WO2024098653A1 - 用于六氟磷酸锂制备的自动化采样分析系统及其方法 - Google Patents

用于六氟磷酸锂制备的自动化采样分析系统及其方法 Download PDF

Info

Publication number
WO2024098653A1
WO2024098653A1 PCT/CN2023/086450 CN2023086450W WO2024098653A1 WO 2024098653 A1 WO2024098653 A1 WO 2024098653A1 CN 2023086450 W CN2023086450 W CN 2023086450W WO 2024098653 A1 WO2024098653 A1 WO 2024098653A1
Authority
WO
WIPO (PCT)
Prior art keywords
gas
feature
gas chromatogram
denoised
feature map
Prior art date
Application number
PCT/CN2023/086450
Other languages
English (en)
French (fr)
Inventor
张志强
华杭州
吴远胜
陈东林
Original Assignee
福建省龙德新能源有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 福建省龙德新能源有限公司 filed Critical 福建省龙德新能源有限公司
Publication of WO2024098653A1 publication Critical patent/WO2024098653A1/zh

Links

Classifications

    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02E60/10—Energy storage using batteries

Definitions

  • the present application relates to the field of sampling and analysis of lithium hexafluorophosphate, and more specifically, to an automated sampling and analysis system and method for preparing lithium hexafluorophosphate.
  • Lithium-ion batteries have the characteristics of high energy density, high specific power, good cycle performance, no memory effect, and no pollution. They are currently widely used in electronic digital products and are also an ideal choice for future electric vehicle energy.
  • the battery electrolyte is composed of an organic solvent and a lithium salt electrolyte.
  • the commonly used lithium salt electrolytes include lithium perchlorate, lithium hexafluorophosphate, lithium tetrafluoroborate, etc.
  • lithium hexafluorophosphate has good conductivity and electrochemical stability and is currently the most widely used lithium salt electrolyte.
  • the embodiment of the present application provides an automated sampling and analysis system and method for the preparation of lithium hexafluorophosphate, which uses an artificial intelligence detection algorithm based on deep learning to extract multi-scale implicit feature distribution information of different sizes in the gas chromatogram of the sampled gas after noise reduction, and uses the attention mechanism to further strengthen the detection of small-scale phosphorus oxyfluoride content in the gas chromatogram and filter out the influence of useless interference features, so as to improve the accuracy of phosphorus oxyfluoride content detection.
  • the phosphorus oxyfluoride content can be intelligently and accurately detected, thereby improving the quality inspection accuracy of the prepared lithium hexafluorophosphate, so as to improve the quality of the electrolyte and the battery performance used in lithium-ion batteries.
  • an automated sampling and analysis system for the preparation of lithium hexafluorophosphate which includes: a sampling gas data acquisition module, used to obtain a gas chromatogram of the sampling gas; a denoising module, used to pass the gas chromatogram through a denoising module based on an automatic codec to obtain a denoised gas chromatogram; a chromatographic feature extraction module, used to pass the denoised gas chromatogram through a convolutional neural network model including multiple mixed convolutional layers to obtain a denoised gas chromatogram feature graph; a coding compensation module, used to perform defocusing fuzzy optimization of feature clustering on the denoised gas chromatogram feature graph to obtain an optimized denoised gas chromatogram feature graph; a feature distribution enhancement module, used to pass the optimized denoised gas chromatogram feature graph through a residual double attention mechanism module to obtain an enhanced gas chromatogram feature graph as a decoding feature graph; and an analysis result generation module, used to pass the decoded feature graph through
  • an automated sampling and analysis method for the preparation of lithium hexafluorophosphate includes: obtaining a gas chromatogram of the sampled gas; passing the gas chromatogram through a denoising module based on an automatic codec to obtain a denoised gas chromatogram; passing the denoised gas chromatogram through a convolutional neural network model comprising a plurality of mixed convolutional layers to obtain a denoised gas chromatogram characteristic graph; performing defocusing fuzzy optimization of feature clustering on the denoised gas chromatogram characteristic graph to obtain an optimized denoised gas chromatogram characteristic graph; passing the optimized denoised gas chromatogram characteristic graph through a residual dual attention mechanism module to obtain an enhanced gas chromatogram characteristic graph as a decoding characteristic graph; and passing the decoded characteristic graph through a decoder to obtain a decoding value, wherein the decoding value is used to represent the content value of phosphorus oxyfluoride in the sampled gas.
  • an electronic device comprising: a processor; and a memory, wherein computer program instructions are stored in the memory, and when the computer program instructions are executed by the processor, the processor executes the automated sampling and analysis method for the preparation of lithium hexafluorophosphate as described above.
  • a computer-readable medium on which computer program instructions are stored.
