Generating Images with Physics-Based Rendering for an Industrial Object Detection Task: Realism versus Domain Randomization

Limited training data is one of the biggest challenges in the industrial application of deep learning. Generating synthetic training images is a promising solution in computer vision; however, minimizing the domain gap between synthetic and real-world images remains a problem. Therefore, based on a real-world application, we explored the generation of images with physics-based rendering for an industrial object detection task. Setting up the render engine’s environment requires a lot of choices and parameters. One fundamental question is whether to apply the concept of domain randomization or use domain knowledge to try and achieve photorealism. To answer this question, we compared different strategies for setting up lighting, background, object texture, additional foreground objects and bounding box computation in a data-centric approach. We compared the resulting average precision from generated images with different levels of realism and variability. In conclusion, we found that domain randomization is a viable strategy for the detection of industrial objects. However, domain knowledge can be used for object-related aspects to improve detection performance. Based on our results, we provide guidelines and an open-source tool for the generation of synthetic images for new industrial applications.

Control, computing and communication in industrial cyberphysical systems with feedback

The cyber-physical production base technologies are control, computing and connection being applied altogether in the industrial object technological processes automatic regulation systems. They use control technologies in multi-loops and multi-channel regulation systems forming a hierarchy structure. Automatics functional elements unite in the intermediary regulation scheme stabilizing hierarchy control objects, which states are detected with sensors. Computing technologies are used in the cyber-physical production imitation virtual environment and provide processes simulation based on control models and cyber-physical systems digital twins structured in hierarchy levels. The regulation accuracy increase is provided with physical and model processes results comparison detecting non-definition factors acting production processes accuracy. The communication technology is used for the cyber-physical systems net information exchange given with the Internet of Things parameters. The cyber-physical systems continuous work in the non-ideal communication net regulation interval is provided with preventive control signals compensating operation and information delay of automatic conveyor lines. There is a scheme given of multi-loop and multi-channel automatic cyber-physical production using control objects hierarchy based on cyber-physical systems hierarchy and item manufacturing technological tasks.

PRINCIPLES OF REVITALIZATION OF WORKED OUT QUARRIES, BASED ON EXAMPLE OF KRYVYJ RIH

Problem statement. A quarry is classified as an architectural object that has its own structure. As each industrial facility has an expiration date, the question of its restoration and inclusion in the planning structure of the city increasingly arises. Quarries that are not used for their intended purpose pose a real threat to the dispersal of housing and public buildings in the city, complicating public relations. Although mining is necessary for the development of the country's economy, the development of deposits has a negative impact on the environment. Firstly − the change of natural relief, different in depth and other parameters, and secondly − not the same type of design solution during the revitalization of the worked quarry. The presence of analysis and principles of reorganization of space and functional purpose in the quarries can improve the affected areas of the city. Purpose − development of scientifically substantiated proposals and recommendations of the principles of revitalization of quarries as spent industrial facilities in the urban environment in accordance with the existing functional content of the city in order to create an ecologically balanced and comfortable human environment with a positive impact on the urban ecosystem. Conclusions. The study of the worked quarry not as a mining object for mining, but as an engineering complex industrial object, which in turn is able to meet the main groups of human processes, expands the criteria for analysis and improvement of the revitalization of the territory. Restoration of these facilities is aimed at improving the architectural, environmental and emotional state of the city. Using the necessary principles for a single point of integration (career), the urban area becomes more balanced in terms of functional content.

ON THE METHODICS OF MATHEMATICAL MODELLING OF POLLUTION OF HYDROLITHOSPHERE BY PETROLEUM PRODUCTS (EXEMPLIFIED BY KURSK OIL TERMINAL)

A potential threat of pollution of hydrolithosphere with petroleum and its products due to accidental and technological leaks on a certain industrial object has been analyzed. The author's method of mathematical modelling is described that allows to perform an adequate dynamic modelling of a petroleum products lens in an aquifer. Execution of the resulting model allows to forecast behavior of petroleum products in the conditions of a real geological and hydro-geological cross-section.

