Laboratory of neural nets and profound studying of MIPT was one of the participants of the program Global Academic Partnership of Laboratories Facebook Artificial Intelligence Research Laboratory (FAIR), engaged in research in the field of artificial intelligence. The purpose of the affiliate program of Facebook Company is to accelerate the development of science, focus on big, unsolved problems.
Currently, we can see a huge variety of studies being carried out in the field of artificial intelligence, and we believe that providing talented research teams with the necessary technologies will accelerate progress and innovation in the industry, - the Director of FAIR Yang Li Kun says.
Along with MIPT, 13 academic and research institutions from seven different countries, including France, Germany, Great Britain, engaged in the creation of innovations in the field of artificial intelligence and machine studying, qualified the selection. Facebook has long been using artificial intelligence systems, so there is DeepText AI technology, which enables to define the meaning of a post or comment. One of its possible applications in the future is blocking of offensive messages. In general, artificial intelligence has a huge potential in implementation, from improving the quality of machine translation to stock exchange speculation.
As a member of the program of the global academic partnership, MIPT laboratory will have at its disposal high-performance computers and modern servers, and its employees, in their turn, will publish the articles, developed algorithms and results obtained in the course of research in public domains.
Our laboratory focuses on the areas, in which the possibility of making breakthrough discoveries is critically dependent on the speed of processing large amounts of data.
The new equipment will enable us to conduct experiments more rapidly, thats why, the process from the start of the study to obtaining basic scientific results, and further implementation in the market products will take less time. As a result, this will lead to the fact that the new possibilities of artificial intelligence will soon become available in the daily life of every person, - the head of the Laboratory of neural nets and profound studying of MIPT, Mikhail Burtsev, commented on the prospects of cooperation with the Facebook.
Machine studying with support is another direction of research of MIPT laboratory. This studying differs from the neural net studying with a teacher in the fact that there are no examples of solving the problem, there is only an estimate of the success degree of the algorithm action. A program or a robot, controlled by this algorithm, tries different strategies to get the desired (positive support) or to avoid the punishment (negative support). As a result, people will understand intelligent systems better and the letter ones will help the mankind more efficiently, besides, these systems will be more independent, be able to plan their actions and decisions.
Meanwhile, the research team of NRTU MISA, led by Professor Alexander Karabutov developed a unique national device for laser-ultrasonic diagnosis of materials capable of detecting the tiniest internal defects up to hundredths of a millimeter.
The development of techniques leads to the increase of mechanical and thermal stress on the most critical elements of the bearing structures, which requires use of new strategic materials such as composites and coatings additives. They are unique in strength and lightness, but structurally they are heterogeneous, so the development of a critical destruction in them occurs more rapidly than in traditional metals.
For example, the smallest pores and defects in parts of aircraft mechanisms, when exposed to all the variables of stress in the air, may become the main, crack, that is, comparable to the size of the functional parts of the aircraft, and cause a crash. To prevent damage, the most significant components of the liner must pass a three-dimensional examination of the internal structure, which detects defective centers at an early stage.
The scientific team of the new NRTU MISA laboratory, created in 2015 under the program of improving the competitiveness, under the guidance of the doctor of physical and mathematical sciences, professor Alexander Karabutov, introduced a new technology of laser-ultrasound structuroscopy with unique opportunities for 3D-control of materials, parts and products of aviation technology, which doesnt have analogues in the world, - the Rector NRTU MISA, Alevtina Chernikov, said.
Laser-ultrasonic structurescope developed in NRTU MISA, has a high spatial resolution (50 microns), while maintaining a large probing depth (up to 40 mm), and enables to characterize defects in their mechanical characteristics and shape, actually getting 3D-model of the part.
Unlike traditional ultrasound, laser-ultrasound in structuroscopy uses the laser to create probing pulses, and UWB piezoelectric hydrophones for recording the echoes. This enables to increase the spatial resolution of ultrasound monitoring by 6-10 times, as well as its sensitivity, to obtain quantitative information on the state of the material - elastic modules, porosity, stress states, to distinguish between hard and soft heterogeneity. Measurements can be automated, which improves the accuracy of control, - the project manager, Professor Alexander Karabutov comments.
Ultrasonic beam generated by laser pulse has a number of unique characteristics that are virtually impossible to implement by traditional means, that is, ultrashort pulses, maintaining a clear waveform, a narrow ultrasonic beam without lateral interferences. This makes it possible to carry out control of parts, when other methods are inefficient, for example, to detect and distinguish between each successive layer of composite material layering.
The experimental prototype of the developed automated system of nondestructive control has been successfully tested for the control of carbon fiber wing boxes of a new Russian short and medium passenger liner, including the ones being stress tested. The first innovative laser ultrasonic fault detector set of is already running in the small-scale production and ready for use in the domestic aerospace industry.