TY - JOUR AU - Szűcs, Balázs TI - Data Integration Framework to Collect Data from OT/IT Systems JF - ACTA TECHNICA JAURINENSIS J2 - ACTA TECH JAURIN PY - 2023 SN - 1789-6932 UR - https://m2.mtmt.hu/api/publication/33668415 ID - 33668415 AB - Industry 4.0 and industrial data processing, due to its inherent possibilities, is gaining more and more emphasis in production companies these days. In a corporate environment, the age of equipment is extremely heterogeneous, in addition to state-of-the-art equipment, legacy systems can also be found in the machine park, which do not have appropriate communication protocols. Also, with the increase in the number of data sources, the management of data is becoming more and more challenging. Not only the operational technology, but also the connection of different IT systems and the extraction of data pose challenges. The different data processing use-cases using partly or entirely the same data sources, so it is necessary to extract and transmit the data to the target systems in a standard way, and avoiding an increase in the number of point-to-point interfaces. In this work we present a possible framework, to solve the above mentioned problems in industrial environment, with the introduction of standardized naming conventions, OT/IT gateways, data integration and distribution layers. LA - English DB - MTMT ER - TY - BOOK AU - Szűcs, Balázs TI - Machine learning based optimization of tool replacement strategy in machine tools PY - 2021 UR - https://m2.mtmt.hu/api/publication/33668428 ID - 33668428 LA - English DB - MTMT ER - TY - JOUR AU - Szűcs, Balázs AU - Ballagi, Áron TI - Autoencoder based flexible industrial supervision system for process- and quality monitoring JF - ACTA CYBERNETICA J2 - ACTA CYBERN-SZEGED PY - 2021 SN - 0324-721X UR - https://m2.mtmt.hu/api/publication/31795590 ID - 31795590 AB - In the industrial process- and quality control systems the measurement data like press-in, fitting, screwing and other curves are often supervised with window functions. One drawback of these window functions are that they are cannot handle the deviation of the measurement data, thus cannot adjust dynamically or to as small as possible the window size and detect low magnitude anomalies inside the window function. The other drawback of the above mentioned technique is that it cannot detect anomalies outside the predefined window functions. In this paper we present a neural network based method for the monitoring of measurement data. The new method, in contrast with the classical approaches, which are using envelope test or window functions, the autoencoder based approach is capable to detect unexpected events and anomalies, which are cannot be pre-programmed. By applying the above mentioned method, a higher level of quality assurance can be achieved. LA - English DB - MTMT ER - TY - BOOK AU - Szűcs, Balázs AU - Ballagi, Áron TI - Challenges of the application of machine learning in the serial production PY - 2021 UR - https://m2.mtmt.hu/api/publication/31795573 ID - 31795573 AB - The industrial applications of the machine learning methods have promising results in the field of smart manufacturing and quality assurance; however, the development of these models have challenges. In this paper we present the classical steps of model creation and the challenges and questions of industrial applications in each step, especially for series production. Finally, we present a possible workflow for model development in the manufacturing. LA - English DB - MTMT ER - TY - CONF AU - Szűcs, Balázs AU - Ballagi, Áron TI - An Industrial Application of Autoencoders for Force-Displacement Measurement Monitoring T2 - The 12th Conference of PhD Students in Computer Science PB - Szegedi Tudományegyetem (SZTE) C1 - Szeged PY - 2020 SP - 28 UR - https://m2.mtmt.hu/api/publication/31795561 ID - 31795561 AB - The applications of artificial intelligence and neural networks in the industrial process monitoring and supervision are on the rise. One potential use case of these technologies are the anomaly detection in processes and measurements, without the need of pre-programming well defined patterns and supervision functions, thus unexpected events can be detected dynamically. In this paper we present a novel, neural network based method for the monitoring of press-in and joining processes. The new method, in contrast with the classical approaches, which are using