Double-probe electron density diagnosis method and system based on long short-term memory network

Abstract

The application provides a double-probe electron density diagnosis method and system based on a long short-term memory network, and relates to the field of plasmas. The method comprises the following steps: building a plasma simulation vacuum chamber, obtaining double-probe I-V characteristic curve data, and calculating three-probe electron density at the same position; constructing a bidirectional long short-term memory network model, taking the double-probe I-V characteristic curve data as input features, taking the three-probe electron density as labels, training the bidirectional long short-term memory network model, and obtaining a trained bidirectional long short-term memory network model; inputting double-probe I-V characteristic curve data to be tested into the trained bidirectional long short-term memory network model, and outputting a diagnosis result. e The application uses double-probe I-V characteristic curve data at the same position as features, three-probe N e as labels to train the bidirectional long short-term memory network model, solves the problem that the double probe cannot directly measure N e due to the fact that the double probe can only collect high-energy part of electrons in the plasma, and the number of points required to collect the double-probe I-V characteristic curve is less.

Description

Technical Field

[0001] This invention belongs to the field of plasma technology, and particularly relates to a dual-probe electron density diagnostic method and system based on long short-term memory networks. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Electrostatic probe diagnostics is a widely used plasma diagnostic technique, characterized by its simplicity, ability to obtain numerous parameters, and reliable results. Currently, electrostatic probe diagnostics mainly includes single-probe, dual-probe, and triple-probe techniques.

[0004] In the 1920s, Langmuir and Mott-Smith proposed a single probe capable of measuring a variety of plasma parameters. The Langmuir probe was very simple to operate; it only required inserting a conductor into the plasma, applying a scanning voltage to the metal electrodes using a variable power supply, and measuring the current on the electrodes to obtain the IV curve, thereby obtaining the plasma parameters.

[0005] However, using a single probe for plasma diagnostics has many limitations. For example, when the scanning voltage applied to the probe is higher than the plasma potential, the probe attracts electrons from the plasma, forming an electron saturation current. Due to the difference in mass between electrons and ions, the electron saturation current is much larger than the ion saturation current, causing relatively greater interference with the plasma environment. In a sense, it alters the plasma being diagnosed. Furthermore, Langmuir probes are susceptible to contamination, leading to inaccurate diagnostic results. Additionally, when a Langmuir single probe is used, a large-area counter electrode is required to compensate for the current collected by the probe. This is easily achieved in a laboratory environment, for example, by using a plasma container as a counter electrode. However, this is difficult to achieve in a space environment, especially when carried on small aircraft: it is often difficult to find a conductor with a sufficiently large area to serve as a counter electrode.

[0006] Compared to single-probe systems, dual-probe systems operate in a suspended state, eliminating the need for counter electrodes. Furthermore, the loop current of a dual-probe system never exceeds the ion saturation current collected by any one probe. Therefore, dual-probe systems significantly reduce the interference of the collection current on the plasma environment. However, in traditional dual-probe diagnostics, each probe can only collect the high-energy electrons from the plasma, making it impossible to detect N... e Direct measurement can be performed. Therefore, if the dual-probe diagnostic method can be improved to allow direct measurement of N, it would be beneficial. e This will make dual probes the preferred method for achieving multi-parameter diagnosis of plasma, replacing single probes. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, this invention provides a dual-probe electron density diagnostic method and system based on long short-term memory networks, using dual-probe IV characteristic curve data from the same location as features and a three-probe N... e As a label-trained bidirectional long short-term memory network model, different N e The dual-probe IV characteristic curve data under the given conditions are input into the trained bidirectional long short-term memory network model to obtain N. e The predicted value solves the problem that dual probes cannot directly measure N because they can only collect high-energy electrons from the plasma. e The problem is that the number of points required to collect the dual-probe IV characteristic curve is relatively small.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0009] The first aspect of this invention provides a dual-probe electron density diagnostic method based on long short-term memory networks.

