Column on Optimal Design of Mechanical Parameter Sensors for Harsh Environments
Zheng Yongqiu;
<正>Mechanical parameters are important parameters for characterizing performance indicators in the equipment test process. Due to harsh environments such as high temperature and high pressure, conventional mechanical sensors face technical challenges of "unable to measure", "inaccurate measurement" and "poor measurement". Therefore, the acquisition of mechanical parameters in harsh environments has always been a difficult problem for equipment testing. With the rapid development of multidisciplinary cross-integration such as materials, semiconductors, and micro-nano processing technology, sensor technology based on new principles, new methods, and new materials has also made great progress, especially mechanical parameter sensors for harsh environments, which are developing in the direction of integration, intelligence, low cost, green and low carbon.
Chang Hanlin, Xie Lin, Yang Xiaopeng, Fan Zhiqiang;
According to the piezoelectric principle of flexible piezoelectric film, a composite piezoelectric effect sensor was designed for shock wave signal measurement. In order to study the influence of the sensor substrate configuration on its measurement performance, substrates with different aperture deformation zones were prepared respectively, and the deformation mode of the piezoelectric film was adjusted to control its shock wave measurement performance. The experimental results show that by increasing the aperture of the polyvinyl chloride (PVC) substrate, the sensitivity coefficient of the polyvinylidene fluoride (PVDF) film sensor can be greatly increased, and with the increase of the diameter of the deformation zone, the pulse width of the signal also increases. Finally, through the explosion experiment, the influence of the PVC substrate on the PVDF film sensor signal was studied, and the accuracy of the PVDF sensor with a deformation zone diameter of 8 mm in the explosion shock wave measurement was verified. This provides a reference for the selection of the PVC substrate deformation zone of the PVDF film sensor in the future.
Wang Lühao; Cao Yonghong;
The output of the pressure sensor is difficult to be stable within a large pressure range. The traditional pressure sensor adopts a membrane structure, which will produce strong nonlinearity under high pressure, and complex conversion is required to convert the measured data into linear results. Based on the piezoresistive effect of graphene, this paper designs a high-pressure pressure sensor through a cylindrical sensitive structure, which can achieve linear output of pressure values within the range of 400 MPa. The sensor structure is mainly composed of an elastic metal diaphragm, compressible silicone oil, a cylindrical structure and a graphene sensitive membrane. The sensor structure is optimized by theoretical and simulation analysis. By analyzing the thickness of the diaphragm, the size structure of the metal diaphragm that only produces elastic strain within the pressure range of 400 MPa is designed; the overall structure of the sensor is simulated and analyzed by the fluid-solid coupling method, and the relationship between the size of the cylindrical structure and the strain of the outer wall and the relationship with the natural frequency of the overall structure are clarified. The reasonable position of the graphene sensitive membrane installation is obtained by finding the maximum strain position; finally, the linear relationship between the strain of the outer wall of the cylindrical structure and the pressure measured by the sensor is explored. The results show that the designed sensor structure can work within a pressure of 400 MPa, and the appropriate size of the sensor cylindrical structure is 1.2 mm long and 0.08 mm thick. At this size, the natural frequency of the sensor is 127 kHz, the maximum strain of the outer wall of the cylindrical structure can reach 2 082με, and the graphene sensitive film is installed at the highest strain point, and the sensitivity of the sensor structure can reach 5.205με/MPa; the sensor linearity structure has good linearity in the range of 0 to 400 MPa, and the structural linearity relative error is only 0.37%.
Song Gaoying;Gao Jingwu;Yang Zheyi;Qiu Yongfeng;
Based on the principle of thermal temperature difference, this paper simulates the working conditions of thermal wind sensor in two control modes, constant power and temperature balance, using finite element analysis software ANSYS, and analyzes the influence of different characteristic size parameters of chip structure on sensor output performance. The research results show that the output of the sensor increases with the increase of chip side length and thickness, and heating element length, while the output decreases when the distance between the heating element and the temperature measuring element increases. After optimizing the structural dimensions, the wind direction error of the sensor is less than ±2° in constant power mode, and the error is less than 10% when the wind speed is 0-15 m/s; in temperature balance mode, the wind direction error is ±3°, the effective wind speed measurement range can reach 0-20 m/s, the error is less than 5%, and the measurement accuracy is improved.
