The accuracy of the algorithm is 92%
![New Neuraleta will help scientists to develop heavy-duty materials 3516_1](/userfiles/22/3516_1.webp)
A group of Russian researchers has developed a new neural network capable of selecting metals and alloys to create heavy-duty materials and structures. This became known from the press service of the Perm National Research Polytechnic University (PNIPU).
![New Neuraleta will help scientists to develop heavy-duty materials 3516_2](/userfiles/22/3516_2.webp)
It is known that for the selection of the optimal combination of metals and alloys, scientists had to conduct a series of experiments to measure their qualities. The authors of the new study decided to simplify the search for durable materials, creating a special neural network, analyzing digital images of samples to determine promising types of materials.
![New Neuraleta will help scientists to develop heavy-duty materials 3516_3](/userfiles/22/3516_3.webp)
The algorithm can recognize the properties of materials, relating each of them to one of the hardness classes. In the work of the neuralition, real, and non-synthesized data, which makes it possible to ensure the depth of the technology. The accuracy of the results of the analysis of the neural network is 92.1%. Experts also noted that a special study made it possible to determine the number of erroneously marked pictures of potential materials capable of affecting the accuracy of the result.
The authors of the new development intend to continue work on its improvement. In the future, they intend to add new criteria by which the neural network could select promising metals and alloys to create heavy-duty materials and products.
Earlier, the Central News Service reported to overcome the gap between quantum simulators and quantum computers.