Search

In recent years, semantic segmentation based on deep convolutional neural networks has developed rapidly. However, it is still a challenge to balance the computing cost and segmentation performance for the current semantic segmentation methods. This paper proposes a lightweight real-time semantic segmentation model Balanced Sample Distribution Network (BSDNet), to solve this problem. In BSDNet, we introduce the Balanced Sample Distribution Module (BSDModule) to balance the sampling distribution of convolution and obtain features with a larger receptive field. To optimize the segmentation effect, we introduce a Shuffle Channel Attention Module (SCAModule) to enhance the interaction of channel features at the cost of a small number of parameters. BSDModule and SCAModule are lightweight and flexible and can adapt to different types of network structures. Extensive experiments on CityScapes and CamVid show that the proposed method can balance the computing cost and segmentation performance. Specifically, on the CityScapes with $512\times1024$ resolution, BSDNet-Xception39 achieves 68.3% MIoU and 84.6 FPS with only 1.2M parameters.

To satisfy the requirement of the RF output in mega-watts level of the 170GHz gyrotrons. A design of a two-stage depressed collector with the radial magnetic field for 1MW, 170GHz gyrotrons for increasing the efficiency and the power capacity of the tube has been presented. The simulation results indicate that the total efficiency of the tube with the design could increase from 31% to 89% significantly.

A 140GHz megawatt-class open resonator cavity gyrotron for electron cyclotron heating and current drive in the International Thermonuclear Experimental Reactor (ITER) is under development. TE 22,6 mode open resonator cavity has been chosen as the operating mode of gyrotron in our scheme. This report presents the effect of length and radius of three sections and angular of two slopes in cavity from oscillation power and frequency by self-consistent simulation. The calculation result has a good agreement with the preliminary test results recently.

A water cooling system has been designed aiming to insure the 1 MW 170GHz gyrotron stable operates under continuous wave (CW) operating condition. In this work, thermal loss power of main parts is calculated by theory analysis and software simulation. Considered to maximum allowable heat-flux, the water cooling system controls the maximum temperature below the boiling temperature and the deformations of the geometry are significantly small enough not to influence the performance heavily.

A study of a novel directional coupler for measuring the output power of a W-band gyro-TWT has been presented. The designed directional coupler is a three ports device, in which, the multi-apertures with different diameters have been arranged in two rows. By using it, the rapid power measurement for the W-band TE01 mode gyro-TWT has been achieved. The measurement result indicates that the coupling coefficient of the directional coupler is between −50dB and −56dB with the bandwidth of 6GHz, which agrees well with the simulation result. In additional, the influence of the arrangement method of the apertures on the coupling coefficient has also been analyzed.

Stress Induced Voiding of a nose type Single Via is studied. Typical nose type single via stress voiding structure has top and bottom wide metal plates originating from the pads and tapering to a narrow metal line connected by single via enabling the Kelvin type of measurements. In this paper, effect of removal of either of the top or bottom metal plates on the formation of stress induced voids at an elevated temperature is studied and results are characterized. It is found from the study that presence of bottom metal plates is vital for the void accumulation resulting in stress voiding failure.

A double-anode magnetron injection gun (MIG) has been designed and taken place of the single­anode magnetic injection gun used in the first version 140GHz gyrotron. The others parameters of a 140GHz TE22.6-mode gyrotron are same as the first version. And the second version gyrotron has been fabricated and tested. With pulse width 300-us and repetition frequency of 5 Hz, an output power of ~430kW with frequency of 140.2GHz is achieved in operating voltage 68kV, electron beam 28A, corresponding to 22.6% of efficiency.

A study of a Ka-band broadband mode converter for gyro-TWT has been presented. By using this mode converter, a TE 01 cylindrical waveguide mode which is used as the operation mode for the gyro-TWT can be obtained from a TE 10 rectangular waveguide mode. The measurement result indicates that -3dB bandwidth of 6GHz has been obtained, which agrees well with the simulation results and doubles the level of the previous design. In addition, the influences of the critical dimensions on the transmission performance have also been analyzed. The mode converter is suitable for the broadband gyro-TWT.

