X-ray tube and current control method thereof

Abstract

The present application relates to the field of intelligent control, and specifically discloses an X-ray tube and a current control method thereof, which adopts artificial intelligence control technology, and uses a convolutional neural network model as a feature extractor to extract implicit features of local correlation with the primary current drive signal and the secondary current drive signal of the filament transformer, as well as the feedback sampling voltage signal and the control voltage signal output by the system control unit. In this way, when controlling the output control voltage of the system control unit, it is possible to pay attention to the implicit correlation flow feature information between the data, so as to control the control voltage more accurately, thereby keeping the secondary current drive signal stable.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control, and more specifically, to an X-ray tube and a current control method thereof. Background Art

[0002] An X-ray tube is a vacuum diode operating at high voltage. It is a vacuum electronic device that can generate X-rays by using high-speed electrons to hit a metal target. The X-ray tube contains two electrodes: one is the cathode of the filament for emitting electrons, and the other is the anode of the target for receiving electron bombardment. Both electrodes are sealed in a high vacuum glass or ceramic shell.

[0003] With the continuous development of power electronics technology, X-ray tubes are widely used in medicine and industry. However, during the operation of X-ray tubes, each time a different high voltage or different tube (including tubes of the same model from the same manufacturer) is matched, the system control unit needs to find the corresponding relationship table between the filament current, tube current and tube voltage of different tubes, and the table provided by the tube manufacturer itself has a large error range. Under the same tube voltage, during the initial period of heating the tungsten filament, the filament current of the same tube also corresponds to different tube currents. Since X-ray tubes have the characteristics of general electron tubes, they are easily affected by factors such as temperature and lifespan. The existing technology only performs feedback control on the filament current, which is one of the factors affecting it, when controlling the X-ray tube current, resulting in poor accuracy and stability of the output tube current.

[0004] Therefore, an optimized X-ray tube is desired to keep the final driving current stable by controlling the current gain. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an X-ray tube and a current control method thereof, which adopts artificial intelligence control technology and uses a convolutional neural network model as a feature extractor to extract implicit features of local correlation of the primary current drive signal and the secondary current drive signal of the filament transformer, as well as the feedback sampling voltage signal and the control voltage signal output by the system control unit, so that when controlling the output control voltage of the system control unit, it is possible to pay attention to the implicit correlation flow feature information between the data, so as to more accurately control the control voltage, thereby keeping the final drive current stable.

[0006] According to one aspect of the present application, there is provided an X-ray tube, comprising:

[0007] A current signal acquisition module, used to acquire a primary current drive signal and a secondary current drive signal acquired by a current transformer;

[0008] A voltage signal acquisition module is used to acquire the feedback sampling voltage signal acquired by the pulse modulator and the control voltage signal output by the system control unit;

[0009] A current signal encoding module, used for respectively passing the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal through a first convolutional neural network as a feature extractor to obtain a primary current feature vector and a secondary current feature vector;

[0010] A first transfer module, used for calculating a current transfer matrix of the secondary current eigenvector relative to the primary current eigenvector;

[0011] A voltage signal encoding module, used for respectively passing the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal through a second convolutional neural network as a feature extractor to obtain a feedback sampled voltage feature vector and a control voltage feature vector;

[0012] A second transfer module, used for calculating a voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector;

[0013] A feature dimension reduction module, used for performing feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector;

[0014] a feature fusion module, configured to fuse the transpose of the circuit transfer feature vector and the product of the voltage transfer feature vector to obtain a decoding feature matrix; and

[0015] The decoding module is used to decode and regress the decoding feature matrix through a decoder to obtain a decoding value, and the decoding value is used to represent the adjustment voltage value of the output voltage signal.

[0016] In the above-mentioned X-ray tube, the current signal encoding module is further used for: each layer of the first convolutional neural network performs respectively in the forward transmission of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network is the primary current feature vector and the secondary current feature vector, and the input of the first layer of the first convolutional neural network is the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal.

[0017] In the above X-ray tube, the first transfer module is further used to: calculate the current transfer matrix of the secondary current characteristic vector relative to the primary current characteristic vector using the following formula;

[0018] Wherein, the formula is:

[0019] M1=S1*F1

[0020] Wherein F1 represents the secondary current characteristic vector, M1 represents the current transfer matrix, and S1 represents the primary current characteristic vector.

[0021] In the above-mentioned X-ray tube, the voltage signal encoding module is further used for: each layer of the second convolutional neural network performs respectively in the forward transmission of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the second convolutional neural network is the feedback sampled voltage feature vector and the control voltage feature vector, and the input of the first layer of the second convolutional neural network is the waveform diagram of the feedback sampled voltage signal and the waveform diagram of the control voltage signal.

[0022] In the above X-ray tube, it is further used to: calculate the voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector by the following formula;

[0023] Wherein, the formula is:

[0024] M2=S2*F2

[0025] Wherein F2 represents the feedback sampling voltage characteristic vector, M2 represents the voltage transfer matrix, and S2 represents the control voltage characteristic vector.

[0026] In the above X-ray tube, the feature dimension reduction module includes: a first row vector characterization unit, which is used to calculate, for each row vector in the current transfer matrix, a logarithmic function value of the difference between the sum of the natural exponential function values ​​of the eigenvalues ​​at each position in the row vector raised to the power and the sum of the eigenvalues ​​at all positions in the row vector divided by the logarithmic function value of the difference between the sum of the eigenvalues ​​at all positions in the row vector minus one to obtain the eigenvalues ​​corresponding to each row vector in the current transfer matrix; and a first vector construction unit, which is used to arrange the eigenvalues ​​corresponding to each row vector in the current transfer matrix in a longitudinal direction to obtain the eigenvalues ​​corresponding to each row vector in the current transfer matrix. A circuit transfer characteristic vector; a second row vector characterization unit, used to calculate, for each row vector in the voltage transfer matrix, the logarithmic function value of the difference between the sum of the natural exponential function values ​​raised to the power of the eigenvalues ​​at each position in the row vector and the sum of the eigenvalues ​​at all positions in the row vector, divided by the logarithmic function value of the difference between the sum of the eigenvalues ​​at all positions in the row vector minus one, to obtain the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix; and a second vector construction unit, used to vertically arrange the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix to obtain the circuit transfer characteristic vector.

[0027] In the above X-ray tube, the feature fusion module is further used to: calculate the product between the transpose of the circuit transfer feature vector and the voltage transfer feature vector using the following formula to obtain the decoding feature matrix;

[0028] Wherein, the formula is:

[0029]

[0030] Among them, V1 represents the circuit transfer feature vector, V2 represents the voltage transfer feature vector, and M represents the decoding feature matrix.