[0010] The dual-probe electron density diagnostic method based on long short-term memory networks includes the following steps:

[0011] A plasma simulation vacuum chamber was built to obtain the dual-probe IV characteristic curve data and calculate the electron density of the three-probe at the same location.

[0012] A bidirectional long short-term memory network model was constructed. The dual-probe IV characteristic curve data was used as input features and the three-probe electron density was used as labels. The bidirectional long short-term memory network model was trained to obtain a trained bidirectional long short-term memory network model.

[0013] The dual-probe IV characteristic curve data to be tested is input into the trained bidirectional long short-term memory network model, and the diagnostic results are output.

[0014] A second aspect of the present invention provides a dual-probe electron density diagnostic system based on a long short-term memory network.

[0015] A dual-probe electron density diagnostic system based on long short-term memory networks includes:

[0016] The data acquisition module is configured to: build a plasma simulation vacuum chamber, acquire dual-probe IV characteristic curve data, and calculate the electron density of the three-probe at the same location;

[0017] The model training module is configured to: construct a bidirectional long short-term memory network model, use the dual-probe IV characteristic curve data as input features and the three-probe electron density as labels, train the bidirectional long short-term memory network model, and obtain a trained bidirectional long short-term memory network model.

[0018] The diagnostic result acquisition module is configured to input the dual-probe IV characteristic curve data to be tested into the trained bidirectional long short-term memory network model and output the diagnostic results.

[0019] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the dual-probe electron density diagnostic method based on a long short-term memory network as described in the first aspect of the present invention.

[0020] A fourth aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the dual-probe electron density diagnostic method based on a long short-term memory network as described in the first aspect of the present invention.

[0021] The above one or more technical solutions have the following beneficial effects:

[0022] The dual probe can only collect high-energy electrons from the plasma and cannot directly measure N. e To address this problem, this invention provides a dual-probe electron density diagnostic method and system based on long short-term memory networks, using dual-probe IV characteristic curve data from the same location as features and a three-probe N... e Using the labels to train the BLSTM model, after training, it can achieve N-level mapping using fewer data points from the dual probes. e High-precision diagnosis.

[0023] This invention solves the problem that dual probes can only collect high-energy electrons from plasma, which prevents direct measurement of N. e The problem is that the number of points required to collect the dual-probe IV characteristic curve is relatively small. Applying the diagnostic method provided by this invention to ionospheric plasma diagnosis can reduce the amount of data collected by the probe and improve the spatial resolution of ionospheric detection, which has a very good application prospect.

[0024] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0026] Figure 1 This is a schematic diagram of the dual-probe diagnostic principle in the first embodiment.

[0027] Figure 2This is a typical dual-probe IV characteristic curve for the first embodiment.

[0028] Figure 3 This is a basic unit structure diagram of the first embodiment of LSTM.

[0029] Figure 4 This is a diagram of the BLSTM architecture for the first embodiment.

[0030] Figure 5 This is a schematic diagram of the plasma simulation vacuum chamber structure for the first embodiment.

[0031] Figure 6 The flowchart shows the training process of BLSTM in the first embodiment.

[0032] Figure 7 For the first embodiment N e A comparison chart of test results and actual voltage.

[0033] Figure 8 This is a flowchart of the method in the first embodiment.

[0034] Figure 9 This is a system structure diagram of the second embodiment. Detailed Implementation

[0035] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0036] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0037] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0038] Example 1

[0039] Before introducing the dual-probe electron density diagnostic method based on long short-term memory network in this embodiment, we will first introduce the principles of dual-probe diagnostic technology, LSTM principle and BLSTM principle.

[0040] (I) Principle of Dual-Probe Diagnostic Technology

[0041] like Figure 1 As shown, two single probes are inserted into the plasma to form a double probe. A scanning voltage V from negative to positive is applied between the two probes by a DC voltage source. The current of the entire probe circuit is I. Probe 1 collects electron current and probe 2 collects ion current. Figure 2This is a typical IV characteristic curve for a dual-probe system. Because the net current collected by the dual probes is zero, the currents collected by the two probes are equal in magnitude and opposite in direction. The electron current collected by probe 1 will not exceed the ion saturation current collected by probe 2, thus the interference to the probed plasma is very low. However, the probes in the dual-probe system can only collect electrons from the high-energy portion of the plasma, and these high-energy electrons do not represent the bulk electron distribution in the plasma. Therefore, the dual-probe system cannot directly measure N. e .

