Silicon imine production method adopting continuous method

Abstract

The invention relates to a silicon imine production method adopting a continuous method. The method comprises the following steps: by taking liquid ammonia as a liquid medium, turbulence the liquid ammonia from top to bottom along the inner wall of the falling-film tubular reactor; spraying silicon tetrachloride in a mist form to form gaseous silicon tetrachloride; carrying out gas-liquid reaction on the gaseous silicon tetrachloride and liquid ammonia to obtain a silicon imine solid reaction product; in the aging device, gaseous silicon tetrachloride which is not completely reacted continuously reacts with liquid ammonia; the filter is used for filtering a silicon imine solid reaction product; the separator is used for separating ammonia gas and ammonium chloride. In this way, waste discharge is reduced through continuous production. Then, in combination with a deep learning algorithm, an implicit change mode of real-time pressure data in the reactor can be mined, consumption of reactants and accumulation of products in the reactor can be learned and understood from the implicit change mode, and intelligent spraying of gaseous silicon tetrachloride is realized based on the condition.

Description

Technical Field

[0001] The present application relates to the technical field of silicium imide, and in particular to a continuous method for producing silicium imide. Background Art

[0002] With the rapid development of modern industry and technology, the demand for new materials is growing. As an important organosilicon compound, silanimide has shown great potential in the fields of chemical engineering, material science, electronics and medicine due to its unique chemical properties and wide application prospects.

[0003] However, the traditional silicium imine production method has many limitations, such as low efficiency, serious environmental pollution, unstable product quality, etc. Therefore, an optimized silicium imine production method is expected. Summary of the invention

[0004] This summary is provided to introduce concepts in a brief form that will be described in detail in the detailed description below. This summary is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0005] In a first aspect, the present application provides a continuous method for producing silanimide, the method comprising:

[0006] Using liquid ammonia as the liquid medium, the liquid ammonia is turbulently flowed from top to bottom along the inner wall of the falling film tube reactor;

[0007] Spraying silicon tetrachloride in a mist form to form gaseous silicon tetrachloride;

[0008] The gaseous silicon tetrachloride and the liquid ammonia undergo a gas-liquid reaction to obtain a silimidate solid reaction product;

[0009] An aging device connected to the falling film tubular reactor is provided, wherein the incompletely reacted gaseous silicon tetrachloride continues to react with the liquid ammonia in the aging device;

[0010] A filter connected to the aging device is provided, wherein the filter filters the silimido solid reaction product;

[0011] A separator connected to the filter is provided, wherein the separator is used to separate ammonia and ammonium chloride.

[0012] Optionally, the gaseous silicon tetrachloride and the liquid ammonia undergo a gas-liquid reaction to obtain a silicon imide solid reaction product, including: obtaining a time series of real-time pressure in a reactor collected by a pressure sensor; performing first-order time series correlation feature extraction on the time series of the real-time pressure in the reactor to obtain a sequence of time series correlation implicit feature vectors of the real-time pressure in the reactor; inputting the sequence of time series correlation implicit feature vectors of the real-time pressure in the reactor into an anchor-centered feature distribution reverse selection enhancement network to obtain a sequence of time series correlation implicit feature vectors of the real-time pressure in the enhanced reactor; performing second-order time series correlation feature extraction on the sequence of time series correlation implicit feature vectors of the real-time pressure in the enhanced reactor to obtain a time series correlation representation vector of the real-time pressure in the full time domain reactor; and determining a control instruction based on the time series correlation representation vector of the real-time pressure in the full time domain reactor, the control instruction being used to indicate whether to stop spraying the gaseous silicon tetrachloride.

[0013] Optionally, a first-order time series correlation feature extraction is performed on the time series of the real-time pressure in the reactor to obtain a sequence of implicit feature vectors of the time series correlation of the real-time pressure in the reactor, including: inputting the time series of the real-time pressure in the reactor into a first-order sequence encoder based on a 1D-CNN model to obtain a sequence of implicit feature vectors of the time series correlation of the real-time pressure in the reactor.

[0014] Optionally, the sequence of implicit feature vectors associated with the real-time pressure series in the reactor is input into an anchor-centered feature distribution reverse selection enhancement network to obtain a sequence of implicit feature vectors associated with the real-time pressure series in the reactor, including: calculating the semantic correlation coefficient between each implicit feature vector associated with the real-time pressure series in the reactor and the anchor-centered feature vector associated with the real-time pressure series in the reactor in the sequence of implicit feature vectors associated with the real-time pressure series in the reactor to obtain a sequence of semantic implicit correlation coefficients; taking the reciprocal of each semantic implicit correlation coefficient in the sequence of semantic implicit correlation coefficients to obtain a sequence of semantic anti-correlation coefficients; and performing Normalization processing is performed to obtain a sequence of semantic anti-correlation energy coefficients; each semantic anti-correlation energy coefficient in the sequence of semantic anti-correlation energy coefficients is used as a weight to weight each real-time pressure time series correlation implicit feature vector in the sequence of real-time pressure time series correlation implicit feature vectors in the reactor to obtain a sequence of real-time pressure time series correlation implicit feature vectors in reverse inhibition reactor; the positional difference between the sequence of real-time pressure time series correlation implicit feature vectors in the reactor and the sequence of real-time pressure time series correlation implicit feature vectors in reverse inhibition reactor is calculated to obtain a sequence of real-time pressure time series correlation implicit feature vectors in the enhanced reactor.