Mechanical and Power Engineering
Liu Yang;Ma Guichun;Zhang Huabin;Wang Lei;Guo Sibo;
The uniformity of the temperature field distribution in the autoclave during the curing process has an important influence on the structural properties and strength of composite materials. This paper takes the frame-type mold widely used in the aerospace manufacturing industry for making complex large-scale structural parts as the research object, establishes the unsteady three-dimensional temperature field of the autoclave frame-type mold during the curing process in COMSOL, and studies the influence of different rib plate thicknesses and ventilation hole sizes of the mold and the presence or absence of semicircular heat dissipation holes under the mold surface on the uniformity of the mold surface temperature field. By comparing the temperature standard deviation curves of the molds with support rib thickness of 10, 8, 6 and 4 mm, it is found that reducing the thickness of the support rib can increase the uniformity of the surface temperature field, but the overall stability of the mold decreases rapidly as the thickness of the rib becomes thinner, and it is more likely to cause compressive instability; by comparing the temperature standard deviation curves of the molds with longitudinal ventilation hole sizes of 130 mm×220 mm, 150 mm×220 mm, 170 mm×220 mm and 190 mm×220 mm, it is found that appropriately widening the rib ventilation hole size can improve the uniformity of the surface temperature field, but the curves of 170 mm and 190 mm width basically overlap, and the impact is no longer obvious; similarly, the existence of heat dissipation holes under the surface can improve the uniformity of the surface temperature field.
Hu Hongjun; Yang Xiwang; Huang Jinying;
Aiming at the problems of few plunger pump fault samples, weak fault signals under noise interference and traditional deep learning relying on a large number of training samples, a few-sample plunger pump fault diagnosis method based on model agnostic meta-learning (MAML) is proposed. Firstly, the improved fully integrated empirical mode decomposition with adaptive noise (ICEEMDAN) method is used to decompose the collected one-dimensional vibration signal to obtain the IMF components of the intrinsic mode function, and the sensitive components with rich fault information are screened to enhance the feature information in the vibration signal. Secondly, a multi-channel one-dimensional convolution model is established, which constructs a channel interaction feature encoder with an efficient channel attention mechanism, aiming to focus on the interactive fault information between different channels, and then effectively extract the common diagnostic knowledge of multiple diagnostic meta-tasks. Finally, the one-dimensional convolution model is used as the base model, and the optimal model initialization parameters are obtained by training with the MAML method; the optimal initialization model can quickly adapt to a small number of plunger pump fault samples under new working conditions, thereby realizing plunger pump fault diagnosis under few samples. The performance of the model is verified by plunger pump experimental data. The results show that the diagnostic accuracy of the proposed method for various diagnostic tasks under few-sample conditions is more than 90%.
Automation and computers
Shi Zitong; Shi Zhibin; Liu Dongming; Shi Qiong; Gong Xiaoyuan;
With the widespread application of blockchain technology, the security issues of smart contracts have attracted widespread attention. In view of the fact that the conversion of smart contract source code to bytecode will lose some semantic information, and the existing deep learning vulnerability detection methods cannot detect reentrancy vulnerabilities and timestamp vulnerabilities well, this paper proposes a smart contract source code vulnerability detection method based on feature modulation graph neural network (GNN-film). Firstly, the characteristics of reentrancy vulnerabilities and timestamp vulnerabilities are analyzed, and the graph structure is constructed and simplified using the smart contract source code; secondly, a graph neural network model based on feature-level linear modulation is built, and the powerful feature modulation ability of the network model is used to accurately represent the contract vulnerability characteristics; finally, the simplified graph structure data is input into the built model to obtain the detection results. The experimental results show that the accuracy of the proposed method for reentrancy vulnerabilities and timestamp vulnerabilities is 91.00% and 91.64%, which is 4.20 percentage points and 9.70 percentage points higher than that of the method based on graph neural network, respectively, proving that the ability of the proposed method to detect related vulnerabilities is better than other detection tools.
