Marmara University (MarUn)
MInD-NET is a research lab founded at the Computer Engineering department, Marmara University. MInD-NET conducts research in two main areas, which are "Networking/Internet of Things" and "Artificial Intelligence/Machine Learning", and is in strong collaboration with both industry and academia to propose innovative and promising solutions for real-world problems.
MarUn, MinD-NET’s capabilities mainly consist of (1) Wireless Networks and Internet of Things (IoT) including V2X communications, (2) Machine Learning (ML) / Artificial Intelligence (AI). MInD-NET has participated various EU and national funded research projects. MInD-NET works on in different layers of the communication protocol stack and optimize solutions for wired/wireless communication technologies. MInD-NET has competences in communication and IoT related solutions as well as modelling and simulation to provide efficient and reliable data communication for operational area considering requirements and challenges in industrial domains. In V2X communication, MInD-NET is the leading research organization in the national side providing innovative solutions. MInD-NET also follows Edge and Cloud Computing paradigms to generate real-time response and big data analysis according to standards, and infrastructure. MInD-NET builds applications and software tools for reliable data transfer, connectivity monitoring and network performance measurement.
In terms of Machine Learning (ML) / Artificial Intelligence (AI) research area, MInD-NET models, devises, implements and analyses complex and intelligent systems employing not only existing state-of-the-art algorithms but also novel ones of comparable or outperforming performance in the fields of pattern analysis (PA) and machine intelligence (MI). MInD-NET pursues intensive research on modelling, analysis and identification of multi-variate, sequential dynamic systems and make, in their research, extensive use, in particular, of clustering and dimensionality reduction, neural networks, Markov models both with a specific or variable order, learning automata and, whenever necessary, evolutionary algorithms to mention a few in PA and MI.
In the scope of iRel4.0, MInD-NET’s roles are in three folds; (1) designing and building a data chain for reliable and efficient data transfer with low latency requirements to be processed by AI/ML based solutions at the edge and cloud, (2) developing AI/ML based algorithms to improve reliability in manufacturing, (3) leading the dissemination, exploitation and standardization activities in iRel40. Data and its acquisition methods are the limiting factors of the learning-based cognition and prediction methods for reliability. Efficient data chain and infrastructure will be built considering the several constraints (such as data-rate/bandwidth, latency, confidentiality/privacy, security, data complexity and incompatibility) and data communication methods will be optimized to meet the requirements. Multiple sensor data in the UC test-bed will be synchronized via time-stamping to ensure relevance and correlation.
Transferred data will be processed at edge/cloud for AI/ML intensified reliability. Learning behaviour patterns of a dynamic system from collected sensor data is a key to achieve Industry 4.0 related concepts like anomaly detection, condition monitoring and predictive maintenance. With the help of Artificial Intelligence (AI) and Machine Learning (ML) based methods, system modes can be represented with appropriate mathematical models that are constructed automatically and directly from the data. These algorithms will be realized on the production line of Arcelik, the leader of IP9. Also, IP9's current fault detection system will be improved in many aspects. Thus, the fault detection rate of this production line is aimed to increase, so the system will be more reliable.
- Designing and building the data chain and infrastructure for efficient, reliable and low-latency data transfer/acquisition
- A new test environment will be designed and built that merge/integrates multiple test-beds into one test-bed with additional features
- New and additional sensors and devices will be deployed to the test-bed for collecting and efficiently transferring data to analyse and apply developed AI/ML based algorithms
- Data transfer&communication methods will be optimized to meet requirements such as reliability, efficiency and low-latency.
- Multiple sensor data having timestamps will be synchronized to a time reference ensuring relevance and correlation
- Essential contribution is AI/ML based algorithm development and implementation.
- With the help of Markov models, important characteristics of the system behaviour can be captured more efficiently.
- On the production line, the state-of-the-art even novel mode detection methods will be used, such as syntactic pattern recognition.
- Leading the dissemination, exploitation and standardization activities within the project consortium and Pan-Europe.