Teaching - I have been involved in the Digit-T Project, an online platform and associated e-book aimed for training industrialists in Digital Manufacturing concepts. I developed the content for… read more
Anne McLaren Fellow - Adaptable Artificial Intelligece for the Manufacturing of the Future
Vision: To create AI solutions that do not require any programming or advanced data science knowledge to be created or maintained, making the technology accessible to everyone. This will ultimately widen the uptake of AI in the manufacturing industry and other sectors of the economy.
Introduction: The digitalisation of manufacturing is a key enabler in the UK Government's drive to raise the level of industrial productivity that is needed to remain competitive and resilient in a global market. As part of this digitalisation, Artificial Intelligence promises to be a game changer at every level of the value chain. Across all sectors, AI is estimated to help achieve 79% of the Sustainable Development Goals of the United Nation's Sustainability Agenda2. In manufacturing, it is estimated that the greatest value will be created using Machine Learning (ML), a sub-set of AI techniques, for predictive maintenance, which is estimated to be $0.5 Trillion dollars across the world's businesses3. Although the UK is considered a global power in AI research, the COVID-19 pandemic and the withdrawal from the European Union have shown the need for UK's industry to increase its adoption of AI. This has not happened in part due to off-the-shelf ML approaches not being ready for dealing with the dynamic nature of manufacturing systems, combined with the limited data availability that comes with legacy systems and a high mix of technologies. Simultaneously there is a skills gap and lack of confidence on the use of the emerging technologies. The UK has a National AI Strategy1 that aims at increasing the technology uptake through the upskilling of the workforce, fostering more data availability, and providing support to areas with low adoption. Although this is an essential course of action, there should be more supporting activities in addition to this strategy to increase the uptake of AI.
In this fellowship I propose to tackle these challenges through a different perspective. I propose to reduce the skill needed to use AI by developing the first mechanisms inspired on the mammalian brain that will underpin the realisation of Industrial "Plug and Play" AI. New dynamic models will continuously learn (Continual AI) and adapt to changing environments and to data variability and availability with no human AI specialist knowledge needed. These mechanisms will push the boundaries of AI and will enable the accessibility and scalability needed to reach the efficiency levels that the new Sustainability Agenda demands.
Research Fellow in Data Science (Institute for Advanced Manufacturing (IfAM), University of Nottingham)
During my time as post-doctoral researcher, I have been involved in a variety of projects and activities that have strengthened my experience in the application of Artificial Intelligence (AI) and ML in manufacturing,
Automated Feature Extraction and Deep Learning - Through the EU Horizon 2020-funded MC-Suite project, I developed a new approach for tool condition monitoring using image-encoded sensor data and deep learning that enabled in-process real-time condition monitoring with little manual data preparation.
Data Analytics Industrial Collaboration - I have led a project with BMW, where I developed a data analytics and visualisation tool to monitor machine performance based on quality data. The tool was deployed at the Birmingham Hams Hall Plant for most of 2020 and a reduction of 97% of produced scrap by the monitored machines compared to the previous year was achieved.
Automated Data Visualisation and SME collaboration - I contributed to the EPSRC Digital Manufacturing on a Shoestring project together with the University of Cambridge on the development of low-cost digital solutions targeted for SMEs. My work has focused on the implementation of modular data capture, data visualisation and data analysis "building blocks" that are used to develop and deploy solutions in the factory using a no code approach, meaning solutions are generated in a drag-and-drop manner without the need of any programming. Through this project I have worked closely with manufacturing SMEs developing 2 pilots and 3 demonstrators. This has helped me develop an understanding of the current challenges of manufacturing, particularly in SMEs given their levels of digitalisation and their perspectives on the benefits of digital technologies such as ML.
Semantic Modelling - In addition to my experience in Machine Learning applied to manufacturing, I currently work on the implementation of semantic models for the capture of manufacturing systems capabilities as a researcher co-investigator of the EPSRC Elastic Manufacturing Systems project. These models allow understanding of whether an existing production system has the capability as well as the capacity to manufacture a new product or a new volume of an existing product supporting decision making and enabling responsiveness and resilience of a manufacturing system. This work will be continued and extended through the EPSRC Research Centre for Connected Factories project where I am currently researcher co-investigator.
AHMADIEH KHANESAR, MOJTABA, BANSAL, RIDHI, MARTINEZ-ARELLANO, GIOVANNA and BRANSON, DAVID T., 2020. XOR Binary Gravitational Search Algorithm with Repository: Industry 4.0 Applications APPLIED SCIENCES-BASEL. 10(18),
MARTINEZ-ARELLANO, GIOVANNA, TERRAZAS, GERMAN and RATCHEV, SVETAN, 2019. Tool wear classification using time series imaging and deep learning INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY. 104(9-12), 3647-3662
MARTINEZ-ARELLANO, GIOVANNA and RATCHEV, SVETAN, 2019. TOWARDS AN ACTIVE LEARNING APPROACH TO TOOL CONDITION MONITORING WITH BAYESIAN DEEP LEARNING PROCEEDINGS OF THE 33RD INTERNATIONAL ECMS CONFERENCE ON MODELLING AND SIMULATION (ECMS 2019). 33(1), 223-229
TERRAZAS, GERMAN, MARTINEZ-ARELLANO, GIOVANNA, BENARDOS, PANORIOS and RATCHEV, SVETAN, 2018. Online Tool Wear Classification during Dry Machining Using Real Time Cutting Force Measurements and a CNN Approach JOURNAL OF MANUFACTURING AND MATERIALS PROCESSING. 2(4),
Teaching - I have been involved in the Digit-T Project, an online platform and associated e-book aimed for training industrialists in Digital Manufacturing concepts. I developed the content for several online sessions that are now being used as part of the Digital Manufacturing module in the Manufacturing Engineering BEng degree course (MMME4111).
As a post-doctoral researcher, I've had opportunities to deliver seminars, lectures an invited talks in digital manufacturing, data mining for manufacturing and machine learning as part of the Digital Manufacturing Module MMME4111, and events such as the UK Symposium on Knowledge Discovery and Data Mining 2021.
During my position as a Lecturer in Computing at the Nottingham Trent University, I lead modules on topics involving Software Design and Implementation, Systems Analysis and Design, Python Programming ( BSc program ) and Machine Learning Applications (MSc program).