To investigate the efect of possible changes to decarbonise the economy, a detailed picture of the current production system is needed. Material/energy flow analysis (MEFA) allows for building such a model. There are, however, prohibitive barriers to the integration and use of the diverse datasets necessary for a system-wide yet technically-detailed MEFA study. Here we present a demonstration of the “Physical Resources Observatory” (PRObs) system, which exploits Semantic Web technologies to integrate and reason on top of this diverse production system data. We demonstrate how this system supports domain experts to load datasets, map them to the ontology, and query the extended data to meet several modelling use cases.
A whole-systems understanding of production systems is essential to navigating
the necessary rapid transition to a zero-carbon economy. Understanding how
we produce and consume physical resources is fundamental to understanding
the impacts human activity has, and opportunities to increase eficiency. This
understanding relies on access to data about the production and consumption of
physical resources (materials, energy, etc.) and their associated environmental
impacts. Due to the economy-wide yet detailed nature of these questions, they
must be based on many pieces of data from diverse sources. Moreover, this data is
incomplete, and defined using inconsistent categorisations of the types of resource
and activities. In addition, the lack of well-defined data models for this type of
data is limiting to data reuse and holding back academic research [
We propose and develop a solution using a domain ontology and the RDFox triple store to eficiently implement Datalog rules integrating diverse data points into a consistent structure. This forms part of the “Physical Resources Observatory” (PRObs) system, being developed within the UK FIRES research programme3, where it supports a wider research agenda on resource eficiency and decarbonisation in UK industrial strategy. 2
To allow quantified data points on resource use to be expressed in RDF, we
build on the data model proposed by Pauliuk et al. [
The ontology and the RDFox scripts implementing the rules and algorithm
are available online at [
The PRObs system is intended to be used to support modelling by domain experts who are not familiar with semantic web technologies. As such, they should be supported to enter information (e.g. about materials of interest and their hierarchical structure) and retrieve results without becoming experts in RDF and complex SPARQL queries. The system should be usable (as far as possible) on typical researchers’ computers, and integrate with typical modelling workflows involving e.g. Python scripts and notebooks. Further, because defining the system is subjective, the implementation should support users in clearly documenting this input information. 3 https://ukfires.org Raw Data PRObs TBox
Pre-processing
Ontology Conversion
Clean Data
OFN Ontology
External Ontologies
PRObs Ontology
Queries Conversion
Reasoning
Answers
Specifically, our system is composed of a frontend interface for defining
and documenting the system definitions as input RDF data, and the backend
implementation based on RDFox [
To make the use of the PRObs system accessible to domain experts we adopt a literate programming approach to produce code (RDF) and documentation (HTML) from a single source.
This allows for full documentation-writing features within Python executable notebooks. They are easy-to-use, interactive, and widely used by scientists in various fields; therefore, they are the ideal platform for a system that aims to support the needs of domain experts. To streamline use in a MEFA analysis, we have developed Python wrappers that manage the process of setting up the RDFox scripting language commands to load the relevant datasets, and running RDFox as part of a wider workflow to answer queries accessing the relevant information which form the input to subsequent modelling and analysis steps. The PRObs system runs RDFox scripts to load the input data and answer queries, supported by Python utilities to embed this within a testing or analysis workflow. The input data consists of system definitions in RDF, with external datasets provided in the form of tabular data files and mapping scripts which are read during processing by RDFox. 4
To illustrate the use of the system, we describe a case study of mapping flows
through the UK production system. This case study will form the main part of
our demo. In particular, we will demonstrate several use cases within this case
study displaying particular intricacies of the PRObs system. A working example
of the PRObs system containing this case study and the material that will be
used for our demo is available online at [
UK production system This case study forms part of the ongoing research within the UK FIRES programme, motivated by seeking opportunities for innovation in manufacturing processes. The goal is to obtain a detailed understanding of how supply chains are dependent on diferent manufacturing processes, and where scrap is currently arising within the system, in order to quantify the benefits of innovation in diferent areas . To this end, a MEFA model is used to define the structure of supply chains and estimate the pattern of flows through the system which best matches the available measurements. The role of the PRObs ontology described here is to provide access to data from a diverse set of external datasets in a coherent structure aligned to the required inputs of the optimisation model.
Since there is no standard system definition of UK manufacturing supply
chains at the level of technical detail required for this analysis, a major element
of the project is to describe a suitable set of Processes and Objects to which the
available data can be mapped, and which describe entities of relevance to the
study’s research questions. These are defined and documented using the system
described in [
Example use cases will be demonstrated through the Jupyter Book7, with the opportunity to demonstrate live queries. During our demo we are going to show the full pipeline of the PRObs system, from loading the source data to the reasoning and querying phases. The demo will encompass sample datasets loaded and mapped into the ontology, as well as use cases which include: retrieving data points with their system context, retrieving inferred observations at aggregated object level, and identifying provenance of observations. Concrete details about the challenges of our scenario and the lessons learned during this work will be provided as well.
Specifically, we will introduce the data sources mentioned above, and we will describe loading and retrieval details for some objects in the BGS dataset using the PRObs system. Then, we will show why retrieving the original data points in not enough in this scenario and how we can derive new information about the observations thanks to the modelling and reasoning capabilities of our system. We will focus on (i) the aggregation of observations, useful when we want to derive all possible measurements of an object that can be classified into further components; (ii) the propagation of observations, useful when we want to analyse equivalent objects; and (iii) the provenance of observations, useful for explainability and quality of information. Finally, we will present more complex real-world examples and explain why this inferred knowledge is useful for domain experts and for further analyses. 4 Statistics on the production of goods and materials within the EU (Eurostat).
https://ec.europa.eu/eurostat/web/prodcom 5 Statistics on international trade (United Nations). https://comtrade.un.org 6 British Geological Survey Minerals Yearbook : data tables for production of various minerals. https://www2.bgs.ac.uk/mineralsuk/download/ukmy/UKMY2015.pdf 7 https://jupyterbook.org