Qin Jiahao, Qin Pinle, Chai Rui, Chen Zuojun, Gao Yipeng, Wang Bao;
In view of the complexity of noise in weak light environment and the difficulty in establishing noise model, this paper proposes a noise image generation method based on diffusion model. First, various types of noise are added to the clear image through the formula in the forward process of the diffusion model, and then it is input into the network together with the clear conditional image. Finally, the sampling algorithm of the cold diffusion model is used to generate the noise image in a loop iteration, so as to establish a more realistic noise model. The experimental results on the weak light dataset show that compared with other algorithms, the noise image generated by the algorithm in this paper has a KL divergence value of 0.068 in objective indicators, which is 0.001 lower than the best existing method. The subjective quality is higher, closer to the image in the weak light environment, and the established noise model is more accurate. The method in this paper can establish a high-quality noise model in a weak light environment, which provides a new idea for noise modeling in a weak light environment, and applies the diffusion model to the field of weak light noise modeling.
Image Processing and Computational Imaging
Li Linqi; Chang Min; Hou Xiaoyu; Jia Caiqin; Pang Min;
In order to solve the problem of coupling error in RGB-D data, the problem of edge point mis-extraction in existing feature extraction methods, and the problem of poor tracking stability of constant-speed motion model in VSLAM using RGB-D cameras, a CEP-SLAM algorithm is proposed based on the ORB-SLAM2 framework. The algorithm uses the constant acceleration motion model to set the initial pose of the frame to be tracked; the optimized pose is used to calculate the visual odometer between frames and update the constant acceleration motion model, and the pose offset is estimated based on the acquisition time difference of the RGB image and the depth image. The epipolar geometry constraint is constructed based on the pose offset, and the position of the feature point corresponding to the pixel point in the depth image is found by using the binary search method. The depth of the feature point is adjusted to alleviate the influence of the coupling error of RGB-D data on VSLAM; a key frame edge point removal algorithm based on the joint method is proposed, which uses the neighborhood information of the feature point in the depth image to judge and remove the bad edge points in the key frame to be inserted. The CEP-SLAM algorithm proposed in this paper is used to conduct experiments on the TUM public dataset. The results show that the algorithm can better remove the bad edge points and has better robustness, tracking stability and higher positioning accuracy than the classical algorithm.
Wang Zhankui, Qin Pinle, Zeng Jianchao;
In view of the large scale variation, arbitrary direction and dense distribution of targets in remote sensing images, the existing detection methods pay little attention to the dense edge information directly and the targets cannot obtain a suitable receptive field, resulting in poor remote sensing detection effect. This paper proposes a multi-scale feature enhancement network (LKCSFP-NET) based on large kernel convolution and dense target refinement for remote sensing image detection. Firstly, the network adds a dilated convolution to form a large kernel convolution block (LKB) on the basis of SKNET, so as to obtain the best receptive field of small targets and improve the adaptability and accuracy of the network to multiple scales; secondly, the centralized spatial feature pyramid CSFP module is added on the basis of FPN, and the problem of low detection efficiency of remote sensing images caused by dense target distribution and complex background is solved by combining global semantic information with local semantic information. Experimental results show that on the DOTA and HRSC2016 public datasets, the average detection accuracy of the proposed algorithm on the two datasets is 75.96% and 96.60%, respectively, which is 1.36 percentage points and 0.63 percentage points higher than the baseline network, and is better than most existing models. The proposed LKCSFP-NET performs stably in two public datasets, and has good detection effects on small targets and densely arranged targets. Its detection accuracy is higher than that of most existing models, and it can be well applied to the detection of remote sensing targets.
Gong Yiyun, Yu Haibo, Wang Yun, Kang Li, Zeng Jianchao;
In order to enhance the performance of agent model-assisted evolutionary algorithm for solving high-dimensional expensive optimization problems, a dual-model driven multi-preference strategy adaptive differential evolution algorithm is proposed. The algorithm is based on global and local agent modeling methods, and organically integrates three evolutionary strategies with different optimization preferences. In each iteration, by utilizing the online iteration feedback information of the optimal solution during the optimization process, the calling frequency of different evolutionary strategies is adaptively adjusted in a sequential manner to efficiently balance the global exploration and local exploitation of the algorithm. In order to promote the sharing of excellent information among individuals in the population, an elite individual-driven differential perturbation strategy is designed to increase the optimal sample prior of the potential optimal solution area. By processing 26 high-dimensional benchmark test problems of different scales, the results show that the convergence performance and optimization efficiency of the proposed algorithm are absolutely superior to those of four advanced algorithms of the same type in at least 17 test problems.
