As the proportion of composite materials used in aircraft continues to increase, the electromagnetic Shielding Effectiveness (SE) of these materials becomes a critical factor in the electromagnetic safety design of ai...As the proportion of composite materials used in aircraft continues to increase, the electromagnetic Shielding Effectiveness (SE) of these materials becomes a critical factor in the electromagnetic safety design of aircraft structures. The assessment of electromagnetic SE for Slotted Composite Structures(SCSs) is particularly challenging due to their complex geometries and there remains a lack of suitable models for accurately predicting the SE performance of these intricate configurations. To address this issue, this paper introduces SCS-Net, a Deep Neural Network (DNN) method designed to accurately predict the SE of SCS. This method considers the impacts of various structural parameters, material properties and incident wave parameters on the SE of SCSs. In order to better model the SCS, an improved Nicolson-Ross-Weir (NRW) method is introduced in this paper to provide an equivalent flat structure for the SCS and to calculate the electromagnetic parameters of the equivalent structure. Additionally, the prediction of SE via DNNs is limited by insufficient test data, which hinders support for large-sample training. To address the issue of limited measured data, this paper develops a Measurement-Computation Fusion (MCF) dataset construction method. The predictions based on the simulation results show that the proposed method maintains an error of less than 0.07 dB within the 8–10 GHz frequency range. Furthermore, a new loss function based on the weighted L1-norm is established to improve the prediction accuracy for these parameters. Compared with traditional loss functions, the new loss function reduces the maximum prediction error for equivalent electromagnetic parameters by 47%. This method significantly improves the prediction accuracy of SCS-Net for measured data, with a maximum improvement of 23.88%. These findings demonstrate that the proposed method enables precise SE prediction and design for composite structures while reducing the number of test samples needed.展开更多
基金supported by the National Natural Science Foundation of China(Nos.62101020 and 62141405)the Special Scientific Research Project of Civil Aircraft,China(No.MJZ5-2N22).
文摘As the proportion of composite materials used in aircraft continues to increase, the electromagnetic Shielding Effectiveness (SE) of these materials becomes a critical factor in the electromagnetic safety design of aircraft structures. The assessment of electromagnetic SE for Slotted Composite Structures(SCSs) is particularly challenging due to their complex geometries and there remains a lack of suitable models for accurately predicting the SE performance of these intricate configurations. To address this issue, this paper introduces SCS-Net, a Deep Neural Network (DNN) method designed to accurately predict the SE of SCS. This method considers the impacts of various structural parameters, material properties and incident wave parameters on the SE of SCSs. In order to better model the SCS, an improved Nicolson-Ross-Weir (NRW) method is introduced in this paper to provide an equivalent flat structure for the SCS and to calculate the electromagnetic parameters of the equivalent structure. Additionally, the prediction of SE via DNNs is limited by insufficient test data, which hinders support for large-sample training. To address the issue of limited measured data, this paper develops a Measurement-Computation Fusion (MCF) dataset construction method. The predictions based on the simulation results show that the proposed method maintains an error of less than 0.07 dB within the 8–10 GHz frequency range. Furthermore, a new loss function based on the weighted L1-norm is established to improve the prediction accuracy for these parameters. Compared with traditional loss functions, the new loss function reduces the maximum prediction error for equivalent electromagnetic parameters by 47%. This method significantly improves the prediction accuracy of SCS-Net for measured data, with a maximum improvement of 23.88%. These findings demonstrate that the proposed method enables precise SE prediction and design for composite structures while reducing the number of test samples needed.