[0031] In the above X-ray tube, the decoding module is further used to: use the decoder to perform decoding regression on the decoding feature matrix to obtain the decoding value according to the following formula; wherein the formula is: Where X is the decoded feature matrix, Y is the decoded value, and W is the weight matrix, Represents matrix multiplication.

[0032] According to another aspect of the present application, a current control method for an X-ray tube includes:

[0033] Acquire a primary current drive signal and a secondary current drive signal collected by a current transformer;

[0034] Acquire a feedback sampling voltage signal collected by a pulse modulator and a control voltage signal output by a system control unit;

[0035] Passing the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal through a first convolutional neural network as a feature extractor to obtain a primary current feature vector and a secondary current feature vector respectively;

[0036] Calculating a current transfer matrix of the secondary current eigenvector relative to the primary current eigenvector;

[0037] Passing the waveform of the feedback sampling voltage signal and the waveform of the control voltage signal through a second convolutional neural network as a feature extractor to obtain a feedback sampling voltage feature vector and a control voltage feature vector;

[0038] Calculating a voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector;

[0039] Performing feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector;

[0040] fusing the transpose of the circuit transfer feature vector and the product of the voltage transfer feature vector to obtain a decoding feature matrix; and

[0041] The decoding characteristic matrix is ​​decoded and regressed through a decoder to obtain a decoding value, and the decoding value is used to represent the adjustment voltage value of the output voltage signal.

[0042] In the above-mentioned current control method for an X-ray tube, the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal are respectively passed through a first convolutional neural network as a feature extractor to obtain a primary current feature vector and a secondary current feature vector, including: each layer of the first convolutional neural network performs in the forward transmission of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network is the primary current feature vector and the secondary current feature vector, and the input of the first layer of the first convolutional neural network is the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal.

[0043] In the above-mentioned current control method of the X-ray tube, calculating the current transfer matrix of the secondary current characteristic vector relative to the primary current characteristic vector includes: calculating the current transfer matrix of the secondary current characteristic vector relative to the primary current characteristic vector by the following formula;

[0044] Wherein, the formula is:

[0045] M1=S1*F1

[0046] Wherein F1 represents the secondary current characteristic vector, M1 represents the current transfer matrix, and S1 represents the primary current characteristic vector.

[0047] In the above-mentioned current control method for an X-ray tube, the waveform of the feedback sampling voltage signal and the waveform of the control voltage signal are respectively passed through a second convolutional neural network as a feature extractor to obtain a feedback sampling voltage feature vector and a control voltage feature vector, including: each layer of the second convolutional neural network performs the following steps in the forward transmission of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on a local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the second convolutional neural network is the feedback sampling voltage feature vector and the control voltage feature vector, and the input of the first layer of the second convolutional neural network is the waveform of the feedback sampling voltage signal and the waveform of the control voltage signal.

[0048] In the above-mentioned current control method of the X-ray tube, calculating the voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector includes: calculating the voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector using the following formula;

[0049] Wherein, the formula is:

[0050] M2=S2*F2

[0051] Wherein F2 represents the feedback sampling voltage characteristic vector, M2 represents the voltage transfer matrix, and S2 represents the control voltage characteristic vector.

[0052] In the above-mentioned current control method for an X-ray tube, the current transfer matrix and the voltage transfer matrix are subjected to feature dimensionality reduction to obtain a circuit transfer feature vector and a voltage transfer feature vector, including: for each row vector in the current transfer matrix, calculating a logarithmic function value of a difference between a sum of natural exponential function values ​​raised to power by the eigenvalues ​​at each position in the row vector and a sum of the eigenvalues ​​at all positions in the row vector divided by a logarithmic function value of a difference between the sum of the eigenvalues ​​at all positions in the row vector minus one to obtain a feature value corresponding to each row vector in the current transfer matrix; and dividing the logarithmic function value of a difference between the sum of the eigenvalues ​​at all positions in the row vector and one by one. The eigenvalues ​​of each row vector in the matrix are arranged vertically to obtain the circuit transfer eigenvector; for each row vector in the voltage transfer matrix, the logarithmic function value of the difference between the sum of the natural exponential function values ​​raised to the power of the eigenvalues ​​at each position in the row vector and the sum of the eigenvalues ​​at all positions in the row vector is calculated, divided by the logarithmic function value of the difference between the sum of the eigenvalues ​​at all positions in the row vector and one, to obtain the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix; and the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix are arranged vertically to obtain the circuit transfer eigenvector.

[0053] In the above-mentioned current control method of the X-ray tube, fusing the transpose of the circuit transfer feature vector and the product between the voltage transfer feature vector to obtain a decoding feature matrix includes: calculating the product between the transpose of the circuit transfer feature vector and the voltage transfer feature vector by the following formula to obtain the decoding feature matrix;

[0054] Wherein, the formula is:

[0055]

[0056] Among them, V1 represents the circuit transfer feature vector, V2 represents the voltage transfer feature vector, and M represents the decoding feature matrix.

[0057] In the above-mentioned current control method of the X-ray tube, the decoding characteristic matrix is ​​decoded and regressed by a decoder to obtain a decoding value, and the decoding function is used to represent the adjustment voltage value of the output voltage signal, including: using the decoder to decode and regress the decoding characteristic matrix according to the following formula to obtain the decoding value; wherein the formula is: Where X is the decoded feature matrix, Y is the decoded value, and W is the weight matrix, Represents matrix multiplication.

[0058] Compared with the prior art, the X-ray tube and current control method thereof provided by the present application adopt artificial intelligence control technology and use a convolutional neural network model as a feature extractor to perform local correlated implicit feature extraction on the primary current drive signal and the secondary current drive signal of the filament transformer, as well as the feedback sampling voltage signal and the control voltage signal output by the system control unit. In this way, when controlling the output control voltage of the system control unit, attention can be paid to the implicitly correlated flow characteristic information between the data, so as to more accurately control the control voltage, thereby keeping the final drive current stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0060] Figure 1A Schematic diagram of the structure of an X-ray tube.

[0061] Figure 1B The figure is a schematic diagram of a current control system of an existing X-ray tube.

[0062] Figure 1C This is a diagram of an application scenario of an X-ray tube according to an embodiment of the present application.

[0063] Figure 2 is a block diagram of an X-ray tube according to an embodiment of the present application.

[0064] Figure 3 4 is a block diagram of a feature dimension reduction module in an X-ray tube according to an embodiment of the present application.

[0065] Figure 4 Flow chart of a current control method for an X-ray tube according to an embodiment of the present application.