Chemical and Environmental Engineering
Ma Yao, Yuan Zhiguo, Li Yulong, Fu Yiting, Zhang Yaqian, Jia Zhengna, Kong Zhe;
Zero-dimensional and multi-dimensional barium strontium titanate (Ba1-xSrx TiO3, BST) powders have the advantages of high specific surface area and high aspect ratio, respectively. In order to improve the material performance and expand the industrial application of raw materials in electronic components, and to improve the uneven distribution and long time consumption in the powder preparation process, the BST powder with glycerol added was prepared by impinging stream-rotating packed bed (IS-RPB) combined with hydrothermal method, and the influence of hydrothermal temperature, hydrothermal time and glycerol addition on the morphology of BST powder was investigated. The morphology, crystal phase and composition of BST powder were characterized and analyzed by X-ray diffractometer (XRD), scanning electron microscope (SEM) and transmission electron microscope (TEM). The results show that under the conditions of supergravity factor of 40, initial impact velocity of 10.61 m/s, hydrothermal temperature of 180℃ and hydrothermal time of 12 h, when a small amount of glycerol is added, it acts as a dispersant and the prepared BST powder is a nano-spherical particle with good uniformity and dispersion; when more glycerol is added, it acts as a solvent and grows along the unit crystal plane under hydrothermal conditions to obtain flake and flower-like morphologies. The supergravity-hydrothermal method and the addition of glycerol proposed in this paper provide a convenient and fast way to explore the industrial morphology controllable preparation of BST powder.
Niu Bingkang, Li Ruirui, Yuan Bo, Chao Zhengyi, Kang Hongyuan, Guo Jing;
Atomic layer deposition can achieve precise control of the surface of micro-nano particles in the sub-nanometer or even smaller scale range, but it is difficult to achieve the modification of a large number of micro-nano particles in traditional static atomic layer deposition reactors. To address this problem, this paper independently built a vibrating fluidized atomic layer deposition reactor with an inner diameter of 26 mm and a height of 350 mm to achieve batch and precise modification of micro-nano particles. Micro-nano particles are easy to agglomerate and difficult to fluidize. To achieve uniform deposition, the problem of uniform fluidization of micro-nano particles must be solved first. In this paper, TiO_2 and SiO_2 particles with a particle size of 20 nm and resin particles with a particle size of 600μm were selected to study their fluidization behavior in a self-made vibrating fluidized bed reactor. When the initial bed height is 15 mm, due to the difference in viscosity between agglomerates, the pressure drop of the TiO_2 bed is high and the bed expansion rate is low during stable fluidization, which is bubbling fluidization, while the pressure drop of the SiO_2 bed is low and the bed expansion rate is high, which is bulk fluidization. Through the SiO_2 particle tracer fluidization experiment, it is inferred that the nanoparticle agglomerates are in a dynamic change of continuous rupture and aggregation during the fluidization process. Vibration can improve the channeling and agglomeration phenomena in the fluidization process of nanoparticles, promote the breakup of agglomerates, reduce the minimum fluidization velocity of particles, and help improve the gas-solid contact efficiency. In order to test the coating performance of the homemade atomic layer deposition reactor on nanoparticles, TiCl_4 and H_2O were used as precursors and TiO_2 nanoparticles were used as substrates under the optimal fluidization conditions. Atomic layer deposition was performed at 80°C to obtain a dense and uniform amorphous TiO_2 film to shield the photocatalytic activity of the pigment TiO_2 and improve its weather resistance. After 30 atomic layer deposition cycles, the thickness of the amorphous TiO_2 film reached 3.11 nm, and its photocatalytic activity was inhibited by about 90% compared with that of anatase TiO_2, and the photocatalytic activity shielding effect was significant. The experimental results show that the homemade vibration fluidized atomic layer deposition reactor can achieve precise modification of micro-nanoparticles and has good application prospects.
