Carbon nanotube cold cathode miniature vacuum sensor and field emission current response detection method thereof

Abstract

The invention provides a carbon nanotube cold cathode miniature vacuum sensor and a field emission current response detection method thereof, and relates to the technical field of micro-nano sensing and vacuum metrology, the sensor comprises an insulating substrate, a cold cathode array is arranged on the insulating substrate, an anode electrode is arranged on the insulating substrate through an insulating pillar corresponding to the cold cathode array, and the anode electrode is connected with the cold cathode array. A glass cover plate is arranged on the insulating substrate, the cold cathode array and the anode electrode are covered by the glass cover plate to form a vacuum cavity, and the cold cathode array and the anode electrode are connected with a signal acquisition and processing module; the method comprises the steps of applying pulse voltage, collecting field emission current, carrying out noise reduction processing on the collected field emission current, carrying out temperature compensation on the field emission current after noise reduction processing, and carrying out vacuum degree calculation based on the compensated field emission current. The vertical electric field configuration design and the deep learning algorithm are fused, the performance boundary of a traditional vacuum sensor is broken through, a good technical effect is achieved, and use is convenient.

Description

Technical Field

[0001] This invention relates to the fields of micro-nano sensing technology and vacuum metrology technology, and in particular to a carbon nanotube cold cathode micro vacuum sensor and its field emission current response detection method. Background Technology

[0002] Traditional vacuum sensors primarily rely on the principle of thermionic cathode ionization gauges, which ionize gas molecules by emitting electrons from a heated tungsten filament and then inverting the vacuum level by measuring the ion flow. However, this technology suffers from three major bottlenecks: first, thermionic cathodes require maintaining high temperatures, resulting in power consumption of 1-5 watts, making them difficult to deploy in energy-constrained environments such as spacecraft; second, thermionic cathodes are susceptible to oxidation and ion bombardment, leading to failure, with a lifespan typically less than 5000 hours, and their performance degrades rapidly in highly polluted environments; and third, their slow response speed makes them unable to capture transient vacuum fluctuations. Although subsequent cold cathode sensors (such as the Penning gauge) have reduced power consumption, their reliance on strong magnetic fields results in large sizes, and their measurement limit is only 10... -2 Pascal. In recent years, carbon nanotube (CNT) cold cathodes have become a research hotspot due to their significant field enhancement effect and low turn-on electric field. However, existing CNT vacuum sensors still face fundamental challenges: First, the field emission current is susceptible to electromagnetic interference and Johnson noise, with a signal-to-noise ratio generally below 20 dB, especially in industrial environments with multiple devices coexisting, where measurement errors exceed ±10%; second, gas adsorption at the CNT tip causes work function drift, resulting in current decay over time, requiring frequent calibration; third, temperature changes alter the electron tunneling probability, and traditional linear temperature drift compensation models experience a sharp increase in error to >5% below -20℃ or above 60℃.

[0003] Therefore, there is an urgent need for a carbon nanotube cold cathode micro vacuum sensor and its field emission current response detection method. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the purpose of this invention is to provide a carbon nanotube cold cathode micro vacuum sensor and its field emission current response detection method.

[0005] To achieve the above objectives, the present invention provides the following solution: This invention provides a carbon nanotube cold cathode miniature vacuum sensor, comprising: an insulating substrate, a cold cathode array, an anode electrode, a glass cover plate, and a signal acquisition and processing module. The cold cathode array is disposed on the insulating substrate, and the anode electrode is disposed on the insulating substrate corresponding to the cold cathode array via an insulating support. The anode electrode is positioned directly above the cold cathode array. The glass cover plate is disposed on the insulating substrate, and the glass cover plate covers the cold cathode array and the anode electrode inside, forming a vacuum cavity. The cold cathode array and the anode electrode are connected to the signal acquisition and processing module.

[0006] Preferably, the insulating substrate is an Al2O3 ceramic or a silicon-based SiO2 insulating layer with a thickness of 0.5-1 mm.

