Abstract:In the field of TSV 3D integration, due to the miniaturization of internal defects and the challenges associated with non-contact detection, identifying a non-destructive, sensitive, and efficient method for internal defect detection is crucial. In response to this challenge, a TSV internal defect detection method based on a temperature sensor array has been proposed. Internal defects influence the external temperature distribution of TSV 3D packaged chips, displaying regular patterns of change. Each type of defect causes different deviations in the external temperature distribution. By utilizing a temperature sensor array to measure these distribution changes, effective identification and classification of defects can be achieved. A detection system, based on the temperature sensor array, was designed to reveal the internal defect information based on the thermal signals generated by chips under operational conditions. Through theoretical analysis and simulation modeling, a model simulating the temperature distribution and thermal changes of chips under working conditions was developed. In the experiments, based on the preparation of chip samples and the setup of a testing platform, effective classification of internal defects was achieved using a classification recognition model, reaching an accuracy rate of up to 99.17%. This detection method provides a cost-effective and efficient new approach for the reliability analysis and fault diagnosis of high-density and miniaturized chips.

Abstract:Ultrasonic microscopy measurement technology is a non-destructive testing method widely used in various fields such as industry, aerospace, and medicine. In order to reduce dependence on imported equipment, particularly for high-frequency ultrasonic excitation and reception devices operating above 100 MHz, research into methods for high-frequency signal excitation and reception is crucial. This paper introduces an ultrasonic excitation circuit utilizing the avalanche breakdown characteristics of a transistor and the voltage multiplication principle of the MARX topology circuit to minimize the excitation pulse width, thus enhancing the bandwidth coverage. Additionally, research on high-frequency ultrasonic signal reception circuits and the integration of an ultrasonic excitation reception instrument are presented. The research realized the localization replacement of the equipment, filling the gap of 100~500 MHz high-frequency pulse transceiver in the domestic ultrasonic field. Experimental results demonstrate that the designed high-frequency excitation signal has a peak-to-peak amplitude of not less than 128 V, a rise time of less than 0.47 ns, and a pulse width of less than 3.5 ns, with a system operating bandwidth covering 500 MHz, meeting the requirements of high-frequency ultrasonic measurement systems.

Abstract:In the field of precision electrical measurement, there are few special coaxial scanner for AC automatic test systems. Aiming at this, a 4-wire-15 channel all-coaxial scanner device is designed. The device is based on the precision motion control of the stepper motor, and drags the fast-plug-in coaxial axis component to do X-Y bidirectional positioning motion, realizing the reliable automatic switching of the coaxial measurement channel. By independently designing the flexible docking mechanism, the risk of channel docking error and irreversible physical damage caused by movement misalignment are eliminated. By analyzing the generation mechanism of parasitic parameters such as channel thermal electromotive force, leakage impedance, parasitic capacitance and inductive coupling potential, the key design parameters affecting the electrical characteristics of scanner are defined, and then the targeted optimization design method is proposed. The test results demonstrate that the thermal electromotive force in the scanner channel is less than ±50 nV, the leakage impedance of the channel to the ground reaches 1014 Ω, and the parasitic capacitance between the channels is less than 1×10-5 pF, additionally, the inductive coupling coefficient between channels is less than 1.2×10-10 H.

Abstract:This article proposes an online identification method for a electro-optical tracking system based on the Zynq platform, addressing the cumbersome and inefficient nature of traditional identification. By designing data acquisition and driver programs for A/D and D/A modules, real-time collection and control are achieved, and running the VxWorks operating system on Zynq platform. The use of FFT technology enables the online measurement of frequency characteristics, and a hybrid improved quantum-behaved particle swarm optimization algorithm with natural selection parameters is employed for system identification. Experimental results demonstrate that this method possesses high-precision frequency characteristic measurement and accurate system identification capabilities, providing effective means for real-time control and optimization of electrooptical tracking systems.

Abstract:A spoof surface plasmon polaritons (SSPPs) sensor based on liquid switch control is proposed to detect metal groove filling density. The sensor consists of liquid switch controllers and power division structure. The liquid switch consists of a plastic pipe and syringe with water, which is fixed on the branch structure of SSPPs sensor. By filling the plastic pipe with water, the branch cannot transmit signal, which is called “OFF” state. In contrast, when there is no water in the plastic tube, it is called “ON” state. By switching the “ON” or “OFF” state of sensor’s branches, it can control the sensing area of SSPPs sensor. The detection method is based on the change of reflection coefficient of SSPPs sensor caused by the change of groove filling quality. By switching the states of branches in turn, it can detect whether whether there are filling quality problems at each branch. The experimental results show that the SSPPs sensor can effectively detect the filling changes of 10 mg, which can be used for rapid detection of workpiece with mass production.