[0042] (II) LSTM Principle

[0043] The emergence of Long Short-Term Memory (LSTM) networks has solved the problem of long-term information retention in traditional neural network models. For example... Figure 3 As shown, LSTM includes a forget gate, an input gate, and an output gate. It is processed by a fully connected layer through three Sigmoid activation functions to obtain the values ​​of the forget gate, the input gate, and the output gate, and the values ​​of the three gates can be controlled within the range of (0,1).

[0044] Forgotten Gate F t Determine the state C of the previous unit t-1 The degree of forgetting is calculated as shown in equation (1):

[0045] F t =σ(X) t W xf +H t-1 W hf +b f (1)

[0046] Where W and b are the weight parameter and bias parameter, respectively.

[0047] Input Gate I t The decision on whether to retain the current data is calculated as shown in equation (2):

[0048] I t =σ(X) t W xi +H t-1 W hi +b i (2)

[0049] Forget gate, input gate, and candidate memory C t After element-wise multiplication and addition, we obtain the memory element C. t The calculation formulas are shown in equations (3) and (4):

[0050] C t =tanh(X) t Wxc +H t-1 W hc +b c (3)

[0051] C t =F t *C t-1 +I t *C t (4)

[0052] Output gate O t The degree of external visibility within the control unit is calculated using the formula shown in equation (5):

[0053] O t =σ(X) t W xo +H t-1 W ho +b io (5)

[0054] Hidden state H t It is the medium through which the output gate functions, and the calculation formula is shown in equation (6):

[0055] H t =O t *tanh(C t (6)

[0056] (III) BLSTM Principle

[0057] The architecture of BLSTM is as follows Figure 4 As shown, the input layer is fed into both a forward LSTM and a backward LSTM, and the final output is a combination of the two LSTM outputs. Compared to LSTM, BLSTM can uncover the correlation between historical and future information, thus making fuller use of data and improving model performance.

[0058] This embodiment discloses a dual-probe electron density diagnostic method based on long short-term memory networks.

[0059] like Figure 8 As shown, the dual-probe electron density diagnostic method based on long short-term memory networks includes the following steps:

[0060] A plasma simulation vacuum chamber was built to obtain the dual-probe IV characteristic curve data and calculate the electron density of the three-probe at the same location.

[0061] A bidirectional long short-term memory network model was constructed. The dual-probe IV characteristic curve data was used as input features and the three-probe electron density was used as labels. The bidirectional long short-term memory network model was trained to obtain a trained bidirectional long short-term memory network model.

[0062] The dual-probe IV characteristic curve data to be tested is input into the trained bidirectional long short-term memory network model, and the diagnostic results are output.

[0063] like Figure 5 As shown, the data used in this method were collected and calculated by dual-probe and triple-probe systems in a plasma simulation vacuum chamber. After argon gas was introduced into the chamber, a DC glow discharge plasma was generated by controlling the plasma source through a control unit, forming a plasma environment with a gradient density distribution. With the same filament current and gas pressure parameters, the plasma density inside the chamber changed with the distance from the plasma source; changing the filament current could significantly adjust the plasma density within the vacuum chamber. A two-dimensional motor platform was installed inside the vacuum chamber, and three probes were fixed to the platform. The motor control system controlled the platform to move the probes in the X or Y direction of the two-dimensional plane, moving 20 mm at a time. Two probes were designated as dual-probe systems. Data from the triple-probe system at each location was collected using a source meter, and the IV characteristic curve of the dual-probe system was scanned, thus obtaining dual-probe and triple-probe data with continuously distributed plasma density.