[0066] Figure 5 FIG. 4 is a schematic diagram of the structure of a current control method for an X-ray tube according to an embodiment of the present application. DETAILED DESCRIPTION

[0067] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0068] Scenario Overview

[0069] As mentioned above, an X-ray tube is a vacuum diode operating at high voltage. It is a vacuum electronic device that can generate X-rays by using high-speed electrons to hit a metal target. The X-ray tube contains two electrodes: one is the cathode of the filament for emitting electrons, and the other is the anode of the target for receiving electron bombardment. Both electrodes are sealed in a high vacuum glass or ceramic shell.

[0070] Figure 1A The figure shows the schematic diagram of the structure of the X-ray tube. Figure 1A As shown, the X-ray tube includes a cathode filament 01, an anode 02 and a vacuum tube 03. After power is applied, the cathode filament 01 generates heat to generate thermal electrons. The cathode filament 01 is made of tungsten wire, and the anode 02 is made of high melting point metal. Electrons are accelerated by a high voltage ranging from tens of thousands of volts to hundreds of thousands of volts, and the electron beam bombards the anode 02. X-rays 04 are emitted from the target of the anode 02. Among them, the current that heats the cathode filament 01 is called the filament current, and the electron beam formed by the high voltage acceleration is called the X-ray tube current.

[0071] Figure 1B The figure shows a schematic diagram of a current control system of an existing X-ray tube. Figure 1B As shown, the control voltage signal is output to the pulse width modulator 06 through the system control unit 05. The pulse width modulator 06 performs processing such as limiting and noise elimination on the control voltage signal, and then outputs a stable control voltage signal to the filament inverter 07 to generate the primary current of the filament transformer 08. The secondary current of the filament transformer 08 drives the X-ray tube 09. At the same time, a current transformer 010 is connected in series to indirectly measure the primary current of the filament transformer 08, and feedback sampling is performed and output to the pulse width modulator 06. The pulse width modulator 06 compares the feedback sampling voltage with the control voltage signal output by the system control unit 05, and adjusts the size of the output voltage signal to form a small closed loop of gain control, so that the secondary output of the filament transformer provides a controlled and relatively stable filament current to the filament of the X-ray tube.

[0072] With the continuous development of power electronics technology, X-ray tubes are widely used in medicine and industry. However, during the operation of X-ray tubes, each time a different high voltage or different tube (including tubes of the same model from the same manufacturer) is matched, the system control unit needs to find the corresponding relationship table between the filament current, tube current and tube voltage of different tubes, and the table provided by the tube manufacturer itself has a large error range. Under the same tube voltage, during the initial period of heating the tungsten filament, the filament current of the same tube also corresponds to different tube currents. Since X-ray tubes have the characteristics of general electron tubes, they are easily affected by factors such as temperature and lifespan. The existing technology only performs feedback control on the filament current, which is one of the factors affecting it, when controlling the X-ray tube current, resulting in poor accuracy and stability of the output tube current.

[0073] Therefore, an optimized X-ray tube is desired to keep the final driving current stable by controlling the current gain.

[0074] Accordingly, in the technical solution of the present application, the inventor of the present application considered that if one wants to control the tube current of the X-ray so that the output tube current has better accuracy and stability, that is, the stability of the secondary current is better, the key lies in adjusting the output control voltage of the system control unit.

[0075] After research, the inventor of the present application found that: in the existing current control scheme, the primary current drive signal of the filament transformer is collected by the current transformer, but considering that the current transformer itself will also bring errors and noise to the current signal collection, that is, there is a certain difference between the primary current drive signal collected by the current transformer and the real current drive signal. At the same time, with the use of the X-ray tube, the performance of its internal components will change. For example, with the increase of temperature, the voltage transformation performance of the filament transformer will change. Therefore, in order to accurately measure the change in the performance of the filament transformer, it is also necessary to collect the secondary current drive signal. For example, the secondary current drive signal is also collected by the current transformer. Similarly, the current transformer itself will also bring errors and noise to the current signal collection. In order to adjust the output control voltage of the system control unit, it is also necessary to measure the difference between the sampled voltage signal collected and fed back by the pulse modulator and the control voltage signal output by the system control unit to control the output control voltage.

[0076] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields.

[0077] In recent years, the development of deep learning and neural networks has provided solutions and plans for the current control of X-ray tubes.

[0078] In order to eliminate the deviation introduced by the sensor when collecting signals and the noise existing in the signal itself as much as possible, and at the same time, in order to capture the performance changes of the device as the working process unfolds, in the embodiment of the present application, a first convolutional neural network model with excellent performance in implicit feature extraction is used as a feature extractor to process the waveform of the primary current drive signal and the waveform of the secondary current drive signal respectively, so as to extract the local implicit feature information in the waveform of the primary current drive signal and the waveform of the secondary current drive signal respectively, so as to obtain the primary current feature vector and the secondary current feature vector. In addition, a second convolutional neural network model with excellent performance in implicit feature extraction is used as a feature extractor to process the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal respectively, so as to extract the local implicit feature information in the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal respectively, so as to obtain the feedback sampled voltage feature vector and the control voltage feature vector.

[0079] Since the primary current drive signal and the secondary current drive signal, as well as the feedback sampling voltage signal and the control voltage signal output by the system control unit, all have corresponding correlation relationships. Therefore, in order to extract this specific correlation feature, the current transfer matrix of the secondary current feature vector relative to the primary current feature vector is further calculated respectively, and the current transfer matrix is ​​used to represent the dynamic change characteristics of the real-time voltage transformation performance of the filament transformer. And the voltage transfer matrix of the feedback sampling voltage feature vector relative to the control voltage feature vector is calculated, and the voltage transfer matrix is ​​used to represent the feature representation of the difference between the feedback sampling voltage signal and the control voltage signal in the high-dimensional feature space.

[0080] In this way, the current transfer matrix and the voltage transfer matrix can be combined to obtain a decoding feature representation for decoding the control voltage value. However, for the current transfer matrix, for example, denoted as M1 and the voltage transfer matrix, for example, denoted as M2, since the transfer matrix calculates the vector inner product of the eigenvector in the row direction and arranges the eigenvalue sample dimension of the output vector in the column direction, the feature distribution of the eigenvalues ​​at each position in the row direction needs to pay more attention to the integrity than in the column direction. Therefore, in order to improve the accuracy of semantic expression, the semantic reasoning information of the matrices M1 and M2 is explicitly generalized to be converted into feature vectors, which are expressed as:

[0081]

[0082]

[0083] in and represents the eigenvalues ​​of each position of the current transfer matrix M1 and the voltage transfer matrix M2, V1 and V2 represent the circuit transfer eigenvector and the voltage transfer eigenvector, respectively, and ∑ j (·) means to sum the eigenvalues ​​of each row of the matrix to obtain the column vector, Indicates difference by position.

