Abstract:Self-supervised monocular depth estimation methods trained on sequences of monocular images have received considerable attention in recent years by using the photometric consistency loss between adjacent frames instead of depth labels as the supervisory signal for network training. The photometric consistency constraint follows the static world assumption， but the moving objects in the monocular image sequence violate this assumption， which affects the camera pose estimation accuracy and the calculation accuracy of the photometric loss function during the self-supervised training process. By detecting and removing the moving target area， the camera pose decoupled from the target motion can be obtained， and the in?uence of the moving target area on the calculation accuracy of the photometric loss can be discarded. To this end， this paper proposes a self-supervised monocular depth estimation network based on semantic assistance and depth temporal consistency constraints. First， an offline instance segmentation network is used to detect dynamic category objects that may violate the static world assumption， and the corresponding region input pose network is removed to obtain a camera pose decoupled from object motion. Secondly， based on semantic consistency and photometric consistency constraints， the motion status of dynamic category targets is detected so that the photometric loss in the moving area does not affect the iterative update of network parameters.Finally， depth temporal consistency constraints are imposed in non-motion areas， and the estimated depth value of the current frame is explicitly aligned with the projected depth value of adjacent frames to further refine the depth prediction results. Experiments on the KITTI， DDAD and KITTI Odometry datasets verify that the proposed method has better performance than previous self-supervised monocular depth estimation methods.

Abstract:Edge detection is the basis of many computer vision tasks. Current techniques mainly rely on deep learning， but most models improve the accuracy of predicted edges using Non-Maximum Suppression in the evaluation stage. These models only focus on the accuracy of predicted edges without considering the coarseness and fineness of the edges. To address this issue， this paper proposes a new feature fusion strategy based on the dense extreme inception network. This strategy incorporates top-down multi-scale fusion edge detection with spatial co-location rule and retains the multi-network structure based on the traditional deep learning edge detection side output. The proposed strategy can better integrate the high semantic characteristic of high-layer information with the high-resolution texture characteristic of low-layer information， thereby suppressing pixel confusions in backgrounds and lines that are predicted incorrectly in edge detection. In the feature connection， Concat block is used to replace the single operation of Concat， to better fuse semantic information in different scales. Lastly， a simple attention fusion block is used to fuse outputs of multiple networks. Also， different output prediction maps at different scales are deeply supervised combining the tracing loss. This model is independent of Non-Maximum Suppression. By fully utilizing the multi-scale and multi-level information of the target image， this model improves the accuracy of prediction along with improving images’ edges. The experimental results show that without the morphological Non-Maximum Suppression， on the BIPED data set， the proposed model on ODS， OIS， and AP are 0.891， 0.895， 0.900， respectively； with the morphological Non-Maximum Suppression， the proposed model on ODS， OIS， and AP are 0.894、0.899、0.931， respectively， which is superior to all comparison algorithms involved in this article. Also， on the MDBD data set， optimal results were also achieved.

Abstract:For the problem of insufficient accuracy of traditional fuzzy C-means clustering （FCM） algorithm for noise-containing multi-spectral remote sensing image segmentation， an FCM multi-spectral remote sensing image segmentation algorithm combining adaptive local fuzzy subspace and enhanced KL is proposed. Firstly， the local fuzzy factor is used to automatically eliminate the noise interference and extract the local spatial information of the image by similarity metric and adaptive constraint parameters without relying on parameters. Secondly， the original image information and the local spatial information processed by the fuzzy factor are unified and integrated into the fuzzy subspace clustering， and the multiple channels of the image are adaptively weighted to enhance the segmentation accuracy. Finally， the KL information is introduced into the FCM objective function in the form of regular terms for clustering calculation， and the outliers in the membership matrix are removed by ESD （Extreme Studentized Deviate） detection model to enhance the KL prior information and reduce the ambiguity of the membership. The experiments of real multi-spectral remote sensing image segmentation show that in the simulation of noise environments， the algorithm in this paper can suppress the noise and can guarantee the segmentation accuracy better at the same time. In addition， the algorithm in this paper outperforms several other variant FCM algorithms in terms of evaluation indexes such as segmentation accuracy， fuzzy coefficient， and peak signal-to-noise ratio.

