Accepted
Finite element prediction and device performance of piezoelectric fiber composite based smart sensor
, , Received Date: 2024-09-30
Abstract +
Macro fiber composite (MFC) is extensively utilized in aviation, aerospace, civilian, and military domains due to its high piezoelectricity, flexibility, and minimal loss. Nevertheless, existing research on MFC sensors has focused on material applications, with a conspicuous lack of systematic investigation into the simulation and modeling of MFC sensor devices. In this study, three models, namely, a representative volume element (RVE) model, a direct model, and a Hybrid model are established to analyze the finite element models of MFC, covering the scales from micro to macro. On the one hand, the equivalent RVE model contributes to an understanding of the internal electric field distribution in MFC, thereby establishing a theoretical foundation for force-electric coupling. On the other hand, the application of the direct model and hybrid model accords with the boundary conditions in MFC applications, which lays a theoretical foundation for the stress sensing and resonance sensing mechanisms of MFC. These models constitute effective tools for predicting the sensing performance of MFC smart element sensors. The simulation outcomes indicate that resonant sensors exhibit significantly superior performance compared with patch sensors. Under the conditions where the excitation acceleration is 5 m/s² and the cantilever substrate length is 80 mm, the simulated resonant frequency of the MFC resonant sensor is 67 Hz, with an output voltage of 4.17 V. Experimental results confirm these findings. It is reported that the resonant frequency is 74 Hz and the output voltage is 3.59 V for the MFC sensor. The remarkable consistency between the simulation results and experimental data of the MFC sensor deserves to be emphasized. In addition, the MFC sensor shows excellent sensing sensitivity at low frequencies, with a sensitivity of 7.35 V/g. Obviously, MFC shows remarkable sensing characteristics at low-frequency resonance. The three finite element models established in this work can well predict the sensing performance of MFC sensors, thus ensuring reliable prediction of the performance of such sensors.
, , Received Date: 2024-12-05
Abstract +
Metamaterials can freely control terahertz waves by designing the geometric shape and direction of the unit structure to obtain the desired electromagnetic characteristics, so they have been widely used in sensing, communication and radar stealth technology. The traditional design of terahertz metamaterial absorber usually requires continuous structural adjustment and a large number of simulations to meet the expected requirements. The process largely relies on the experience of researchers, and the physical modeling and simulation solution process is time-consuming and inefficient, greatly hindering the development of metamaterial absorbers. Therefore, due to its powerful learning ability, deep learning has been used to predict the structural parameters or spectra of metamaterial absorbers. However, when designing a new structure, it is necessary to prepare a large number of training samples again, which is both time-consuming and not universal. Particle swarm optimization algorithm can quickly converge to the optimal solution through the sharing and cooperation of individual information in the group, with no need for prior preparation. Therefore, a method of fast designing terahertz metamaterial absorber is proposed based on multi-objective particle swarm optimization algorithm in this work. Taking a new center symmetric absorber structure composed of four Ls for example, the structure parameters are optimized to achieve rapid and automatic design of metamaterial absorber. The multi-objective particle swarm optimization algorithm takes the absorptivity and quality factor as independent targets to design the structure parameters of the absorber, realizing the dual-objective optimization of the absorber, and overcoming the shortcoming of the multi-objective conflicts that cannot be solved by PSO. When used for refractive index sensing, the optimally-designed absorber achieves perfect absorption at 1.613 THz with a quality factor of 319.72 and a sensing sensitivity of 264.5 GHz/RIU. In addition, the reasons of absorption peaks are analyzed in detail through impedance matching, surface current, and electric field distribution. By studying the polarization characteristics of the absorber, it is found that the absorber is not sensitive to polarization, which is more stable in practical application. In summary, the multi-objective particle swarm optimization algorithm can realize the design according to the requirements, reduce the experience requirement of researchers in the design of metamaterial absorber, thereby improving design efficiency and performance, and has great potential for application in the design of terahertz functional devices.