[0007] Preferably, the method for fabricating the cold cathode array is as follows: Cathode electrode fabrication: A Cr / Pt metal layer with a thickness of 200 nm–500 nm is sputtered on an insulating substrate, and a strip-shaped cathode electrode pattern is formed by photolithography. CNT growth: Vertical multi-walled carbon nanotube arrays with heights of 10 μm–50 μm and densities of 10-1 were grown on the cathode electrode surface using CVD. 5 -10 7 tubes / mm 2 The thickness unevenness is controlled by screen printing process, so that the CNT bundles form a gradient distribution structure, in which the top is sinusoidal wavy to enhance the field emission efficiency.

[0008] Preferably, the anode electrode is a Ni mesh structure with a light transmittance greater than 80% and a spacing of 50μm–200μm.

[0009] Preferably, a glass solder layer is provided at the edge of the insulating substrate, and the glass cover plate is aligned with the glass solder layer and placed above the insulating substrate, and heated to achieve hermetically sealed packaging.

[0010] This invention also provides a method for detecting the field emission current response of a carbon nanotube cold cathode microvacuum sensor, applied to the aforementioned carbon nanotube cold cathode microvacuum sensor, comprising: Step 1: Apply pulse voltage; Step 2: Collect field emission current; Step 3: Perform noise reduction processing on the collected field emission current; Step 4: Perform temperature compensation on the field emission current after noise reduction; Step 5: Calculate the vacuum level based on the compensated field emission current.

[0011] Preferably, in step 1, applying a pulse voltage specifically involves: A -200 V pulse voltage with a pulse width of 20 μs and a frequency of 5 kHz is applied to the cold cathode array, with the anode grounded.

[0012] Preferably, in step 3, the collected field emission current is subjected to noise reduction processing, specifically as follows: The noise reduction method based on NGO-VMD and wavelet thresholding is used to denoise the acquired field emission current.

[0013] Preferably, in step 4, temperature compensation is performed on the field emission current after noise reduction, specifically as follows: A temperature sensor is placed at the bottom of an insulating substrate, and a dynamic compensation network is constructed based on an LSTM model to compensate for the temperature of the field emission current after noise reduction.

[0014] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: This invention provides a carbon nanotube cold cathode micro-vacuum sensor and its field emission current response detection method. The sensor includes an insulating substrate, a cold cathode array, an anode electrode, a glass cover plate, and a signal acquisition and processing module. The cold cathode array is disposed on the insulating substrate, and the anode electrode is disposed on the insulating substrate corresponding to the cold cathode array via an insulating support. The anode electrode is positioned directly above the cold cathode array. The glass cover plate is disposed on the insulating substrate, covering the cold cathode array and the anode electrode inside to form a vacuum cavity. The cold cathode array and the anode electrode are connected to the signal acquisition and processing module. The method includes applying a pulse voltage, acquiring the field emission current, performing noise reduction processing on the acquired field emission current, performing temperature compensation on the noise-reduced field emission current, and calculating the vacuum degree based on the compensated field emission current. This invention breaks through the performance boundaries of traditional vacuum sensors by integrating vertical electric field configuration design with deep learning algorithms: At the structural level, it adopts a parallel layout of CNT cold cathode array and transparent nickel mesh anode integrated on an Al2O3 substrate, combined with a microcavity formed by Pyrex glass cover plate and glass solder bonding, laying the hardware foundation for high signal-to-noise ratio current acquisition; At the signal processing level, it pioneers a joint noise reduction mechanism based on Northern Eagle optimized variational mode decomposition (NGO-VMD) and wavelet thresholding method, which suppresses power frequency interference, thermal noise and background noise caused by CNT emission fluctuations by adaptively selecting IMF components and optimizing the threshold function, thereby improving the signal-to-noise ratio; At the environmental adaptability level, it develops a dynamic temperature compensation model based on long short-term memory network (LSTM), inputting four-dimensional parameters of current, temperature, voltage and historical vacuum degree, and transmitting nonlinear temperature change characteristics through hidden layer states, compressing measurement errors across the entire temperature range, and significantly improving accuracy compared to traditional linear models. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a structural diagram of the field emission current response detection method for a carbon nanotube cold cathode micro vacuum sensor provided in an embodiment of the present invention; Figure 2A schematic diagram illustrating the VMD algorithm optimization process for NGOs; Figure 3 This is a schematic diagram of the zero-sequence current noise reduction method for single-phase grounding faults based on the joint noise reduction of NGO-VMD and wavelet thresholding. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The purpose of this invention is to provide a carbon nanotube cold cathode micro vacuum sensor and its field emission current response detection method. By integrating vertical electric field configuration design with deep learning algorithms, it breaks through the performance boundaries of traditional vacuum sensors, achieves better technical results, and is easy to use.