Abstract:To address the issue of texture detail blurring in single-scale sampled compressed sensing image reconstruction at low sampling rates and achieve a lightweight reconstruction network， this paper proposes a filter pruning based multi-scale compressed sensing image reconstruction network. In the sampling phase， the image is linearly decomposed by convolution， and then fused with the input image and different scale decomposition features to obtain the compressed sensing measurements. In the reconstruction phase， a coordinate attention based multi-scale dilated residual module is designed， which incorporates positional information into channel attention to enhance the feature learning ability of the network. Moreover， by calculating the entropy of the feature map to judge the importance of the filters， the less important filters is pruned to achieve the purpose of compressing the model. Training and testing are carried out on datasets such as DIV2K， Set5， BSDS68 and Urban100. The experimental results show that the algorithm proposed improves the Peak Signal-to-Noise Ratio （PSNR）， Structural Similarity（SSIM）， and image visual effects. For instance， with a sampling rate of 4% and a test set of Set14， the proposed algorithm improves the PSNR of the reconstructed image by 4.17 dB and 2.39 dB， respectively， compared with CSNet+ and FSOINet， resulting in clearer texture details. Under the premise of slightly reducing the reconstruction effect， a lighter model was obtained， which improved the reconstruction speed.

Abstract:This paper designs and implements an improved personal location beacon （PLB） for maritime search and rescue. The PLB has the functions of Beidou navigation and positioning， Beidou short message communication， human vital signs monitoring and reporting， and water touch trigger. What’s more， the proposed PLB supports the COSPAS-SARSAT system. To extend the working time of the PLB， this paper introduces a dynamic power management method to conduct the power channels of different functional circuits. Meanwhile， this paper designs an ultra-low power consumption standby circuit and low power consumption software control scheme to reduce power dissipation. Finally， the prototype of the PLB is tested， respectively， in the adjacent sea areas of Weihai City in Shandong Province and Ningbo to Zhoushan City in Zhejiang Province. The test results of the prototype show that in the standby state， the current of the PLB is 2.3 A， with a standby time exceeding 5 years； in the state of receiving the Beidou short message， the current is 79 mA； and in the state of transmitting rescue messages， the 7 000 mAh 7.4 V battery can send 200 distress messages. Compared to existing products， the proposed PLB in this paper for maritime rescue has lower standby power consumption.

Abstract:When the global navigation satellite system （GNSS） signal is unavailable， the performance of GNSS/inertial navigation system （INS） integrated navigation system significantly degrades， which leads to the rapid divergence of INS errors of UAV swarms. At present， the neural network model is used to predict the position and speed instead of GNSS navigation information to correct the positioning error of the INS. However， this method suffers from high positioning errors and a sharp decline in prediction accuracy when the trajectory changes suddenly. Therefore， a position and velocity prediction method based on convolution neural networks （CNN） -bidirectional long short-term memory network （BiLSTM） joint residual compensation model is proposed to compensate for inertial navigation errors and improve position and velocity positioning accuracy. Firstly， aiming at the problem of high positioning error of GNSS/INS integrated navigation system after GNSS denial， a time series prediction network is formed by fusing CNN and BiLSTM to train and establish the relationship between inertial measurement unit （IMU） dynamics measurement and GNSS information， so as to realize position and speed prediction. Secondly， aiming at the problem that the prediction effect drops sharply when the trajectory changes abruptly，CNN-BiLSTM is used again to mine the relationship between the IMU dynamics measurement， prediction value and prediction residual， and to predict and compensate the prediction residual. Simulation results show that the proposed model outperforms traditional CNN-LSTM and LSTM network models in terms of prediction accuracy， effectiveness， and stability.