, , Received Date: 2024-12-02
Abstract +
Surface enhanced Raman spectroscopy (SERS) can provide rich molecular structure information about ultra-sensitive, non-destructive, and rapid detection, with accuracy down to the single-molecule level. It has been widely applied to physics, chemistry, biomedicine, environmental science, materials science and other fields. Combining the advantages of metals and two-dimensional (2D) nanomaterials, various 2D metal composite structures have been proposed for SERS. However, the contribution of 2D nanomaterials in Raman enhancement is often limited. In this work, vertically aligned MoS2 nanosheet composite with silver nanoparticles (Ag NPs) is proposed for SERS detection. Large-area vertically aligned MoS2 nanosheets, which are grown directly on molybdenum (Mo) foil by using hydrothermal method, can effectively enhance molecular adsorption, light absorption, and provide dual electromagnetic and chemical enhancement. Furthermore, annealing treatment of the MoS2 nanosheets significantly improves the efficiency of charge transfer between Ag NPs and MoS2, thereby increasing the chemical contribution to SERS. The results demonstrate that the annealed MoS2/Ag substrate exhibits outstanding SERS performance, with a detection limit for R6G molecules as low as 10–12 M, which is four orders of magnitude lower than that of the unannealed substrate. The enhancement factor (EF) is calculated to be approximately 1.08×109, approaching the sensitivity required for single-molecule detection. Additionally, the substrate has high signal reproducibility at low concentrations, enabling ultra-sensitive detection of pesticide residues in aquatic products.
, , Received Date: 2024-12-02
Abstract +
To further understand the patterns and mechanisms of total ionizing dose (TID) radiation damage in carbon nanotube field-effect transistor (CNTFET), the total dose effects of 10 keV X-ray irradiation on N-type and P-type CNTFETs are investigated in this work. The irradiation dose rate is 200 rad(Si)/s, with a cumulative dose of 100 krad(Si) for N-type devices and 90 krad(Si) for P-type devices. The differences in TID effect between N-type and P-type CNTFETs under the conditions of floating gate bias and on-state bias, the influence of irradiation on the hysteresis characteristics of N-type CNTFETs, and the influence of channel sizes on the TID effects of N-type CNTFETs are also explored. The results indicate that both types of transistors, after being irradiated, exhibit the threshold voltage shift, transconductance degradation, increase in subthreshold swing, and decrease in saturation current. In the irradiation process, N-type devices under floating gate bias suffer more severe damage than those under on-state bias, while P-type devices under on-state bias experience more significant damage than those under floating gate bias. The hysteresis widths of N-type devices decrease after being irradiated, and the TID damage becomes more severe with the increase of channel dimensions. The main reason for the degradation of device parameters is the trap charges generated in the irradiated process. The gate bias applied during irradiation affects the capture of electrons or holes by traps in the gate dielectric, resulting in different radiation damage characteristics for different types of devices. The reduction in the hysteresis width of N-type devices after being irradiated may be attributed to the negatively charged trap charges generated during irradiation, which hinders the capture of electrons by water molecules, OH groups, and traps in the gate dielectric. Moreover, the channel dimensions of the transistors also influence their radiation response: larger channel dimensions result in more trap charges generated in the gate dielectric and at the interface during irradiation, leading to more severe transistor damage.
, , Received Date: 2024-11-29
Abstract +
In recent years, electrochromic materials have been extensively utilized in smart windows, displays, and dimmable devices. WO3, as a typical electrochromic material has received significant attention. Existing researches indicate that the concentration and distribution of oxygen vacancies in WO3 are both important in determining electrochromic effect. However, it has been reported that traditional preparation methods such as annealing can significantly reduce the ability to modulate the crystallinity and optical performance. Hence, proposing a novel approach to enhance the electrochromic properties of WO3 films holds important research significance and application prospects. In this work, the electrochromic properties of WO3 thin films are enhanced by increasing the oxygen vacancy concentration and forming its gradient distribution on the surface through plasma treatment. Firstly, the oxygen vacancy concentration and distribution of the film are optimized by regulating the power and duration of the plasma treatment. Secondly, the structure and optical properties of the plasma treated WO3 films are analyzed using X-ray diffraction (XRD), scanning electron microscopy (SEM), and UV-Vis spectroscopy. Finally, the stability and response speed of each film during the electrochromic cycle are evaluated via electrochemical tests. Through plasma treatment, the concentrations of oxygen vacancies on the surfaces of WO3 films are all significantly increased, and a gradient distribution is formed, which is conducive to enhancing the ability to inject and extract electrons. The treated WO3 films demonstrate better electrochemical stability and chromic stability during the electrochromic cycle, and their transparencies and electrochromic response speeds are also significantly enhanced. Additionally, by increasing the concentration of oxygen vacancies through plasma treatment, the band gap of the film decreases and the electrical conductivity increases, which further validates the effectiveness of modulating concentration of oxygen vacancies on the electrical conductivity of WO3 film. Overall, these results indicate that plasma treatment is an emerging method of significantly improving the electrochromic properties of WO3 films.