[0015] Optionally, calculating the semantic correlation coefficient between each implicit feature vector of the real-time pressure time series association in the reactor and the anchor center feature vector of the real-time pressure time series association in the reactor to obtain a sequence of semantic implicit correlation coefficients, including: calculating the position mean vector of the sequence of implicit feature vectors of the real-time pressure time series association in the reactor to determine the anchor center feature vector of the real-time pressure time series association in the reactor; calculating the semantic correlation coefficient between the implicit feature vector of the real-time pressure time series association in the reactor and ... semantic correlation coefficient Perform positional dot multiplication between the central feature vectors, and perform element-by-element summation of the vectors obtained by the positional dot multiplication to obtain the local time series correlation feature heap value of the network traffic constrained by the anchor center; calculate the element-by-element sum of the squares of the real-time pressure time series correlation implicit feature vector in the i-th reactor, and the element-by-element sum of the squares of the real-time pressure time series correlation anchor center feature vector in the reactor, and calculate the square root of the sum of the two values ​​obtained by the element-by-element sum of the squares, divide the local time series correlation feature heap value of the network traffic constrained by the anchor center by the value obtained by the square root to obtain the i-th semantic correlation coefficient.

[0016] Optionally, the sequence of semantic anti-correlation coefficients is normalized to obtain a sequence of semantic anti-correlation energy coefficients, including: using a Softmax function to normalize the sequence of semantic anti-correlation coefficients to obtain a sequence of semantic anti-correlation energy coefficients.

[0017] Optionally, a second-order time series correlation feature extraction is performed on the sequence of implicit feature vectors of the real-time pressure time series correlation in the enhanced reactor to obtain a real-time pressure time series correlation representation vector in the full time domain reactor, including: inputting the sequence of implicit feature vectors of the real-time pressure time series correlation in the enhanced reactor into a second-order sequence encoder based on a Bi-LSTM model to obtain a real-time pressure time series correlation representation vector in the full time domain reactor.

[0018] Optionally, based on the real-time pressure timing association representation vector in the full-time domain reactor, a control instruction is determined, wherein the control instruction is used to indicate whether to stop spraying the gaseous silicon tetrachloride, including: inputting the real-time pressure timing association representation vector in the full-time domain reactor into a classifier-based spray controller to obtain the control instruction.

[0019] The above technical solution is adopted, by using liquid ammonia as the liquid medium, the liquid ammonia is turbulently flowed from top to bottom along the inner wall of the falling film tube reactor; silicon tetrachloride is sprayed down in the form of mist to form gaseous silicon tetrachloride; the gaseous silicon tetrachloride reacts with liquid ammonia to obtain a solid reaction product of silicimide; in the aging device, the gaseous silicon tetrachloride that has not reacted completely continues to react with liquid ammonia; the filter filters the solid reaction product of silicimide; the separator is used to separate ammonia and ammonium chloride. In this way, the discharge of waste is reduced through continuous production. Then, the deep learning algorithm can be combined to mine the implicit change pattern of the real-time pressure data in the reactor, learn and understand the consumption of reactants and the accumulation of products in the reactor, and based on this, the intelligent spraying of gaseous silicon tetrachloride can be realized.

[0020] Other features and advantages of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale. In the drawings:

[0022] Figure 1 The present invention is a flow chart showing a continuous method for producing silanimide according to an exemplary embodiment.

[0023] Figure 2 is based on Figure 1 The illustrated embodiment shows a flow chart of step S103 of a continuous method for producing silanimide.

[0024] Figure 3 It is a schematic diagram of a falling film shell-and-tube gas-liquid reaction process according to an exemplary embodiment.

[0025] Figure 4 is a block diagram of a continuous silane production system according to an exemplary embodiment.

[0026] Figure 5 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0027] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0028] It should be understood that the various steps described in the method implementation of the present application can be performed in different orders and / or performed in parallel. In addition, the method implementation may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0029] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0030] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0031] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0032] The names of the messages or information exchanged between multiple devices in the embodiments of the present application are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0033] In order to solve the above problems, the present application provides a continuous method for producing silicium imide, which uses liquid ammonia as a liquid medium, and turbulently flows the liquid ammonia from top to bottom along the inner wall of a falling film tube reactor; sprays silicon tetrachloride in a mist form to form gaseous silicon tetrachloride; gaseous silicon tetrachloride reacts with liquid ammonia to obtain a solid reaction product of silicium imide; in an aging device, the incompletely reacted gaseous silicon tetrachloride continues to react with liquid ammonia; a filter filters the solid reaction product of silicium imide; a separator is used to separate ammonia and ammonium chloride. In this way, waste emissions are reduced through continuous production. Then, the deep learning algorithm can be combined to mine the implicit change pattern of the real-time pressure data in the reactor, learn and understand the consumption of reactants and the accumulation of products in the reactor, and based on this, the intelligent spraying of gaseous silicon tetrachloride can be realized.