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] This invention provides a carbon nanotube cold cathode micro vacuum sensor, comprising: an insulating substrate, a cold cathode array, an anode electrode, a glass cover plate, and a signal acquisition and processing module. The cold cathode array is disposed on the insulating substrate, and the anode electrode is disposed on the insulating substrate corresponding to the cold cathode array via insulating pillars (SiO2 micropillars, spaced 50-200 μm). The anode electrode is positioned directly above the cold cathode array, and the two are arranged in parallel and spaced apart to form a vertical electric field configuration. The glass cover plate is disposed on the insulating substrate, and the glass cover plate covers the cold cathode array and the anode electrode inside, forming a vacuum cavity. The cold cathode array and the anode electrode are connected to the signal acquisition and processing module.

[0021] The insulating substrate is an Al2O3 ceramic or a silicon-based SiO2 insulating layer with a thickness of 0.5-1mm, which serves as the mechanical support platform for the entire sensor.

[0022] The method for fabricating the cold cathode array is as follows: Cathode electrode fabrication: A Cr / Pt metal layer with a thickness of 200 nm–500 nm is sputtered on an insulating substrate, and a strip-shaped cathode electrode pattern is formed by photolithography. CNT growth: Vertical multi-walled carbon nanotube arrays with heights of 10 μm–50 μm and densities of 10-1 were grown on the cathode electrode surface using CVD. 5 -10 7tubes / mm 2 The thickness unevenness is controlled by screen printing process, so that the CNT bundles form a gradient distribution structure, in which the top is sinusoidal wavy to enhance the field emission efficiency.

[0023] The anode electrode is a Ni mesh structure with a light transmittance greater than 80% and a spacing of 50μm–200μm. The mesh structure can reduce electron path obstruction and improve electron transmittance.

[0024] A glass solder layer is disposed at the edge of the insulating substrate. The glass cover plate is aligned with the glass solder layer and placed above the insulating substrate. The substrate is then heated (to 450°C under N2 protection, where the glass solder melts to achieve a hermetically sealed enclosure, forming a 0.1–5 mm layer). 3 The vacuum chamber is used to achieve hermetically sealed packaging. The glass cover can be made of Pyrex 7740 glass with a thickness of 10–50 μm.

[0025] like Figure 1 As shown, the present invention also provides a field emission current response detection method for a carbon nanotube cold cathode micro vacuum sensor, applied to the aforementioned carbon nanotube cold cathode micro vacuum sensor, comprising: Step 1: Apply pulse voltage; Step 2: Collect field emission current; Step 3: Perform noise reduction processing on the collected field emission current; Step 4: Perform temperature compensation on the field emission current after noise reduction; Step 5: Calculate the vacuum level based on the compensated field emission current.

[0026] In step 1, a pulse voltage is applied, specifically as follows: Applying a -200 V pulse voltage with a pulse width of 20 μs and a frequency of 5 kHz to the cold cathode array, with the anode grounded, the pulse mode can suppress cathode poisoning caused by gas adsorption and reduce power consumption.

[0027] In step 2, the field emission current is collected, specifically as follows: The anode current I is acquired in real time and sent to the signal acquisition and processing module for further processing.