[0064] Since the data collected under the same filament current and pressure parameters is limited, data collection in this embodiment was conducted under different current conditions, with dual-probe and triple-probe data collected under each condition. The scanning voltage range for dual-probe data acquisition was -8V to +8V, with a sampling interval of 1V. One IV characteristic curve contained 17 sampling points. The current values ​​at these 17 sampling points were used as input features to train and predict the BLSTM network. The electron density calculated from the triple-probe data at the same location was used as the label for network training.

[0065] This method standardizes the data to eliminate the influence of different dimensions among the indicators. The data is divided into training, validation, and test sets. The training set is used to fit the data samples and train the weight parameters; the validation set is used for preliminary evaluation of the model, adjusting the hyperparameters to prevent overfitting; and the test set is used to test the model's generalization ability.

[0066] The training process of the BLSTM model is as follows: Figure 6 As shown, the training set is input into the network for one training iteration; the validation set is used to perform a preliminary evaluation of the model's performance to prevent overfitting, and the network parameters that perform best on the validation set are saved; the test set is input into the saved network, and the prediction results are output to evaluate the model.

[0067] This method uses BLSTM to process N. eTo make predictions, the internal connections of the entire sequence are extracted from both positive and negative directions. To prevent overfitting due to excessive extraction of internal data information, a random dropout layer is added after the BLSTM layer. The BLSTM consists of two LSTM layers, one positive and one negative. All outputs of the LSTMs enter a fully connected layer, which uses a linear activation function. To ensure that the final output matches the number of labels, the number of neurons in the fully connected layer is set to 1.

[0068] Because the input features have a low dimensionality, too many network layers can easily lead to overfitting. Therefore, a single-layer network was chosen for both forward and reverse directions, resulting in a total of two layers for both directions. To obtain more accurate prediction results, networks with different numbers of neurons were trained within a certain range, and the training loss of the networks was compared to select the optimal network structure. Ultimately, 35 neurons were chosen for each layer.

[0069] The learning rate is a very important hyperparameter in BLSTM network training. In order to find a suitable learning rate, the maximum number of update steps is fixed, and the initial learning rate and target learning rate are combined within a certain range. The evaluation index under different learning rate conditions is compared, and finally the initial learning rate is determined to be 0.0001 and the target learning rate is 0.00001.

[0070] The training and validation sets are input into the model for training and validation. After the model is trained, the test set is used to test the model. After 200 training and validation iterations, the test set is input into the model for testing.

[0071] like Figure 7 As shown, the test set contains 238 points. As the number increases, the data acquisition location gradually moves closer to the plasma source. e This also increases accordingly. The initial point in the test set is far from the plasma source, N e The current signal is very weak and easily affected by electromagnetic interference from surrounding electronic instruments, resulting in a large prediction error. The overall prediction accuracy of the test set is relatively high, with a loss value of approximately 0.005.

[0072] Example 2

[0073] This embodiment discloses a dual-probe electron density diagnostic system based on long short-term memory networks.

[0074] like Figure 9 As shown, the dual-probe electron density diagnostic system based on long short-term memory networks includes:

[0075] The data acquisition module is configured to: build a plasma simulation vacuum chamber, acquire dual-probe IV characteristic curve data, and calculate the electron density of the three-probe at the same location;

[0076] The model training module is configured to: construct a bidirectional long short-term memory network model, use the dual-probe IV characteristic curve data as input features and the three-probe electron density as labels, train the bidirectional long short-term memory network model, and obtain a trained bidirectional long short-term memory network model.

[0077] The diagnostic result acquisition module is configured to input the dual-probe IV characteristic curve data to be tested into the trained bidirectional long short-term memory network model and output the diagnostic results.

[0078] Example 3

[0079] The purpose of this embodiment is to provide a computer-readable storage medium.

[0080] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps in the dual-probe electron density diagnostic method based on a long short-term memory network as described in Embodiment 1 of this disclosure.

[0081] Example 4

[0082] The purpose of this embodiment is to provide an electronic device.

[0083] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the dual-probe electron density diagnostic method based on a long short-term memory network as described in Embodiment 1 of this disclosure.