[0084] Here, the semantic reasoning information is explicitly generalized by explicitly generalizing the semantic concepts corresponding to the eigenvalues ​​into groups of eigenvalues ​​along the row direction to perform information decoupling based on grouping instances. In this way, the obtained eigenvector can represent the information plasticity of the high-dimensional manifold corresponding to the feature along the column direction in the high-dimensional semantic space by improving the accuracy of the semantic expression of the transfer matrix through the information-based reasoning of the feature semantics in the row direction.

[0085] Furthermore, the product of the transpose of the circuit transfer feature vector and the voltage transfer feature vector is calculated to fuse the implicit feature information in the circuit transfer feature vector and the voltage transfer feature vector, thereby obtaining a decoding feature matrix. The decoding feature matrix is ​​then decoded and regressed through a decoder to obtain a decoding value for representing the adjusted voltage value of the output voltage signal.

[0086] Based on this, the present application proposes an X-ray tube, which includes: a current signal acquisition module, used to obtain a primary current drive signal and a secondary current drive signal collected by a current transformer; a voltage signal acquisition module, used to obtain a feedback sampled voltage signal collected by a pulse modulator and a control voltage signal output by a system control unit; a current signal encoding module, used to pass the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal through a first convolutional neural network as a feature extractor to obtain a primary current feature vector and a secondary current feature vector; a first transfer module, used to calculate a current transfer matrix of the secondary current feature vector relative to the primary current feature vector; a voltage signal encoding module, used to convert the waveform diagram of the feedback sampled voltage signal and the waveform diagram of the secondary current drive signal into a primary current feature vector; a first transfer module, used to calculate a current transfer matrix of the secondary current feature vector relative to the primary current feature vector; a second ... The waveform diagram of the control voltage signal is respectively passed through a second convolutional neural network as a feature extractor to obtain a feedback sampling voltage feature vector and a control voltage feature vector; a second transfer module is used to calculate the voltage transfer matrix of the feedback sampling voltage feature vector relative to the control voltage feature vector; a feature dimension reduction module is used to perform feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector; a feature fusion module is used to fuse the transpose of the circuit transfer feature vector and the product of the voltage transfer feature vector to obtain a decoding feature matrix; and a decoding module is used to decode and regress the decoding feature matrix through a decoder to obtain a decoding value, and the decoding function is used to represent the adjusted voltage value of the output voltage signal.

[0087] Figure 1C The figure shows an application scenario diagram of an X-ray tube according to an embodiment of the present application. Figure 1C As shown, in this application scenario, the X-ray tube includes a system control unit 05 , a pulse width modulator 06 , a filament inverter 07 , a filament transformer 08 , and an X-ray tube 09 .

[0088] First, through the current transformer (for example, Figure 1C T1, T2) shown in the figure) collect the filament transformer (for example, Figure 1C 08) and the primary current drive signal and the secondary current drive signal, and through a pulse modulator (for example, Figure 1C 06) collects the sampled voltage signal fed back and the voltage transformer (for example, Figure 1C V1) system control unit (e.g., as shown in FIG. Figure 1C Then, the acquired primary current drive signal and the secondary current drive signal, the feedback sampling voltage signal and the control voltage signal output by the system control unit are input into a server deployed with an X-ray tube algorithm (for example, Figure 1C The cloud server S shown in the figure) is capable of processing the primary current drive signal and the secondary current drive signal as well as the feedback sampling voltage signal and the control voltage signal output by the system control unit using an X-ray tube algorithm to generate a decoded value for representing the adjustment voltage value of the output voltage signal.

[0089] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0090] Exemplary Systems

[0091] Figure 2 FIG. 1 is a block diagram of an X-ray tube according to an embodiment of the present application. Figure 2As shown, according to the embodiment of the present application, the X-ray tube 200 includes: a current signal acquisition module 210, which is used to acquire the primary current drive signal and the secondary current drive signal acquired by the current transformer; a voltage signal acquisition module 220, which is used to acquire the feedback sampled voltage signal acquired by the pulse modulator and the control voltage signal output by the system control unit; a current signal encoding module 230, which is used to pass the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal through the first convolutional neural network as a feature extractor to obtain the primary current feature vector and the secondary current feature vector; a first transfer module 240, which is used to calculate the current transfer matrix of the secondary current feature vector relative to the primary current feature vector; a voltage signal encoding module 250, which is used to convert the waveform diagram of the feedback sampled voltage signal and the waveform diagram of the control voltage signal are respectively passed through a second convolutional neural network as a feature extractor to obtain a feedback sampling voltage feature vector and a control voltage feature vector; a second transfer module 260 is used to calculate the voltage transfer matrix of the feedback sampling voltage feature vector relative to the control voltage feature vector; a feature dimension reduction module 270 is used to perform feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector; a feature fusion module 280 is used to fuse the transpose of the circuit transfer feature vector and the product of the voltage transfer feature vector to obtain a decoding feature matrix; and a decoding module 290 is used to decode and regress the decoding feature matrix through a decoder to obtain a decoding value, and the decoding function is used to represent the adjusted voltage value of the output voltage signal.

[0092] Specifically, in the embodiment of the present application, the current signal acquisition module 210 and the voltage signal acquisition module 220 are used to obtain the primary current drive signal and the secondary current drive signal collected by the current transformer, and to obtain the feedback sampling voltage signal collected by the pulse modulator and the control voltage signal output by the system control unit. As mentioned above, in the technical solution of the present application, considering that if the tube current of the X-ray is to be controlled so that the output tube current has better accuracy and better stability, it can be adjusted according to the output control voltage of the system control unit. Specifically, in the technical solution of the present application, the primary current drive signal and the secondary current drive signal of the filament transformer are collected by the current transformer. It should be understood that in the embodiment of the present application, the primary current drive signal of the filament transformer is collected by the current transformer, but considering that the current transformer itself will also bring errors and noises in the current signal collection, that is, there is a certain difference between the primary current drive signal collected by the current transformer and the real current drive signal. In order to reduce the difference, the secondary current drive signal of the filament transformer is further collected by the same type of current transformer. Furthermore, in order to adjust the output control voltage of the system control unit, it is also necessary to collect the feedback sampling voltage signal and the control voltage signal output by the system control unit through a pulse modulator to assist in controlling the output control voltage.

