Search

Spiking neural networks (SNNs), the models inspired by the mechanisms of real neurons in the brain, transmit and represent information by employing discrete action potentials or spikes. The sparse, asynchronous properties of information processing make SNNs highly energy efficient, leading to SNNs being promising solutions for implementing neural networks in neuromorphic devices. However, the nondifferentiable nature of SNN neurons makes it a challenge to train them. The current training methods of SNNs that are based on error backpropagation (BP) and precisely designing surrogate gradient are difficult to implement and biologically implausible, hindering the implementation of SNNs on neuromorphic devices. Thus, it is important to train SNNs with a method that is both physically implementatable and biologically plausible. In this paper, we propose using augmented direct feedback alignment (aDFA), a gradient-free approach based on random projection, to train SNNs. This method requires only partial information of the forward process during training, so it is easy to implement and biologically plausible. We systematically demonstrate the feasibility of the proposed aDFA-SNNs scheme, propose its effective working range, and analyze its well-performing settings by employing genetic algorithm. We also analyze the impact of crucial features of SNNs on the scheme, thus demonstrating its superiority and stability over BP and conventional direct feedback alignment. Our scheme can achieve competitive performance without accurate prior knowledge about the utilized system, thus providing a valuable reference for physically training SNNs., Comment: 20 pages, 8 figures, 2 tables

The quality of power grid equipment forms the material foundation for the safety of the large power grid. Ensuring the quality of equipment entering the grid is a core task in material management. Currently, the inspection of incoming materials involves the generation of sampling plans, sampling, sealing, sample delivery, and testing. Due to the lack of a comprehensive control system and effective control measures, it is not possible to trace the execution process of business operations afterward. This inability to trace hampers the investigation of testing issues and risk control, as it lacks effective data support. Additionally, a significant amount of original record information for key parameters in the testing process, which is based on sampling operation standards, has not been effectively utilized. To address these issues, we conduct researches on key monitoring technologies in the typical material inspection process based on the Internet of Things (IoT) and analyze the key parameters in inspection results. For purpose of complete the above tasks, this paper investigates the use of Long Short-Term Memory (LSTM) algorithms for quality prediction in material equipment based on key inspection parameters. In summary, this paper aims to provide professional and reliable quality data support for various business processes within the company, including material procurement, engineering construction, and equipment operation.

The quality of power grid equipment forms the material foundation for the safety of the large power grid. Ensuring the quality of equipment entering the grid is a core task in material management. Currently, the inspection of incoming materials involves the generation of sampling plans, sampling, sealing, sample delivery, and testing. Due to the lack of a comprehensive control system and effective control measures, it is not possible to trace the execution process of business operations afterward. This inability to trace hampers the investigation of testing issues and risk control, as it lacks effective data support. Additionally, a significant amount of original record information for key parameters in the testing process, which is based on sampling operation standards, has not been effectively utilized. To address these issues, we conduct researches on key monitoring technologies in the typical material inspection process based on the Internet of Things (IoT) and analyze the key parameters in inspection results. For purpose of complete the above tasks, this paper investigates the use of Long Short-Term Memory (LSTM) algorithms for quality prediction in material equipment based on key inspection parameters. In summary, this paper aims to provide professional and reliable quality data support for various business processes within the company, including material procurement, engineering construction, and equipment operation.

As one of the key equipment in the distribution system, the distribution transformer directly affects the reliability of the user power supply. The probability of accidents occurring in the operation of transformer equipment is high, so it has become a focus of material inspection in recent years. However, the large amount of raw data from sample testing is not being used effectively. Given the above problems, this paper aims to mine the relationship between the unqualified distribution transformer inspection items by using the association rule algorithm based on the distribution transformer inspection data collected from 2017 to 2021 and sorting out the key inspection items. At the same time, the unqualified judgment basis of the relevant items is given, and the internal relationship between the inspection items is clarified to a certain extent. Furthermore, based on material and equipment inspection reports, correlations between failed inspection items, and expert knowledge, the knowledge graph of material equipment inspection management is constructed in the graph database Neo4j. The experimental results show that the FP-Growth method performs significantly better than the Apriori method and can accurately assess the relationship between failed distribution transformer inspection items. Finally, the knowledge graph network is visualized to provide a systematic knowledge base for material inspection, which is convenient for knowledge query and management. This method can provide a scientific guidance program for operation and maintenance personnel to do equipment maintenance and also offers a reference for the state evaluation of other high-voltage equipment.

Accurate perception of the combustor thermal environment is crucial for thermal protection design of a rotating detonation combustor (RDC). In this study, a comprehensive analysis method is established to calculate the non-uniform heat flux distribution of the RDC by utilizing the measured temperature distribution of the combustor outer wall obtained by the high-speed infrared thermal imager. Firstly, in order to determine the inverse heat flux solving method, a physical model based on the geometric characteristics of the RDC is constructed and its thermal conductive process is simulated, given by different heat flux boundary conditions. Then the wall heat fluxes are inversely calculated by the Levenberg-Marquardt (L-M) method based on the above numerical data. Results show that the L-M method can obtain more accurate heat flux distribution even in the zones with large heat flux gradients, considering the axial heat conduction within the combustor outer wall caused by the non-uniform heat flux. Finally, the wall heat flux distribution is analyzed coupling the L-M method together with the experimental measurements in kerosene two-phase RDC. The analyses show that the highest temperature of combustor outer surface and the highest wall heat flux occurs within the region of 20mm from the combustor head, which corresponds to nearly 14% of the combustor length. With the increase of axial distance, the heat flux is rapidly reduced, and then the heat flux distribution is more uniform at the downstream region of the combustor. The heat flux peak and thermal heat rate are positively correlated with the combustor equivalence ratio in the range between 0.44 and 0.64., Comment: 26 pages, 21 figures