[0034] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0035] Figure 1 is a flow chart of a continuous method for producing silanimide according to an exemplary embodiment. Figure 1 As shown, the method includes:

[0036] Step S101, using liquid ammonia as a liquid medium, and causing the liquid ammonia to flow turbulently from top to bottom along the inner wall of a falling film tubular reactor;

[0037] Step S102, spraying silicon tetrachloride in a mist form to form gaseous silicon tetrachloride;

[0038] Step S103, the gaseous silicon tetrachloride and the liquid ammonia undergo a gas-liquid reaction to obtain a silylimide solid reaction product;

[0039] Step S104, providing an aging device connected to the falling film tubular reactor, wherein in the aging device, the gaseous silicon tetrachloride that has not completely reacted continues to react with the liquid ammonia;

[0040] Step S105, setting a filter connected to the aging device, wherein the filter filters the silimido solid reaction product;

[0041] Step S106: Setting a separator connected to the filter, wherein the separator is used to separate ammonia and ammonium chloride. In this way, waste discharge is reduced through continuous production.

[0042] Considering that the reaction between silicon tetrachloride and liquid ammonia is a gas-liquid reaction, its efficiency is affected by the contact area and reaction time. In actual application scenarios, it is necessary to control the spraying of silicon tetrachloride to ensure that silicon tetrachloride is evenly dispersed in liquid ammonia in the form of mist, increase the gas-liquid contact area, and ensure the reaction efficiency. However, the traditional control method is highly dependent on the experience and knowledge level of relevant technicians and operators. This control method may be affected by human subjective bias and lead to inconsistent control results. Therefore, an optimized solution is expected.

[0043] In response to the above technical problems, the technical concept of the present application is: combining a deep learning algorithm to mine the implicit change pattern of the real-time pressure data in the reactor, learn and understand the consumption of reactants and the accumulation of products in the reactor, and based on this, realize the intelligent spraying of the gaseous silicon tetrachloride.

[0044] Figure 2 is based on Figure 1 The embodiment shown is a flow chart of step S103 of a continuous method for producing silanimide, as shown in FIG. Figure 2 As shown, step S103, the gaseous silicon tetrachloride and the liquid ammonia undergo a gas-liquid reaction to obtain a silimidate solid reaction product, comprising:

[0045] Step S1031, obtaining a time series of real-time pressure in the reactor collected by a pressure sensor;

[0046] Step S1032, performing first-order time series correlation feature extraction on the time series of the real-time pressure in the reactor to obtain a sequence of implicit feature vectors of the time series correlation of the real-time pressure in the reactor;

[0047] Step S1033, inputting the sequence of implicit feature vectors associated with the real-time pressure series in the reactor into the anchor point centralized feature distribution reverse selection enhancement network to obtain a sequence of implicit feature vectors associated with the real-time pressure series in the enhanced reactor;

[0048] Step S1034, performing second-order time series correlation feature extraction on the sequence of implicit feature vectors of real-time pressure time series correlation in the enhanced reactor to obtain a representation vector of real-time pressure time series correlation in the full-time domain reactor;

[0049] Step S1035: determining a control instruction based on the real-time pressure time series correlation representation vector in the full-time reactor, wherein the control instruction is used to indicate whether to stop spraying the gaseous silicon tetrachloride.

[0050] Based on this, in the technical solution of the present application, the specific implementation process of the gaseous silicon tetrachloride and the liquid ammonia to obtain a silimido solid reaction product includes: first, obtaining a time series of real-time pressure in the reactor collected by a pressure sensor. Among them, the pressure change in the reactor can reflect the dynamic process of the gas-liquid reaction. By monitoring the pressure change, the consumption of the reactants and the accumulation of the products can be indirectly understood, thereby judging the progress of the reaction. Specifically, the pressure sensor can collect pressure data in the reactor in real time and provide continuous feedback information. This real-time monitoring capability makes the production process more controllable, and the reaction conditions can be adjusted in time to ensure the stability and efficiency of the reaction.

[0051] Since the pressure change in the reactor is usually time-dependent, that is, the current pressure change may be affected by the previous pressure change. In the technical solution of the present application, the time series of the real-time pressure in the reactor is input into a first-order sequence encoder based on the 1D-CNN model to obtain a sequence of implicit feature vectors associated with the time series of the real-time pressure in the reactor. Among them, the 1D-CNN model is good at extracting local features from one-dimensional data. In this way, the first-order sequence encoder based on the 1D-CNN model can identify important local patterns and changing trends from the time series of the real-time pressure in the reactor to understand the dynamic process of the reaction.

[0052] In one embodiment of the present application, a first-order time series correlation feature extraction is performed on the time series of the real-time pressure in the reactor to obtain a sequence of implicit feature vectors of the time series correlation of the real-time pressure in the reactor, including: inputting the time series of the real-time pressure in the reactor into a first-order sequence encoder based on a 1D-CNN model to obtain a sequence of implicit feature vectors of the time series correlation of the real-time pressure in the reactor.

[0053] Considering that there may be noise information in the time series of the real-time pressure in the reactor, and the data at different positions in the time series of the real-time pressure in the reactor have different degrees of importance for characterizing and describing the reaction characteristics in the reactor. For example, the peak pressure in the reaction process usually corresponds to the moment when the reaction rate is the fastest or the reactant consumption is the highest. In the technical solution of the present application, in order to enhance the model's ability to capture key information and suppress those features that have low relevance to the target task, the sequence of implicit feature vectors associated with the real-time pressure time series in the reactor is further input into the anchor-centered feature distribution reverse selection enhancement network to obtain a sequence of implicit feature vectors associated with the real-time pressure time series in the enhanced reactor.