Abstract:Now，almost every processor architecture has added support for SIMD （single instruction multiple data） instructions. SIMD instructions can perform the same operation on a set of data simultaneously， enhancing the processing performance of the processor through data parallelism. However， most dynamic binary translators ignore the use of native SIMD instructions and instead simulate floating-point computations in software languages. This paper proposes a framework called FP-QEMU， based on QEMU translation system. FP-QEMU adopts SIMD instructions to optimize and replace floating-point calculation instructions， and completes a complete floating-point implementation on X86 and ARM benchmark platforms. The framework can identify the optimization opportunities of floating-point computation acceleration in dynamic binary translation system and use SIMD instructions to achieve the effect of improving the translation performance of dynamic binary translation system. Using SPEC 2006 as the benchmark， experiments show that compared with QEMU， FP-QEMU cross-platform ARM applications running on X86 computers can achieve a maximum speedup of 51.5% and an average speedup of 37.42%.

Abstract:Aiming at the speed control problem of the train under external disturbance and uncertain dynamics， a composite control scheme combining integral back-stepping （IBS） method and linear active disturbance rejection control （LADRC） is designed. Firstly， considering the strong coupling of the train， a multi-particle model with time-varying coefficients is established to better conform to the real longitudinal dynamic characteristics and force conditions of the train. Secondly， to reduce the difficulty of parameter adjustment， the tracking differentiator （TD） and the extended state observer （ESO） are in linear form. TD is used to obtain the differential signal and has a filtering effect. The problem of differential explosion in back-stepping method can be solved by using the TD to derive the virtual control quantity. ESO is used to estimate the total disturbance in real-time. In addition， the IBS method is used to improve the error feedback control law， and an integral back-stepping linear active disturbance rejection control （IBS-LADRC） algorithm is designed. Finally， the convergence of the observation error and the stability of the closed-loop system are proved. Combined with the parameters of AH electric multiple units of Hangzhou Metro Line 6 and the actual line data， the simulation comparison is carried out， and the IBS-LADRC is compared with the back-stepping method， LADRC algorithm and PID control.The results show that under the IBS-LADRC method， the velocity error of each power unit is within ±0.04 km/h， the acceleration is within ±1 m/s2， and the acceleration and velocity error change smoothly. The coupler force is the smallest and the change is the most gentle compared with the other three methods， and the maximum coupler force is only 2 320 N. The proposed control strategy has high tracking accuracy for the expected speed of the train， which is conducive to ensuring the safety of the coupler， preventing the coupler from breaking， and improving the safety， stability and passenger comfort of the train operation.

Abstract:To address the difficulty in evaluating and quantifying the value of personalized customized fuel cell test bench （FCTB）， a method is proposed to establish a comprehensive evaluation index system of FCTB， based on the principle of FCTB technology and the actual production process. The fuzzy hierarchical analysis method is used to assign weights to the indexes， taking into account the user’s personalized customization needs. This creates a quantitative value evaluation model， which converts the weighting problem into a constrained optimization problem. An improved harmonic search algorithm is then proposed to solve this problem. By designing a solution vector generation mechanism and a parameter adaptive adjustment strategy， the algorithm improves the solution efficiency and search capability of the traditional harmonic search algorithm. Simulation results show that the proposed method has obvious advantages in terms of computational efficiency and accuracy. It is able to achieve quantitative value assessment of different FCTB scenarios according to the user’s personalized needs.