Chai Meng, Lou Huifang, Ban Yao, Niu Bingkang, Guo Jing;
Increasing the exposure ratio of high-energy {001} crystal planes can effectively improve the photocatalytic activity of TiO_2. Crystal plane controllers are usually used to expose high-energy {001} crystal planes, but the morphology of synthesized TiO_2 is unstable due to various factors, and the particle size is large. It is necessary to achieve stable synthesis of TiO_2 nanocrystals on the basis of reducing the particle size. Based on this, this paper uses potassium titanate nanowires as precursors to stably synthesize TiO_2 nanocrystals. First, the octahedral bipyramidal morphology of TiO_2 is ensured by acid treatment, and the particle size of TiO_2 nanocrystals is reduced by reducing the K+ content in potassium titanate nanowires. Studies have shown that TiO_2-H7 synthesized with potassium titanate nanowires as precursors and acid treated for 7 h has the strongest photocatalytic activity, and its particle size is nearly 86% smaller than that of the untreated TiO_2. The activity of degrading methylene blue is 7.4 times that of the original anatase TiO_2. On this basis, (NH_4)_2CO_3 was used as a morphology control agent, and the exposure ratio of the {001} crystal plane was adjusted by changing the (NH_4)_2CO_3 concentration to further improve the photocatalytic activity of TiO_2 nanocrystals. The photocatalytic activity of the (NH_4)_2CO_3 series without acid treatment can be effectively improved after the crystal plane adjustment. When the concentration of (NH_4)_2CO_3 was 0.14 mmol/L, the synthesized TiO_2-0.14 was 2.6 times the degradation activity of the original anatase TiO_2. After the introduction of (NH_4)_2CO_3 to adjust the crystal plane of potassium titanate nanowires treated with acid for 7 h, the photocatalytic activity was inhibited. This may be due to the insufficient exposure ratio of the {001} crystal plane, and the particle size increased with the increase of (NH_4)_2CO_3 concentration, and the oxygen vacancy content decreased accordingly. This study found that acid treatment of potassium titanate nanowire precursors can synthesize octahedral bipyramidal TiO_2 nanocrystals and reduce the particle size, thereby improving the photocatalytic activity of TiO_2. However, the degree of acid treatment will affect the subsequent crystal face regulation {001} crystal face exposure ratio and nanocrystal crystallinity. In order to further improve the photocatalytic activity of TiO_2 nanocrystals, it is necessary to further explore the structure-activity relationship between acid treatment and crystal face regulation.
Liu Xuan; Yang Shaoqiang; Fu Yuping;
In order to select the optimal spatial interpolation method for radon detection of hidden fire sources at a local observation scale, this paper takes the sample data of the fire area of Beiyan Coal Mine as the research object. Based on the Arc GIS geostatistical analysis module, the interpolation model with the smallest error under four interpolation methods is quantitatively screened out through cross-checking, and the spatial distribution characteristics of different models are qualitatively compared. The optimal interpolation method is determined and verified by drilling. The results show that the interpolation model with the smallest error among the four methods is the power exponent p=1 of the inverse distance weighted method, the thin plate spline function model of the radial basis function method, and the Gaussian function model of the universal kriging method and the ordinary kriging method. According to the interpolation and extrapolation surface distribution characteristics of the spatial distribution map and the drilling verification results, the interpolation methods are ranked as ordinary kriging method (Gaussian function) > universal kriging method (Gaussian function) > radial basis function method (thin plate spline function) > inverse distance weighted method (p=1), and the ordinary kriging method (Gaussian function) is preferred for interpolation to obtain the difference information of surface radon distribution in the coal spontaneous combustion fire area.
Electronics and Electronic Information
Yang Yang;Du Hongmian;Guo Jinjie;Wang Ruhao;
The explosion shock wave parameters are one of the main bases for evaluating the power of ammunition. However, in the actual test process, the test system may be damaged by fragments or other factors, making it impossible to capture the complete signal, affecting the subsequent damage assessment. To address this problem, this paper proposes a method based on a bidirectional long short-term memory network (BiLSTM) fused with a multi-head self-attention module to construct the integrity of the incomplete shock wave signal. The BiLSTM is used to analyze the local temporal dependency of the shock wave signal, and the multi-head self-attention module is used to capture the frequency information in the signal. Finally, the fusion of the timing signal and the frequency information is realized to obtain a complete shock wave signal. In an information collection process, the measured signal data is usually only dozens of groups, which leads to the problem of small samples. This paper establishes a GAN network with LSTM units as generators to expand the complete shock wave signal and enhance the data set capacity. The experimental results based on the expanded data set show that the MSE and MAE between the complete signal constructed by the proposed method and the original signal are 0.006 8 and 0.146 2 respectively, which are better than LSTM, BiLSTM, CNN+BiLSTM and other methods. The proposed method can meet the actual needs of constructing incomplete shock wave signals.