[0054] In one embodiment of the present application, the sequence of implicit feature vectors associated with the real-time pressure series in the reactor is input into an anchor-centered feature distribution reverse selection enhancement network to obtain a sequence of implicit feature vectors associated with the real-time pressure series in the reactor, including: calculating the semantic correlation coefficient between each implicit feature vector associated with the real-time pressure series in the reactor and the anchor-centered feature vector associated with the real-time pressure series in the reactor in the sequence of implicit feature vectors associated with the real-time pressure series in the reactor to obtain a sequence of semantic implicit correlation coefficients; taking the reciprocal of each semantic implicit correlation coefficient in the sequence of semantic implicit correlation coefficients to obtain a sequence of semantic anti-correlation coefficients; and The sequence of real-time pressure-series-associated implicit feature vectors is normalized to obtain a sequence of semantic anti-correlation energy coefficients; each semantic anti-correlation energy coefficient in the sequence of semantic anti-correlation energy coefficients is used as a weight to weight each real-time pressure-series-associated implicit feature vector in the sequence of real-time pressure-series-associated implicit feature vectors in the reactor to obtain a sequence of real-time pressure-series-associated implicit feature vectors in reverse inhibition reactors; the positional difference between the sequence of real-time pressure-series-associated implicit feature vectors in the reactor and the sequence of real-time pressure-series-associated implicit feature vectors in reverse inhibition reactors is calculated to obtain a sequence of real-time pressure-series-associated implicit feature vectors in the enhanced reactors.

[0055] Furthermore, in one embodiment of the present application, the semantic correlation coefficient between each implicit feature vector of the real-time pressure time series association in the reactor and the anchor center feature vector of the real-time pressure time series association in the reactor is calculated to obtain a sequence of semantic implicit correlation coefficients, including: calculating the positional mean vector of the sequence of implicit feature vectors of the real-time pressure time series association in the reactor to determine the anchor center feature vector of the real-time pressure time series association in the reactor; calculating the positional mean vector of the sequence of implicit feature vectors of the real-time pressure time series association in the reactor and ... The positional dot product is performed between the feature vectors of the pressure-time series association anchor center, and the vectors obtained by the positional dot product are added element by element to obtain the local time series association feature heap value of the network traffic constrained by the anchor center; the element-by-element sum of the squares of the real-time pressure-time series association implicit feature vectors in the i-th reactor and the element-by-element sum of the squares of the real-time pressure-time series association anchor center feature vectors in the reactor are calculated, and the square root of the sum of the two values ​​obtained by the element-by-element sum of the squares is calculated, and the local time series association feature heap value of the network traffic constrained by the anchor center is divided by the value obtained by the square root to obtain the i-th semantic correlation coefficient.

[0056] Furthermore, in one embodiment of the present application, the sequence of semantic anti-correlation coefficients is normalized to obtain a sequence of semantic anti-correlation energy coefficients, including: using a Softmax function to normalize the sequence of semantic anti-correlation coefficients to obtain a sequence of semantic anti-correlation energy coefficients.

[0057] Specifically, the sequence of the real-time pressure time series associated implicit feature vectors in the reactor is input into the anchor point centralized feature distribution reverse selection enhanced network using the following enhanced network formula to obtain a sequence of the real-time pressure time series associated implicit feature vectors in the enhanced reactor; wherein the enhanced network formula is:

[0058] X={x (1) , x (2) , ..., x (N)}

[0059]

[0060]

[0061]

[0062] Y={y (1) ,y (2) , ..., y (N )},y (i) =w (i) ·x (i)

[0063] Z={z (1) , z (2) , ..., z (N )},z (i) =x (i) -y (i)

[0064] Where X is the sequence of implicit feature vectors associated with the real-time pressure sequence in the reactor, x (1) 、x (2) 、x (i) and x (N) are respectively the first implicit feature vector of real-time pressure time series association in the reactor, the second implicit feature vector of real-time pressure time series association in the reactor, the i-th implicit feature vector of real-time pressure time series association in the reactor, and the N-th implicit feature vector of real-time pressure time series association in the reactor, wherein the value of N is the number of implicit feature vectors of real-time pressure time series association in the reactor in the sequence of implicit feature vectors of real-time pressure time series association in the reactor, and V r is the anchor center feature vector of the real-time pressure time series correlation in the reactor, and are the jth eigenvalue of the real-time pressure time series associated implicit feature vector in the ith reactor and the real-time pressure time series associated anchor center feature vector in the ith reactor, the value of L is the length of the real-time pressure time series associated implicit feature vector in the reactor, r (i) , (i) and w (i) are the ith semantic implicit association coefficient, semantic anti-association coefficient and semantic anti-association energy coefficient, respectively; Y is the sequence of the real-time pressure temporal association implicit feature vectors in the reverse inhibition reactor; y (1) ,y (2) ,y (i) and (N) are respectively the first implicit feature vector of the real-time pressure time series association in the reverse inhibition reactor, the second implicit feature vector of the real-time pressure time series association in the reverse inhibition reactor, the i-th implicit feature vector of the real-time pressure time series association in the reverse inhibition reactor and the N-th implicit feature vector of the real-time pressure time series association in the reverse inhibition reactor, Z is the sequence of the implicit feature vector of the real-time pressure time series association in the enhanced reactor, z (1) 、z (2) 、z (i) and z (N) are respectively the implicit feature vector of the real-time pressure time series association in the first enhanced reactor, the implicit feature vector of the real-time pressure time series association in the second enhanced reactor, the implicit feature vector of the real-time pressure time series association in the i-th enhanced reactor and the implicit feature vector of the real-time pressure time series association in the N-th enhanced reactor, e(·) represents an exponential function with a natural constant as the base, and ⊙ represents dot product.