[0093] Specifically, in the embodiment of the present application, the current signal encoding module 230 and the first transfer module 240 are used to pass the waveform of the primary current drive signal and the waveform of the secondary current drive signal through the first convolutional neural network as a feature extractor to obtain the primary current feature vector and the secondary current feature vector, and calculate the current transfer matrix of the secondary current feature vector relative to the primary current feature vector. That is, in the technical solution of the present application, then, the first convolutional neural network model with excellent performance in implicit feature extraction is used as a feature extractor to process the waveform of the primary current drive signal and the waveform of the secondary current drive signal respectively, so as to extract the local implicit feature information in the waveform of the primary current drive signal and the waveform of the secondary current drive signal respectively, so as to obtain the primary current feature vector and the secondary current feature vector. And, it should be understood that, considering that for the primary current drive signal and the secondary current drive signal, they both have a corresponding association relationship. Therefore, in the technical solution of the present application, in order to extract this specific correlation feature, the current transfer matrix of the secondary current feature vector relative to the primary current feature vector is further calculated. Here, the current transfer matrix can be used to represent the implicit correlation information of the secondary current signal and the primary current signal in the high-dimensional feature space.

[0094] Correspondingly, in a specific example, each layer of the first convolutional neural network performs in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network is the primary current feature vector and the secondary current feature vector, and the input of the first layer of the first convolutional neural network is the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal.

[0095] More specifically, in the embodiment of the present application, the first transfer module is further used to: calculate the current transfer matrix of the secondary current characteristic vector relative to the primary current characteristic vector using the following formula;

[0096] Wherein, the formula is:

[0097] M1=S1*F1

[0098] Wherein F1 represents the secondary current characteristic vector, M1 represents the current transfer matrix, and S1 represents the primary current characteristic vector.

[0099] Specifically, in the embodiment of the present application, the voltage signal encoding module 250 and the second transfer module 260 are used to pass the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal through the second convolutional neural network as a feature extractor to obtain the feedback sampled voltage feature vector and the control voltage feature vector, and calculate the voltage transfer matrix of the feedback sampled voltage feature vector relative to the control voltage feature vector. That is, in the technical solution of the present application, it is also necessary to further use the second convolutional neural network model with excellent performance in implicit feature extraction as a feature extractor to process the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal respectively, so as to extract the local implicit feature information in the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal respectively, so as to obtain the feedback sampled voltage feature vector and the control voltage feature vector. It should be understood that for the feedback sampled voltage signal and the control voltage signal output by the system control unit, they both have a corresponding association relationship. Therefore, in the technical solution of the present application, in order to extract this specific correlation feature, the voltage transfer matrix of the feedback sampled voltage feature vector relative to the control voltage feature vector is further calculated. Here, the voltage transfer matrix is ​​used to represent the implicit correlation information of the feedback sampled voltage signal and the control voltage signal in the high-dimensional feature space.

[0100] Correspondingly, in a specific example, each layer of the second convolutional neural network performs in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local feature matrix on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the second convolutional neural network is the feedback sampled voltage feature vector and the control voltage feature vector, and the input of the first layer of the second convolutional neural network is the waveform graph of the feedback sampled voltage signal and the waveform graph of the control voltage signal.

[0101] More specifically, in the embodiment of the present application, the second transfer module is further used to: calculate the voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector using the following formula;

[0102] Wherein, the formula is:

[0103] M2=S2*F2

[0104] Wherein F2 represents the feedback sampling voltage characteristic vector, M2 represents the voltage transfer matrix, and S2 represents the control voltage characteristic vector.

[0105] Specifically, in an embodiment of the present application, the feature dimension reduction module 270 is used to perform feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector. It should be understood that for the current transfer matrix, for example, denoted as M1 and the voltage transfer matrix, for example, denoted as M2, since the transfer matrix calculates the vector inner product of the feature vector in the row direction and is arranged along the eigenvalue sample dimension of the output vector in the column direction, the feature distribution of the eigenvalues ​​at each position in the row direction needs to pay more attention to integrity than in the column direction. Therefore, in the technical solution of the present application, in order to improve the accuracy of semantic expression, the semantic reasoning information of the matrices M1 and M2 is explicitly generalized to be converted into feature vectors. It should be understood that here, the explicit generalization of the semantic reasoning information is performed by explicitly generalizing the semantic concepts corresponding to the eigenvalues ​​into groups of eigenvalues ​​along the row direction to perform information decoupling based on grouping instances. In this way, the obtained feature vector can represent the information plasticity along the column direction of the high-dimensional manifold corresponding to the feature in the high-dimensional semantic space by informatized reasoning on the feature semantics in the row direction, thereby improving the accuracy of the semantic expression of the transfer matrix and further improving the accuracy of classification.

[0106] More specifically, in an embodiment of the present application, the feature dimension reduction module includes: first, for each row vector in the current transfer matrix, the logarithmic function value of the difference between the sum of the natural exponential function value of the eigenvalue of each position in the row vector and the sum of the eigenvalues ​​of all positions in the row vector is calculated, divided by the logarithmic function value of the difference between the sum of the eigenvalues ​​of all positions in the row vector minus one, to obtain the eigenvalues ​​corresponding to each row vector in the current transfer matrix. Next, the eigenvalues ​​corresponding to each row vector in the current transfer matrix are arranged vertically to obtain the circuit transfer eigenvector. Then, for each row vector in the voltage transfer matrix, the logarithmic function value of the difference between the sum of the natural exponential function value of the eigenvalue of each position in the row vector and the sum of the eigenvalues ​​of all positions in the row vector is calculated, divided by the logarithmic function value of the difference between the sum of the eigenvalues ​​of all positions in the row vector minus one, to obtain the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix. Finally, the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix are arranged vertically to obtain the circuit transfer eigenvector. That is, in a specific example, the formula for converting the current transfer matrix and the voltage transfer matrix into feature vectors by performing explicit generalization of semantic reasoning information is:

[0107]

[0108]

[0109] in and represents the eigenvalues ​​of each position of the current transfer matrix M1 and the voltage transfer matrix M2, V1 and V2 represent the circuit transfer eigenvector and the voltage transfer eigenvector, respectively, and ∑ j (·) means to sum the eigenvalues ​​of each row of the matrix to obtain the column vector, Indicates difference by position.