, , Received Date: 2024-11-19
Abstract +
,
Abstract +
Rydberg atoms are a type of atoms characterized by high principal quantum numbers. Due to their large polarizability, electric field sensors based on Rydberg atoms have attracted considerable attention. However, research on direct current (DC) electric fields or lowfrequency electric fields utilizing Rydberg atoms is currently scarce, primarily due to the shielding effects associated with atomic vapor cells in regard to low-frequency electric fields, which make precise measurements of the electric field extremely challenging.
In this paper, we construct a Rydberg ladder configuration using probe laser at 852 nm and coupling laser at 510 nm in a room temperature cesium vapor cell with integrated electrode plates. Thereby this enables the realization of a Floquet-EIT (electromagnetically induced transparency) spectrum dressed by a radio frequency (RF) field in the presence of a DC electric field, and we further study the influence from DC electric field on spectral characteristic. In experiments, it has been observed that when only the RF electric field is applied, the EIT spectrum displays solely even-order sidebands. Furthermore, when both the RF field and the DC electric field are simultaneously present, the first-order sideband signal of the Floquet-EIT are observed. As the intensity of the DC electric field increases, the amplitude of the firstorder sidebands gradually increases. However, increasing the DC electric field to a sufficient magnitude induces sidebands interference, which results in a reduction of the sideband amplitudes. Furthermore, increasing the RF frequency can alleviate the interference effects induced by the DC electric field on the first-order sidebands. Finally, comparing the relative standard deviation of the sideband amplitudes of the Floquet-EIT spectra with the frequency shifts of the DC-Stark spectra under weak DC electric fields, we find that the measurement accuracy of the former is significantly superior to the latter.
This work make use of a Cs atomic vapor cell with an integrated electrode to avoid shielding effects. By observing Floquet-EIT spectra, the response of the spectra to DC electric fields is investigated. This experiment offers novel insights for quantum sensing measurements of DC and low-frequency electric fields.
In this paper, we construct a Rydberg ladder configuration using probe laser at 852 nm and coupling laser at 510 nm in a room temperature cesium vapor cell with integrated electrode plates. Thereby this enables the realization of a Floquet-EIT (electromagnetically induced transparency) spectrum dressed by a radio frequency (RF) field in the presence of a DC electric field, and we further study the influence from DC electric field on spectral characteristic. In experiments, it has been observed that when only the RF electric field is applied, the EIT spectrum displays solely even-order sidebands. Furthermore, when both the RF field and the DC electric field are simultaneously present, the first-order sideband signal of the Floquet-EIT are observed. As the intensity of the DC electric field increases, the amplitude of the firstorder sidebands gradually increases. However, increasing the DC electric field to a sufficient magnitude induces sidebands interference, which results in a reduction of the sideband amplitudes. Furthermore, increasing the RF frequency can alleviate the interference effects induced by the DC electric field on the first-order sidebands. Finally, comparing the relative standard deviation of the sideband amplitudes of the Floquet-EIT spectra with the frequency shifts of the DC-Stark spectra under weak DC electric fields, we find that the measurement accuracy of the former is significantly superior to the latter.
This work make use of a Cs atomic vapor cell with an integrated electrode to avoid shielding effects. By observing Floquet-EIT spectra, the response of the spectra to DC electric fields is investigated. This experiment offers novel insights for quantum sensing measurements of DC and low-frequency electric fields.