[0065] In particular, the anchor-centered feature distribution reverse selection enhancement network first obtains a sequence of semantic latent correlation coefficients by calculating the semantic correlation coefficients between the implicit feature vectors associated with the real-time pressure time series in each reactor and the anchor-centered feature vector associated with the real-time pressure time series in the reactor (for example, the position mean vector of the sequence of implicit feature vectors associated with the real-time pressure time series in the reactor). Then, the feature weights with a high correlation with the reaction characteristics in the reactor are reduced by taking the inverse of the sequence of semantic latent correlation coefficients, in preparation for subsequent reverse screening. Subsequently, the anchor-centered feature distribution reverse selection enhancement network uses the Softmax function to normalize the obtained semantic anti-correlation coefficients to ensure that the sum of all coefficients is 1. Among them, the introduction of the Softmax function makes the weight distribution smoother, avoids the appearance of extreme values, and thus improves the stability and rationality of the weight distribution. In the weighted reverse inhibition stage, the anchor-centered feature distribution reverse selection enhancement network uses the normalized semantic anti-correlation energy coefficient as a weight to weight the input feature vector, that is, the real-time pressure time series associated implicit feature vector in each reactor, and calculates the difference between the original input feature vector sequence and the weighted reverse inhibition feature vector sequence, thereby obtaining the sequence of the real-time pressure time series associated implicit feature vector in the enhanced reactor. That is, firstly, the noise component or interference component in the feature distribution is made relatively significant by weighting, and then the part of the original feature with a higher correlation with the target task is strengthened by the difference between the feature vectors, while the part with a lower correlation is suppressed, thereby achieving the enhancement of the feature distribution.

[0066] Next, the sequence of the implicit feature vectors of the real-time pressure time series association in the enhanced reactor is input into the second-order sequence encoder based on the Bi-LSTM model to obtain the representation vector of the real-time pressure time series association in the reactor in the entire time domain. That is, the second-order sequence encoder is constructed using the Bi-LSTM model to further deepen the understanding and prediction of the pressure changes in the reactor. In particular, the Bi-LSTM model can process past and future information at the same time, which enables the model to more comprehensively understand the time series association in the time series data. This two-way understanding is crucial for predicting pressure changes during the reaction process.

[0067] In one embodiment of the present application, a second-order time series correlation feature extraction is performed on the sequence of implicit feature vectors of the real-time pressure time series correlation in the enhanced reactor to obtain a real-time pressure time series correlation representation vector in the full time domain reactor, including: inputting the sequence of implicit feature vectors of the real-time pressure time series correlation in the enhanced reactor into a second-order sequence encoder based on a Bi-LSTM model to obtain the real-time pressure time series correlation representation vector in the full time domain reactor.

[0068] Furthermore, the real-time pressure time series correlation representation vector in the full-time reactor is input into a spray controller based on a classifier to obtain a control instruction, wherein the control instruction is used to indicate whether to stop spraying the gaseous silicon tetrachloride.

[0069] In one embodiment of the present application, a control instruction is determined based on a real-time pressure timing association representation vector in the full-time domain reactor, and the control instruction is used to indicate whether to stop spraying the gaseous silicon tetrachloride, including: inputting the real-time pressure timing association representation vector in the full-time domain reactor into a classifier-based spray controller to obtain the control instruction.

[0070] In the technical solution of the present application, each of the implicit feature vectors of real-time pressure time series association in the reactor in the sequence of implicit feature vectors of real-time pressure time series association in the reactor respectively represents the local time series association characteristics of the real-time pressure in the reactor. Considering the volatility of the real-time pressure in the reactor, the time series of the real-time pressure in the reactor has a relatively significant instability in the data source domain, which makes the sequence of implicit feature vectors of real-time pressure time series association in the reactor have a relatively significant feature distribution manifold inconsistency. After the feature distribution reverse selection enhancement of the anchor point centralized feature distribution reverse selection enhancement network, the difference of the feature manifold in the sequence of implicit feature vectors of real-time pressure time series association in the enhanced reactor will be amplified, which will cause the full-time domain real-time pressure time series association representation vector of the reactor obtained by sequence encoding the sequence of implicit feature vectors of real-time pressure time series association in the enhanced reactor by the second-order sequence encoder based on the Bi-LSTM model to have a posterior probabilistic causal association missing relative to the source domain time series distribution of the real-time pressure in the reactor, thereby affecting the accuracy of the control instruction.