Abstract:A linear negative feedback power amplifier （PA） based on gallium arsenide （GaAs） heterojunction bipolar transistor （HBT） in the C-band is proposed. The design employs a three-stage common emitter （CE） structure and utilizes two different active linear biases to enhance the linearity of the PA. Simultaneously， an RLC negative feedback network is incorporated to improve stability and broaden the operational bandwidth. Addressing the issue of gain reduction in traditional feedback network structures， the RLC negative feedback network can effectively mitigate the impact of gain reduction induced by feedback by adjusting the inductance values within the feedback network. Test results demonstrate that， at room temperature， within the frequency range of 5.1 GHz to 7.4 GHz， a gain exceeding 28 dB is achieved. In the linear operating frequency range of 5.9 GHz to 7.1 GHz， the average gain is 29.5 dB， and both S11 and S22 are less than -10 dB. Complying with the wireless LAN standard 802.11a， a 20 MHz 64-QAM signal is utilized and the output power achieving an EVM of -30 dB ranges from 18.9 dBm to 22.5 dBm. Between 5.9 GHz and 6.2 GHz， the saturated output power exceeds 30 dBm， and the maximum power added efficiency （PAE） is greater than 35%.

Abstract:Aiming at the development trend of integrated circuits （ICs） with rapid increase in scale and power consumption， a novel capacitor-less low-drop-out linear voltage regulator （CL-LDO） has been designed to provide a wide range of load currents. To solve the issues of stability and transient response caused by the requirements of a wide range of load currents and no off-chip capacitor， a dynamic zero-point compensation method and a transient enhancement circuit structure are proposed， which not only ensures the stability of the whole circuit in the full load range， but also achieves excellent transient characteristics. Based on 0.11 μm CMOS technology， the circuit design， layout design and simulation are completed. The simulation results show that the overall loop gain can reach 68 dB with a minimum phase margin of 56° within the load range of 0 mA to 500 mA， the output overshoot and undershoot are 56 mVand 141 mV， and their settling time is 2 μs and 0.78 μs， respectively， when load current transients range from 1 mA to 500 mA （Δt=500 ns）， the power supply rejection （PSR） is -67.2 dB@1 kHz， and the load regulation rate is 0.137 μV/mA.

Abstract:The research on PV aggregation scenarios is the basis and premise for realizing the joint planning of distributed PV and energy storage in regional power grids. It plays a significant role in promoting the full consumption of distributed PV in regional power grids. By analyzing the historical PV data based on the improved K-means++ algorithm aggregation， the typical PV output scenarios were generated. Starting from the distribution network economy， environmental protection and reliability， a PV and energy storage siting and capacity planning model considering the aggregation of photovoltaic scenarios was established. And an analytic hierarchy process （AHP） was introduced to transform the multi-objective optimization problem into a single-objective problem. The multi-objective optimization problem was transformed into a single-objective problem and solved by particle swarm optimization （PSO）. The results of the case study show that the proposed PV-scenario aggregation model more accurately depicts the uncertainty of PV output， verifying the effectiveness and feasibility of the model.

Abstract:Proton exchange membrane fuel cells （PEMFCs） can achieve higher power by expanding the reaction area of the stack active area. However， for stacks with enlarged reaction areas， there is a tendency for increased non-uniformity in the distribution of membrane electrode assembly （MEA） stress， leading to a decline in the electrochemical performance of fuel cells. In this study， four different structural sizes of PEMFC stacks were designed. Using a combination of the equivalent stiffness model and finite element software， the impact of stack structures with expanded reaction areas on the uniformity of stress distribution of the MEA was analyzed. Furthermore， the installation position of steel belts within the stack was optimized to enhance the uniformity of internal contact pressure distribution. The research results indicate that the uniformity of MEA contact pressure distribution is particularly sensitive to changes in the reaction area width. When the size of the active area is widened， the standard deviation of the average stress within the stack’s internal active region increases by 23.2%. Conversely， when the active area is lengthened， or both lengthened and widened simultaneously， adding a corresponding bundled steel belt reduces the standard deviation of the average stress within the stack’s internal active area by 8.6% and 8.7%， respectively. This suggests that appropriately increasing the number of bundled steel belts can improve the uniformity of contact pressure distribution within the stack. Additionally， optimization results for belt positioning show that the more the outer steel belt is positioned closer to the end plate side， the more uniform the stress distribution within the stack’s internal active region.