3D Convolutional Neural Networks Initialized from Pretrained 2D Convolutional Neural Networks for Classification of Industrial Parts

Deep learning methods have been successfully applied to image processing, mainly using 2D vision sensors. Recently, the rise of depth cameras and other similar 3D sensors has opened the field for new perception techniques. Nevertheless, 3D convolutional neural networks perform slightly worse than other 3D deep learning methods, and even worse than their 2D version. In this paper, we propose to improve 3D deep learning results by transferring the pretrained weights learned in 2D networks to their corresponding 3D version. Using an industrial object recognition context, we have analyzed different combinations of 3D convolutional networks (VGG16, ResNet, Inception ResNet, and EfficientNet), comparing the recognition accuracy. The highest accuracy is obtained with EfficientNetB0 using extrusion with an accuracy of 0.9217, which gives comparable results to state-of-the art methods. We also observed that the transfer approach enabled to improve the accuracy of the Inception ResNet 3D version up to 18% with respect to the score of the 3D approach alone.

On the methodics of mathematical modelling of pollution of hydrolithosphere by petroleum products: Exemplified by Kursk oil terminal

A potential threat of pollution of hydrolithosphere with petroleum and its products due to accidental and technological leaks on a certain industrial object has been analyzed. The author's method of mathematical modelling is described that allows to perform an adequate dynamic modelling of a petroleum products lens in an aquifer. Execution of the resulting model allows to forecast behavior of petroleum products in the conditions of a real geological and hydro-geological cross-section.

THE EXPERIENCE IN USING LASER SCANNING AND BUILDING INFORMATION MODELING FOR ENGINEERING DATA MANAGEMENT DURING THE LIFE CYCLE OF AN INDUSTRIAL OBJECT

An industrial object accumulates a great deal of information about its assets throughout its whole development period. This information is provided in the different drawings, passports, regulations, and other technical documentation. One of the common problems of most industrial objects is the disor-ganized storage of technical documentation on assets and its limited access to different industrial ser-vices. This greatly complicates the retrieval of information about the assets to ensure the steady opera-tion of the industrial object. As a consequence, one of the ongoing important tasks becomes the crea-tion of a unified source of up-to-date information about the object’s assets and the facilitation of the access to that data for all the participants of the project, construction, and operation process. Exactly these issues are tackled in the article alongside with the solutions based on using BIM and terrestrial laser scanning. This article also describes the types of BIM and detailed differences between them, the methods to form a BIM, as well as how the methods change at different stages of the life cycle. As well, the typology of tasks for which BIM solutions are applicable. TLS technology is described as a source of initial data for the formation of BIM. This article describes the results of the combined use of BIM and TLS at the stages of design, construction, and operation of an industrial object based on the implemented project. The article provides the result of clash detection in design documentation. The result of clash detection between designed and existing structures is also given. The article also provides the deviations at the construction stage of industrial objects, which were discovered. The ac-quired results demonstrated the effectiveness of using terrestrial laser scanning and BIM in engineering solutions.

Variation in leaf morphological traits of Populus tremula L. under the conditions of technogenic load

The article presents the study of leaf morphological features of Populus tremula L. Growing under the conditions of the influence emissions from Burshtynska TPP. The methods of linear morphometry were used to analyze the variability of the morphological parameters of leaves at high levels of pollutants. In the zone of influence of the industrial object there is a significant decrease in the foliar parameters. Uneven variability of morphological parameters between the study plots of industrial pollution and with minimized technogenic load are revealed. Levels of variability of morphological parameters are estimated using coefficient of variation, which for most of the analyzed morphological features at the polluted plot load is characterized by high and average values. The most varied ones were the distance between the first and second veins and the length of the second vein from the base of the leaf vein. For bioindication of anthropogenic load, an index of fluctuating asymmetry of P. tremul leaf blades has been tested. The analysis of the fluctuating asymmetry was based on leaf linear measurements and area of ​​ the halves of leaf blades. The results indicated an increase in fluctuating asymmetry index in plant growing at the polluted plot, both in linear measurements and in area indices. The comparison of the two approaches to the estimation of the level of fluctuating asymmetry has been carried out. As a result of this study, those foliar morphometric parameters of P. tremula, which exhibit high morphological reactivity under the influence of anthropogenic load, can be used as bioindicatory markers for assessing the state of the environment. A reliable indicator of technological pressure is the level of fluctuating asymmetry that can be recommended for use in bioindicative studies.