Method for predicting shielding effectiveness of metal box body based on LSTM (Long Short Term Memory)

Abstract

The invention provides a metal box shielding effectiveness prediction method based on LSTM (Long Short Term Memory). The method comprises the following steps: building a memory unit structure of the LSTM; building a metal shielding box example model; constructing a shielding cavity equivalent circuit based on a Robinson algorithm, and solving the shielding effectiveness; and predicting the shielding effectiveness of the metal cavity by using the LSTM. According to the method, the prediction technology based on the LSTM network is innovatively introduced, breakthrough of the shielding effectiveness of the metal box in the field of low-frequency to high-frequency data prediction is achieved, the prediction accuracy and efficiency are greatly improved, and remarkable optimization and improvement are brought to technical progress and application practice of related fields.

Description

Technical Field

[0001] The present invention relates to the technical field of computational electromagnetics, and in particular to a method for predicting the shielding effectiveness of a metal box based on LSTM. Background Art

[0002] In the field of computational electromagnetics, it is very difficult to obtain accurate high-frequency data, whether it is experimental or finite element simulation. In high-frequency simulation, each simulation will cost a huge simulation cost. Therefore, how to obtain high-reliability computational electromagnetic high-frequency prediction data through low-cost means is a key scientific problem that needs to be solved urgently. Some scholars have proposed the Cauchy method, in which the frequency response is modeled as the ratio of two polynomials. This method is applied to the prediction of smooth frequency response, but it cannot predict the uncertainty of the response, and the accuracy of this method depends on the number of sampling points in the narrow band. If the sampling technology misses the pole or zero frequency point, the prediction may be wrong. Later, another scholar proposed another prediction method, which can predict the broadband response by fitting early and low-frequency data as a finite sum of orthogonal polynomials. This type of method has low computational efficiency and has not been promoted.

[0003] The feature selection validation (FSV) method is a standard for electromagnetic simulation validation, which verifies electromagnetic models by quantifying the consistency between reference and numerical results. FSV is a technique that combines amplitude-based (or trend-based) comparisons with feature-based (fast-changing features or peaks and troughs) comparisons to give an overall indication of the consistency between two sets of data. FSV can be used to evaluate the ADM, FDM, and GDM of the prediction results. ADM is an evaluation of the trend of the data, FDM evaluates the periodic characteristics, and GDM is a global evaluation. The evaluation indicators are shown in Table 1.

[0004] Table 1 FSV evaluation index

[0005]

[0006] Summary of the invention

[0007] According to the technical problems mentioned in the above background technology, a method for predicting the shielding effectiveness of a metal box based on LSTM is provided.

[0008] The technical means adopted by the present invention are as follows:

[0009] A method for predicting the shielding effectiveness of a metal box based on LSTM, comprising the following steps:

[0010] Step 1: Build the memory unit structure of LSTM;

[0011] Step 2: Build a metal shielding box calculation model;

[0012] Step 3: Construct the shielding cavity equivalent circuit based on the Robinson algorithm and calculate the shielding effectiveness;

[0013] Step 4: Use LSTM to predict the shielding effectiveness of the metal cavity.

[0014] Furthermore, in step 1, the memory unit structure of the LSTM includes: an input gate i that determines the amount of information input into the memory unit, an output gate o that determines the output amount of the LSTM, and a forget gate f that determines the forgetting or passing of historical information;

[0015] The input gate i randomly generates a value between 0 and 1 according to the current input and the previous hidden layer state, that is, it controls how much information flows into the memory unit;

[0016] The forget gate f randomly generates a value between 0 and 1 according to the importance of the input of the input gate i for a given future sample, that is, how much information the memory unit forgets;

[0017] The output gate o determines how much information the memory cell outputs by the following formula:

[0018] o t =σ(w xo x t +w ho h t-1 +b o );

[0019] Among them, t represents the output information at time t; w xo represents the input weight; x t represents the input information at time t; w ho represents the state weight of the memory unit; h t-1 represents the output information of the hidden layer at time t-1; b o Represents the bias term.

