Abstract:This study aimed to explore an anti-eavesdropping technique based on time-frequency feature design, focusing on how to dynamically modify the temporal and frequency aspects to enhance human speech interference within specific frequency ranges. The paper conducted research on existing speech interference techniques, comparing them with standard noise injection methods. The research methods included theoretical analysis and experimental validation. By testing and evaluating the interference signals based on time-frequency feature extraction on an actual prototype, the effectiveness of the interference in disrupting speech recognition systems was assessed.The experimental results showed that when the signal-to-noise ratio (SNR) was lower than 0 dB, the proposed method′s text recognition error rate (WER) was over 60%. Moreover, when the SNR was 0 dB, the WER of the algorithm in this paper was higher than that of current jamming algorithms by more than 20% on average. Additionally, when the jamming system maintained the same distance from the recording device, the SNR produced by this paper′s algorithm on the recording device was lower than that of the current algorithm by almost 2 dB. This demonstrates the high energy utilization efficiency of the proposed algorithm.Therefore, the findings of this research have significant implications for improving communication security and protecting privacy, especially in environments that require a high level of confidentiality.

Abstract:Aiming at the current market mainstream weldment surface defects model detection accuracy is not high, the model is complex and does not meet the real-time monitoring and other issues, a new detection model of weldment surface defects based on the improvement of YOLOV7-tiny obtained KThin-YOLOV7 is proposed.Firstly, the EMA-BasicRFBC module, which is based on simulating the human visual sensory field, was designed to replace the spatial feature pyramid SPP module of the YOLOV7-tiny model, so as to enhance the performance of the model feature expression.Secondly, the ThinNeck structure is designed based on the SlimNeck design paradigm structure, and it is used to replace the NECK feature fusion part of YOLOV7-tiny, which reduces the number of parameters and computation of the model and improves the average detection accuracy of the model at the same time. Finally, the K-means++algorithm is introduced to find out the appropriate anchor frame and replace the LOSS of the original model with the FEIOU loss function, which further helps the model to optimize the position and size of the target frame.The mAP of the KThin-YOLOV7 is improved by 7.11% to 87.64% compared to the original YOLOV7-tiny model, while the number of parameters and computation of the model are decreased by 11.14% and 15.5%, respectively. decreased by 11.14% and 15.26%, respectively. Experimental results show that KThin-YOLOV7 can efficiently and accurately locate and detect defects on the surface of welded parts.

Abstract:Network management is an essential foundational function in modern automobiles. With the rise of intelligent connected vehicles, the increasing number of Electronic Control Units has led to higher demands and complexities in in-vehicle network systems. To ensure the security and stability of network management, comprehensive testing solutions are required. This study proposes a testing method specifically for AUTOSAR CAN network management, based on a deep understanding of the AUTOSAR CAN network management specification. Firstly, a testing system is constructed using the standard TTCN-3 scripts provided by AUTOSAR. The system is designed to generate variable-length test messages and conduct multi-scenario testing of CAN network management. Secondly, a new syntax format is defined based on the Backus-Naur Form to impose time constraints on TTCN-3, thereby enhancing the accuracy of monitoring CAN network messages. Using this method, three abnormal messages were identified in practical tests on the MPC5745B platform, enabling the evaluation of the security and stability of AUTOSAR CAN network management. Experimental results demonstrate that this method enables comprehensive testing of the consistency of AUTOSAR CAN network management while conserving testing software and hardware resources.

Abstract:To cope with the increasingly complex communication environment requirements of modern communication systems and fully use limited spectrum resources, this paper proposes a design that uses a meta-surface structure combined with active devices PIN diodes and varactor diodes to achieve dual polarization functional frequency independence—reconfigurable multi-modal features. Different from the traditional multi-mode structure, this design uses PIN diodes to control the transmission and reflection states of dual-polarized waves independently and uses varactor diodes to achieve dynamic frequency modulation of the dual-polarized wave transmission window, enabling switching of 9 working modes. In the static state, this design can independently control the transmission and reflection of TE and TM polarized waves; switching to the dynamic state and regulating the polarized waves adds an independent tuning TE or TM polarized wave transmission window offset. function. Each mode is independent of each other and has good mode compatibility. In addition, this design shows good angular stability in the incident angle range of 0°~45°. Experimental verification using the electromagnetic simulation software CST proves the reliability of this design and shows that it has broad application prospects in radomes and signal multi-frequency transmission.