Abstract:In light of the current research on Battery Management System （BMS） circuit boards for electric vehicles， which only takes into account the temperature field and heat dissipation effects of individual functional module but lacks consideration of the interdependence and cooperative effects under the temperature and force fields among multiple functional modules of the BMS， a certain commercial BMS circuit board is taken as the research object， the software ANSYS is used to construct a thermal-stress coupling numerical simulation analysis model that characterizes the synergistic action of multiple BMS modules， and the effectiveness of the model is verified. On this basis， a numerical simulation study was carried out on the temperature field and thermal deformation behavior of each functional module of the BMS circuit board. The results show that the temperature distribution of the BMS circuit board is uneven， and the maximum temperature difference reaches 20.5 ℃. The balancing module generates heat， and the temperature reaches 54.4 ℃， which causes thermal expansion and deformation of circuit board components. At the same time， thermal stress concentration occurs at the chip resistors on the edges of the balancing module and the power supply module due to the circuit board constraint， which causes the balancing module and the power supply module of the BMS to produce protruding and warping deformation. The Z-axis thermal deformation increases with the temperature rising， and the maximum deformation amount reaches 9.5 μm. Therefore， the heat dissipation optimization design should be carried out for the heat-generating modules on the BMS circuit board.

Abstract:In engineering practice， the parameters of the electric vehicle mounting system inevitably have certain uncertainties and correlations. Firstly， a polygonal convex set model based on principal component analysis is introduced to effectively deal with the complex situation where the uncertain parameters of the system are correlated and independent at the same time； then， an uncertainty analysis method for the inherent characteristics of the mounting system is proposed in combination with the Monte Carlo method， and the specific analysis steps of the method are given； finally， the method is applied to the analysis of the mounting system of an electric passenger car to verify the effectiveness of the method. The results of numerical analysis show that the response range of the inherent characteristics of the system obtained by the proposed method is more reasonable than that of the interval method without considering the parameter correlation； compared with the analysis method based on the multidimensional parallel hexahedron model， the proposed method can more effectively deal with the case of irregular boundary distribution of the uncertain parameter samples of the system； for the studied model， the correlation of the stiffness parameters of the right and front mounting points has a greater influence on the inherent characteristics， which should be paid attention to in the design research process.

Abstract:To realize real-time evaluation of the steering pump motor’s operating performance， a physical model of the steering pump motor is firstly constructed in this paper， which considers the operating mechanism of the steering pump motor. By solving the pump motor’s external characteristic equation， hydraulic circuit balance equation， and internal flow calculation equation， etc.， the preliminary evaluation and calculation of the internal unmeasurable parameters can be realized. Then， the influence mechanism of various parameters on the performance of the steering pump motor is obtained. At the same time， LSTM deep learning method is used to build a data-driven model of the steering pump motor to learn and predict the variation rule of the volume efficiency of the steering pump motor under different working conditions. The volume efficiency is then input into the physical model to form a mixed model of steering pump motor performance evaluation， so as to improve the accuracy and applicability of steering pump motor performance evaluation under different working conditions. The example analysis shows that the hybrid model can calculate and evaluate the internal and external parameters such as internal pressures， flow rate， speed and torque of the steering pump motor with high accuracy under complex working conditions， and the model error is less than 10%.

