PNIPU: Technology for automating the creation of maps of material cutting at production facilities

Main article: Neural networks (neural networks)

2024: Presentation of the algorithm for creating cutting maps in industries

The development of scientists from the Perm Polytechnic University automates the creation of cutting maps at production facilities. The university announced this on July 16, 2024.

source = PNIPU

In many areas of activity, saving resources is an important problem and necessity. This is especially true for industry. The reduction of production waste leads to a decrease in the cost of the final product, and, therefore, to an increase in profit. To make parts from raw materials, it is often necessary to cut from metal sheet material into free-form blanks, for example, in mechanical engineering, the automotive industry, in metal processing plants and other industries. To save money, it is necessary to rationally place parts on the canvas. Such a process is known as the creation of a material cut map. The use of expensive raw materials imposes even more stringent requirements on the quality and methods of solving the problem. PNIPU scientists have developed a combined algorithm for optimal arrangement of figures on a sheet based on artificial neural network technology. This method minimizes production waste when cutting out parts, which will save raw materials and increase production efficiency.

Certificate No. 2023661038 has been issued for development. The article was published in the VSU Bulletin. System Analysis and Information Technologies "Nº1, 2024. The study was carried out with the financial support of the Russian Science Foundation, project No. 19-07-00895.

There are two types of figurative cut: regular and irregular. In the first, all geometric objects have the same shape and orientation. Solving the second type problem is almost impossible by accurate methods due to the arbitrariness of the forms and, therefore, the large amount of input data.

The main problem is that some methods lead to impractical time costs due to overhauling of objects, others are limited to only one of the best options, and not the best (local results that do not correspond to global).

Scientists of the Perm Polytechnic University have developed an algorithm for solving the problem of two-dimensional (flat) irregular cutting of material using neural network technology.

In the first step, the operator sets the size and shape of the sheet. Next, it makes a selection from standard shapes such as a circle, square, triangle, etc., and indicates their dimensions. If you want to decompose non-standard shaped parts, you can specify them using coordinates or by uploading them from the database. Then the training starts, and then the program performs the cutting itself.

To improve efficiency, the algorithm is supplemented by gravitational compaction of the cutting map - random forces act on the figures inside a physically simulated medium. With such an impact on objects, they are located more tightly: by analogy with the real world, if you shake a closed box with various objects, then they will be located in the most optimal way for themselves. If used, the filling result improves to 22%.

The algorithm was based on the idea of ​ ​ modeling the learning process, by analogy with the real world, when a person learns a skill from scratch. The neural network is faced with the goal of learning to find an acceptable solution to the cutting-packaging problem. In order to reduce the load and reduce the convergence time, an environment is designed according to real physical laws. It completely excludes cases of mutual intersection of figures and their going beyond the boundaries of the region of cutting, - said Sergei Zykin, senior lecturer at the Department of Technical Disciplines of the Lysven branch of PNIPU.

The work of the neural network takes place in episodes. An episode is one attempt to get a solution to a problem. The neural network studied for 30,000 such episodes, while their number can increase depending on the number of figures in the open.

The development of polytechnics is implemented in the form of a program. With it, in production, you can automatically sketch workpieces with contours of cut parts and get recommendations for rational placement of complex geometric objects on a sheet. As of July 2024, the program is being tested at both city-forming and small enterprises that use cut material in their production.