Abstract:With the increasing resolution of highresolution satellite payload CCD cameras, the amount of image data they obtain has sharply increased. How to transmit payload data to backend devices for processing at high speed and reliability is a problem that must be solved. This article conducts research on the high-speed serial interface chip TLK2711 and the same source clock, and analyzes the transmission errors and other issues that may occur in the high-speed data transmission link of the onboard TLK2711. A high-peed data transmission interface design based on the low complexity CRC algorithm is proposed, and the reliability analysis of high-speed data transmission is carried out from both hardware and logic aspects. In terms of hardware, it is based on the same origin clock, providing reference clocks for FPGA and TLK2711 at the sending and receiving ends. In terms of logic, it uses the FB-SC-CRC verification method to provide technical support for data monitoring and error correction during high-speed transmission, reducing resource consumption during data transmission. Through experimental verification, the data transmission interface has achieved reliable data transmission through the use of homologous clocks, with a bit rate of up to 1 600 Mbit/s and an error rate of 0. The use of logical resources has been reduced by about 2/3 compared to traditional CRC.

Abstract:In order to meet the demand of measuring liquid pressure in narrow chamber, this paper designs a non-contact Fabry-Perot microcavity based optical liquid pressure measuring system. The optical microcavity is fabricated by two-photon 3D printing technology, with a compact size of 350 μm. The diaphragms with a thickness of 4 μm and 6 μm were designed. The system is based on combination of the embedded design and diffraction grating spectral module, which can acquire the optical interference signal, realizing high-resolution spectrum dynamic demodulation. The experiments show that the sensing sensitivity of the 4 μm-thickness device can reach 398 pm/kPa at room temperature (25 ℃), and the system shows a high resolution of 35.5 Pa and good repeatability, which has potential in biomedical intraocular pressure sensing applications.

Abstract:A wearable, integrated vehicular temperature and vibration testing system has been devised to streamline equipment setup processes during vehicle testing, thereby bolstering testing efficiency. The system′s testing nodes, governed primarily by STM32 microcontrollers, are affixed using magnetic suction mechanisms. Leveraging a multifunctional adapter developed for the purpose, synchronization and concurrent configuration of diverse testing nodes are achievable, ensuring operation within a unified temporal framework. Data interchange with the host computer is facilitated through USB flash drive connectivity, facilitating seamless retrieval of test data. Furthermore, the multifunctional adapter offers ancillary functionalities such as charging, parameter configuration, data retrieval, and storage, augmenting the overall efficacy of the equipment. Extensive experimentation encompassing laboratory validation and real-world vehicular trials attests to the system′s convenient and dependable installation modality, markedly curtailing equipment setup durations. The testing system exhibits vibration acquisition accuracy within a relative error margin of 1%, while temperature acquisition errors are confined within ±0.1 ℃. Synchronization time discrepancies among multiple devices are constrained to the microsecond level. This system adeptly captures and archives temperature and vibration data pertinent to vehicular operations, effectively catering to the exigencies of operational state parameter acquisition for vehicles.

Abstract:Aiming at the problem that the traditional Maximum power point tracking method is easy to fall into local optimal and lead to large power oscillation range in photovoltaic power generation system under local shade, an improved algorithm based on Coot chicken was proposed. The photovoltaic MPPT method of ICOOT. Based on the traditional Coot bird optimization algorithm, the algorithm introduced the logistics-sine-cosine chaotic mapping factor into the chain movement of the population follower, so that the chain movement became chaotic movement. The algorithm has the ability to jump out of the local optimal solution; The current optimal position after each optimization is modified by Cauchy, and the optimal update and replacement before and after the optimization are compared to increase the global search ability of the algorithm. Under four lighting modes, ICOOT and MPPT of the other three algorithms were simulated and analyzed. The results show that the tracking speed of the proposed algorithm is 0.14 s, 1.13 s, 0.13 s, 1.07 s, and the system stability rate is 99.43%, 99.34%, 98.73%, 98.80%. In summary, ICOOT can effectively solve the problem that traditional algorithm is easy to fall into the local maximum power point, which leads to slow optimization speed and large power oscillation when used for local concealed MPPT of photovoltaic power generation.