An inverse Magnetron Injection Gun (IMIG) has been designed for the KIT 2 MW, 170 GHz coaxial cavity gyrotron. The ambition is the possibility for implementation of a larger emitter ring compared to the “conventional” MIGs used in today's fusion gyrotrons. Considering the fundamental beam parameters, in the theoretical analysis an excellent beam quality has been achieved which results in a very low velocity spread. In order to ensure a stable operation under hot condition, the thermomechanical behavior and material compositions were investigated and optimized.

The design of a single-anode magnetron injection gun (MIG) for the 140-GHz TE 226 -mode gyrotron are presented. The electron beam produced by the MIG can operate at accelerating voltage of 80-kV and current of 40-A. The guiding center radius of the electron beam located at the cavity is about 7.98 mm, corresponding to 0.512 times of the radius of the cavity. The perpendicular and axial velocity spreads are predicted to be 1.9% and 4.2% separately at the nominal velocity pitch factor of 1.47. The single-anode MIG is under manufactured and will be adopted in the Fabrication of the 140-GHz gyroton in the future.

In the experimental study a periodic dielectric loaded structure is used as the high-frequency structure of the W-band Gyro-TWT. The experimental measurement is carried out when the accelerating voltage is 63kV and the electron beam current is 6A. A peak output power of 120 kW, 3dB bandwidth of 6 GHz are obtained.

This paper presents the design of a W-band low loss TE 10 -TE 01 mode converter, which is mainly used in the input system of W-band gyro-TWT. Based on the design, the 3-D model has been optimized and simulated in ANSOFT HFSS. A TE 10 -TE 01 broad band mode converter with a loss of less than 0.14 dB, maximum VSWR of 1.388, and a band width of 10GHz was obtained through the simulation. Then a pair of manufactured mode converters was tested in the experiment, and the loss was about 1 dB.

The design of a low voltage W-Band gyro-TWT amplifier operating in the circular TE 01 mode is presented. The interaction circuit in the design consisted of a two stage distributed-loss structure. Results of simulations with nonlinear analysis and particle-in-cell code show that the amplifier can generate a saturated peak power of 42.7kW with a gain of 54dB at 93.8GHz and a 3dB bandwidth of 3.7GHz when operated at a 35kV, 4A, v ⊥ /v 1 =1.0 electron beam from a magnetron injection gun. The simulations also show a stable operation of the gyrotron in the operating mode.

The experimental study of a Q-band gyrotron-travelling-wave tube (gyro-TWT) amplifier operating in the circular TE01 mode at fundamental cyclotron harmonic is presented. A periodic dielectric loaded structure is adopted in the interaction circuit of the gyro-TWT to suppress spurious modes TE11 and TE21. A saturated peak power of 175 kW is obtained corresponding to saturated gain of 53 dB and an efficiency of 38% with -3dB bandwidth of 3 GHz(~7%). The gyro-TWT is found to be zero-drive stable at these operating points.

The direct electroplating of copper film on an ultrathin cobalt film in the alkaline bath was investigated. The plating bath consists of CuSO 4 ·5 H 2 O and ethylenediamine. It is found that the ethylenediamine can well stop the cobalt reaction with copper ion and make the copper deposition on cobalt possible. The experimental results show that the copper films on the cobalt have a high preferential (111) orientation, very low surface roughness and low resistivity.

CoMo alloys as copper diffusion barriers were investigated in this work. The thermal stability was studied after annealing, which was measured by FPP, XRD, SEM and AES. According to the electrical test, we carried out a new p-cap structure to measure. The breakdown electrical field and C-V properties were measured at 150°C. Both the thermal test and electrical results show CoMo is a potential diffusion barrier.

The design of a W-band gyrotron-travelling-wave tube (gyro-TWT) operating in the circular TE 01 mode at fundamental cyclotron harmonic are presented. A dielectric-loaded structure is adopted in the interaction circuit of the gyro-TWT to suppress spurious modes TE 11 and TE 21 . A saturated peak power of 180 kW is simulated with saturated gain of 62 dB, efficiency of 28.5% in operating voltage 70kV, electron beam current 9A and bandwidth of 5 GHz (∼7%).

This paper presents the design of the interaction cavity for a 94 GHz, CW Gyrotron. A weakly tapered interaction cavity has been designed to excite TE03 operation mode at the fundamental harmonic for 94 GHz, 93 kW gyrotron. The cold cavity has been optimized for the operating mode TE 03 mode. The beam-wave interaction was carried out using commercially available PIC code MAGIC. An axial output power above 93 kW has been achieved at guiding cavity magnetic field 3.56 T and corresponding efficiency is about 31%.