Abstract:Autonomous driving test method is an important part of autonomous driving research systems, and it is also one of the current research hotspots. In this paper, an automatic driving test method based on risk scenario search is proposed. The swap scenario is extracted through the NGSIM (Next Generation Simulation) dataset, and the scope and characteristics of the scene parameter space are determined. To improve the efficiency of scene generation, the search process of scene parameters is enhanced, constraining the lanechanging scene of the preceding vehicle, optimizing the simulation time of a single scene, and adding a memory module to prevent the same scene from repeated testing. A three-lane leading vehicle lane-changing scenario is constructed based on the PreScan/Simulink platform, and 439 scenarios are generated and tested, among which 266 scenarios result in vehicle collisions. It is found that collision failures are caused by the delayed action of the measuring system and the failure to trigger emergency braking.

Abstract:State of health (SOH) predictions are critical for battery management systems. Due to the complexity of battery health assessment modeling and large prediction errors, accurate SOH prediction still needs to be improved. In this paper, Improved sparrow search algorithm (ISSA)-Convolutional neural network (CNN)-Bidirectional Gated Recurrent Unit (BiGRU)Attention mechanism for lithium battery health status assessment is proposed by combining capacity increment analysis (ICA) and differential voltage analysis (DVA) methods. Firstly, the capacity increment (IC) curve and differential voltage (DV) curve are processed by Gaussian filtering to avoid the influence of noise. A set of new battery aging features were extracted from the filtered IC and DV curves through the center for advanced life cycle engineering Advanced Life Cycle Engineering (CALCE) data processing. The Pearson correlation coefficient between the four aging features and SOH was above 0.9. ISSA-CNN-BiGRU-Attention method was used to construct a prediction model of battery SOH, and the proposed method was compared with CNN, BiGRU, CNN-BIGRU and other methods. Experimental results showed that the maximum MAE and RMSE errors of the proposed method were 0.005 44 and 0.007 17, respectively. Compared with other models, it has excellent robustness and accuracy, and has better practical use value.

Abstract:Seamless steel pipe production, as a typical representative of high-energy-consuming industries, has always been a focus of energy-saving and consumption reduction. By predicting power consumption, enterprises can identify effective ways to save energy, thereby reducing electricity consumption in the production process and improving production efficiency. In order to improve the accuracy of electricity consumption prediction for seamless steel pipe continuous rolling, an improved Stacking ensemble learning model is adopted to predict power consumption. Firstly, the collected power consumption data is preprocessed, and XGBoost and LightGBM are used for feature selection based on embedding method. Then, a combination of random search and Bayesian optimization is used to optimize the hyperparameters of the base learners. In the first layer of the Stacking ensemble model, LightGBM, ET, and MLP are selected as the base learners. Finally, based on the predictive performance of the base learners on the data, they are assigned corresponding weights, and the original dataset is also included in the training of the meta-learner. The results show that the improved Stacking ensemble learning model has the best prediction effect, with an R2 of 0.975 6. The prediction accuracy is higher than that of single base learners and traditional Stacking ensemble learning models, demonstrating the effectiveness of the proposed method.

Abstract:In order to enhance the accuracy of road traffic flow prediction, this paper proposes a traffic flow prediction model based on the self-attention mechanism TCN-BiGRU. Firstly, the predictive model utilizes the convolutional property of time convolutional network (TCN) to extract temporal correlations within traffic flow data across different time steps. Secondly, bidirectional BiGRU is employed to comprehensively capture time-related characteristics of traffic flow by updating and resetting gates. Recognizing that bidirectional gated recurrent unit has limited parallelism and may not capture certain features during bidirectional calculation, the introduction of self-attention mechanism allows the model to focus on global correlation between different inputs, overcoming limitations posed by sequence length in feature capture and maximizing feature retention for improved robustness. Finally, predicted values for traffic flow are obtained. To validate the applicability of the model, real traffic data is selected for multiple prediction and comparison experiments against benchmark models and ablation experimental models across various road segments. The results demonstrate that single or multiple section predictions using multiple features based on self-attention mechanism TCN-BiGRU yield mean MAE values of 15.91 and 19.62 respectively; MAPE/% values of 10.89 and 13.53 respectively; as well as R2 values of 0.976 and 0.982 respectively-indicating strong predictive performance.