,
Abstract +
Photo-oriented liquid crystal technology utilizes polarized light illumination to achieve the directional alignment of liquid crystal molecules. This technology can be developed into polarization volume gratings (PVG), which possess polarization and volume holographic selectivity characteristics. It also has a broad application prospect as an optical coupling element in optical waveguides and for pupil expansion output. This paper reports on the fabrication of a liquid crystal polarization volume holographic cylindrical lens (PVLS) with a beam diameter of 2cm using photo-oriented technology combined with a polarization off-axis holographic optical path. During the experiment, the exposure angle can be controlled to achieve the desired grating period variation range, enabling the diffraction angles of red, green, and blue (RGB) light incident on different grating periods to be the same. The experimental results show that within the grating period variation range of 1721.2 nm to 5346.5 nm, when RGB light is incident on grating periods of 3147 nm, 2649.1 nm, and 2275.6 nm respectively, the measured diffraction angles are all 11.59°, with an error between the actual and theoretical diffraction angles within ±0.5°; under 532nm right-handed circularly polarized light, the diffraction efficiency for 18 normal incidence reaches 90.6%, and the diffraction efficiency for oblique incidence satisfying the Bragg condition is 84.4%; simultaneously, beam expansion in one-dimensional direction is achieved, preliminarily verifying the feasibility of PVLS application in the field of color waveguides.
,
Abstract +
Accurate atmospheric wind field measurements are critical for understanding global climate dynamics and facilitating space exploration. Doppler Asymmetric Spatial Heterodyne Interferometers (DASH) measure atmospheric wind speed by detecting the phase changes in interferograms induced by Doppler shifts of airglow emission lines. However, environmental temperature fluctuations and mechanical vibrations often cause imaging plane shifts, introducing phase deviations that degrade measurement accuracy. In this study, we propose a novel global fitting-based imaging shift monitoring method. By etching periodic notches on the diffraction grating surface, the method models and fits the notch patterns formed on the detector plane to achieve precise imaging shift detection and correction. The optimization of notch signal modeling significantly reduces the number of fitting parameters, improving computational efficiency and detection precision. Through extensive simulations, we analyze the impact of SNR and model parameter variations on detection accuracy. Results indicate that when the SNR exceeds 11, the detection uncertainty remains below 6.5 nm. Sensitivity analysis reveals that the detection error stays within acceptable limits when the notch number and notch width variation are controlled within 40% and 0.7%, respectively, while the edge smoothness parameter of notch pattern has negligible influence. To validate the method’s performance, a thermal stability test using a near-infrared DASH prototype was conducted. The experimental results demonstrate a strong correlation between interferogram phase shifts, imaging plane shifts, and environmental temperature variations. After applying the proposed correction method, local phase fluctuations in the interferogram are significantly reduced, improving phase stability. Further, artificially applied imaging shifts were accurately detected with errors consistently below 9.96 nm, confirming the method’s reliability and precision. In conclusion, the proposed method effectively detects and corrects imaging plane shifts caused by temperature variations, enhancing interferogram phase stability and ensuring high-precision wind speed measurements. This approach offers a robust and computationally efficient solution for mitigating imaging shifts in DASH systems, with significant potential for atmospheric wind field measurement and space-based observational applications.
,
Abstract +
This paper proposes a phaseless radiation source reconstruction method based on singular value decomposition (SVD) regularization and fast iterative shrinkage-thresholding algorithm (FISTA), aimed at efficiently identifying electromagnetic interference (EMI) sources in integrated circuits (ICs). The method acquires electromagnetic field data through near-field scanning and reconstructs an equivalent dipole array on the surface of the radiation source using the source reconstruction method (SRM). During the reconstruction process, the SVD regularization term enhances the algorithm's stability and noise resistance, while FISTA accelerates the convergence speed.
To validate the effectiveness of the proposed method, dipole array reconstruction was first performed using near-field data at a height of 5mm for a patch antenna simulation model, followed by an analysis of magnetic field data at a 10mm validation plane. At the 35th iteration, the total relative error of the reconstruction was 1.21%. The influence of the regularization parameter α on the results was then investigated, and it was found that α = 0.05 yielded the smallest error. The method was also tested under different Gaussian white noise conditions, with relative errors remaining below 5%, demonstrating strong robustness.
Finally, chip experiments were conducted to verify the method. The proposed method converged stably within 35 iterations, with a relative error of 2.3% in the reconstruction results. The total iteration time is 61.7% of the single-layer phaseless interpolation algorithm, and the relative error is reduced by 52% compared to the double-layer phasless iteration algorithm.The experimental results demonstrate that the proposed method can efficiently and accurately reconstruct phaseless radiation sources, with good noise robustness, making it suitable for EMI analysis in integrated circuits.