Wan Chao;Xie Rui;
Aiming at the large error of high-speed signal state judgment of fuze control system and the shortcomings of BP (Back Propagation, BP) neural network in the state judgment process, such as poor accuracy and easy to fall into local optimal solution, other optimization algorithms are used to improve the shortcomings of BP neural network to reduce the error of high-speed signal state judgment. This paper uses genetic algorithm to optimize BP neural network to construct a model, and establishes a classification model with the high-speed signal time and voltage of the fuze as input indicators. It is used to judge the state of high-speed signals to improve the recognition accuracy, accelerate the convergence speed, reduce the error, and understand the state of the fuze control system at each moment according to the high-speed signal to judge whether the system is normal and reliable. The simulation analysis results show that the proposed method has the characteristics of excellent recognition results, fast convergence speed and small error in the judgment of the high-speed signal state of the fuze. Its accuracy rate reaches 99.6%, which is better than 88.6% of BP neural network and 98.7% of convolutional neural network. At the same time, the mean absolute error is reduced to 0.012 10, the mean square error is reduced to 0.043 68, the root mean square error is reduced to 0.209 01, and the evolutionary generation is 23 generations, which is better than 0.168 42, 0.319 85, 0.564 75, 51 generations of BP neural network and 0.022 63, 0.060 5, 0.245 97, 25 generations of convolutional neural network. The continuous experimental results show that the improved model has better robustness. The Wilcoxon rank sum test results also show that the improved model has better recognition effect than BP neural network and convolutional neural network, and has better generalization ability. The model meets the requirements of high-speed signal state judgment.
Li Ning; An Kun; Guo Lishan; Li Sen; Meng Jiang;
Giant magnetostrictive material (GMM) is a new type of functional material. It is widely used in energy harvesting, micro-displacement drive, precision positioning control and other fields because of its advantages such as large magneto-mechanical coupling coefficient, fast response speed and good frequency response characteristics. However, the complex hysteresis nonlinearity of the material affects the positioning accuracy of its actuator. In order to identify the hysteresis nonlinearity in giant magnetostrictive materials, this paper proposes a new Hammerstein model modeling method. The advantage of this method is that the model can better approximate the hysteresis nonlinearity, provide higher accuracy, and reduce the workload of parameter identification in the series link. First, an extreme learning machine model based on the hysteresis operator expansion space of hyperbolic function is constructed to represent the static nonlinear part in the new Hammerstein model. Secondly, the weights and bias parameters of the fully connected layer of the extreme learning machine model are extracted to construct the state space equation of the dynamic linear part in the new model, which reduces the model parameter identification work of the series link in the traditional model. Finally, a new Hammerstein model that can describe the hysteresis characteristics of giant magnetostrictive materials is established. The relative error of the new Hammerstein model is 0.86% to 3.69%, and the average absolute error is 2.63%, which is about 0.8μm lower than the root mean square error of the traditional Hammerstein model, and the average absolute error is increased by nearly 4%. The simulation results prove the effectiveness of the new Hammerstein model in modeling the complex hysteresis characteristics of giant magnetostrictive materials.
Zhang Yanfei;Yao Shuncai;Wang Jianfei;
The measurement results of embedded multi-channel radiation dosimeters are easily disturbed by the system itself and environmental noise, which in turn affects its measurement accuracy. This paper proposes a fuzzy Kalman small-range adaptive filtering method to improve the measurement accuracy. This method first extracts the filter residual of the Kalman filter as the filter relative error, and sets a reasonable threshold for the filter relative error by analyzing the environmental noise characteristics of the radiation dosimeter. On this basis, a fuzzy controller is designed. The difference between the filter relative error and the error threshold is used as the input of the fuzzy controller, and the change of the process noise covariance Q of the Kalman filter is used as the output. The basic domain of fuzzy logic input and output is set, and the membership function is established. Fuzzy control rules are formulated based on measurement experience to achieve small-range adaptive adjustment of the process noise covariance Q with the filter relative error in the dynamic radiation stage and the constant radiation stage, thereby improving the measurement accuracy of the dosimeter in the two stages. The dosimeter is placed in an environment with a dose rate of 0μSv/h~15 Sv/h. The maximum errors of the dynamic and constant radiation dose rates measured by the method described in this paper are 4.3% and 8.3% respectively, and the cumulative dose error of 180 s is about 10%. Compared with the sliding average filtering method and the Kalman filtering method based on residual judgment, the fuzzy Kalman small-range adaptive filtering method proposed in this paper has more advantages in embedded multi-channel dosimeters.
2024 Issue 05 v.45;No.