[0084] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0085] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0086] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A dual-probe electron density diagnostic method based on long short-term memory networks, characterized in that, Includes the following steps: A plasma simulation vacuum chamber was built to obtain the dual-probe IV characteristic curve data and calculate the electron density of the three-probe at the same location. A bidirectional long short-term memory network model was constructed. The dual-probe IV characteristic curve data was used as input features and the three-probe electron density was used as labels. The bidirectional long short-term memory network model was trained to obtain a trained bidirectional long short-term memory network model. The dual-probe IV characteristic curve data to be tested is input into the trained bidirectional long short-term memory network model, and the diagnostic results are output. A two-dimensional motor operating platform was set up inside a plasma simulation vacuum chamber. Three probes were set up on the two-dimensional motor operating platform. After argon gas was introduced into the vacuum chamber, plasma was generated by controlling the plasma source. The motor operating platform was then controlled to move the probes on a two-dimensional plane. Two of the probes were designated as dual probes. Data of the three probes and the IV characteristic curves of the dual probes were collected at each position. The electron density of the three probes was calculated based on the collected data.

2. The dual-probe electron density diagnostic method based on long short-term memory networks as described in claim 1, characterized in that, Under different current conditions, I-V characteristic curves of three probes and two probes were collected respectively.

3. The dual-probe electron density diagnostic method based on long short-term memory networks as described in claim 1, characterized in that, The bidirectional long short-term memory network model uses forward LSTM and backward LSTM as input layers. Both forward and backward LSTM are single-layer networks with 35 neurons per layer. The output layer is connected to a fully connected layer with 1 neuron. The initial learning rate is 0.0001 and the target learning rate is 0.00001.

4. The dual-probe electron density diagnostic method based on long short-term memory networks as described in claim 3, characterized in that, The final output of the bidirectional long short-term memory network model is a combination of the outputs of the forward LSTM and the backward LSTM.

5. The dual-probe electron density diagnostic method based on long short-term memory networks as described in claim 1, characterized in that, The dataset consists of the dual-probe IV characteristic curve data and the electron density of the three probes at the same location. The dataset is divided into a training set, a validation set, and a test set. The bidirectional long short-term memory network model is trained using the training set, the hyperparameters of the bidirectional long short-term memory network model are adjusted using the validation set, and the trained bidirectional long short-term memory network model is tested using the test set.

6. The dual-probe electron density diagnostic method based on long short-term memory networks as described in claim 5, characterized in that, A well-trained bidirectional long short-term memory network model is obtained when the loss of the bidirectional long short-term memory network model does not decrease or reaches the maximum training period.

7. A dual-probe electron density diagnostic system based on long short-term memory networks, characterized in that: include: The data acquisition module is configured to: build a plasma simulation vacuum chamber, acquire dual-probe IV characteristic curve data, and calculate the electron density of the three-probe at the same location; The model training module is configured to: construct a bidirectional long short-term memory network model, use the dual-probe IV characteristic curve data as input features and the three-probe electron density as labels, train the bidirectional long short-term memory network model, and obtain a trained bidirectional long short-term memory network model. The diagnostic result acquisition module is configured to: input the dual-probe IV characteristic curve data to be tested into the trained bidirectional long short-term memory network model and output the diagnostic results; A two-dimensional motor operating platform was set up inside a plasma simulation vacuum chamber. Three probes were set up on the two-dimensional motor operating platform. After argon gas was introduced into the vacuum chamber, plasma was generated by controlling the plasma source. The motor operating platform was then controlled to move the probes on a two-dimensional plane. Two of the probes were designated as dual probes. Data of the three probes and the IV characteristic curves of the dual probes were collected at each position. The electron density of the three probes was calculated based on the collected data.

8. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by a processor, the program implements the steps in the dual-probe electron density diagnostic method based on long short-term memory networks as described in any one of claims 1-6.

9. An electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the dual-probe electron density diagnostic method based on long short-term memory networks as described in any one of claims 1-6.

Images