[0028] In step 3, the collected field emission current is subjected to noise reduction processing, specifically as follows: A noise reduction method based on a combination of NGO-VMD and wavelet thresholding is used to denoise the acquired field emission current. This method is described in detail below. First, let's introduce the principle of VMD: The principle of VMD is to decompose a non-stationary signal f into K modal component sub-signals u. kEach component has a defined finite bandwidth and a center frequency ω. k The variational problem can be expressed as finding K modal functions u k (t), minimizing the sum of estimated bandwidths for each mode, with the constraint that the sum of modal components equals the input signal f. During construction, the modal function components u are first... k (t) Perform Hil-bert transform to obtain its analytic signal, and then combine the analytic signals of each modal function component with e -jωkt By mixing the components and modulating the spectrum of each modal component to the corresponding predicted fundamental frequency band, the bandwidth is estimated by calculating the squared L2 norm of the shifted signal gradient. The final expression obtained from the entire variational problem construction process is as follows: (1) In the formula, {u k}={u1, u2, ..., u K}、{ω k}={ω1, ω2, ..., ω K} represents all modes and their center frequencies; To find the partial derivative with respect to t; δ(t) is the Dirac distribution; Let convolution be the expression. Introducing a quadratic penalty factor α and a Lagrange multiplier λ transforms the above equation into an unconstrained variational problem. Here, α ensures the reconstruction accuracy of the signal even in the presence of Gaussian noise, and the Lagrange multiplier keeps the constraints strict. The augmented Lagrange function expression is: (2) Alternate updates using the alternating direction method of multipliers (ADMM) , and Solving the above equation for the optimal solution yields the modal component u. k Center frequency ω k The formulas for solving for λ are: (3) (4) (5) In the formula, and f(t) and u are respectively l n (t), λ n (t) and u k n+1 Fourier transform of (t); τ is the update parameter; n is the number of iterations.

[0029] Next, we will introduce the NGO principle. The Northern Goshawk Optimization Algorithm optimizes parameters by simulating the behavior of the Northern Goshawk during the hunting process. The algorithm principle is mainly divided into two stages: the prey identification and attack stage and the pursuit and escape stage. The mathematical model for each stage is as follows: (1) Prey identification and attack stage. In this stage, the Northern Goshawk randomly selects a prey and then quickly attacks it. Due to the random selection of prey in the search space, this stage increases the search capability of the algorithm. This stage is a global search, and its purpose is to identify the optimal region. The mathematical model formulas for this stage are shown in Equations (6) to (8): (6) (7) (8) In the formula, P i Let F be the location of the prey of the i-th Northern Goshawk; Pi Its objective function value; k is a random natural number in the interval [1, N]; X i new，P1 The new state for the i-th solution; X i，j new，P1 For its j-th dimension; F i new，P1 is the objective function value for the first stage; r is a random number in the interval [0, 1]; I is a random number, which can be 1 or 2; parameters r and I are random numbers used to generate random behavior during search and update; (2) The pursuit and escape stage: After the Northern Goshawk attacks the prey, the prey tries to escape. Therefore, during a chase, the Northern Goshawk continues to pursue the prey. Due to the high speed of the Northern Goshawk, it can chase the prey and eventually hunt it in almost any situation. The simulation of this behavior improves the algorithm's ability to search the local search space. This stage is a local search, and the goal is to find the optimal solution. The mathematical model formula for this stage is shown in Equations (9) to (11). (9) (10) (11) In the formula, t is the number of iterations; T is the maximum number of iterations; X i new，P2 The new state for the i-th solution; X i，j new，P2 For its j-th dimension; F i new，P2 This represents the objective function value for the second stage. After updating all population parameters according to the mathematical model of the NGO algorithm, the algorithm iteration is completed. At this time, all population parameter values, objective function and current optimal solution are determined. Then the algorithm enters the next iteration. Population members continue to update according to equations (6) to (11) until the last iteration is completed. The optimal solution obtained in the whole iteration process is used as the solution of the given optimization problem.