Abstract:In order to enable radar to deal with interference scenario and improve LPI performance of signal, a novel resource allocation scheme is proposed in this paper, and the specific jamming type studied is deceptive jamming. Firstly, the CRLB for deceptive distance in three dimensions is derived. Then, the resource allocation problem under anti-interference is established according to the CRLB. Specifically, the total power consumption is taken as the objective function, and the CRLB is taken as the performance constraint. The total power value is suppressed under the premise that the radar meets the predetermined performance constraint by allocating sensor, power and bandwidth resources. In order to solve the proposed optimization problem, this paper firstly solves the sensor resource allocation problem belonging to integer programming, and then adopts the SQP algorithm in cyclic form to solve the joint power and bandwidth allocation problem. The final simulation results show that the power consumption of the new scheme is less than 50.0% compared with the optimal power allocation scheme, which verifies the feasibility of the new scheme in reducing the total power. In addition, the advantages and disadvantages of the proposed algorithm are explained by comparing with the nonlinear programming genetic algorithm in the simulation experiment. Finally, the power comparison results and analysis at the interception receiver are given, which show that the maximum power component effectively decreases, so as to improve the LPI performance of the signal.

Abstract:In order to improve the accuracy and speed of multi-sensor data fusion, a convex combination fusion algorithm and covariance cross fusion algorithm were combined, and the fusion coefficient of the covariance cross fusion algorithm was optimized using the fruit fly optimization algorithm. An improved covariance cross fusion algorithm was proposed, which achieved fast and accurate fusion of multi-sensor data. The simulation results show that the root mean square error of the proposed algorithm for data fusion on the x-axis and y-axis is about 3 m, and the fusion time is about 0.44 s. Compared with data fusion algorithms such as multi Bayesian estimation, fuzzy clustering, and maximum likelihood estimation, it has significant advantages and improves the accuracy and speed of multi-sensor data fusion.

Abstract:Carbon fiber composites have excellent properties such as high specific strength, high specific stiffness, corrosion resistance and fatigue resistance, which are ideal construction materials for large-span structural cables. In this paper, a simulation study of broken strands and debonding in carbon fiber cables was carried out using the finite element method based on the electromagnetic tomography technique, and an 8-coil circumferential sensor array was designed to investigate the relationship between the sensor dimension, the magnitude of the excitation current and the magnetic induction intensity. The effects of three reconstruction algorithms, LBP algorithm, Tikhonov algorithm and Landweber algorithm, on the quality of reconstructed images of different cables defects were investigated based on the simulated electromagnetic signals. The results show that as the diameter of the sensor coils decreases and the excitation current increases within the studied range, the magnetic field intensity progressively increases. When the diameter of the sensor coils is 5 mm and the excitation current is 1 A, the magnetic field intensity reaches its maximum value.Compared with other image reconstruction methods, the Tikhonov is suitable for the reconstruction process of defect detection images in carbon fiber cables due to Tikhonov algorithm can better balance the image reconstruction effect and imaging quality. In addition, the effect of different projection angles on defect imaging was analyzed by rotating the cable defect with a fixed sensor array. The study aims to provide a reference basis for the design of sensor arrays and defect detection imaging of carbon fiber cables under experimental conditions.

Abstract:Laser point cloud target detection based on deep learning has become an important research field. This article uses a SOTA deep learning network based on spherical projection and 2D images to achieve rapid detection of 3D laser point cloud targets. Firstly, a single frame 3D point cloud from the Semantic KITTI data set is transformed into a 2D RGB three channel image through spherical projection. The pixel position of the image plane depends on the three-dimensional coordinates of the point cloud, and the grayscale values of the R, G, and B channels depend on the normalized reflection intensity, distance, and height of the point cloud. Secondly, the overlapping distribution of spherical projections at different resolutions and their technical impact on image quality were analyzed. Finally, using the semantic segmentation model DeepLab-V3+network, simulation results show that this method has good performance in segmentation accuracy and speed, and has high application value.This paper presents a method of license plate character recognition based on the combination of Zernike moment and wavelet transformation features.