[0110] Figure 3 FIG. 1 is a block diagram of a feature dimension reduction module in an X-ray tube according to an embodiment of the present application. Figure 3As shown, the feature dimension reduction module 270 includes: a first row vector characterization unit 271, which is used to calculate, for each row vector in the current transfer matrix, a logarithmic function value of the difference between the sum of the natural exponential function values ​​of the eigenvalues ​​at each position in the row vector raised to the power and the sum of the eigenvalues ​​at all positions in the row vector divided by the logarithmic function value of the difference between the sum of the eigenvalues ​​at all positions in the row vector minus one to obtain the eigenvalues ​​corresponding to each row vector in the current transfer matrix; a first vector construction unit 272, which is used to vertically arrange the eigenvalues ​​corresponding to each row vector in the current transfer matrix to obtain the eigenvalues ​​corresponding to each row vector in the current transfer matrix. circuit transfer eigenvector; a second row vector characterization unit 273, used to calculate, for each row vector in the voltage transfer matrix, the logarithmic function value of the difference between the sum of the natural exponential function values ​​raised to the power of the eigenvalues ​​at each position in the row vector and the sum of the eigenvalues ​​at all positions in the row vector, divided by the logarithmic function value of the difference between the sum of the eigenvalues ​​at all positions in the row vector minus one to obtain the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix; and a second vector construction unit 274, used to vertically arrange the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix to obtain the circuit transfer eigenvector.

[0111] Specifically, in an embodiment of the present application, the feature fusion module 280 and the decoding module 290 are used to fuse the product of the transpose of the circuit transfer feature vector and the voltage transfer feature vector to obtain a decoding feature matrix, and the decoding feature matrix is ​​decoded and regressed through a decoder to obtain a decoding value, and the decoding function is used to represent the adjusted voltage value of the output voltage signal. That is, further, the product of the transpose of the circuit transfer feature vector and the voltage transfer feature vector is calculated to fuse the implicit feature information in the circuit transfer feature vector and the voltage transfer feature vector, thereby obtaining a decoding feature matrix. Then, the decoding feature matrix is ​​decoded and regressed through a decoder to obtain a decoding value for representing the adjusted voltage value of the output voltage signal. In a specific example, the decoder is used to decode and regress the decoding feature matrix using the following formula to obtain the decoding value; wherein the formula is: Where X is the decoded feature matrix, Y is the decoded value, and W is the weight matrix, Represents matrix multiplication.

[0112] More specifically, in the embodiment of the present application, the feature fusion module is further used to: calculate the product between the transpose of the circuit transfer feature vector and the voltage transfer feature vector using the following formula to obtain the decoding feature matrix;

[0113] Wherein, the formula is:

[0114]

[0115] Among them, V1 represents the circuit transfer feature vector, V2 represents the voltage transfer feature vector, and M represents the decoding feature matrix.

[0116] In summary, the X-ray tube 200 based on the embodiment of the present application is explained, which adopts artificial intelligence control technology and uses a convolutional neural network model as a feature extractor to perform local correlated implicit feature extraction on the primary current drive signal and the secondary current drive signal of the filament transformer, as well as the feedback sampling voltage signal and the control voltage signal output by the system control unit. In this way, when controlling the output control voltage of the system control unit, it is possible to pay attention to the implicit correlation flow characteristic information between the data, so as to more accurately control the control voltage, thereby keeping the final drive current stable.

[0117] As described above, the X-ray tube 200 according to the embodiment of the present application can be implemented in various terminal devices, such as a server of an X-ray tube algorithm. In one example, the X-ray tube 200 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the X-ray tube 200 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the X-ray tube 200 can also be one of the many hardware modules of the terminal device.

[0118] Alternatively, in another example, the X-ray tube 200 and the terminal device may also be separate devices, and the X-ray tube 200 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0119] Exemplary Methods

[0120] Figure 4 The figure shows a flow chart of the current control method of the X-ray tube. Figure 4As shown, according to the current control method of the X-ray tube of the embodiment of the present application, the method comprises the following steps: S110, acquiring a primary current drive signal and a secondary current drive signal acquired by a current transformer; S120, acquiring a feedback sampled voltage signal acquired by a pulse modulator and a control voltage signal output by a system control unit; S130, passing the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal through a first convolutional neural network as a feature extractor to obtain a primary current feature vector and a secondary current feature vector; S140, calculating a current transfer matrix of the secondary current feature vector relative to the primary current feature vector; S150, passing the waveform diagram of the feedback sampled voltage signal and the control voltage signal output by a system control unit; The waveform diagram of the voltage signal is respectively passed through a second convolutional neural network as a feature extractor to obtain a feedback sampling voltage feature vector and a control voltage feature vector; S160, calculating the voltage transfer matrix of the feedback sampling voltage feature vector relative to the control voltage feature vector; S170, performing feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector; S180, fusing the transpose of the circuit transfer feature vector and the product of the voltage transfer feature vector to obtain a decoding feature matrix; and, S190, decoding and regressing the decoding feature matrix through a decoder to obtain a decoding value, and the decoding function is used to represent the adjusted voltage value of the output voltage signal.

[0121] Figure 5 FIG. 1 is a schematic diagram of the architecture of a current control method for an X-ray tube according to an embodiment of the present application. Figure 5 As shown, in the network architecture of the current control method of the X-ray tube, first, the waveform diagram of the primary current driving signal (for example, as shown in FIG. Figure 5 P1 shown in FIG. 1 ) and the waveform diagram of the secondary current drive signal (eg, as shown in FIG. Figure 5 P2) are respectively passed through a first convolutional neural network as a feature extractor (e.g., Figure 5 CNN1 as shown in FIG. 1 ) to obtain the primary current feature vector (e.g., Figure 5 VF1) and the secondary current characteristic vector (e.g., as shown in Figure 5 VF2 as shown in FIG); then, calculating the current transfer matrix of the secondary current eigenvector relative to the primary current eigenvector (for example, as Figure 5 Then, the waveform diagram of the feedback sampling voltage signal obtained (for example, as shown in FIG. Figure 5 Q1 shown in FIG. 1 ) and the waveform diagram of the control voltage signal (eg, as shown in FIG. Figure 5 Q2) are respectively passed through a second convolutional neural network as a feature extractor (e.g., Figure 5CNN2 shown in FIG. 2) to obtain the feedback sample voltage feature vector (for example, Figure 5 VF3) and the control voltage characteristic vector (e.g., as shown in Figure 5 Then, the voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector is calculated (for example, as Figure 5 MF2 shown in FIG); then, the current transfer matrix and the voltage transfer matrix are subjected to feature dimension reduction to obtain a circuit transfer feature vector (for example, Figure 5 V1 as shown in FIG) and the voltage transfer characteristic vector (e.g., Figure 5 V2 as shown in FIG); then, the product of the transpose of the circuit transfer feature vector and the voltage transfer feature vector is fused to obtain a decoding feature matrix (for example, Figure 5 MF as shown in FIG); and, finally, the decoded feature matrix is ​​passed through a decoder (e.g., Figure 5 D) shown in the figure performs decoding regression to obtain a decoding value, and the decoding value is used to represent the adjusted voltage value of the output voltage signal.