[0071] In a preferred embodiment, inputting the real-time pressure time series correlation representation vector in the full-time reactor into the classifier-based spray controller to obtain the control instruction comprises:

[0072] Obtaining a real-time pressure time series correlation representation vector in the full-time reactor and inputting it into a spraying controller based on a classifier to obtain a stop probability value corresponding to the stop of spraying the gaseous silicon tetrachloride;

[0073] Probabilistically converting each eigenvalue of the real-time pressure time series correlation representation vector in the full-time reactor based on a probability activation function, such as a sigmoid function, to obtain a probabilistic real-time pressure time series correlation representation vector in the full-time reactor;

[0074] Determine the real-time pressure time series correlation coding cognitive symbol value in the full-time reactor based on the comparison between each characteristic value of the probabilistic real-time pressure time series correlation representation vector in the full-time reactor and the stop probability value, wherein the real-time pressure time series correlation coding cognitive symbol value in the full-time reactor is equal to one, zero and negative one in response to the characteristic value of the probabilistic real-time pressure time series correlation representation vector in the full-time reactor being greater than, equal to and less than the stop probability value, respectively;

[0075] Calculating the mean of all eigenvalues ​​of the probabilistic full-time-domain reactor real-time pressure time series correlation representation vector to obtain the full-time-domain reactor real-time pressure time series correlation encoding overall phase shift value;

[0076] Each eigenvalue of the probabilistic full-time-domain reactor real-time pressure time series association representation vector is multiplied by the real-time pressure time series association coding cognition symbol value and the real-time pressure time series association coding overall phase shift value to obtain a probabilistic full-time-domain reactor real-time pressure time series association coding cognition eigenvector and a probabilistic full-time-domain reactor real-time pressure time series association coding phase shift eigenvector;

[0077] The probabilistic full-time reactor real-time pressure time series correlation coding cognitive feature vector and the probabilistic full-time reactor real-time pressure time series correlation coding phase shift feature vector are respectively multiplied by a first weight and a second weight as hyperparameters, and then the difference is calculated to obtain a probabilistic full-time reactor real-time pressure time series correlation coding differential feature vector;

[0078] Taking the absolute value of each eigenvalue of the probabilistic full-time-domain reactor real-time pressure time series correlation encoding differential eigenvector to obtain an optimized full-time-domain reactor real-time pressure time series correlation representation vector;

[0079] The optimized real-time pressure time series correlation representation vector in the full-time reactor is input into the classifier-based spray controller to obtain the control instruction.

[0080] Specifically, the optimization steps are expressed as:

[0081]

[0082]

[0083] Among them, f i is the ith eigenvalue of the probabilistic full-time reactor real-time pressure time series correlation representation vector, f' i is the i-th eigenvalue of the optimized real-time pressure time series correlation representation vector in the full-time reactor, α and β are the first weight and the second weight, respectively. is the overall phase shift value of the real-time pressure timing correlation code in the full-time reactor, sgn(f i ,p) is the real-time pressure time series associated coding recognition symbol value in the full-time reactor, p is the stop probability value, and |·| represents the absolute value.

[0084] Accordingly, in the above-mentioned preferred embodiment of the present application, the class-cognitive phase transformation response of the real-time pressure time series association representation vector in the full time domain reactor is obtained by comparing the probabilistic amplitude of the eigenvalue of the real-time pressure time series association representation vector in the full time domain reactor with the conforming probability representation, and the phase shift response of the eigenvalue of the real-time pressure time series association representation vector in the full time domain reactor relative to the class probability representation of the feature as a whole is subjected to a feature distribution sequence invariance transformation based on the class probability differential distribution expansion, so as to realize the causal constraint of the posterior class probability of the real-time pressure time series association representation vector in the full time domain reactor on its prior feature distribution representation, so as to improve the accuracy of the control instructions obtained by the real-time pressure time series association representation vector in the full time domain reactor when it is input into the spray controller based on the classifier.

[0085] In summary, the above scheme is adopted in combination with a deep learning algorithm to mine the implicit change pattern of the real-time pressure data in the reactor, from which the consumption of reactants and the accumulation of products in the reactor are learned and understood, and based on this, the intelligent spraying of the gaseous silicon tetrachloride is realized.

[0086] Figure 3 is a schematic diagram of a falling film tube-in-tube gas-liquid reaction process according to an exemplary embodiment. Figure 3 In another embodiment of the present application, a method for continuously producing silanimide is provided, which comprises the steps of:

[0087] 1. The falling film shell-and-tube gas-liquid reaction process technology is mature and suitable for large-scale exothermic reactions, such as the industrial production of alkylbenzene sulfonic acid.

[0088] 2. Liquid ammonia is used as the liquid medium, and the inner wall of the tube is leaked from top to bottom. Silicon tetroxide is sprayed down in the form of mist to form a salty gas. The gas waves react at the interface to produce silicimide and ammonium amide, and release heat.

[0089] 3. Excess liquid ammonia and circulating liquid ammonia in the jacket can keep the reaction temperature at the set temperature. The reaction progress can be controlled by adjusting the injection amount of silicon tetrachloride. At the same time, due to the closed environment, it can withstand partial pressure and the reaction is relatively stable.

[0090] 4. The ammonium oxide generated by the reaction dissolves in liquid ammonia, and the generated silicic acid imide solid powder floats or suspends in the liquid ammonia. The clear flow of liquid ammonia from top to bottom can carry away the silicic acid imide particles without forming precipitation or blockage.

[0091] 5. Set up an aging device to allow the incomplete reaction of silicon tetroxide and liquid ammonia to continue to react. Under normal circumstances, nitrogen and silicon tetroxide react violently, and the previous stage should have reacted completely. The aging device only plays a role in stabilizing the voltage.

[0092] 6. A pressure-stabilizing check chamber and a bursting disc are set on the aging device. The check chamber can be connected to the gas treatment device to absorb the escaped ammonia or recover the gas. The rear end of the bursting disc is connected to the gas treatment device to prepare for safe pressure release in case of emergency.