Abstract:In order to achieve autonomous driving in cities, it is necessary to be able to efficiently detect the on-site command gestures of traffic police. Aiming at the problems of low recognition accuracy, slow detection speed, and difficulty in dealing with complex road environments in existing gesture recognition algorithms, an improved YOLOX-tiny traffic police gesture recognition algorithm is proposed. Firstly, an improved GhostNet network was used to replace the original backbone network, and a Coordinate Attention mechanism was inserted to comprehensively extract input image features, improving the detection accuracy of the network and enhancing the detection performance for small and medium-sized targets; Secondly, the decoupling head was improved by designing the SCDE Head structure, which reduces computational complexity while filtering redundant information, making the decoupling head more efficient. The decoupling head also integrates multi-scale features, improving the accuracy of object detection; Finally, applying SIoU to localization loss accelerates network convergence and improves regression accuracy. Tested on a self-made traffic police command gesture dataset, the experimental results showed that compared with the YOLOX-tiny model, the improved algorithm reduced the number of parameters by 27.97%, the model′s computational complexity by 33.31%, and the average detection accuracy increased by 2.31%, with a 45% increase in detection speed, which is more suitable for the practical needs of autonomous driving and hardware deployment.

Abstract:In harsh environments such as high temperature, high pressure, and strong magnetic fields, pointer instruments have shown superior performance compared to digital instruments. Therefore, the research on pointer instrument reading recognition has significant practical significance. In recent years, the combination of deep learning and computer vision has become a key focus in the research of pointer instrument reading recognition technology. This paper first outlines the process of pointer instrument reading recognition, and then discusses the research status and progress of pointer instrument reading recognition technology from 3 aspects: image preprocessing, dial region detection, and reading recognition. The paper elaborates on both traditional machine learning methods and deep learning methods. Finally, it introduces publicly available pointer instrument datasets and application scenarios, and provides prospects and suggestions for future research from the aspects of deep learning algorithms, inspection robot characteristics, and the establishment of public datasets.

Abstract:In order to solve the problems of large computation and low localization accuracy of ZNCC template matching algorithm, this paper proposes a ZNCC displacement extraction algorithm based on improved edge information. First, the sub-pixel accuracy of the ZNCC displacement extraction algorithm based on ZNCC is improved by surface fitting method. Then, the important information features in each frame of the video are extracted using the filtered Sobel operator edge detection algorithm to create a new template and perform ZNCC correlation value calculation. In order to reduce the number of matching points and operations, different step sizes are set according to the number of information points. Finally, the ZNCC displacement extraction algorithm based on improved edge information is used to extract the displacement waveform graphs of the moving objects in the experimental video for analysis and verification. The experimental results show that the measurement speed is improved by 77.46% and the measurement error is 0.234 4%, which proves that the ZNCC displacement extraction algorithm based on improved edge information not only retains the search accuracy and stability advantages of the original algorithm, but also achieves a significant improvement in speed.

Abstract:Aiming at image perspective distortion and surface defect detection of solar photovoltaic cells, a method based on a virtual camera for perspective correction and an improved YOLOv5s neural network model for defect detection are proposed. Firstly, a virtual camera with a horizontal orientation is constructed based on camera extrinsics to establish a perspective mapping relationship between the original image and the virtual camera, by which perspective correction of the original image is achieved. Secondly, a dynamic head is employed to enhance the representation capacity of the YOLOv5s head, and a receptive field expansion (RFI) module is added into the bottleneck of the C3 module to enhance the receptive field for small targets. Finally, the localization loss of YOLOv5s is fused with the normalized weighted distance (NWD) loss to compensate for the positional deviation of small targets. Experimental results demonstrate that the perspective correction based on the virtual camera can achieve significant improvements in correction effectiveness with shorter runtime. Moreover, the average accuracy of the improved YOLOv5s model can be increased up to 6.1%, 27.7%, and 1.1% than YOLOv5s, YOLOv7, and YOLOv8 respectively, which exhibits the practical value in surface quality inspection of solar photovoltaic cells.