[0122] More specifically, in step S110 and step S120, the primary current drive signal and the secondary current drive signal collected by the current transformer are obtained, and the feedback sampling voltage signal collected by the pulse modulator and the control voltage signal output by the system control unit are obtained. It should be understood that in the technical solution of the present application, considering that if the tube current of the X-ray is to be controlled so that the output tube current has better accuracy and stability, it can be adjusted according to the output control voltage of the system control unit. Specifically, in the technical solution of the present application, the primary current drive signal and the secondary current drive signal of the filament transformer are collected by the current transformer. It should be understood that in the embodiment of the present application, the primary current drive signal of the filament transformer is collected by the current transformer, but considering that the current transformer itself will also bring errors and noises in the current signal collection, that is, there is a certain difference between the primary current drive signal collected by the current transformer and the real current drive signal. In order to reduce the difference, the secondary current drive signal of the filament transformer is further collected by the same type of current transformer. Furthermore, in order to adjust the output control voltage of the system control unit, it is also necessary to collect the feedback sampling voltage signal and the control voltage signal output by the system control unit through a pulse modulator to assist in controlling the output control voltage.

[0123] More specifically, in step S130 and step S140, the waveform of the primary current drive signal and the waveform of the secondary current drive signal are respectively passed through the first convolutional neural network as a feature extractor to obtain the primary current feature vector and the secondary current feature vector, and the current transfer matrix of the secondary current feature vector relative to the primary current feature vector is calculated. That is, in the technical solution of the present application, then, the first convolutional neural network model with excellent performance in implicit feature extraction is used as a feature extractor to process the waveform of the primary current drive signal and the waveform of the secondary current drive signal respectively, so as to extract the local implicit feature information in the waveform of the primary current drive signal and the waveform of the secondary current drive signal respectively, so as to obtain the primary current feature vector and the secondary current feature vector. And, it should be understood that, considering that for the primary current drive signal and the secondary current drive signal, they both have a corresponding correlation relationship. Therefore, in the technical solution of the present application, in order to extract this specific correlation feature, the current transfer matrix of the secondary current feature vector relative to the primary current feature vector is further calculated respectively. Here, the current transfer matrix can be used to represent the implicit correlation information between the secondary current signal and the primary current signal in a high-dimensional feature space.

[0124] More specifically, in step S150 and step S160, the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal are respectively passed through the second convolutional neural network as a feature extractor to obtain the feedback sampled voltage feature vector and the control voltage feature vector, and the voltage transfer matrix of the feedback sampled voltage feature vector relative to the control voltage feature vector is calculated. That is, in the technical solution of the present application, it is also necessary to further use the second convolutional neural network model with excellent performance in implicit feature extraction as a feature extractor to process the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal respectively, so as to extract the local implicit feature information in the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal respectively, so as to obtain the feedback sampled voltage feature vector and the control voltage feature vector. It should be understood that for the feedback sampled voltage signal and the control voltage signal output by the system control unit, they both have a corresponding correlation relationship. Therefore, in the technical solution of the present application, in order to extract this specific correlation feature, the voltage transfer matrix of the feedback sampled voltage feature vector relative to the control voltage feature vector is further calculated respectively. Here, the voltage transfer matrix is ​​used to represent the implicit correlation information between the feedback sampling voltage signal and the control voltage signal in the high-dimensional feature space.

[0125] More specifically, in step S170, the current transfer matrix and the voltage transfer matrix are subjected to feature dimensionality reduction to obtain a circuit transfer feature vector and a voltage transfer feature vector. It should be understood that for the current transfer matrix, for example, denoted as M1 and the voltage transfer matrix, for example, denoted as M2, since the transfer matrix calculates the vector inner product of the feature vector in the row direction and is arranged along the eigenvalue sample dimension of the output vector in the column direction, the feature distribution of the eigenvalues ​​at each position in the row direction needs to pay more attention to integrity than in the column direction. Therefore, in the technical solution of the present application, in order to improve the accuracy of semantic expression, the semantic reasoning information of the matrices M1 and M2 is explicitly generalized to be converted into feature vectors. It should be understood that here, the explicit generalization of the semantic reasoning information is performed by explicitly generalizing the semantic concepts corresponding to the eigenvalues ​​into groups of eigenvalues ​​along the row direction to perform information decoupling based on grouping instances. In this way, the obtained feature vector can represent the information plasticity along the column direction of the high-dimensional manifold corresponding to the feature in the high-dimensional semantic space by informatized reasoning on the feature semantics in the row direction, thereby improving the accuracy of the semantic expression of the transfer matrix and further improving the accuracy of classification.

[0126] More specifically, in step S180 and step S190, the product of the transpose of the circuit transfer feature vector and the voltage transfer feature vector is fused to obtain a decoding feature matrix, and the decoding feature matrix is ​​decoded and regressed through a decoder to obtain a decoding value, and the decoding function is used to represent the adjusted voltage value of the output voltage signal. That is, further, the product of the transpose of the circuit transfer feature vector and the voltage transfer feature vector is calculated to fuse the implicit feature information in the circuit transfer feature vector and the voltage transfer feature vector, thereby obtaining a decoding feature matrix. Then, the decoding feature matrix is ​​decoded and regressed through a decoder to obtain a decoding value for representing the adjusted voltage value of the output voltage signal. In a specific example, the decoder is used to decode and regress the decoding feature matrix using the following formula to obtain the decoding value; wherein the formula is: Where X is the decoded feature matrix, Y is the decoded value, and W is the weight matrix, Represents matrix multiplication.

[0127] In summary, the current control method of the X-ray tube based on the embodiment of the present application is explained, which adopts artificial intelligence control technology and uses a convolutional neural network model as a feature extractor to perform local correlated implicit feature extraction on the primary current drive signal and the secondary current drive signal of the filament transformer, as well as the feedback sampling voltage signal and the control voltage signal output by the system control unit. In this way, when controlling the output control voltage of the system control unit, it is possible to pay attention to the implicit correlation flow characteristic information between the data, so as to more accurately control the control voltage, thereby keeping the final drive current stable.