To validate the effectiveness of the proposed method, dipole array reconstruction was first performed using near-field data at a height of 5mm for a patch antenna simulation model, followed by an analysis of magnetic field data at a 10mm validation plane. At the 35th iteration, the total relative error of the reconstruction was 1.21%. The influence of the regularization parameter α on the results was then investigated, and it was found that α = 0.05 yielded the smallest error. The method was also tested under different Gaussian white noise conditions, with relative errors remaining below 5%, demonstrating strong robustness.
Finally, chip experiments were conducted to verify the method. The proposed method converged stably within 35 iterations, with a relative error of 2.3% in the reconstruction results. The total iteration time is 61.7% of the single-layer phaseless interpolation algorithm, and the relative error is reduced by 52% compared to the double-layer phasless iteration algorithm.The experimental results demonstrate that the proposed method can efficiently and accurately reconstruct phaseless radiation sources, with good noise robustness, making it suitable for EMI analysis in integrated circuits.
,
Abstract +
Ternary GaAsSb nanowires (NWs) have shown considerable potential in the applications of infrared optical nanodevices due to their direct bandgap and wavelength-tunable light emission which covers the range from 870 nm to 1700 nm by changing the content of Sb in GaAsSb NWs. Due to the high surface state density, the light emission efficiency of GaAsSb NWs is quite low and the light emission is difficult to observe under room-temperature conditions. The previous studies about the optical properties of GaAsSb NWs were mainly carried out under low-temperature conditions, which has limited their room-temperature optical properties modulation study and room-temperature application. In the present study, we realize optical properties modulation of GaAsSb NWs under room-temperature conditions through the high-pressure strategy, using both photoluminescence (PL) and Raman spectroscopy methods. With increasing the pressure, the PL intensity of GaAsSb NWs presents an obvious enhancement at room temperature and the PL peak position presents a blue-shifted trend. With varying the wavelength (473 nm, 514 nm, and 633nm) of the incident laser, the excitation-wavelength-dependent PL can be observed in GaAsSb NWs. The laser with a longer wavelength (633 nm) will excite the stronger light emission. The Raman spectra of GaAsSb NWs excited by varied lasers (473 nm, 514 nm, and 633 nm) both showed blue shift under compression. We selected four pressure points (0.7 GPa, 1.2 GPa, 1.8 GPa, and 2.5 GPa) for the detailed comparison between the Raman spectra excited by different lasers. Under the excitation of 473 nm laser, the Raman peaks of GaAsSb NWs present evident red-shift compared to those excited by 514 nm or 633 nm laser, which reveals the existence of temperature difference. The relative temperature difference in GaAsSb NWs induced by two different lasers (473 nm and 633 nm) could be estimated up to 200 K. The laser with shorter wavelength will induce a stronger heating effect in GaAsSb NWs and reduce the light-emission efficiency. Under high-pressure condition, the charge transfer between the surface of GaAsSb NWs and pressure transmitting medium can be enhanced, which results in the reduction of surface state density and laser-heating effect. Therefore, the high-pressure strategy provides an efficient route to suppress the high surface state density and optimize optical properties of semiconductor nanostructures.
,
Abstract +
Diamond is a kind of ultimate functional material, which is widely used in industry, science and technology, military defense, medical and health, jewelry and other fields. However, its application in the semiconductor field is still limited, because its electrical transport performance has not yet met the requirements of semiconductor devices. In order to improve the electrical transport performance of diamond as much as possible, the synthesis of diamond single crystal was studied with B2S3 additive in the synthesis system by temperature gradient growth (TGG) method at pressure of 6.5 GPa condition in this work. The growth rates of the synthesized diamond crystals reduced from 2.19 mg/h to 1.26 mg/h, indicating that the growth rate of diamond not only depended on the growth driving force, but also affected by the impurity elements in the synthetic cavity. Additionally, the colors of the synthesized diamond crystals transformed from yellow to baby blue, accompanying with the increase in the amount of additives added. Raman measurement results indicated that the obtained diamond appeared as a single sp3 hybrid phase without the sp3 hybrid graphite phase. However, the corresponding Raman characteristic peaks of the as-grown diamond crystals located at about 1331 cm-1 and consistently tended to move towards low wave number. According to FTIR measurement results, the absorption peaks at 1130 cm-1 and 1344 cm-1 attributed to nitrogen defects. It was found that the nitrogen defect concentrations of the synthesized diamond crystals decreased gradually from about 300 ppm to 60 ppm. Furthermore, the electrical transport performance of the synthesized diamond was characterized by Hall effects measurement. Diamond had an insulating behavior due to the absence of any additives in the synthetic cavity. However, the result showed that there was little difference in carrier hall mobility, but there was a difference of two orders of magnitude in carrier concentration, when B2S3 was introduced into the synthetic system as additive. Furthermore, the resistivity of the synthesized [111]-oriented diamond crystal reduced to 45.4 Ω·cm, due to the addition of B2S3 additive in the synthesis system. However, it is worth noting that the resistivity of the diamond crystal synthesized with 0.002 g B2S3 and Ti/Cu additives in the synthesis system drops sharply to 0.43 Ω·cm. Therefore, the nitrogen defects in diamond will have an important effect on its conductivity. It provides an important experimental basis for the application of diamond in semiconductor field.