[0020] Furthermore, in step 1, assuming that at time t, the input is x t , the hidden layer output at the current moment is h t , the memory cell state is c t ;

[0021] The sigmoid function is used to control the values ​​of the input gate i, output gate o, and forget gate f within the range of [0,1]:

[0022]

[0023] Wherein, σ(x) represents the sigmoid function with respect to x; x represents the input of the sigmoid function;

[0024] At each time t, update the unit gate g t , g t The value of the input x t and the previous hidden layer state h t-1 Related; the update method is:

[0025] g t =tanh(w xg x t +w hg h t-1 +b g );

[0026] Among them, w xg represents the input weight; x t represents the input information at time t; w hg represents the state weight of the memory unit; h t-1 represents the output information of the hidden layer at time t-1; b g represents the bias term;

[0027] Memory cell state c t Information is updated through f t Control how much information and i are forgotten at the last moment t How much information is input at this moment to achieve common control; the state of the memory unit is:

[0028] c t =f t ×c t-1 +i t ×g t ;

[0029] Among them, f t represents the output of the forget gate at time t; c t-1 represents the state of the memory unit at time t-1; i t represents the output of the input gate at time t; g t represents the state of the candidate memory unit at time t;

[0030] The entire hidden layer state h at the previous moment t Through the output gate o t Control memory unit c t How much information is output? The entire hidden layer state h t It is expressed as:

[0031] h t =o t ×tanh(c t );

[0032] Then the output y of the LSTM network is t for:

[0033] y t =σ(w hy h t +b y );

[0034] Among them, w hy represents the output weight; r t represents the output information of the hidden layer at time t; b y Represents the bias term.

[0035] Furthermore, the information flow into the input gate i is:

[0036] i t =σ(w xi x t +w hi h t-1 +b i );

[0037] Among them, w xi represents the input weight; x t represents the input information at time t; w hi Represents the memory unit weight; h t-1 represents the output information of the hidden layer at time t-1; b i Represents the bias term.

[0038] Furthermore, the amount of forgotten information of the forget gate f is:

[0039] f t =σ(w xf x t +w hf h t-1 +b f );

[0040] Among them, w xf represents the input weight; x t represents the input information at time t; w hf Represents the memory unit weight; h t-1 represents the output information of the hidden layer at time t-1; b f Represents the bias term.

[0041] Furthermore, in step 2, the metal shielding box is a rectangular parallelepiped; the inner dimensions of the cavity of the metal shielding box are a×b×d, and the thickness is t; a rectangular hole with dimensions of l×w is provided at the center of the front panel of the metal shielding box;

[0042] There is an excitation plane wave radiation outside the metal shielding box, the excitation plane wave radiation is incident on the rectangular hole cavity surface perpendicular to the panel, and is polarized along the cavity height direction;

[0043] The shielding effectiveness test point P is set along the center line of the panel, with a distance p from the opening.

[0044] Furthermore, in step 3, the incident plane wave is equivalent to a voltage source with a voltage of V0 and an impedance of Z0≈377, and the shielding cavity is equivalent to a rectangular waveguide with a short-circuited terminal. The characteristic impedance and propagation constant of the shielding cavity are Z g and k g , k0 = 2π / λ is the free space wave number, and the rectangular opening is equivalent to a coplanar strip transmission line with short circuits at both ends. The characteristic impedance of the rectangular opening is:

[0045]

[0046] Among them, K and K ′ is the elliptic integral, w e is the effective width of the hole, and the expression is:

[0047]

[0048] Characteristic impedance of aperture Z 0s The approximate expression of is:

[0049]

[0050] Using the transmission line theory, the input impedance at the midpoint of the aperture equivalent transmission line, that is, the aperture impedance, can be obtained as:

[0051]

[0052] Wherein, l / a represents the aperture impedance coefficient introduced to consider the coupling effect between the aperture and the cavity;

[0053] According to the Thevenin theorem, the equivalent voltage at the hole is V1 = V0Z ap / (Z0+Z ap ) and the equivalent source impedance Z1=Z0Z ap / (Z0+Z ap );For TE 10 Propagation mode, waveguide characteristic impedance is The propagation constant is According to the transmission line theory, the equivalent voltage, source impedance and load impedance at the monitoring point X≤0.1 can be obtained:

[0054]

[0055] Z3=jZg tan k g (dp);

[0056] According to Thevenin's theorem, the voltage at point P is V box =V2Z3 / (Z2+Z3); and when there is no shielding cavity, the voltage at point P is V no =V0 / 2; given the frequency f of the excitation plane wave, the electric field shielding effectiveness at point P is calculated as:

[0057]

[0058] Compared with the prior art, the present invention has the following advantages:

[0059] 1. The present invention innovatively introduces prediction technology based on long short-term memory network (LSTM), achieving a breakthrough in the prediction of metal box shielding effectiveness from low-frequency to high-frequency data, greatly improving the accuracy and efficiency of prediction, and bringing significant optimization and improvement to the technological progress and application practice in related fields.

[0060] 2. The LSTM prediction method of the present invention shows excellent prediction performance when processing frequency response sequence data with periodicity and trend, and can produce ideal prediction results.

[0061] In general, the technical solution of the present invention not only overcomes the limitations of the existing technology in terms of prediction response uncertainty and low computational efficiency, but also effectively solves the problem of high-frequency data acquisition in the field of computational electromagnetics, providing strong support for technological progress and applications in related fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0063] Figure 1 Schematic diagram of the memory unit structure of the LSTM of the present invention.

[0064] Figure 2 It is a schematic diagram of a calculation example of a metal shielding box of the present invention.

[0065] Figure 3 Schematic diagram of an equivalent circuit of the shielding cavity of the present invention.

[0066] Figure 4 This is the prediction result of the first solution in the embodiment.

[0067] Figure 5 This is the prediction result of the second solution in the embodiment. DETAILED DESCRIPTION

[0068] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0069] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0070] like Figure 1-3 As shown, the present invention provides a method for predicting the shielding effectiveness of a metal box based on LSTM, comprising the following steps:

[0071] Step 1: constructing a memory cell structure of LSTM; in step 1, the memory cell structure of LSTM includes: an input gate i that determines the amount of information input into the memory cell, an output gate o that determines the output amount of LSTM, and a forget gate f that determines whether historical information is forgotten or passed;

[0072] The input gate i randomly generates a value between 0 and 1 according to the current input and the previous hidden layer state, that is, it controls how much information flows into the memory unit;

[0073] The forget gate f randomly generates a value between 0 and 1 according to the importance of the input of the input gate i for a given future sample, that is, how much information the memory unit forgets;

[0074] The output gate o determines how much information the memory cell outputs by the following formula:

[0075] o t =σ(w xo x t +w hoh t-1 +b o );

[0076] Among them, t represents the output information at time t; w xo represents the input weight; x t represents the input information at time t; w ho represents the state weight of the memory unit; h t-1 represents the output information of the hidden layer at time t-1; b o Represents the bias term.

[0077] As a preferred embodiment, in the present application, in step 1, assuming that at time t, the input is x t , the hidden layer output at the current moment is h t , the memory cell state is c t ;

[0078] The sigmoid function is used to control the values ​​of the input gate i, output gate o, and forget gate f within the range of [0,1]:

[0079]

[0080] Wherein, σ(x) represents the sigmoid function with respect to x; x represents the input of the sigmoid function;

[0081] At each time t, update the unit gate g t , g t The value of the input x t and the previous hidden layer state h t-1 Related; the update method is:

[0082] g t =tanh(w xg x t +w hg h t-1 +b g );

[0083] Among them, w xg represents the input weight; x t represents the input information at time t; w hg represents the state weight of the memory unit; h t-1 represents the output information of the hidden layer at time t-1; b g represents the bias term;

[0084] Memory cell state c t Information is updated through f t Control how much information and i are forgotten at the last moment tHow much information is input at this moment to achieve common control; the state of the memory unit is:

[0085] c t =f t ×c t-1 +i t ×g t ;

[0086] Among them, f t represents the output of the forget gate at time t; c t-1 represents the state of the memory unit at time t-1; i t represents the output of the input gate at time t; g t represents the state of the candidate memory unit at time t;

