability in composition, dynamics in availability and pricing, as well as economic and environmental properties including hazardousness, iii) processing technologies in terms of processing capabilities, but also range of type of inputs, conversion rates, water and energy requirements, range of capacities, emissions as well as fixed and operational costs and environmental effects, and iv) general knowledge and past experience in process design and synthesis. While tacit knowledge associated with all four elements is embedded in the ontology structure, the explicit knowledge associated with particular process models and datasets, as well as particular process and feedstock is enquired during the ontology instantiation with actual data collected from the model, data, process and feedstock owners during the registration. The ontologies are designed using ontology web language and hence prepared to grow and to share. In the current implementation more than 400 different model and datasets have been included on the top of more than 150 different type of feedstock and over 40 different biorefining technologies, all organised in respective repositories.

Purpose designed matchmaker [5] is used to i) provide a fast and accurate discovery of elements in all repositories and enable repository audited updates, ii) enable model and data integration, and iii) formation of processing networks with assessment of economic and environmental benefits. Semantic relevance of integrated models and datasets, as well as integrated processes into a network solution defines their mutual suitability and it is calculated from distance between the two instances in the respective ontology. Explicit relevance is calculated using vector similarity algorithm for respective properties, such as quantity, availability, geographical location and environmental and economical potential etc. More intuitive and complex model and data networks, as well as processing networks are proposed by reclusively repeating matches between two participants which in turn gives an opportunity for even better economic and environmental savings and/or targeted production. Both semantic and explicit matching relevance are aggregated in into a numerical values use for match ranking and providing beter but flexible decision support to the user.

The proposed decision making platform has been implemented as a web service with performance validated verified using several real-life cases for model and data integration, as well as real-life biorefining scenarios.

  1. Bussemaker, M., et al., Ontology Modelling for Lignocellulosic Biomass: Composition and Conversion. Computer Aided Chemical Engineering: 28th European Symposium on Computer A, 2018.
  2. Koo, L., N. Trokanas, and F. Cecelja, A semantic framework for enabling model integration for biorefining. Computers & Chemical Engineering, 2017. 100: p. 219-231.
  3. Raafat, T., et al., An Ontological Approach Towards Enabling Processing Technologies Participation in Industrial Symbiosis. Computers & Chemical Engineering, 2013. 59(1): p. 33-46.
  4. Cecelja, F., et al., Integration of CAPE Models and Data for the Domain of Biorefining: InterCAPEmodel Ontology Design. Computers and Chemical Engineering, 2019. - accepted for publication.
  5. Trokanas, N., F. Cecelja, and T. Raafat, Semantic Input/Output Matching for Waste Processing in Industrial Symbiosis. Computers & Chemical Engineering, 2014. 66(1): p. 259 - 268.
  6. Trokanas, N., F. Cecelja, and T. Raafat, Semantic Approach for Pre-assessment of Environmental Indicators in Industrial Symbiosis. Journal of Cleaner Production, 2015. 96(1): p. 349-361.