,
Abstract +
To meet the requirements of both high thermal conductivity and substantial latent heat storage and release in spacecraft thermal management materials, this study employs a hot-pressing technique to fabricate a multidimensional carbon-based, thermally enhanced microencapsulated phase change composite. This approach addresses the limitations of conventional phase change materials, which exhibit low thermal conductivity and a propensity for liquid leakage. By integrating experimental assessments with finite element numerical simulations, we systematically investigated the effects of varying contents and ratios of microencapsulated phase change materials, flake graphite, and pitch-based carbon fibers on the composite’s thermal properties, specifically thermal conductivity and latent heat. Furthermore, the formation mechanism of the internal multidimensional thermal conduction network was elucidated.
The results indicate that introducing multidimensional thermally conductive materials into the microencapsulated phase change system, coupled with the optimization of component composition and structure, can establish a continuous and dense multidimensional carbon-based conduction network. Leveraging the synergistic effects of these conductive materials and employing a multi-size flake graphite filling strategy significantly enhanced the overall thermal conductivity of the composite, reaching 1.021 W·m-1·K-1, while maintaining a high latent heat of 81.540 J·g-1. These findings provide theoretical and practical guidance for the optimization and application of advanced thermal management materials in spacecrafts.
The results indicate that introducing multidimensional thermally conductive materials into the microencapsulated phase change system, coupled with the optimization of component composition and structure, can establish a continuous and dense multidimensional carbon-based conduction network. Leveraging the synergistic effects of these conductive materials and employing a multi-size flake graphite filling strategy significantly enhanced the overall thermal conductivity of the composite, reaching 1.021 W·m-1·K-1, while maintaining a high latent heat of 81.540 J·g-1. These findings provide theoretical and practical guidance for the optimization and application of advanced thermal management materials in spacecrafts.
,
Abstract +
Magnetic resonance imaging (MRI) is one of the most important imaging modalities used in clinical radiology research and diagnosis due to its non-invasive nature, absence of ionizing radiation, high soft tissue contrast, and diverse imaging capabilities. However, traditional MRI is limited by a relatively low signal-to-noise ratio (SNR), which can be enhanced by increasing the strength of the main magnetic field. Ultra-high field MRI (UHF-MRI), an emerging technology, typically refers to MRI systems with a main magnetic field strength of 7 T or higher. Compared to conventional MRI, UHF-MRI improves image SNR and extends the boundaries of spatial resolution and detection sensitivity. These advancements not only provide clinicians with more detailed and accurate bioimaging data but also open new research avenues in fields such as life sciences and cognitive neuroscience. This paper introduces the historical development and theoretical foundations of UHF-MRI, highlights its advantages over conventional MRI, and summarizes current research on UHF-MRI applications in human brain imaging, with a focus on functional and metabolic studies. Additionally, the challenges of UHF-MRI are discussed, and potential future research directions are proposed.