[0128] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0129] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0130] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0131] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0132] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. An X-ray tube, characterized in that: include: A current signal acquisition module, used to acquire a primary current drive signal and a secondary current drive signal acquired by a current transformer; A voltage signal acquisition module is used to acquire the feedback sampling voltage signal acquired by the pulse modulator and the control voltage signal output by the system control unit; A current signal encoding module, used for respectively passing the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal through a first convolutional neural network as a feature extractor to obtain a primary current feature vector and a secondary current feature vector; A first transfer module, used for calculating a current transfer matrix of the secondary current eigenvector relative to the primary current eigenvector; A voltage signal encoding module, used for respectively passing the waveform of the feedback sampled voltage signal and the waveform of the control voltage signal through a second convolutional neural network as a feature extractor to obtain a feedback sampled voltage feature vector and a control voltage feature vector; A second transfer module, used for calculating a voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector; A feature dimension reduction module, used for performing feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector; A feature fusion module, used for calculating the product between the transpose of the circuit transfer feature vector and the voltage transfer feature vector to obtain a decoding feature matrix; as well as The decoding module is used to decode and regress the decoding feature matrix through a decoder to obtain a decoding value, and the decoding value is used to represent the adjustment voltage value of the output voltage signal.

2. The X-ray tube according to claim 1, characterized in that The current signal encoding module is further used for: each layer of the first convolutional neural network performs the following in the forward transfer of the layer: Perform convolution on the input data to obtain a convolution feature map; Performing mean pooling on the convolution feature map based on a local feature matrix to obtain a pooled feature map; as well as Performing nonlinear activation on the pooled feature map to obtain an activated feature map; Among them, the output of the last layer of the first convolutional neural network is the primary current feature vector and the secondary current feature vector, and the input of the first layer of the first convolutional neural network is the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal.

3. The X-ray tube according to claim 2, characterized in that The first transfer module is further used to: calculate the current transfer matrix of the secondary current characteristic vector relative to the primary current characteristic vector using the following formula; Wherein, the formula is: M1=S1*F1 Wherein F1 represents the secondary current characteristic vector, M1 represents the current transfer matrix, and S1 represents the primary current characteristic vector.

4. The X-ray tube according to claim 3, characterized in that The voltage signal encoding module is further used for: each layer of the second convolutional neural network performs the following in the forward transfer of the layer: Perform convolution on the input data to obtain a convolution feature map; Performing mean pooling on the convolution feature map based on a local feature matrix to obtain a pooled feature map; as well as Performing nonlinear activation on the pooled feature map to obtain an activated feature map; Among them, the output of the last layer of the second convolutional neural network is the feedback sampling voltage feature vector and the control voltage feature vector, and the input of the first layer of the second convolutional neural network is the waveform diagram of the feedback sampling voltage signal and the waveform diagram of the control voltage signal.

5. The X-ray tube according to claim 4, characterized in that The second transfer module is further used to: calculate the voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector using the following formula; Wherein, the formula is: M2=S2*F2 Wherein F2 represents the feedback sampling voltage characteristic vector, M2 represents the voltage transfer matrix, and S2 represents the control voltage characteristic vector.

6. The X-ray tube according to claim 5, characterized in that The feature dimension reduction module comprises: A first row vector characterization unit is used to calculate, for each row vector in the current transfer matrix, a logarithmic function value of a difference between a sum of natural exponential function values ​​raised to a power of the eigenvalues ​​at each position in the row vector and a sum of the eigenvalues ​​at all positions in the row vector divided by a logarithmic function value of a difference obtained by subtracting one from the sum of the eigenvalues ​​at all positions in the row vector to obtain an eigenvalue corresponding to each row vector in the current transfer matrix; A first vector construction unit, configured to vertically arrange the eigenvalues ​​corresponding to each row vector in the current transfer matrix to obtain the circuit transfer eigenvector; A second row vector characterization unit is used to calculate, for each row vector in the voltage transfer matrix, a logarithmic function value of a difference between a sum of natural exponential function values ​​raised to a power of the eigenvalues ​​at each position in the row vector and a sum of the eigenvalues ​​at all positions in the row vector divided by a logarithmic function value of a difference between the sum of the eigenvalues ​​at all positions in the row vector and one, so as to obtain an eigenvalue corresponding to each row vector in the voltage transfer matrix; and The second vector construction unit is used to vertically arrange the eigenvalues ​​corresponding to each row vector in the voltage transfer matrix to obtain the circuit transfer eigenvector.

7. The X-ray tube according to claim 6, characterized in that The feature fusion module is further used to: calculate the product of the transpose of the circuit transfer feature vector and the voltage transfer feature vector using the following formula to obtain the decoding feature matrix; Wherein, the formula is: Among them, V1 represents the circuit transfer feature vector, V2 represents the voltage transfer feature vector, and M represents the decoding feature matrix.

8. The X-ray tube according to claim 7, characterized in that The decoding module is further used to: use the decoder to perform decoding regression on the decoding feature matrix to obtain the decoding value according to the following formula; wherein the formula is: Where X is the decoded feature matrix, Y is the decoded value, and W is the weight matrix, Represents matrix multiplication.

9. A method for controlling current of an X-ray tube, characterized in that: include: Acquire a primary current drive signal and a secondary current drive signal collected by a current transformer; Acquire a feedback sampling voltage signal collected by a pulse modulator and a control voltage signal output by a system control unit; Passing the waveform diagram of the primary current drive signal and the waveform diagram of the secondary current drive signal through a first convolutional neural network as a feature extractor to obtain a primary current feature vector and a secondary current feature vector respectively; Calculating a current transfer matrix of the secondary current eigenvector relative to the primary current eigenvector; Passing the waveform of the feedback sampling voltage signal and the waveform of the control voltage signal through a second convolutional neural network as a feature extractor to obtain a feedback sampling voltage feature vector and a control voltage feature vector; Calculating a voltage transfer matrix of the feedback sampling voltage characteristic vector relative to the control voltage characteristic vector; Performing feature dimension reduction on the current transfer matrix and the voltage transfer matrix to obtain a circuit transfer feature vector and a voltage transfer feature vector; fusing the transpose of the circuit transfer feature vector and the product of the voltage transfer feature vector to obtain a decoding feature matrix; as well as The decoding characteristic matrix is ​​decoded and regressed through a decoder to obtain a decoding value, and the decoding value is used to represent the adjustment voltage value of the output voltage signal.

10. The current control method of an X-ray tube according to claim 9, characterized in that: Calculating a current transfer matrix of the secondary current eigenvector relative to the primary current eigenvector includes: The current transfer matrix of the secondary current characteristic vector relative to the primary current characteristic vector is calculated by the following formula; Wherein, the formula is: M1=S1*F1 Wherein F1 represents the secondary current characteristic vector, M1 represents the current transfer matrix, and S1 represents the primary current characteristic vector.

Images