On July 7,2022, under the guidance of the Center for Science Communication and Achievement Transformation of the National Natural Science Foundation of China, and in accordance with the relevant provisions of the Measures for the Evaluation of Science and Technology of the Ministry of Science and Technology, the third-party evaluation organization of sci-tech results ZKHC follows standards and procedures for the evaluation of sci-tech achievements and adopts the principles of science, independence, objectivity and impartiality, has organized experts to evaluate three sci-tech results jointly completed by Jiangxi KMAX Industrial Co., Ltd., Tsinghua University, Hunan University, Jilin University, Beijing University of Posts and Telecommunications. These three sci-tech results are “Ray Vision multimodal 3D scene modeling system”, “dynamic health benchmark digital twin drive rail vehicle prediction operational key technology”, and “super figure high-order correlation learning digital twin drive rail transit vehicle energy saving driving decision and control”.

The expert panel consists of the Distinguished Young Scholar & Professor Yin Xucheng of Beijing University of Science and Technology, Associate Researcher Jia Leipo of the Committee of National Natural Science Foundation of China, Professor-level Senior Engineer Wen Zhiyong of CRRC Corporation Limited, Professor Xiao Zhenyu of Beihang University, Professor Wang Weida of Beijing Institute of Technology, Professor Chen Wenbo of Beijing Information Science and Technology University, Professor Ma Nan of Beijing University of Technology, and Associate Professor Wang Hongwei of Beijing Jiaotong University.

The experts believed that “Ray Vision multimodal 3D scene modeling system” has the following innovation points: It invents the integrated 3D modeling method of geometry-structure joint learning, develops the digital twin acquisition equipment of Ray Vision, realizes the integrated reconstruction of multimodal data such as vision, lidar and inertial navigation. On the basis of ensuring the accuracy, the project quickly reconstructs the complete visual environment through the integrated framework; invents a structure-guided cross-modal point-cloud data completion method; through structural modeling and registration between 2D images and 3D point clouds, it solves the problem of insufficient robustness of traditional registration methods, realizes the 3D visual data fusion and enhancement; invents a multimodal data fusion retrieval method based on hypergraph calculation, models the complex higher-order correlations of the multimodal visual objects, visual object feature representation via a hypergraph neural network, and realizes efficient retrieval of the content of the digital twin model of rail transit.

The project results have fully independent IPR, which stands at the world-leading level. The multi-modal data fusion and retrieval technology based on hypergraph calculation has reached the international leading level.

The project “dynamic health benchmark digital twin drive rail vehicle prediction operational key technology” has the following innovation points: it invents the construction method of equipment health baseline jointly driven by field knowledge and data, uses the device-principle guided data-driven technology to construct the knowledge graph, improves the efficiency and accuracy of knowledge base construction; invents the digital twin driven fault prediction method based on health baseline; utilizes the technical route of refining the digital twin reference line reflecting the status and trend of the equipment based on the normal operation data; and then conducts abnormal behavior monitoring and fault mode learning under the guidance of the reference line. As a result, the project realizes effective prediction of track equipment failure; invents the fault diagnosis method based on strongly generalized feature extraction; through the self-supervised and unsupervised learning means, it captures the system dynamics as the basis for fault detection, thus improving the robustness of the fault diagnosis.

The project results have fully independent IPR, which stands at the world-leading level. The digital twin drive fault prediction technology based on the health reference line has reached the international leading level.

The project “super figure high-order correlation learning digital twin drive rail transit vehicle energy saving driving decision and control” has the following innovation points: It invents the analytical modeling method of hierarchical constraint driven digital twin model dynamics, realizes the efficient analytical modeling of vehicle many-body system dynamics under ideal constraints and non-ideal constraints; it presents an energy-saving driving target curve planning method based on hypergraph high-order correlation learning, learning speed curve pattern characteristic rules by using excellent driver driving experience data; realizes the efficient calculation and planning of energy-saving driving; it invents the train digital twin model based on analytical binding force and fuzzy set theory; and realizes the precise following and optimal control of trajectory targets under uncertain cognitive conditions.

The project results have fully independent IPR, which stands at the world-leading level. The energy-saving driving target curve planning technology of digital twin drive train based on hypermap high-order correlation learning has reached the international leading level.

‍

Based on the field of industrial digital twin and facing the needs of digital and information development of rail transit industry, the project has formed a number of innovative achievements, which has been applied to the construction of digital twin system of rail transit, thus producing wonderful financial and social benefits and having broad application prospects.

After the expert panel made comprehensive evaluation, they reached consensus that the three projects completed by Jiangxi KMAX Industrial Co., Ltd., Tsinghua University, Hunan University and others have passed the evaluation of sci-tech achievements.