  • the processor executes the automated sampling and analysis method for the preparation of lithium hexafluorophosphate as described above.
  • the present application provides an automated sampling and analysis system and method for preparing lithium hexafluorophosphate, which uses an artificial intelligence detection algorithm based on deep learning to extract multi-scale implicit feature distribution information of different sizes in the gas chromatogram of the sampled gas after noise reduction, and uses the attention mechanism to further enhance the detection of small-scale phosphorus oxyfluoride content in the gas chromatogram and filter out the influence of useless interference features, so as to improve the accuracy of phosphorus oxyfluoride content detection.
  • the phosphorus oxyfluoride content can be intelligently and accurately detected, thereby improving the quality inspection accuracy of the prepared lithium hexafluorophosphate, so as to improve the quality of the electrolyte and the battery performance used in lithium-ion batteries.
  • FIG1 illustrates an application scenario diagram of an automated sampling and analysis system and method for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • FIG. 2 illustrates a block diagram of an automated sampling and analysis system for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • FIG3 illustrates a block diagram of a noise reduction module in an automated sampling and analysis system for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • FIG. 4 illustrates a block diagram of a feature enhancement module in an automated sampling and analysis system for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • FIG5 illustrates a block diagram of a spatial attention unit in an automated sampling and analysis system for lithium hexafluorophosphate preparation according to an embodiment of the present application.
  • FIG. 6 illustrates an automated sampling and analysis system for preparing lithium hexafluorophosphate according to an embodiment of the present application. Block diagram of the mid-channel attention unit.
  • FIG. 7 illustrates a flow chart of an automated sampling and analysis system for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • FIG8 is a schematic diagram illustrating a system architecture of an automated sampling and analysis system for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • FIG. 9 illustrates a block diagram of an electronic device according to an embodiment of the present application.
  • the battery electrolyte is composed of an organic solvent and a lithium salt electrolyte.
  • the commonly used lithium salt electrolytes include lithium perchlorate, lithium hexafluorophosphate, lithium tetrafluoroborate, etc.
  • lithium hexafluorophosphate has good conductivity and electrochemical stability, and is currently the most widely used lithium salt electrolyte.
  • the present application stores a lithium hexafluorophosphate solution sample in a sealed container, and reserves one-fifth to one-third of the volume in the sealed container to accommodate volatile gases. After a period of time, the volatile gas above the lithium hexafluorophosphate solution in the container is sampled, and then a gas chromatograph is used to quantitatively analyze the phosphorus oxyfluoride in the sampled gas components, and the quality of the lithium hexafluorophosphate solution is determined by the phosphorus oxyfluoride content. Therefore, in order to ensure the accuracy of the quality inspection of the prepared lithium hexafluorophosphate, it is necessary to accurately detect the phosphorus oxyfluoride content.
  • deep learning and neural networks have been widely used in computer vision, natural language processing, speech signal processing and other fields.
  • deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields.
  • an artificial intelligence detection algorithm based on deep learning is used to extract the multi-scale implicit feature distribution information of different sizes in the gas chromatogram of the sampled gas after noise reduction, and the attention mechanism is used to further strengthen the detection of small-scale phosphorus oxyfluoride content in the gas chromatogram and filter out the influence of useless interference features to improve the accuracy of phosphorus oxyfluoride content detection.
  • the phosphorus oxyfluoride content can be intelligently and accurately detected, thereby improving the quality inspection accuracy of the prepared lithium hexafluorophosphate, so as to improve the quality of the electrolyte and the battery performance used in lithium-ion batteries.
  • a gas chromatogram of the sampled gas is obtained, and The gas chromatogram is subjected to denoising processing in a denoising module based on an automatic codec to obtain a denoised gas chromatogram.
  • the gas chromatogram is subjected to denoising processing in a denoising module based on an automatic codec to obtain a denoised gas chromatogram.
  • the denoised gas chromatogram is processed by a mixed convolution layer to extract the multi-scale implicit correlation features of the denoised gas chromatogram, thereby obtaining a denoised gas chromatogram feature map.
  • the design of this module includes four branches in parallel, which are composed of a common convolution layer with a convolution kernel size of 3 ⁇ 3 and three dilated convolution layers with a convolution kernel size of 3 ⁇ 3.