[0030] This section introduces the IMF (Information Mode Function) selection process. After VMD decomposition, the frequency bands contained in each IMF component are different. Noise-dominated components contain many anomalous signals, leading to a decreased correlation with the original signal and a smaller correlation coefficient. Conversely, components dominated by useful signal components have a better correlation with the original signal, resulting in a larger correlation coefficient. Therefore, the magnitude of the correlation coefficient between the original signal and each component signal can be used as a basis for selecting effective components. There is a critical threshold for the correlation coefficient between the two types of dominant components. If the correlation coefficient is higher than this threshold, the component is considered to contain a useful signal component; if the correlation coefficient is lower than this threshold, the component is considered to contain noise or anomalous components. The threshold is r. thr The calculation formula is as follows: (12) (13) In the formula, r i X is the correlation coefficient between the i-th IMF component and the original signal; i and Y i These are the IMF component and the original signal, respectively; X i and Y i These represent the mean of the IMF component data points and the mean of the original signal data points, respectively; N is the signal length; the correlation coefficient r is... i Greater than r thr The modal components are considered as effective components, and the correlation coefficient r i Less than r thr The modal components are considered noise components and require further processing. Next, we will introduce the wavelet thresholding method. The essence of wavelet thresholding denoising is signal filtering. After wavelet decomposition of the noisy signal, we obtain the decomposition coefficients of the original signal and the noise. The decomposition coefficients of the original signal are greater than those of the noise. Therefore, a reasonable threshold needs to be selected. Through thresholding, noise is filtered out. Components with decomposition coefficients greater than the threshold are considered to originate from the original signal and are retained. Components with decomposition coefficients less than the threshold are considered to originate from the noise signal and are discarded. Finally, the signal is reconstructed. The key to wavelet thresholding is determining four key parameters: the wavelet basis function, the decomposition scale, the threshold function, and the threshold calculation method. For the wavelet basis function, wavelet systems with higher-order vanishing moments, such as the DBN, SYN, and COIFN wavelet systems, are typically chosen in engineering. For the threshold function, it is generally divided into hard thresholding and soft thresholding. Hard thresholding preserves wavelet coefficients above a threshold while setting those below the threshold to zero. Hard thresholding can better preserve local signal features, but it introduces abrupt changes at the threshold in the wavelet domain, potentially causing new oscillations after signal reconstruction. Soft thresholding, on the other hand, subtracts the threshold from wavelet coefficients whose absolute values ​​are greater than it, and sets coefficients below the threshold to zero. Soft thresholding produces smoother signals, eliminating the local abrupt changes caused by hard thresholding. However, it suffers from constant bias, resulting in some distortion in the amplitude of the reconstructed signal. Currently, the mainstream methods for calculating thresholds include Sqtwolog threshold, Minmax threshold, Rigsure threshold, and Heur-sure threshold.

[0031] The noise reduction method of the present invention will now be described in detail. To improve VMD performance, NGO optimization of VMD parameters is employed. During optimization, an objective function needs to be defined. Envelope entropy reflects the sparsity of the components. After VMD decomposition of a zero-sequence current signal, the more noise the resulting IMF component contains, the smaller the signal sparsity and the larger the envelope entropy value. Conversely, if the IMF component exhibits strong regularity and less noise, the signal sparsity is greater, and the envelope entropy is smaller. Since VMD decomposition yields K components, there will be K envelope entropy values. The smallest of these K values ​​is selected as the local minimum envelope entropy value, min. Ee The objective function is to find the global minimum envelope entropy and the corresponding optimal component combination K and α, where E is the envelope entropy. e The calculation formula is: (14) In the formula, a(j) is the IMF component obtained after VMD decomposition of the signal through Hilbert transform; bj is the normalized form of a(j). The specific steps of the NGO optimization VMD algorithm are as follows, and the flowchart is shown below. Figure 2 as shown (1)Initialization of NGO parameters. The value range of K is [2, 10], the value range of α is [500, 20000], the population size is 30, and the maximum number of iterations is 15; (2)Perform VMD decomposition on the zero-sequence current, and select the objective function as the minimum envelope entropy value minE e , and calculate minE e by substituting different combinations of K and α each time, and then compare and update the current best objective function value; (3)Determine whether to terminate the iteration. If t < T, let t = t + 1 and continue to update. Otherwise, the iteration terminates, and the global minimum envelope entropy and its corresponding parameter combinations K and α are saved; The influence of other parameters on the decomposition effect is relatively small, and they are set to empirical values, that is, the noise tolerance tau = 0, the initial center frequency init = 1, the DC component DC = 0, and the convergence criterion tolerance ε = 1×10 -7 ; Finally, the present invention proposes a zero-sequence current noise reduction method for single-phase grounding faults based on the combination of NGO-VMD and wavelet threshold method. The specific process is as Figure 3 shown