,
Abstract +
Real-time measurement and feedback control of key plasma parameters are critical for future fusion reactor operation, with ion temperature being a vital control target as part of the triple product for fusion ignition. However, plasma diagnostics tends to require complex data analysis. To acquire ion temperature Ti from charge exchange recombination spectroscopy (CXRS), a widely used method is through iterative spectral fitting, which is time-consuming and calls for expert intervention during data analysis. On top of that, frequent human expert intervention is needed in the conventional iterative fitting. Therefore, the conventional method cannot meet the meet the demand for real-time Ti measurement. Neural Networks (NN), which is capable of learning the underlying relationships between the measured spectra and Ti, is a promising approach to cope with this problem. In fact, NN approaches have been widely adopted in the field of magnetic confined plasma. Previous study in JET has achieved a satisfactory accuracy for inferring Ti from CXRS spectra compared to the conventional fitting results. Recently the study of disruption prediction has achieved great progress with the help of deep neural networks. However, these researches are conducted in steadily-operating devices, where for NN models, the data distribution is similar in training set and test set. This is not the case for newly-built tokamak like HL-3, or for future fusion reactors such as ITER. For new devices, there will be a period for the plasma parameters to raise from low to high ranges. In this case, investigating the extrapolation capability of NN models based on low parameter training data is of paramount importance.
A Convolutional Neural Network (CNN)-based model is proposed to accelerate the analysis of spectral data of CXRS, with a focus on investigating the model’s extrapolation capability to much higher Ti ranges. The dataset consists of about 122 thousand pieces of spectral data, along with their corresponding inferred Ti from offline iterative process. The results demonstrate that the CNN-based model provides excellent Ti analysis and reduces the inference time for analyzing a single spectrum to less than 1 ms, which is 100-1000 times faster compared to traditional spectral fitting method. However, the performance of the data-driven neural network model is limited by challenges such as insufficient data and imbalanced data distribution, which further deteriorates the extrapolation capability. Generally, data with higher Ti constitute a small percentage of the total dataset. In the case of our study, only about 5% of the spectra correspond to Ti > 2 keV (among 2-4 keV). Yet they reflect the temperature of central plasma, which is more important for assessing the performance of plasma. To overcome this limitation, the study synthesizes high-temperature data based on experimental data from discharges with Ti in low-temperature range. By incorporating 5% synthetic data into the training set only consisting of data with Ti<2 keV, the model’s extrapolation capability is extended to cover the whole range of Ti < 4 keV. The mean relative error of the mode in 3 keV < Ti < 4 keV is reduced from 35% to below 15%. This approach demonstrates the feasibility of using synthetic data to enhance the performance of artificial intelligence algorithms in the field of magnetic confinement fusion. The findings provide valuable insights for the development of real-time ion temperature measurement and feedback control for future high-parameter fusion devices. Furthermore, the study lays a foundation for research in areas that require high-performance across-device characteristic, such as machine learning-based disruption prediction and tearing mode control.
A Convolutional Neural Network (CNN)-based model is proposed to accelerate the analysis of spectral data of CXRS, with a focus on investigating the model’s extrapolation capability to much higher Ti ranges. The dataset consists of about 122 thousand pieces of spectral data, along with their corresponding inferred Ti from offline iterative process. The results demonstrate that the CNN-based model provides excellent Ti analysis and reduces the inference time for analyzing a single spectrum to less than 1 ms, which is 100-1000 times faster compared to traditional spectral fitting method. However, the performance of the data-driven neural network model is limited by challenges such as insufficient data and imbalanced data distribution, which further deteriorates the extrapolation capability. Generally, data with higher Ti constitute a small percentage of the total dataset. In the case of our study, only about 5% of the spectra correspond to Ti > 2 keV (among 2-4 keV). Yet they reflect the temperature of central plasma, which is more important for assessing the performance of plasma. To overcome this limitation, the study synthesizes high-temperature data based on experimental data from discharges with Ti in low-temperature range. By incorporating 5% synthetic data into the training set only consisting of data with Ti<2 keV, the model’s extrapolation capability is extended to cover the whole range of Ti < 4 keV. The mean relative error of the mode in 3 keV < Ti < 4 keV is reduced from 35% to below 15%. This approach demonstrates the feasibility of using synthetic data to enhance the performance of artificial intelligence algorithms in the field of magnetic confinement fusion. The findings provide valuable insights for the development of real-time ion temperature measurement and feedback control for future high-parameter fusion devices. Furthermore, the study lays a foundation for research in areas that require high-performance across-device characteristic, such as machine learning-based disruption prediction and tearing mode control.
- 1
- 2
- 3
- 4
- 5
- ...
- 21
- 22