Abstract:In order to improve the efficiency and accuracy of bolt positioning and grasping in industrial production, a bolt pose and size detection method based on region of interest was proposed. Firstly, YOLOv5 target recognition algorithm is used to identify the bolt target, and the identified target area is intercepted as the region of interest. Then the ROI region is preprocessed by median filtering and binarization, and the Canny improved algorithm is used to detect the target contour. Then the bolt tilt Angle is calculated based on the best fitting line algorithm, and the bolt center of gravity is calculated by the moment feature algorithm. Finally, the shortest distance algorithm of Hough two straight line segments was used to detect the bolt diameter. After experimental verification, the recognition accuracy of YOLOv5 target recognition algorithm reaches 92.7%, the inspection error of bolt tilt Angle is ±1.2°, and the detection error rate of bolt diameter is ±5.5%, which realizes the identification of bolt pose and size.

Abstract:The convolutional operation is constrained by traversal rules, limiting the extraction of feature information from individual skeletal nodes and preventing effective fusion of feature information between adjacent nodes, resulting in limited expressive power. In response to this issue, a gesture recognition neural network based on a Feature Displacement Module is proposed. This network adopts the architecture of conventional spatiotemporal graph convolutional neural networks and replaces the conventional spatiotemporal convolution module with the Feature Displacement Module to achieve fusion of feature information between adjacent nodes. By reordering the displacement channels through the Feature Displacement Module, global feature information of skeletal nodes is extracted, further enabling efficient and accurate classification of gesture information. The Feature Displacement Module is validated on the public dataset DHG-14/28 and FPHA, achieving classification accuracies of 95.11%, 93.01% and 92.67% for 14-class, 28-class and FPHA gesture datasets. The experimental results demonstrate that this network model can better and more effectively mine global feature information, achieving excellent performance on common gesture recognition datasets.

Abstract:With the continuous development of VLSI circuits, the design of the on-chip power delivery network is becoming increasingly important, and the performance of the design needs to be reflected by calculating the quiescent voltage drop of the power delivery network. However, traditional computational methods are time-consuming, resulting in longer chip design cycles. In order to shorten the cycle of chip design and improve the efficiency of chip design, this paper proposes a fast static voltage drop prediction model based on convolutional neural network—ResCSP-34. The model adopts the encoderdecoder structure, firstly the residual network ResNet34 is modified as the main structure of the encoder, then the feature fusion module is introduced into the decoder, and the attention mechanism module is introduced at the connection of the encoder and the decoder, and finally a loss function combining the mean square error, Pearson correlation coefficient and mean absolute error is proposed to train the model. Experimental results show that on the CircuitNet dataset, the average absolute error of the model prediction results is 0.7 mV, which is less than 1 mV, the average value of the Pearson correlation coefficient is greater than 0.93, close to 1, and the average total time for static voltage drop prediction for an on-chip power supply network design is 7.36 s, and the average inference time of the convolutional neural network is 0.015 s. Experimental results show that the ResCSP-34 model can quickly and accurately predict the quiescent voltage drop.

Abstract:Traffic anomaly detection is a technique used to identify network attacks effectively. In recent years, unsupervised methods have become prevalent in anomaly detection. Aiming at the demand of mining the temporal relationship between existing traffic data and the problem of randomly selecting feature attributes for sample division in iForest, this paper proposed a traffic anomaly detection method based on iForest score extension. Firstly, the paper used the sliding window mechanism and the information entropy property to design an entropic timeseries feature extraction method for network traffic, which was integrated into the feature set to perform significant feature screening. Secondly, the paper constructed an iForest score extension model that utilized the feature set iteration method with the feature importance matrix in the node sample division, integrated the isolated tree features in the set, marked the integrated path length between nodes instead of the original path length, and calculated the anomaly score that better characterized the sample distribution. Finally, by setting the anomaly score threshold, the paper discriminated whether the samples were abnormal. The experimental results on the public dataset show that the anomaly detection model proposed in the paper has obvious advantages over other methods, with good real-time detection performance and lower false alarm rate, which can be effectively used in the anomaly detection of network traffic, and is of great significance for the identification of attack events in real network activities.