Abstract:The problem of reduced comfort in unmanned vessel rides due to the lack of adaptability to uncertain model parameters and external environmental disturbances is addressed. A trajectory tracking sliding mode control algorithm based on Long Short-Term Memory (LSTM) is proposed. LSTM is utilized to compensate for uncertain model parameters and external environmental disturbances, thereby mitigating the jitter phenomenon in sliding mode control. A mathematical model of an unmanned vessel is established based on a recreational boat, and a sliding mode trajectory tracking controller is designed. Additionally, an LSTM neural network is introduced to compensate for uncertainties in the unmanned vessel′s mathematical model and external environmental disturbances. Simulation tests are conducted using MATLAB/Simulink under three different trajectories. The results indicate that the LSTM-based sliding mode control algorithm achieves higher trajectory tracking accuracy compared to the sliding mode control algorithm, with a maximum reduction of 62% in average absolute trajectory error. The LSTM neural network significantly improves the unmanned vessel′s disturbance rejection capability.

Abstract:A two-stage shaft hole assembly trajectory planning strategy was studied for the assembly operation task of circuit breaker shaft hole parts in dual arm robots. In order to improve the flexibility and accuracy of coordinated assembly operations with both arms, a FAIC control algorithm was proposed to achieve accurate force tracking. Firstly, a mechanical model of the contact state between the shaft and the hole is established, and the mechanism of shaft hole assembly is studied. Based on the force balance at the end of the two arms, a dual arm motion trajectory planning method for operating space shaft hole assembly is proposed. Based on impedance control, flexible shaft hole assembly is achieved, and a fuzzy controller is used to identify the optimal impedance control parameters online, improving the quality of assembly operations. The experimental results show that under the action of the dual arm collaborative shaft hole assembly strategy combined with the FAIC algorithm, the indirect contact force of the shaft hole is controlled within 3N, and the assembly efficiency is improved by 36-5%. In summary, the FAIC algorithm and assembly strategy proposed in this article effectively shorten the assembly time of magnetic components, improve assembly efficiency and accuracy in the assembly operation of dual arm collaborative circuit breakers.

Abstract:Landslide displacement prediction is an important task in disaster prevention and mitigation. Aiming at the rationality problem of trend term and period term reconstruction after displacement decomposition as well as the problem of low accuracy of period term displacement prediction, a combined model of improved adaptive noise complete ensemble empirical modal decomposition (ICEEMDAN), sample entropy (SE), and dung beetle optimization algorithm (DBO) optimization of the long- and short-term memory network (LSTM) is presented displacement prediction is performed. Taking the Bazimen landslide as the research object, the cumulative displacement of the landslide was decomposed using the ICEEMDAN method, and the subsequence obtained from the decomposition was characterized by the sample entropy value, which was reconstructed into the trend term and the period term displacements. After that, the LSTM model is used to predict the trend term and the period term displacements. The influence factors of the period term displacement are determined by the method of gray correlation. Considering that the randomness of hyperparameters in the LSTM network affects the model prediction accuracy, the dung beetle optimization algorithm is introduced to obtain the optimal hyperparameters of the LSTM, and finally the predicted trend term and period term displacements are superimposed to obtain the cumulative displacement. The ICEEMDAN-SE-DBO-LSTM model proposed in this paper predicts the period term displacement with the RMSE, MAE and R2 of 1.803 mm, 1.584 mm and 0.988, respectively, which is better than the DBO-BP, LSTM, GRU and BP models, and proves the effectiveness of the model.

Abstract:In response to the current difficulties and high costs in segmenting radial tire defects in China, this paper proposes the following solution: a radial tire defect segmentation algorithm called Swin DAA based on Swin Transformer and attention feature pyramid. Swin Transformer is mainly used as the backbone feature extraction network, and the semantic expression ability is enhanced through the Dynamic Attenuation Attention feature pyramid, Build a software platform written in Python language, cascade the X-ray heavy-duty tire detection system to collect images, and use TCP protocol to communicate with the upper computer and transmit image data. Finally, connect the defect segmentation software system with the MES industrial control system to complete unmanned automated radial tire defect segmentation. The experimental comparison data shows that the Swin DAA network proposed in this article has an accuracy of 82.87%, a recall rate of 85.22%, and a transmission frame rate of 11 per second. The integrated software can effectively meet the actual monitoring requirements of radial tires.

