Use Cases


CNC machines are one of the most commonly used machines in the metal component manufacturing industry’s production line. These machines include numerical control units and mechanical components and during production processes, fluid is used to form the produce components with relatively low power consumption and produce components with dimensional accuracy. Since these machines produce very precious components with tight tolerances for various industries including heating, cooling, automotive, molding and die industries, the repeatability and sensitivity of such machines is extremely important. FL based AI models fed by distributed sensing system data will detect anomalies such as Remaining Useful Lifetime (RUL) of CNC machines and excessive machine heating of the related machine and will ensure uninterrupted 24/7 production. In this application, TORUN and INO will be main contributors for this pilot application.


Nowadays, technology providers are using many different types of systems to monitor passenger flows at airports. However, using events of existant systems cost way much less than new-tech driven tracking solutions. During the use case development processes, to be able to track airport passenger’s movements, TAV Technologies will gather data from three different systems such as; Common Use Passenger Processing System (CUPPS), Travel Document Authorization System (TDAS), and Airport Operation Database System (AODB). To be able to understand use case architecture, it is required to understand the functionality of sub-systems. CUPPS is the last-generation passenger processing system used for passenger check-in and boarding processes as the primary data source. Airlines are operating in the same airport use CUPPS peripherals like workstations, boarding pass printers, and readers jointly. In other words, airlines install their Departure Control System (DCS) software on the workstations which belong to the airport and are connected to the CUPPS then perform passenger processes like printing bag tag, boarding, or reading them. Due to this work method, CUPPS can access some of the passenger data which are flowing through on itself. On the other hand, like CUPPS, the TDAS system is used by police officers to read boarding passes when passengers pass through border control. AODB, as the last data source of the airport scenario within the project’s scope, is an airport software that lists daily, weekly, seasonal flight plans, counter and gate allocations, and flight-related milestones.

When the data gathered from the three systems mentioned above are analyzed together in real-time or for a specific reporting period, many key performance indexes related to airport operation efficiency can be created. AI-based predictions have great importance to ensure efficient management of many intertwined operational processes with a multistakeholder structure. To determine the number of ground service personnel needed the next day in different parts of the airport and different type of operations, it is necessary to predict how many passengers will be in which region in which time zone the next day. Likewise, predictions regarding the flow of passengers within the airport are needed so that the Terminal Operator can purchase electricity at a lower cost with correct consumption estimations. Rental fees defined for the shops at the airport, the toilet cleaning period, the number of security guards in the regions, and the advertisements prepared for the passengers also need the flow predictions of the passengers within the airport. All the data to be collected in order to make the predictions of passenger’s destination airport, destination city, destination country, destination region, flight category (int, dom), airline, flight number, GH company, flight date & time, weekday status, season status, origin holiday status, destination holiday, passenger types should be examined in sub-breakdowns. At this point, the protection of passengers and different stakeholders, the Big Data infrastructure, and network needs required for centralized AI necessitate the federated learning architecture for the solutions like Pax Analyzer.


Mobile devices are not limited to cell phones. In addition to cell phones, there are many other task-specific devices such as tablets, kiosks, and digital signage. These devices often have sensors such as cameras and microphones that can capture data from their environment. For example, speech and video data of customers interacting with kiosks in a retail store can be captured and analyzed with these sensors, and the results can be used to determine customer satisfaction. Furthermore, the federated learning platform to be developed within the project is planned to be used to capture and analyze data from kiosks in different retail stores with these sensors.

‘Clients’ in different retail stores, to which the kiosks are connected, use their existing data to train the predictive neural network in the background with the data from the kiosks and calculate new values of the neural network parameters. The calculated values are sent to the federated learning ‘server’ to calculate the total neural network parameters in the next step. The parameter values from other clients on the server are combined and the new values are distributed to the clients for use in the next step. As a result, FL-based AI models fed with data collected from kiosk sensors in different retail stores mentioned above are analyzed in real-time and can be used to determine customers’ mood satisfaction.

Cognitive AI development for customer tendency analysis requires a lot of annotated data about human faces, which cannot fit into the budgets of innovation projects nor be realized through crowdsourcing activities. Therefore, a systematic collaborative data management and model development system is required in order to improve pre-trained cognitive models in a non-stop fashion. The Turkish company KOCSISTEM and INO will be the main contributors and work together for this application.

#4 (SmartCore & HUFS & DLIT)

A systematic collaborative management system is required when manufacturing products through mutual collaboration in a distributed smart factory environment, or when parts produced in a plurality of different product manufacturing factories are assembled to create finished products. In this case, comprehensive manufacturing management of parts to create finished products and comprehensive quality management of related parts should be performed. In particular, for parts procurement management for general assembly production and overall quality control of related parts, all data related to related parts and finished products must be comprehensively analyzed. Therefore, this Pilot Application requires “distributed and collaborative production” and for small and medium-sized manufacturing-production companies pursuing intelligence, it is to implement a “blockchain-basedfederated learning AI platform” that supports the establishment of an AI-based distributed collaborative production system, and protects and manages therequired technologies and related data of small and medium-sized manufacturing-producing companies. This application will be implemented through acollaboration of SmartCore, HUFS and DLIT.


Sidónios Malhas, SA is a textile company with near 40 years of history. It is specialized in the production of textile meshes in circular looms. At the F4itech, it will be lending a subset of their equipments and provide production and equipment condition data so that federated learning platform can be developed and prototyped. This federated learning platform should help the company in improving energy efficiency of its operations, reduce maintenance costs through predictive maintenance and improve trust downstream the value chain by using blockchain technology. Sidónios will be closely assisted by Sistrade and ISEP, national partners in the project, and contribute to the international project results exploitation.


SAMM Teknoloji is a Turkish company that was founded in 2003. The federated learning platform, which will be developed within the scope of the project, is planned to be used in the distributed fiber optic acoustic sensors (DAS) developed by our company. In this context, a pilot application will be made on distributed fiber optic acoustic sensors installed in Gebze Organized Industrial Zone, Informatics Valley Technology Development Zone and Camlica TV Tower regions in Turkey. Different application areas cause different threat factors. In addition, acoustic data with different characteristics are obtained from each system. With the F4iTECH federated learning system, it will be possible to improve system modeling with data collected from sensors used for different security purposes. SAMM and INO will be main contributors and work together for this application. With the federated learning platform, the distributed fiber optic acoustic sensor system will be improved and solutions that can make artificial intelligence-based alarm classification for critical areas such as border security, military base security, infrastructure security, airport security will be developed.