[0093] 7. Set up a filter, which can be a high-center filter or a traditional filter (more than two traditional filters are required for switching), to separate the solid product silicium imide. Since most of the ammonium chloride is dissolved in liquid ammonia, the ammonium oxide content in silicium imide will be relatively small.

[0094] 8. The filtered liquid ammonia solution containing ammonium oxide enters the separator, and the liquid ammonia is recovered by increasing the temperature to recover the vaporized ammonium.

[0095] 9. This reaction process is a direct gas-liquid reaction between ammonium fluoride and silicon tetroxide. Under the control of appropriate reaction conditions, the traditional organic solvent is eliminated, the reaction interface is expanded, and a continuous large-scale production line can be formed. Compared with the current gas-gas reaction, this reaction makes the reaction conditions milder, and the ammonium chloride content in the obtained product is lower, which is conducive to obtaining high-quality silicon amide products.

[0096] 10. From the perspective of material utilization, all liquid ammonia and silicon tetroxide raw materials in this application have been reacted, and silicic acid imine products and ammonium fluoride by-products have been produced. Under normal circumstances, there is no three-skin emission.

[0097] 11. The main equipment of this application is a falling film tube reactor equipped with an atomizing nozzle, which innovates the traditional gas-phase feed reactor into a liquid-phase reactant feed reactor, greatly expanding the application field of the falling film reactor.

[0098] 12. The distributor of the reactor can be of various types such as orifice plate type, clear flow type, diversion type or spray type. The refrigeration medium of the reaction can be liquid ammonia, propylene and mixed medium, etc. The reaction temperature is controlled at -60°C--35°C.

[0099] 13. All the main equipment of this application needs to be tightly packed and isolated from air, moisture, oxygen and other media. 14. During normal operation, the air tightness of the equipment is first monitored. After the test is qualified, liquid ammonia is used as the medium for circulation, nitrogen is used as the gas phase medium for feeding, and then silicon tetroxide is used for atomization feeding.

[0100] Figure 4FIG. 1 is a block diagram of a continuous silanimide production system according to an exemplary embodiment. Figure 4 As shown, the silimide production system 200 includes:

[0101] The falling film tubular reactor module 201 is used to use liquid ammonia as a liquid medium and to turbulently flow the liquid ammonia from top to bottom along the inner wall of the falling film tubular reactor;

[0102] The silicon tetrachloride spray module 202 is used to spray silicon tetrachloride in a mist form to form gaseous silicon tetrachloride;

[0103] A gas-liquid reaction module 203, used for performing a gas-liquid reaction between the gaseous silicon tetrachloride and the liquid ammonia to obtain a silimido solid reaction product;

[0104] An aging module 204 is used to set an aging device connected to the falling film tube reactor, wherein the incompletely reacted gaseous silicon tetrachloride continues to react with the liquid ammonia in the aging device;

[0105] A filtering module 205 is used to set a filter connected to the aging device, and the filter filters the silimido solid reaction product;

[0106] The separation module 206 is used to set a separator connected to the filter, wherein the separator is used to separate ammonia and ammonium chloride.

[0107] Reference below Figure 5 , which shows a schematic diagram of the structure of an electronic device 600 suitable for implementing the embodiment of the present application. The terminal device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0108] like Figure 5 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0109] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 5 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0110] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present application are executed.

[0111] It should be noted that the computer-readable medium mentioned above in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0112] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0113] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0114] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0115] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0116] The modules involved in the embodiments described in this application can be implemented by software or hardware. The name of the module does not limit the module itself in some cases. For example, the test parameter acquisition module can also be described as a "module for acquiring device test parameters corresponding to the target device".

[0117] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0118] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0119] In the application scenario of the continuous silicon imide production method, first, a time series of real-time pressure in a reactor collected by a pressure sensor is obtained; then, the obtained time series of real-time pressure is input into a server deployed with a continuous silicon imide production algorithm, wherein the server can process the time series of real-time pressure based on the continuous silicon imide production algorithm to determine a control instruction, wherein the control instruction is used to indicate whether to stop spraying the gaseous silicon tetrachloride.

[0120] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

[0121] In addition, although each operation is described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or to be performed in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0122] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely exemplary forms of implementation. With respect to the apparatus in the above-mentioned embodiments, the specific manner in which each module performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

Claims

1. A continuous method for producing silanimide, characterized in that: include: Using liquid ammonia as the liquid medium, the liquid ammonia is turbulently flowed from top to bottom along the inner wall of the falling film tube reactor; Spraying silicon tetrachloride in a mist form to form gaseous silicon tetrachloride; The gaseous silicon tetrachloride and the liquid ammonia undergo a gas-liquid reaction to obtain a silimidate solid reaction product; An aging device connected to the falling film tubular reactor is provided, wherein the incompletely reacted gaseous silicon tetrachloride continues to react with the liquid ammonia in the aging device; A filter connected to the aging device is provided, wherein the filter filters the silimido solid reaction product; A separator connected to the filter is provided, wherein the separator is used to separate ammonia and ammonium chloride.