Abstract:In this paper，a model predictive control （MPC） method integrating feed-forward and state feedback is proposed for the problem of accurate path tracking for intelligent vehicles with dynamic constraints. Firstly， the MPC path tracking base model is established according to the vehicle two-degree-of-freedom model， and then， the modeled steady-state perturbations generated by the road curvature changes on the system in the base model are considered and designed to be eliminated by feed-forward control （FFC）； Furthermore， the proportional integral derivative （PID） controller is used to regulate the system error state feedback； Meanwhile， the model predictive optimal regulation control law （MPC-FF-PID） is verified by integrating the feed-forward and state feedback corner inputs. Finally， the effectiveness of the proposed algorithm is confirmed based on MATLAB/Simulink and Carsim platforms， and a real vehicle test is carried out in the low-speed scenario in the park based on the intelligent driving real vehicle platform， and the maximum lateral and heading errors are 0.128 7 m and 0.063 9 rad， respectively， indicating that the proposed algorithm has higher tracking accuracy and safety.

Abstract:Aiming at the path planning problem of four-wheel independent steering （4WIS） vehicles， a graph search algorithm integrated with mode decision is proposed. Firstly， three motion modes of 4WIS vehicle are modeled， and the motion characteristics of each motion mode are analyzed. Based on this， a multi-mode node expansion strategy is designed to realize the integration of 4WIS vehicle multiple motion modes and path planning. Then， a multi-objective cost function is designed to guide 4WIS vehicles to switch motion modes reasonably and generate a smooth path for optimal node selection and motion mode decision-making. Finally， the simulation experiment is carried out on MATLAB software， and the proposed algorithm is tested in various scenarios to verify the feasibility and effectiveness. The results show that the proposed algorithm considers the optimization combination and mode switching of three motion modes in path planning， and can achieve the optimal motion mode sequence and the shortest path planning. This algorithm has high solving efficiency and excellent planned path， and can fully utilize the high flexibility and trafficability of 4WIS vehicles， effectively solving their path planning problem.

Abstract:A visual inertial localization algorithm integrating semantic information is proposed to address the positioning problems caused by poor GPS signals， dim lighting， limited features， and weak textures in underground parking lots. Firstly， this algorithm fuses visual inertial information through visual odometry and IMU pre-integration. Simultaneously， a panoramic surround view image is constructed using four fisheye cameras， and semantic segmentation algorithms are employed to extract semantic information from the parking environment. Then， the semantic feature projection map is obtained through inverse projection transformation based on the tightly coupled visual inertial pose. Additionally， loop detection and pose graph optimization are employed to reduce accumulated errors and achieve global pose graph optimization， thereby achieving higher localization accuracy. This paper verifies the proposed algorithm through Gazebo simulation and real vehicle testing. The results indicate that this algorithm can fully utilize the semantic information of the environment to construct a complete semantic map and achieve higher vehicle localization accuracy than ORB-SLAM3 based on repeated localization error comparisons.

Abstract:To study the influence of geometric characteristic parameters and forming process parameters of automobile stamping die on the sharp-edged wear and realize the accurate prediction of the die sharp-edged wear， a prediction method of the die sharp-edged wear based on improved SVR algorithm was proposed in this paper. By using the improved Latin hypercube sampling （ILHS） method， the experimental samples of finite element calculation of die sharp-edged wear were obtained， and the input parameter set of the prediction model was then constructed. The chaos theory and dynamic weights were introduced into the grasshopper optimization algorithm （GOA）， and the improved grasshopper optimization algorithm （IGOA） was used to improve key parameters of the SVR algorithm. Based on the IGOA-SVR algorithm， the prediction model of die sharp-edged wear was constructed， which was combined with the particle swarm optimization （PSO） algorithm to establish a multi-objective optimization model so as to realize the high-precision prediction as well as the optimization of geometric characteristic parameters and forming process parameters. Compared with five existing conventional prediction models， the prediction errors of the prediction model based on IGOA-SVR at the sampling point were 8.546%， 8.497%， and 8.473%， respectively， which were 25.9%， 26.2%， and 26.4% higher than the GOA-SVR prediction model， respectively， and the prediction accuracy was also improved to varying degrees compared with other prediction models. The results show that the improved IGOA-SVR has higher accuracy.