Abstract:Currently, in the field of motor imagery decoding, research has focused on two approaches, subject-dependent and subject-independent decoding. However, these two decoding approaches have major limitations in the practical use of brain-computer interface (BCI) systems. Both subject-dependent and subject-independent decoding rely on the same center dataset, and when the decoding model is applied to datasets from other centers, the performance will be significantly degraded, which cannot satisfy the demand for cross-center use of BCI systems. To improve the cross-database decoding performance of motor imagery Electroencephalogram (EEG), a sparse selection model based on Fisher criterion regularization is proposed based on the methodological framework of domain generalization. Based on the least absolute shrinkage and selection operator (LASSO) model, a Fisher criterion regularization term is introduced to explicitly model feature separability during the feature selection process. This helps to improve the representation learning ability for domain generalization, thus enhancing the generalization performance of the classification model on different datasets. The effectiveness of the proposed method is validated using two publicly available motor imagery EEG datasets and two feature extraction methods, filter bank common spatial pattern (FBCSP) and multiple time frequency common spatial pattern (MTFCSP). Further validation through the utilization of self-collected data also confirmed the effectiveness of the proposed method in practical applications. Compared with existing methods, the proposed method achieved the highest average classification accuracy of 67.26%. The experimental results show that the proposed method has better generalization ability, higher feature separability, and better robustness in motion imagery cross-database decoding. The proposed method is expected to facilitate cross-center use of BCI systems and improve generalizability.

Abstract:In view of the issue of high false alarms and missed detections in traditional target detection methods under rainy clutter environments, this paper primarily investigates the joint fractal characteristics of rain clutter spectra and their application in target detection. We propose a joint fractal feature detection method based on the directional blanket covering method has been proposed. Firstly, the fractal dimension and model fitting error features of the echo′s distanceDoppler domain are measured using the blanket covering method. Subsequently, these fractal dimension and model fitting error features are employed as verification statistics to construct a threshold-based detection method with combined features. By optimizing the computational steps of the blanket method, redundant calculations on non-target information are reduced, thereby enhancing the real-time performance of the method. Based on the processing results of the measured data in rainy and cluttered environments, the method demonstrates a significant reduction in false alarms and an improved detection performance for targets compared to traditional target detection algorithms when handling non-stationary data such as rain clutter.

Abstract:Aiming at the problem of low stability and low solution accuracy when using the RANSAC algorithm to solve the essential matrix in the large-scale measurement scene of the monocular system，an improved RANSAC method for solving the essential matrix is proposed．The essential matrix obtained from the points is used to reproject the remaining matching feature points，and use the relative discriminant method to determine whether the current inlier is a high-quality inlier through the value of these errors，and then use the dichotomy method to dynamically adjust the threshold on this basis to find the optimal value from several essential matrices．Finally，this paper designs RANSAC experiments under different mis-matching rates of multiple perspectives．The experiments prove that，compared with traditional and other improved RANSAC algorithms and LMedS algorithms，The improved algorithm in this paper can quickly determine the initial interior points and adaptively adjust the threshold，and at the same time obtain a better essential matrix，which meets the requirements of solution stability and accuracy．

Abstract:In order to solve the problems of low detection accuracy, large environmental interference factors, and difficult deployment in mobile devices with average performance of the existing detection algorithms for hard hats and reflective clothing on small targets and complex weather, an improved detection algorithm for YOLOv8 safety equipment, YOLOV8-DSI, was designed and implemented. Firstly, the DR-SPPF module based on residual idea and parallel cavity convolution is designed to further expand the receptive field without loss of image resolution, and significantly improve the precision of complex weather detection. Secondly, ST-BiFPN is designed in the feature fusion stage to further reduce the number of model parameters and achieve efficient multi-scale feature fusion. Finally, Inner-ShapeIoU loss function is introduced to make bounding box regression more accurate and enhance the detection effect. Compared with the baseline model mAP50 and MAP50:95, the self-built data set increased by 2.1% and 4.7% respectively, while the model parameter number was only 2.4 M and the calculation amount was only 7.3 G, which decreased by 10.9% and 20.0% respectively. Finally, the improved model was deployed to the edge device of Jetson Orin Nano. The actual operation on the development board proved that the improved model of YOLOv8 was effective and applicable in complex scenarios.