[0087] The entire hidden layer state h at the previous moment t Through the output gate o t Control memory unit c t How much information is output? The entire hidden layer state h t It is expressed as:

[0088] h t =o t ×tanh(c t );

[0089] Then the output y of the LSTM network is t for:

[0090] y t =σ(w hy h t +b y );

[0091] Among them, w hy represents the output weight; h t represents the output information of the hidden layer at time t; b y represents the bias term. The information flow into the input gate i is:

[0092] i t =σ(w xi x t +w hi h t-1 +b i );

[0093] Among them, w xi represents the input weight; x t represents the input information at time t; w hi Represents the memory unit weight; h t-1 represents the output information of the hidden layer at time t-1; b i Represents the bias term.

[0094] Furthermore, the amount of forgotten information of the forget gate f is:

[0095] f t =σ(w xf x t +w hf h t-1 +b f );

[0096] Among them, w xf represents the input weight; x t represents the input information at time t; w hf Represents the memory unit weight; h t-1 represents the output information of the hidden layer at time t-1; b f Represents the bias term.

[0097] The key to achieving high-frequency prediction in computational electromagnetics is the forgetting function of LSTM. During the training process, if the data at a certain frequency point is unnecessary for generating a high-frequency response, it will be deleted and forgotten. The previous hidden layer states are retained only when they are useful for generating high-frequency responses. Therefore, information from the earlier part of the training space can enter the later part of the response space. As we provide more and more information to the network, the information added to or deleted from the memory cell state is also gradually increasing, thereby achieving effective high-frequency prediction in computational electromagnetics. It is precisely because of this property of LSTM that it can predict frequency response sequence data with good periodic characteristics and trends.

[0098] Step 2: Build a calculation model of a metal shielding box; in step 2, the metal shielding box is a rectangular parallelepiped; the inner dimensions of the cavity of the metal shielding box are a×b×d, and the thickness is t; a rectangular hole with a size of l×w is provided in the center of the front panel of the metal shielding box;

[0099] There is an excitation plane wave radiation outside the metal shielding box, the excitation plane wave radiation is incident on the rectangular hole cavity surface perpendicular to the panel, and is polarized along the cavity height direction;

[0100] The shielding effectiveness test point P is set along the center line of the panel, and the distance from the opening is p. The specific values ​​of some parameters are a = 300mm, b = 100mm, d = 300mm, t = 1mm, p = 150mm. The rectangular hole in the center of the panel has a length of l = 20mm and a width of w = 10mm. The selected metal is aluminum, and its conductivity is σ = 3.8×10 7 S / m, relative dielectric constant is ε r =1, and the rest of the solution space is also approximately treated as a vacuum.

[0101] Step 3: Construct a shielding cavity equivalent circuit based on the Robinson algorithm and calculate the shielding effectiveness; in step 3, the incident plane wave is equivalent to a voltage source with a voltage of V0 and an impedance of Z0≈377Ω, and the shielding cavity is equivalent to a rectangular waveguide with a short-circuited terminal. The characteristic impedance and propagation constant of the shielding cavity are Z g and k g , k0 = 2π / λ is the free space wave number, and the rectangular opening is equivalent to a coplanar strip transmission line with short circuits at both ends. The characteristic impedance of the rectangular opening is:

[0102]

[0103] Among them, K and K ′ is the elliptic integral, w e is the effective width of the hole, and the expression is:

[0104]

[0105] Characteristic impedance of aperture Z 0s The approximate expression of is:

[0106]

[0107] Using the transmission line theory, the input impedance at the midpoint of the aperture equivalent transmission line, that is, the aperture impedance, can be obtained as:

[0108]

[0109] Wherein, l / a represents the aperture impedance coefficient introduced to consider the coupling effect between the aperture and the cavity;

[0110] According to the Thevenin theorem, the equivalent voltage at the hole is V1 = V0Z ap / (Z0+Z ap ) and the equivalent source impedance Z1=Z0Z ap / (Z0+Z ap );For TE 10 Propagation mode, waveguide characteristic impedance is The propagation constant is According to the transmission line theory, the equivalent voltage, source impedance and load impedance at the monitoring point X≤0.1 can be obtained:

[0111]

[0112] Z3=jZ g tank g (dp);

[0113] According to Thevenin's theorem, the voltage at point P is V box=V2Z3 / (Z2+Z3); and when there is no shielding cavity, the voltage at point P is V no =V0 / 2; given the frequency f of the excitation plane wave, the electric field shielding effectiveness at point P is calculated as:

[0114]

[0115] Step 4: Use LSTM to predict the shielding effectiveness of the metal cavity.