Abstract:To reveal the characteristics of vortex-induced vibration （VIV） responses of an underwater manipulator in a uniform flow， an experimental setup was constructed to test VIV in the uniform flow field of an underwater manipulator by utilizing the relative motion between the moving underwater manipulator and still water. Vibration displalement response at various positions on the underwater manipulator were collected at different reduced velocities. The experimental results indicate that as the reduced velocity increases， both the dominant frequency of vibration and the dimensionless amplitude exhibit an increasing trend. The dimensionless amplitude in the cross-flow direction is less than 0.04D， while the in-line direction shows significant multi-frequency characteristics and higher amplitudes， indicating that the underwater manipulator experiences more noticeable resistance in the in-line direction. The standard deviation of displacement first increases and then decreases with the increase in testing height， and the spatial distribution of the standard deviation of displacement in the cross-flow direction shows more pronounced symmetry compared to the in-line direction， with the first mode being dominant. The VIV in the cross-flow direction follows a strong regularity， generally conforming to the Strouhal pattern observed in cylinder flow.The dimensionless dominant frequency in the in-line direction is between 0.3 and 0.8， which is slightly higher than that in the cross-flow direction， with the Strouhal number being approximately 1 to 1.7 times that of the cross-flow condition. No frequency lock-in phenomenon was observed， but flow velocity and testing height have some influence on the VIV of the underwater manipulator. This study provides a theoretical basis for constructing accurate hydrodynamic models to ensure the precise positioning and control of underwater manipulator.

Abstract:Aiming at the characteristics of the explosive electromagnetic radiation signal of high-energy warhead， such as wide frequency band， large change amplitude， long duration and complex measurement conditions， a measurement system of explosive electromagnetic radiation based on mixing antenna is designed. Four kinds of measuring antennas are designed： short-wave antenna （1.5~30 MHz）， ultra-wideband antenna （30~1 000 MHz）， microstrip antenna （5.9~6.0 GHz， 8.4~8.5 GHz）. By optimizing the antenna structure and adding matching circuits， the measurement efficiency is improved， which can cover all electromagnetic sensitive frequency bands of ordnance equipment. The signal conditioner with many functions is designed by adopting a modular approach， which includes combiner， amplifier， limiter and filter. The amplifier module adopts the structure of "fixed gain amplifier & step attenuator"， which not only realizes the adjustment of 0~30 dB in the full frequency band， but also improves the accuracy of the adjustment system to 0.5 dB. The limiter module limits the sampling signal power to less than 65% of the maximum withstand power of the data acquisition instrument， which improves the adaptability of the measurement system to the electromagnetic signal of explosion field strength. The measurement system can meet all the indexes of high-energy warhead explosion electromagnetic radiation measurement by adopting such technical means as mixed-band antenna combination， high-precision coefficient adjustment， and multi-function signal conditioning.

Abstract:To investigate the influence of altitude on smoke propagation of train fire accident in tunnels，based on the validated numerical simulation method， the sliding grid technology was used to achieve the entire motion process of a train running at a uniform speed in a tunnel， decelerating after a fire accident and eventually stopping in the tunnel. The three-dimensional unsteady N-S equation and RNG k- ε turbulence model were adopted， and the reliability of the numerical simulation method was verified by the datas of the scaled train fire accident test in tunnel. The results indicate that the altitude has a significant impact on the smoke propagation after fired trains stop in the tunnel. Compared with low altitude tunnels at a time， the smoke propagation velocity at the same location as the tunnel vault is faster in high altitude tunnels， and the time when the smoke countercurrents to 50 m upstream of the fire source is earlier. The temperature distribution around the train also shows a rising trend with the increasing altitudes. Compared with the train fire accident in the tunnel at an altitude of 0 m， the time is 102 seconds earlier when the smoke in the tunnel at an altitude of 5 000 m countercurrents to 50 m upstream of the fire source. At the same time， the peak value of temperature at the tunnel vault is increased by 216 K， with an increase of 42.9% when the fired train stopped for 360 seconds.