[0032] In step 4, temperature compensation is performed on the field emission current after noise reduction processing. Specifically: A temperature sensor is set at the bottom of the insulating substrate, and a dynamic compensation network is constructed based on the LSTM model, and then temperature compensation is performed on the field emission current after noise reduction processing; The present invention provides two schemes for temperature compensation. The first is a linear compensation model, which is: (15) The other is to construct a dynamic compensation network based on the LSTM model. The inputs are the original current I, temperature T, voltage V, and historical vacuum degree P t-1 ; the output is the compensated current I comp ; Its network structure code is: class TempCompLSTM(nn.Module): def __init__(self): super().__init__() self.lstm = nn.LSTM(input_size = 4, hidden_size = 32, num_layers = 2) # Input 4-dimensional parameters self.fc = nn.Linear(32, 1) # Output compensation current def forward(self, x): x, _ = self.lstm(x) # x: [batch, seq_len, 4] x = self.fc(x[:, -1, :]) # Get the last time step return x Training data: Calibrate current values ​​at different (P, V) within a temperature range of -40~85℃; In step 5, the vacuum level is calculated based on the compensated field emission current, specifically as follows: The formula is: (16) In the formula, the precalibration parameters B=480, k=1.8, α=0.33.

[0033] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0034] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A carbon nanotube cold cathode micro vacuum sensor, characterized in that, include: The assembly comprises an insulating substrate, a cold cathode array, an anode electrode, a glass cover plate, and a signal acquisition and processing module. The cold cathode array is disposed on the insulating substrate, and the anode electrode is disposed on the insulating substrate corresponding to the cold cathode array via insulating supports. The anode electrode is positioned directly above the cold cathode array. The glass cover plate is disposed on the insulating substrate, and the glass cover plate covers the cold cathode array and the anode electrode inside, forming a vacuum cavity. The cold cathode array and the anode electrode are connected to the signal acquisition and processing module.

2. The carbon nanotube cold cathode micro vacuum sensor according to claim 1, characterized in that, The insulating substrate is an Al2O3 ceramic or a silicon-based SiO2 insulating layer with a thickness of 0.5-1 mm.

3. The carbon nanotube cold cathode micro vacuum sensor according to claim 2, characterized in that, The method for fabricating the cold cathode array is as follows: Cathode electrode fabrication: A Cr / Pt metal layer with a thickness of 200 nm–500 nm is sputtered on an insulating substrate, and a strip-shaped cathode electrode pattern is formed by photolithography. CNT growth: Vertical multi-walled carbon nanotube arrays with heights of 10 μm–50 μm and densities of 10-1 were grown on the cathode electrode surface using CVD. 5 -10 7 tubes / mm 2 The thickness unevenness is controlled by screen printing process, so that the CNT bundles form a gradient distribution structure, in which the top is sinusoidal wavy to enhance the field emission efficiency.

4. The carbon nanotube cold cathode micro vacuum sensor according to claim 3, characterized in that, The anode electrode is a Ni mesh structure with a light transmittance greater than 80% and a spacing of 50μm–200μm.

5. The carbon nanotube cold cathode micro vacuum sensor according to claim 4, characterized in that, A glass solder layer is provided at the edge of the insulating substrate. The glass cover plate is aligned with the glass solder layer and placed above the insulating substrate. It is then heated to achieve an airtight seal.

6. A method for detecting the field emission current response of a carbon nanotube cold cathode micro vacuum sensor, characterized in that, The carbon nanotube cold cathode micro vacuum sensor according to any one of claims 1-5 comprises: Step 1: Apply pulse voltage; Step 2: Collect field emission current; Step 3: Perform noise reduction processing on the collected field emission current; Step 4: Perform temperature compensation on the field emission current after noise reduction; Step 5: Calculate the vacuum level based on the compensated field emission current.

7. The method according to claim 6, characterized in that, In step 1, a pulse voltage is applied, specifically: a -200 V pulse voltage is applied to the cold cathode array with a pulse width of 20 μs and a frequency of 5 kHz, and the anode is grounded.

8. The method according to claim 7, characterized in that, In step 3, the acquired field emission current is denoised, specifically by using a denoising method based on a combination of NGO-VMD and wavelet thresholding.

9. The method according to claim 8, characterized in that, In step 4, temperature compensation is performed on the field emission current after noise reduction. Specifically, a temperature sensor is set at the bottom of the insulating substrate, and a dynamic compensation network is constructed based on the LSTM model to perform temperature compensation on the field emission current after noise reduction.

Images