The design of a 94GHz gyrotron oscillator operating in the circular TE 02 -TE 03 mode at the second cyclotron harmonic is presented. Results of simulations with particle-in-cell code show that the gyrotron can generate saturated peak power of 26.6kW with 26.6% efficiency at 94GHz when utilized a 25kV, 4A, v ⊥ /v ‖ =1.6 electron beam from a magnetron injection gun, and the results of simulations show the gyrotron operates stably in the operating mode.

The effects of shallow trench isolation (STI) induced mechanical stress on hot carrier degradation of 5V NMOSFET with different source(S)/drain(D) areas are studied. It is found that increased compressive STI stress with S/D area reduction can enhance the drain side impact ionization rate due to band gap narrowing effect which results in more hot carrier generation. TCAD simulations are also performed to explain the hot carrier degradation dependence on STI stress.

The broad time-to-failure distribution and bimodality of downstream electromigration (EM) in 45nm technology node are investigated. Liner void and end of line void at wafer edge on downstream EM structure after M1 CMP is a clear physical vapor deposition (PVD) shadowing effect signature caused by poor liner gap fill capability. Furthermore, void growth study during early electromigration stage effectively indicates that the void is initiated at the bottom corner of the via interfacing with cap layer for early failures and slit void is consistently observed at same location on unstressed sample. These pre-existed voids create poor contact either on via bottom interface or trench and cap layer interface or via bottom corner which produce a difference in failure time resulting in poor time to failure (TTF) spread and lower t50. New Cu seed deposition technique eliminates end of line void and liner void and it turns out to improve downstream EM performance. Optimized post etch treatment (PET) chemicals help to reduce Cu oxidation to improve via bottom integrity and eliminate slit void. This effective post etch treatment was demonstrated to improve downstream EM bimodality behavior to tight mono-modal distribution. Via above to metal below overlay is also one of key factors for downstream EM improvement. There is a strong correlation between the TTF and the via to metal overlay and stringent overlay control is beneficial to improve downstream EM sigma and t50.

It is important that selects window function and width to Short-time Fourier Transform(STFT). Especially, when diagnosing fault loudspeaker, the different analysis window function and width can influent the result of analysis on respond signal of loudspeaker. So, it proposed a selecting analysis window function and width method based on the energy correction factor, the maximum side lobe and main lobe peak value in the frequency domain. Through reasonable selecting analysis window function and width, it can reduce the influence of truncation of signal caused by the Gibbs phenomenon and resolve the frequency resolution problem on STFT. Therefore, it can be more accurately to extract the fault feature of loudspeaker, and improve the loudspeaker on-line automatic fault detection accuracy.

The aim of this work is to investigate the hydrogen generation from hydrothermal water by using hydrogen sulfide as a reducer. The effects of reaction temperature, reaction time and the initial pH value of reaction solution on the production of hydrogen from hydrothermal water were investigated. The experimental results showed that hydrogen was produced from water when the reaction temperature was beyond 200 oC. Moreover, the hydrogen production was strongly depended on the initial pH of the solution. High alkaline condition was favorited to generate hydrogen from water. The possible mechanism of hydrogen production was discussed.

This paper presents a new method using principal component analysis (PCA) to eliminate data redundancy in loudspeaker fault detection. It uses wavelet packet transformation (WPT) to decompose the loudspeaker acoustics signal into 32 packet node signals. Then, get the mean, max, standard deviation and correlation coefficient of every node envelopment .With the way of observing, it gets 63 coefficients from 128 ones which are helpful for detection of fault. Using the new way above, 32 coefficients are removed from the 63. The failed loudspeaker can be found with the help of artificial neural network (ANN). It is proved that the method is very effective in experiment.

In this paper, a model-based face recognition approach using a pyramidal Gabor function representation and neural network matching system is proposed. The local feature extraction is based on distortion tolerant Gabor transformation. The system is based on the dynamic link matching. It consists of an image layer and a model layer, which are tentatively simulating the primary visual cortex and infero-temporal cortex. It is inherently invariant with respect to shift, and is robust against many other variations, most notably rotation in depth and deformation. The system requires very little genetic or learned structure, relying essentially on the rules of rapid synaptic plasticity and a priori constraint of preservation of topography to identify matches.