envelope test or window functions, the autoencoder based approach is capable to detect unexpected events and anomalies, which are cannot be pre-programmed. By applying the above mentioned method, a higher level of quality assurance can be achieved. We present the new method through the example of force-displacement monitoring of mounting a sealing ring. LA - English DB - MTMT ER - TY - JOUR AU - Szűcs, Balázs AU - Dr. Ballagi, Áron TI - Artificial Intelligence in Maintenance: The Industrial Application of Natural Language Processing JF - BÁNKI KÖZLEMÉNYEK J2 - BÁNKI KÖZLEMÉNYEK PY - 2020 IS - ESB 2019 Konferencia Közlemény SN - 2560-2810 UR - https://m2.mtmt.hu/api/publication/31128205 ID - 31128205 AB - Text- and natural language processing improved a lot in the recent years. Applications like translators, chatbots and virtual assistants made the everyday life easier, but the industrial use of the technology and its potential in the maintenance planning remained unexploited. In this paper we present two possible ways how to utilize these algorithms to achieve their positive benefits. For humans it’s often impossible to handle historical maintenance data and error logs due to their enormous amount or because the descriptions of errors are subjective. With the analysis of error messages, maintenance and shift logs, the correlations between failures and events can be detected, thus the effectiveness of the maintenance planning and the interventions can be increased. LA - English DB - MTMT ER - TY - JOUR AU - Szűcs, Balázs AU - Ballagi, Áron TI - Reducing pseudo-error rate of industrial machine vision systems with machine learning methods JF - ACTA TECHNICA JAURINENSIS J2 - ACTA TECH JAURIN VL - 12 PY - 2019 IS - 4 SP - 294 EP - 305 PG - 12 SN - 1789-6932 UR - https://m2.mtmt.hu/api/publication/30687828 ID - 30687828 AB - Nowadays machine learning and artificial neural networks are hot topic. These methods gains more and more ground in everyday life. In addition to everyday usage, an increasing emphasis is placed on industrial use. In the field of research and development, materials science, robotics and thanks to the spread of Industry 4.0 and digitalization, more and more machine learning based systems are being introduced in production. This paper gives examples of possible ways of using machine learning algorithms in manufacturing, as well as reducing pseudo-error rate of machine vision quality control systems. Even the simplest algorithms and models can be very effective on real-world problems. With the usage of convolution neural networks the pseudo-error rate of the examined system can be reduced by 83 percent. LA - English DB - MTMT ER - TY - JOUR AU - Szűcs, Balázs TI - Gépi tanulás alkalmazási lehetőségei hibadetektálásra belsőégésű motorok összeszerelésénél JF - BÁNKI KÖZLEMÉNYEK J2 - BÁNKI KÖZLEMÉNYEK PY - 2019 PG - 7 SN - 2560-2810 UR - https://m2.mtmt.hu/api/publication/30393053 ID - 30393053 AB - A gépi tanulás manapság egyre nagyobb teret hódít a hétköznapi életben. Az egészségügyi alkalmazásokon át, gépi látásban, ajánló rendszerekben, különböző virtuális asszisztensekben, beszédfelismerő, beszédszintetizáló, fordító alkalmazásokban, valamint a leghétköznapibb dolgokban is találkozhatunk gépi tanulás és mesterséges intelligencia algoritmusokkal. A mindennapi alkalmazásokon túl egyre nagyobb hangsúly kerül az ipari felhasználásra. A kutatás-fejlesztésben, az anyagtudományban, robotikában, illetve az Ipar 4.0 terjedésének és a digitalizációnak köszönhetően a gyártásban is egyre több gépi tanulásra épülő rendszer kerül bevezetésre. A különböző érzékelők, mérő átalakítók és mérőberendezések által előállított nagy adathalmaz kiváló kiindulási alapot biztosít a berendezések és termékjellemzők vizsgálatára, nem ismert összefüggések feltárásra. A tanulmányban példákon keresztül bemutatásra kerülnek a gépi tanulás algoritmusok lehetséges felhasználási módjai gyártásban, valamint egy hibadetektálási eljárás, melynek következtében a szerelősori kihozatal növelhető, illetve a kritikus szerelési hibák elkerülhetőek. A belsőégésű motorok főtengelyeinek átforgatási nyomaték méréseit gépi tanulás algoritmussal elemezve nem ismert összefüggések kerültek feltárásra, melyek alapján a szerelési folyamat korai szakaszában detektálható az idegen anyag a csapágyakon és a főtengelycsapokon, így elkerülhetőek az utómunka miatt feleslegessé váló további műveletek. LA - Hungarian DB - MTMT ER -