The paper presents a methodology for training neural networks for vision tasks on synthe-sized data on the example of steel defect recognition in automated production control systems. The article describes the process of dataset procedural generation of steel slab defects with a symmetrical distribution. The results of training two neural networks Unet and Xception on a generated data grid and testing them on real data are presented. The performance of these neural networks was assessed using real data from the Severstal: Steel Defect Detection set. In both cases, the neural networks showed good results in the classification and segmentation of surface defects of steel workpieces in the image. Dice score on synthetic data reaches 0.62, and accuracy—0.81.

Improving the tools and mathematical methods to diversify the portfolio of oil and gas assets in the face of limited investment, high market volatility, increasing risks and uncertainties at the current level of technology development is a very urgent task. In order to form an effective investment portfolio, the authors proposed asset diversification using cluster analysis, which implies grouping sample objects according to a set of specific features. The method under consideration involves five stages of asset valuation in order to consider those assets in a three-dimensional space, taking into account the specifics of the oil and gas business, including determination of individual asset trajectory, performing spatial approximation, calculating the clustering coefficient, ranking the resulting pairs, and directly solving the portfolio formation optimization problem. This paper provides a reasonable set of metrics for diversification of the investment portfolio based on cluster analysis: main criterion - ∆PV, additional criteria - ∆Production, ∆OPEX/toe, ∆CAPEX/toe, characterized geological, environmental, social and economic aspects. Thus, the proposed methodology provides an opportunity to identify the most attractive investment projects, thereby allowing large oil and gas companies to diversify their business with minimal risk and maximum return on invested capital.

The effect of high-temperature thermomechanical treatment (HTTT) on abrasive wear resistance of steel 30KHGGSA as the main material to manufacture holder of mining machine cutters is studied. The procedure of HTTT during deformation of cylindrical steel fragments at the preset settling-down rate and subsequent quenching is described. The finite element modeling in DEFORM-3 finds distribution of plastic strain in volume of forged pieces. On the plant based on the upright drilling machine, the abrasive wear resistance of forged pieces is tested on sandstone and siltstone as the medium-hardness and abrasivity co-rocks most frequently met in development heading. It is shown that pre-deformation of the fragments before quenching, i.e. HTTT, imporves the steel hardness from 44 to 55 HRC. It is shown that as against specimens nondeformed before quenching, the abrasive wear resistance of steel subjected to plastic straining up to εΣ ≈ 0.7 increases by 1.9 times in case of both siltstone and sandstone. Considering the obtained results, it is suggested to amend the manufacturing technology of cutters by quenching of fholder forging starting immediately from the forging temperature and by subjecting thereby the holder to high-temperature thermomechanical treatment. The cutting tool cartridge is to be held in the holder by cold pressing. It is assumed that HTTT application can essesntially (two times and more) improve wear resistance of steel holder head and, as a consequence, extend service life of cutters.