[0116] Example 1

[0117] The present invention adopts Robinson algorithm to calculate shielding effectiveness, applies LSTM to predict shielding effectiveness, and performs prediction of two schemes.

[0118] like Figure 4 As shown, the first solution is to calculate the shielding effectiveness from 1 MHz to 2 GHz, take 1700 frequency responses from 1 MHz to 1700 MHz as a training set, and predict the shielding effectiveness of the last 300 frequency points, which is recorded as LSTM (2 GHz).

[0119] Of the total 2000 frequency points, the first 1700 were used as training sets, and the response prediction of the remaining 300 frequency points was performed based on LSTM. Combining the data in Tables 1 and 2, the "GDM", "ADM" and "FDM" evaluations of this set of data are all shown as "Excellent", which confirms the ideality of the prediction effect.

[0120] Table 2 FSV of LSTM (2GHz) and LSTM (20GHz)

[0121]

[0122] Example 2

[0123] like Figure 5 As shown in FIG. 1 , the second solution is to calculate the shielding effectiveness from 1 MHz to 20 GHz, take 17,000 frequency responses from 1 MHz to 17 GHz as a training set, and predict the shielding effectiveness of the last 3,000 frequency points, denoted as LSTM (20 GHz).

[0124] In this solution, a total of 17,000 frequency response points in the range of 1MHz to 17GHz are used as training sets to predict the responses of the next 3,000 frequency points. Thanks to the obvious periodicity and trend shown by the frequency response sequence data, LSTM can effectively capture these key characteristics during the training process. From the data in the table, it can be seen that the prediction values ​​of LSTM (20GHz) are all "Excellent".

[0125] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0126] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0127] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0128] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0129] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0130] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the shielding effectiveness of a metal box based on LSTM, characterized in that: The following steps are involved: Step 1: Build the memory unit structure of LSTM; Step 2: Build a metal shielding box calculation model; Step 3: Construct the shielding cavity equivalent circuit based on the Robinson algorithm and calculate the shielding effectiveness; Step 4: Use LSTM to predict the shielding effectiveness of the metal cavity.

2. According to the method for predicting the shielding effectiveness of a metal box based on LSTM in claim 1, it is characterized in that: In step 1, the memory unit structure of the LSTM includes: an input gate i that determines the amount of information input into the memory unit, an output gate o that determines the output amount of the LSTM, and a forget gate f that determines whether historical information is forgotten or passed; The input gate i randomly generates a value between 0 and 1 according to the current input and the previous hidden layer state, which controls how much information flows into the memory unit; The forget gate f randomly generates a value between 0 and 1 according to the importance of the input of the input gate i for a given future sample, that is, how much information the memory unit forgets; The output gate o determines how much information the memory cell outputs by the following formula: o t =σ(w xo x t +w ho h t-1 +b o ); Among them, t represents the output information at time t; w xo represents the input weight; x t represents the input information at time t; w ho represents the state weight of the memory unit; h t-1 represents the output information of the hidden layer at time t-1; b o Represents the bias term.