Abstract:The low contrast and blurred boundary of chest X-ray images seriously affect the segmentation effect of chest X-ray images. In order to diagnose and treat lung diseases quickly and accurately with chest X-ray images, this paper presents a method of chest X-ray lung image segmentation based on interclass variance and probabilistic error. Based on the pre-processing of chest X-ray images, the method firstly uses the information of human body structure in chest X-ray images for coarse image segmentation. Then, the interclass variances and probabilistic errors between the target class and the background class are calculated respectively for the preprocessed image, and a new segmentation objective function is designed to calculate the optimal threshold after non-dimension processing the interclass variances and probabilistic errors, so as to achieve the image accurate segmentation. Finally, the segmentation results of the coarse and fine segmentation processes are combined and optimized to achieve the image segmentation based on the optimal threshold. The comparative experimental results of chest X-ray images show that The DSC and IOU indicators of the proposed method are 89.5% and 81.1% respectively, and the segmentation of lung regions by the method has good performance in completeness and accuracy. This indicates that the method is effective and feasible, and is suitable for lung image segmentation based on chest X-ray images.

Abstract:Aiming at the problem that the characteristics of micro-resistivity imaging logging tools lead to the regular blank zone of the measured wellbore image, this paper proposes a filling model based on unsupervised learning framework, which integrates multi-scale and multi-level features, and a full-well section filling framework to fill the blank zone. The filling model adopts the UNet architecture, and uses the statistical prior of the non-blank zone resistivity data itself to perform unsupervised training filling based on MAE loss. The model is improved mainly through the following two measures: The multi-scale residual convolution is introduced into the encoder to improve the multi-scale representation ability of the single-layer network; The multi-layer feature fusion module and information guidance module are introduced in the encoding and decoding feature connection link to enrich the feature scale of upsampling and reduce the information loss in the decoding process. The experimental results show that compared with UNet, the visual effect and objective indicators of the model proposed in this paper are significantly improved on the natural scene dataset. PE is reduced by 19.03%, SSIM is increased by 2.9%, and PSNR is increased by 4.66%. The whole well section filling framework applies the filling model to train the filling blank zone resistivity data in sections and then merge them to realize the end-to-end filling of the micro-resistivity imaging logging blank zone of a single well. The filling results have certain robustness and fit the actual production scene.

Abstract:In order to solve the problem of image distortion of defect magnetic flux leakage (MFL) signal caused by unsaturated magnetization of small pipe diameter elbow, and realize intelligent and high-precision identification of elbow distortion magnetic flux leakage defect image. This paper proposed an intelligent image identification method for distortion MFL defects in small-diameter pipe elbows. The MSRCR-HF image restoration algorithm was applied to process the distorted images, which to solve the problem of defective image distortion caused by the weak MFL signal of the elbow. The YOLOv5 network was optimized by integrating the CBAM and the SPD-Conv module to improve the network's feature extraction ability for elbow distortion and MFL defects. Finally, the elbow defect datasets were established through simulation, and it was input into the network for training and testing. The results shown that the MFL signal image of the same defect at the elbow was distorted, and the defect feature information cannot be directly and effectively obtained. The proposed MSRCR-HF algorithm effectively resolved the image distortion problem associated with elbow MFL defects. Additionally, the improved YOLOv5 model achieved high recognition accuracy on the established dataset, with accuracy rates of 95.5% for rectangular groove defects, and 93.0% for hemispherical defects. This method exhibited strong feasibility for intelligent identification of distortion MFL defect in small-diameter pope elbows and can improve the efficiency of pipeline safety inspection.

Abstract:Based on the fiber optic gyroscope, an angular motion integrated measurement sensor can achieve integrated and dynamic precision measurement of various angular motion parameters. However, in practical applications, the fiber optic gyroscope is susceptible to temperature changes, leading to a decrease in measurement accuracy. Addressing this issue, this paper proposes a temperature error compensation technique for the angular motion integrated measurement sensor based on an adaptive wavelet echo state neural network. To advance the progress of temperature error modeling and enhance the approximation capability of traditional neural networks, an adaptive forward linear prediction filter is applied to preprocess temperature drift data from the gyroscopes used for modeling. The paper adopts an adaptive wavelet echo state neural network to establish a temperature drift model, aiming to avoid issues such as the blind design of traditional neural network structures and local optima. This approach enhances the network's learning and generalization abilities. Additionally, an adaptive law is employed to replace neural network gradients during network training, thereby improving the approximation accuracy and convergence speed of the neural network. Experimental results demonstrate that the proposed model can enhance the measurement accuracy and environmental adaptability of angular motion integrated measurement sensor, providing reliable technical support for the performance optimization and practical applications of these sensors.