Abstract:Aiming at the problem of pose feature loss due to up and down sampling in the inference process of head pose estimation, a high-resolution feature maintained soft-stage regression algorithm for head pose estimation is proposed. The algorithm first utilizes the encoder HR-Net to encode multiscale features for high-resolution feature maintaining in raw face images, and TA dimension interaction module joined in its convolutional block to capture more spatial-channel interaction information. The decoder SSR-Net algorithm was then applied to decode the key parameters and soft-stage regression of head pose on the different scale features output from HR-Net, and the Efficient Channel Attention ECA is employed to enhance the information interaction between feature channels and reduce redundant features. The experimental results show that the proposed algorithm has excellent performance on both the public datasets AFLW2000 and BIWI, and its MAE is reduced to 4.19 and 3.00, respectively.

Abstract:Bird nest encroachment on overhead transmission lines can cause safety hazards to the power equipment on the towers, which may indirectly affect the stable operation of the whole power system. Aiming at the current overhead transmission line bird′s nest detection model in the complex scene as well as the small target scene detection accuracy is not high, the detection efficiency is low, the model is complex and other problems. This study proposes a lightweight overhead transmission line bird′s nest detection network based on YOLOv5s framework. Firstly, the YOLOv5s feature extraction network is reconstructed by Fasternet in the backbone part, which reduces the model complexity and improves the operation speed; then the ConvMixer layer is embedded in the feature fusion network part, and the structural design of the ConvMixer layer helps to better capture the relationship between space and channel in the feature information, which improves the model′s detection ability for small targets; finally, the Finally, the ODConv module is introduced in the feature fusion network part, so that the feature map sent to the detection head contains more effective features to improve the detection performance of the model for complex scenes and small targets. The experimental results show that compared with the baseline model YOLOv5s, the computational amount and model volume are reduced by 86% and 72%, the average accuracy reaches 96.4%, and the detection speed reaches 104.2 frames/s, which verifies the effectiveness and feasibility of the improved model in this paper.

Abstract:To solve the problems of complex tea bud detection scenarios in natural environments and the large number of model parameters that cannot be deployed on embedded devices, a lightweight tea bud detection method based on YOLOv8n is proposed. We construct a lightweight backbone network, MFBNet, introducing the MBConv module to significantly reduce model computation. Simultaneously, we incorporate the CBAM attention module into the backbone network to suppress irrelevant information, thereby enhancing the model′s detection accuracy. Furthermore, the introduction of the AKConv module improves the VoVGSCSPC structure, proposing the innovative AVCStem module, which replaces the C2f module in the neck network, further reducing model parameters and enhancing the efficiency of embedded device deployment. Finally, we employ the GSConv module to replace all Conv modules in the neck network structure, facilitating fast model computation and increasing the detection speed of tender tea buds. The results indicate that the proposed model in this paper outperforms the original YOLOv8n model with a 3.5% improvement in mAP50, 55.6% increase in FPS, and a 14.3% reduction in parameters. The model demonstrates strong robustness, meeting the requirements for lightweight and rapid detection of tender tea buds in complex scenarios.

Abstract:To address the challenges in detecting small targets with dense distribution and large-scale variations in UAV aerial images, a cross-scale target detection model for UAV aerial images, named CS-YOLOv5s, is proposed. Firstly, based on YOLOv5s, micro-object detector is utilized to improve the model ability for capturing small targets. Then, the max-pooling branch is embedded into the context augment model, extracting and enhancing deep feature maps at the tail of the backbone network. The PANet is injected to achieve effective fusion of deep and shallow features with enhancing the cross-scale detection capability. Furthermore, the down-sampling convolution module is replaced with the SPDConv module to achieve efficient detection of dense objects in UAV aerial images. Experiments demonstrate that CS-YOLOv5s achieves 42.0% mAP0.5 on the VisDrone2019 dataset, which is increased by 9.8% than that of the baseline model. Our model enhances the network ability to recognize small targets in UAV aerial images effectively, which provides a new way for intelligent targets recognition of UAV.

Abstract:To overcome the limitations of the Hall sensor, a hybrid control strategy based on improved least squares method and integral sliding mode observer is proposed to achieve more accurate position detection using low-resolution signals. Firstly, the reliable start-up and smooth operation at low speeds are achieved using the improved least squares method. Secondly, upon reaching the specified switching speed, the system switches operating states, continuously correcting the integral sliding mode observer using the improved least squares method to reduce system lag and cumulative errors while making the output position signal as continuous as possible. Finally, a comparative experiment is conducted between the improved position detection strategy and the traditional first-order acceleration position detection strategy. The results show that the proposed improved position detection strategy increases the position detection accuracy by over 30% during motor start-up, reduces speed error to within 0.5% at medium to high motor speeds, and reduces position detection error to within 1%, demonstrating a higher level of position detection accuracy.