3. A method for predicting shielding effectiveness of a metal box based on LSTM according to claim 1 or 2, characterized in that: In step 1, assuming that at time t, the input is x t , the hidden layer output at the current moment is h t , the memory cell state is c t ; The sigmoid function is used to control the values ​​of the input gate i, output gate o, and forget gate f within the range of [0, 1]: Wherein, σ(x) represents the sigmoid function with respect to x; x represents the input of the sigmoid function; At each time t, update the unit gate g t , g t The value of the input x t and the previous hidden layer state h t-1 Related; the update method is: g t =tanh(w xg x t +w hg h t-1 +b g ); Among them, w xg represents the input weight; x t represents the input information at time t; w hg represents the state weight of the memory unit; h t-1 represents the output information of the hidden layer at time t-1; b g represents the bias term; Memory cell state c t Information is updated through f t Control how much information and i are forgotten at the last moment t How much information is input at this moment to achieve common control; the state of the memory unit is: c t =f t ×c t-1 +i t ×g t ; Among them, f t represents the output of the forget gate at time t; c t-1 represents the state of the memory unit at time t-1; i t represents the output of the input gate at time t; g t represents the state of the candidate memory unit at time t; The entire hidden layer state h at the previous moment t Through the output gate o t Control memory unit c t How much information is output? The entire hidden layer state h t It is expressed as: h t =o t ×tanh(c t ); Then the output y of the LSTM network is t for: y t =σ(w hy h t +b y ); Among them, w hy represents the output weight; h t represents the output information of the hidden layer at time t; b y Represents the bias term.

4. The method for predicting shielding effectiveness of a metal box based on LSTM according to claim 3 is characterized in that: The information flow of the input gate i is: i t =σ(w xi x t +w hi h t-1 +b i ); Among them, w xi represents the input weight; x t represents the input information at time t; w hi represents the memory unit weight; h t-1 represents the output information of the hidden layer at time t-1; b i Represents the bias term.

5. The method for predicting shielding effectiveness of a metal box based on LSTM according to claim 3 is characterized in that: The amount of forgotten information of the forget gate f is: f t =σ(w xf x t +w hf h t-1 +b f ); Among them, w xf represents the input weight; x t represents the input information at time t; w hf represents the memory unit weight; h t-1 represents the output information of the hidden layer at time t-1; b f Represents the bias term.

6. The method for predicting shielding effectiveness of a metal box based on LSTM according to claim 1, characterized in that: In step 2, the metal shielding box is a rectangular parallelepiped; the inner dimensions of the cavity of the metal shielding box are a×b×d, and the thickness is t; a rectangular hole with dimensions of l×w is provided at the center of the front panel of the metal shielding box; There is an excitation plane wave radiation outside the metal shielding box, the excitation plane wave radiation is incident on the rectangular hole cavity surface perpendicular to the panel, and is polarized along the cavity height direction; The shielding effectiveness test point P is set along the center line of the panel, with a distance p from the opening.

7. The method for predicting shielding effectiveness of a metal box based on LSTM according to claim 1, characterized in that: In step 3, the incident plane wave is equivalent to a voltage source with a voltage of V0 and an impedance of Z0≈377Ω, and the shielding cavity is equivalent to a rectangular waveguide with a short-circuited terminal. The characteristic impedance and propagation constant of the shielding cavity are Z g and k g , k0 = 2π / λ is the free space wave number, and the rectangular opening is equivalent to a coplanar strip transmission line with short circuits at both ends. The characteristic impedance of the rectangular opening is: Among them, K and K′ are elliptic integrals, w e is the effective width of the hole, and the expression is: Characteristic impedance of aperture Z 0s The approximate expression of is: Using the transmission line theory, the input impedance at the midpoint of the aperture equivalent transmission line, that is, the aperture impedance, can be obtained as: Wherein, l / a represents the aperture impedance coefficient introduced to consider the coupling effect between the aperture and the cavity; According to the Thevenin theorem, the equivalent voltage at the hole is V1 = V0Z ap / (Z0+Z ap ) and the equivalent source impedance Z1=Z0Z ap / (Z0+Z ap );For TE 10 Propagation mode, waveguide characteristic impedance is The propagation constant is According to the transmission line theory, the equivalent voltage, source impedance and load impedance at the monitoring point X≤0.1 can be obtained: Z3=jZgtank g (dp); According to Thevenin's theorem, the voltage at point P is V box =V2Z3 / (Z2+Z3); and when there is no shielding cavity, the voltage at point P is V no =V0 / 2; given the frequency f of the excitation plane wave, the electric field shielding effectiveness at point P is calculated as:

Images