Ontology and rule-based reasoning for intelligent manufacturing digital twin

Background:

In the transformed industrial systems as marked by the industry 4.0 and 5.0 era, cyber physical systems, in integration with intelligent analytics, can define a framework of big data integration to capitalize the knowledge from multiple sensor installations and historical data logs in order to make predictive intelligent decisions. For such systems researchers have defined communications protocols such as the MTConnect which can enable data acquisition of multiple sensors simultaneously. The big data utilisation requires the streams of data to be processed and analysed through predictive analytics in real-time. Given the extent of the information researchers have explored deep on the capability of algorithms to spot trends in the data and make sense of the big data in order to further optimise the manufacturing process. Such data mining techniques can be used in various processes across the supply chain from where extensive data is being collected at every instant from deployed sensors and machines, such as dirty data from machines for reliability centred maintenance, predictive manufacturing for maximal optimisation and reduction in production times, etc. As part of this system, manufacturing intelligence is a type of software that uses all available manufacturing data for analysis, prediction of future system states and the creation of graphic summaries. Manufacturing intelligence is used to support manufacturing decisions. Leveraging manufacturing intelligence enables manufacturers to query their manufacturing sites and perform activities like reporting, modelling, experimental design, system alert, recommendation, extrapolation, prediction, optimisation or simulation.

Aim:

The aim of this project is to develop, evaluate and demonstrate an intelligent predictive model for manufacturing purposes. The project will address the need for knowledge extraction and pattern recognition from manufacturing data which is enabled by machine learning and data mining techniques. The research will make use of data from a real workshop case study.

The predictive model will emerge from the successful completion of the following objectives:

Objectives :

Extract and identify relevant manufacturing data within the manufacturing shop floor domain

Cluster and build relationships among key semantics

Identify the quantifiable properties of the key performance indicators (KPI) and key semantics

Build a manufacturing ontology to align the raw SCADA data with the shop-floor key performance indicators

Define the data manipulation rules for calculation of properties from SCADA

Develop an interface between Protégé and manufacturing database in order to populate the ontology in an automatic and efficient way

Build the manufacturing digital twin for creating the initial data conditions for production prediction and scheduling