2. The continuous method for producing silanimide according to claim 1, characterized in that: The gaseous silicon tetrachloride and the liquid ammonia undergo a gas-liquid reaction to obtain a silimidate solid reaction product, comprising: Acquire a time series of real-time pressure in the reactor collected by a pressure sensor; Performing first-order time series correlation feature extraction on the time series of the real-time pressure in the reactor to obtain a sequence of implicit feature vectors of the time series correlation of the real-time pressure in the reactor; Inputting the sequence of implicit feature vectors associated with the real-time pressure sequence in the reactor into the anchor point centralized feature distribution reverse selection enhancement network to obtain a sequence of implicit feature vectors associated with the real-time pressure sequence in the enhanced reactor; Performing second-order time series correlation feature extraction on the sequence of implicit feature vectors of real-time pressure time series correlation in the enhanced reactor to obtain a real-time pressure time series correlation representation vector in the full-time domain reactor; Based on the real-time pressure time series correlation representation vector in the full-time reactor, a control instruction is determined, wherein the control instruction is used to indicate whether to stop spraying the gaseous silicon tetrachloride.

3. The continuous method for producing silanimide according to claim 2, characterized in that: The first-order time series correlation feature extraction is performed on the time series of the real-time pressure in the reactor to obtain a sequence of implicit feature vectors of the time series correlation of the real-time pressure in the reactor, including: The time series of the real-time pressure in the reactor is input into a first-order sequence encoder based on a 1D-CNN model to obtain a sequence of implicit feature vectors associated with the time series of the real-time pressure in the reactor.

4. The continuous method for producing silanimide according to claim 3, characterized in that: Inputting the sequence of implicit feature vectors associated with the real-time pressure sequence in the reactor into the anchor point central feature distribution reverse selection enhancement network to obtain a sequence of implicit feature vectors associated with the real-time pressure sequence in the enhanced reactor, including: Calculating the semantic correlation coefficient between each implicit feature vector of the real-time pressure time series association in the reactor and the anchor center feature vector of the real-time pressure time series association in the reactor in the sequence of implicit feature vectors of the real-time pressure time series association in the reactor to obtain a sequence of semantic implicit correlation coefficients; Taking the reciprocal of each semantic latent association coefficient in the sequence of semantic latent association coefficients to obtain a sequence of semantic anti-association coefficients; Normalizing the sequence of semantic anti-correlation coefficients to obtain a sequence of semantic anti-correlation energy coefficients; Using each semantic anti-correlation energy coefficient in the sequence of semantic anti-correlation energy coefficients as a weight, weighting each reactor real-time pressure time series correlation implicit feature vector in the sequence of reactor real-time pressure time series correlation implicit feature vectors to obtain a sequence of reverse inhibition reactor real-time pressure time series correlation implicit feature vectors; The positional difference between the sequence of implicit feature vectors associated with the real-time pressure time series in the reactor and the sequence of implicit feature vectors associated with the real-time pressure time series in the reverse inhibition reactor is calculated to obtain the sequence of implicit feature vectors associated with the real-time pressure time series in the enhanced reactor.

5. The continuous method for producing silanimide according to claim 4, characterized in that: Calculating the semantic correlation coefficient between each implicit feature vector of the real-time pressure time series association in the reactor and the anchor center feature vector of the real-time pressure time series association in the reactor in the sequence of implicit feature vectors of the real-time pressure time series association in the reactor to obtain a sequence of semantic implicit correlation coefficients, including: Calculating the position mean vector of the sequence of implicit feature vectors associated with the real-time pressure time series in the reactor to determine the anchor center feature vector associated with the real-time pressure time series in the reactor; Calculate the positional dot product between the real-time pressure time series associated implicit feature vector in the i-th reactor and the anchor center feature vector associated with the real-time pressure time series in the reactor, and add the vectors obtained by the positional dot product element by element to obtain the local time series associated feature heap value of the network flow constrained by the anchor center; Calculate the element-by-element sum of the squares of the implicit feature vector of the real-time pressure timing association in the i-th reactor and the element-by-element sum of the squares of the anchor center feature vector of the real-time pressure timing association in the reactor, and calculate the square root of the sum of the two values ​​obtained by the element-by-element sum of the squares, and divide the local timing association feature heap value of the network traffic constrained by the anchor center by the value obtained by the square root to obtain the i-th semantic correlation coefficient.

6. The continuous method for producing silanimide according to claim 5, characterized in that: Normalizing the sequence of semantic anti-correlation coefficients to obtain a sequence of semantic anti-correlation energy coefficients includes: The sequence of semantic anti-correlation coefficients is normalized using a Softmax function to obtain a sequence of semantic anti-correlation energy coefficients.

7. The continuous method for producing silanimide according to claim 6, characterized in that: The second-order time series correlation feature extraction is performed on the sequence of the real-time pressure time series correlation implicit feature vectors in the enhanced reactor to obtain the real-time pressure time series correlation representation vector in the full time domain reactor, including: The sequence of the real-time pressure time series associated implicit feature vectors in the enhanced reactor is input into a second-order sequence encoder based on the Bi-LSTM model to obtain the real-time pressure time series associated representation vector in the full-time domain reactor.

8. The continuous method for producing silanimide according to claim 7, characterized in that: Based on the real-time pressure time series correlation representation vector in the full-time reactor, a control instruction is determined, wherein the control instruction is used to indicate whether to stop spraying the gaseous silicon tetrachloride, including: The real-time pressure time series correlation representation vector in the full-time reactor is input into the classifier-based spray controller to obtain the control instruction.

Images