DEEP LEARNING IN NUCLEAR INDUSTRY: A SURVEY
Abstract : As a high-tech strategic emerging comprehensive industry, the nuclear industry is committed to the research, production, and processing of nuclear fuel, as well as the development and utilization of nuclear energy. Nowadays, the nuclear industry has made remarkable progress in the application fields of nuclear weapons, nuclear power, nuclear medical treatment, radiation processing, and so on. With the development of artificial intelligence and the proposal of “Industry 4.0”, more and more artificial intelligence technologies are introduced into the nuclear industry chain to improve production efficiency, reduce operation cost, improve operation safety, and realize risk avoidance. As a high-tech strategic emerging comprehensive industry, the nuclear industry is committed to the research, production, and processing of nuclear fuel, as well as the development and utilization of nuclear energy. Nowadays, the nuclear industry has made remarkable progress in the application fields of nuclear weapons, nuclear power, nuclear medical treatment, radiation processing, and so on. With the development of artificial intelligence and the proposal of “Industry 4.0”, more and more artificial intelligence technologies are introduced into the nuclear industry chain to improve production efficiency, reduce operation cost, improve operation safety, and realize risk avoidance
? At the same time, more and more deep neural networks used to analyze different data types (e.g., image, video, text, speech, and multi-modal data) have been proposed with the deepening of deep learning research
? With the expansion of problem scale, more and more variants of CNNs
(e.g., VGGNet, GoogLeNet, ResNet, DenseNet) have been proposed and successfully applied to object detection, semantic segmentation, image caption, and other complex CV application scenarios[36, 75, 76]. Applying CNNs-based models in the nuclear industry are able to quickly and effectively analyze the graphics, images, and videos, which can strengthen the automatic operation monitoring of key equipment and improve the safety of operation
? It is worth mentioning that Tsinghua University defines the “industrial intelligence” as the use of AI and other theories and methods to solve the technical problems of operation, management, research and development, production, and service in the process of industrial manufacturing .
? The low efficiency and subjective problem of feature engineering in traditional ML methods, as well as the generalization dilemma of the single static model are solved well by DL methods[
? so as to improve the security of management, as well as solve the problems of multi-level access control and industrial robot authentication.
? In the nuclear equipment manufacturing of midstream segment, DL methods can be used to analyze and process the massive structured, unstructured, and semi-structured data generated by design, production, and operation of the nuclear island, conventional island, auxiliary system, and instrument control equipment, so as to provide intelligent analysis and decision-making system . Moreover, in nuclear equipment manufacturing, high labor costs and low labor productivity exist side by side, and the working and operation environments of equipment production are difficult to control.
? The simulation of NPPs in the personnel training plan must reproduce the situation faced by operators in the actual operation. One purpose of the simulation is to support the design of new control systems or ergonomic evaluation of existing control systems, so as to improve man-machine interaction and operation safety
? In the field of nuclear medicine, nuclear technology can be used to diagnose and treat diseases by the common imagining equipment
? The United States, Germany, China, and other countries have used AI technologies to realize industrial intelligence, in order to improve the competitiveness of national industry through industrial intelligence, so as to take the lead in the new round of the industrial revolution, i.e., the “Industry 4.0
? . As an important technical means of AI, DL methods have also been widely used in the nuclear industry with the upsurge of industrial intelligence concept recently
? Furthermore, DL methods based on multi-objective optimization can be used to optimize the parameters of the processing operation for improving the yield and quality of nuclear fuel elements, as well as the manufacturing efficiency
|