With all the data acquired during a day of production, the possibilities for improvement are endless. All you need to do is find a way of interpreting the data. TOPAZ aims to investigate how a combination of data mining, machine learning and optimization can aid manufacturing companies. This will be done by transforming heterogeneous data derived from multiple sources, and potentially from different part of the value chain, into prescriptive actions that support long-term goals of profitability, sustainability and stability.
Manufacturing companies collect vast quantities of data at various levels of production. Such data comes in different formats and is stored in different types of database systems, which makes analysis difficult. TOPAZ adopts a popular analytics framework that defines different analytics stages in order to transform data into decisions.
In order to transform data into decisions, TOPAZ adopts a popular framework to define different stages of analytics:
- descriptive analytics
- predictive analytics
- prescriptive analytics.
In the context of TOPAZ
Descriptive analytics refers to the use of data integration methods to convert data into information; predictive analytics refers to the use of structured data mining and machine learning techniques to convert information into knowledge; and prescriptive analytics refers to the use of simulation and optimization to convert knowledge into wisdom. Thus, the research approach in TOPAZ contextualizes the Data-Information-Knowledge-Wisdom (DIKW) hierarchy in terms of an analytics framework.
Data mining and machine learning techniques
TOPAZ will focus on structured data, i.e. data with predefined formats, contained in relational or non-relational databases. A variety of data-models are used in manufacturing industries for managing complex data concerning the product, process, flow, logistics, etc. In this project, structured data mining and machine learning techniques will be applied to multi-source, multi-format heterogeneous data to predict how various scenarios captured by the data will evolve. These predictive models will then be used to improve the scenarios by optimizing the corresponding control parameters, thus leading to performance gains and cost savings.
Project partner companies
TOPAZ will establish new collaborations between the University of Skövde and two major Swedish manufacturing companies. The project partners will not only benefit from each other’s case studies, but will also gain exposure to ongoing work in the INFINIT research environment.
Project contribution to research and education
The project will contribute to research and education activities within INFINIT in the areas of Virtual Production Development and Data Science. Primarily, it will complement the methods and algorithms being developed in the KKS-funded Profile2017 project Virtual Factories with Knowledge-Driven Optimization virtualfactories.se. In education, TOPAZ will lead to multiple student thesis projects and contribute to the development of existing advanced level courses in the Master’s programs and new freestanding courses for industries through the Knowledge Foundation funded project Expertkompetens.