Abstract:Due to the non-stationary operating conditions and harsh working environment of wind turbines, the vibration pulse characteristics of wind turbine bearing faults are easily overwhelmed by random noise interference, which poses a challenge to accurately detect rolling bearing faults. In order to reduce the impact of random interference on subsequent feature extraction and algorithm complexity, an improved multi head self attention mechanism (IMHSA)-multi-scale convolutional network (MSCNN)-bidirectional long short-term memory network (BiLSTM) wind turbine bearing fault diagnosis method is proposed. Firstly, the IMHSA composed of periodic cavity self attention and local self attention enhances the features to reduce the impact of random interference and the time consumption during the feature enhancement process; Then, the MSCNN-BiLSTM network is used to extract spatial features and long-term dependency features from the fault signal; Finally, the fault diagnosis results of the fan bearings were output through the fully connected layer and Softmax layer, and the actual operating data of the rolling bearings on the experimental platform was used for numerical analysis. The effectiveness and superiority of the proposed method were verified by comparing it with other similar methods in the field.

Abstract:In order to obtain a complete and high-resolution image of wind turbine blade cracks, image stitching technology is used to stitch multiple high-resolution images into a complete image. Aiming at the problems of difficulty in features detection, low matching rate, and poor stitching quality in wind turbine blade crack images, an image stitching method based on AKAZE algorithm and PROSAC algorithm is proposed. The AKAZE algorithm is used to detect image feature points and generate binary feature point descriptors. Use Hamming distance as a similarity measure to perform brute force matching on feature points. The PROSAC algorithm is used to optimize the feature matching results and calculate the image transformation matrix. The fade-in and fade-out fusion algorithm is used to eliminate splicing traces and obtain a complete blade crack image. The test results show that the method in this paper can detect a rich number of feature points, the matching correct rate is above 95%, the splicing accuracy is about 0.7 pixels, and the splicing speed is improved by 17% compared with the SIFT method. The AKAZE+PROSAC method can better meet the needs of high-resolution wind turbine blade crack images stitching.

Abstract:The spacecraft power system is one of the key subsystems, and its operational status directly affects the lifespan and performance of the entire spacecraft system. Therefore, employing advanced technology for fault diagnosis of the power system to improve the reliability and safety of spacecraft in orbit has become a research focus in the field of fault diagnosis. Methods based on deep learning have advantages such as strong fitting ability and rich feature extraction, making them the mainstream approach in the field of fault diagnosis. However, in the field of fault diagnosis of spacecraft power systems, mainstream fault diagnosis methods cannot capture the long-term dependency of sequences and are limited to modeling in the time dimension, severely affecting the performance of fault diagnosis methods. Therefore, this paper proposes a method based on spatio-temporal self-attention mechanism for efficient and accurate fault diagnosis of spacecraft power system. The method adopts a Transformer-based encoder structure to extract high-dimensional features from spacecraft telemetry data, optimizes the self-attention mechanism therein, uses temporal convolution to extract temporal feature information, and employs both temporal and spatio bidirectional self-attention mechanisms to extract spatio-temporal features from the data. Finally, the features extracted by the model are mapped to obtain the fault diagnosis results of spacecraft power system. Relevant experiments are conducted on the spacecraft power system dataset. The experimental results show that compared with commonly used methods in the field of fault diagnosis, the proposed method has stronger fault representation extraction ability, which can effectively improve the fault diagnosis capability of spacecraft power systems.

Abstract:In response to the shortcomings and deficiencies of the improved back propagation network in aviation equipment fault diagnosis, a hybrid algorithm is formed by combining the adaptive genetic algorithm and the improved back propagation algorithm to train an artificial neural network. Taking the improvement of the initial weight space of the back propagation network as the starting point, a multi-point adaptive genetic optimization is carried out using the improved genetic operation. Based on this, the improved back propagation algorithm is used to carry out local precise search and ultimately achieve global optimization. Taking the fault diagnosis of a certain aircraft electrical control box and a certain aircraft autopilot flight control box as examples, the proposed algorithm was simulated and studied. The simulation results showed that the combination of adaptive genetic algorithm and improved back propagation algorithm has fast convergence speed and high diagnostic accuracy, and has good diagnostic results for engineering samples with complex input-output relationship.