  • the input feature map is operated respectively, and the expansion rates of the three branches of the dilated convolution are set to 2, 3, and 4 respectively.
  • Image information of different receptive domains can be obtained by setting different expansion rates, that is, feature maps of different scales can be obtained. While expanding the receptive field, downsampling loss information is avoided. Then, the four branch feature maps are fused to make the sampling more intensive, with both high-level features and no additional parameters.
  • the characteristic data of the gas chromatogram after noise reduction is further enhanced to obtain an enhanced gas chromatogram characteristic graph.
  • the denoised gas chromatogram feature map is passed through a residual dual attention mechanism module to obtain an enhanced gas chromatogram feature map as a decoding feature map.
  • a residual dual attention mechanism module to obtain an enhanced gas chromatogram feature map as a decoding feature map.
  • a channel attention mechanism and a spatial attention mechanism are introduced, and by introducing a residual structure, it is combined with the proposed dual attention network to construct a residual dual attention model.
  • This model combines spatial attention and channel attention in parallel, so that different types of effective information in the gas chromatogram are captured in large quantities, which can effectively enhance the feature recognition learning ability.
  • the task processing system is more focused on finding useful information in the input data that is significantly related to the current output, thereby improving the quality of the output, and the increasing attention module will bring continuous performance improvement.
  • the residual dual attention mechanism module uses a parallel combination of spatial attention and channel attention, so that different types of effective information in the denoised gas chromatographic characteristic graph are captured in large quantities, which can effectively enhance the feature recognition learning ability, thereby improving the detection of the content value of phosphorus trifluoride in the sampled gas.
  • the decoded characteristic graph is passed through a decoder to obtain a decoded value, and the decoded value is used to represent the content value of phosphorus oxyfluoride in the sampled gas. That is, the enhanced gas chromatography characteristic graph is used as the decoded characteristic graph for decoding regression to obtain a decoded value for representing the content value of phosphorus oxyfluoride in the sampled gas. In this way, the content of phosphorus oxyfluoride in the sampled gas can be accurately detected, thereby improving the quality inspection accuracy for the preparation of lithium hexafluorophosphate.
  • the hybrid convolution layer obtains multiple branch feature maps from the denoised gas chromatogram through branches with different expansion rates, and then fuses each branch feature map by point addition to finally obtain the denoised gas chromatogram feature map.
  • the applicant of the present application considers that since each branch of the hybrid convolution layer has its own expansion rate, the fusion of the multiple branch feature maps by point addition will conform to the Gaussian distribution in the natural state, that is, the local area with average correlation between the feature values of each point addition and average fusion effect has the highest probability density, while the local area with relatively high and relatively low correlation and relatively high and relatively low fusion effect has a lower probability density.
  • the obtained denoised gas chromatogram feature map may have a poor clustering effect on the global correlation relationship between the feature values, and the spatial attention mechanism and channel attention mechanism of the residual dual attention mechanism module will further reduce the clustering effect on the global correlation relationship by strengthening the attention weight of the local distribution, which leads to the poor representation of the key global feature distribution that affects the decoding result in the decoding feature map, affecting the decoding accuracy of the decoding feature map.
  • ⁇ and ⁇ are the mean and standard deviation of the feature set fi,j,k ⁇ F', respectively, and fi,j,k is the feature value of the (i,j,k)th position of the gas chromatogram characteristic graph after denoising.
  • the defocusing fuzzy optimization of the feature clustering is to compensate for the dependency similarity of the high-frequency distribution features following the Gaussian point distribution with respect to the homogenized representation of the overall feature distribution by indexing the focus stack representation used to estimate the clustering metric value based on statistical information, thereby avoiding the focus fuzziness of the overall feature distribution caused by the low dependency similarity.
  • the feature clustering effect of the denoised gas chromatographic feature graph is improved, and the decoding accuracy of the decoding feature graph obtained by the residual dual attention mechanism module of the denoised gas chromatographic feature graph is optimized.
  • the content of phosphorus oxyfluoride can be intelligently and accurately detected, thereby improving the quality inspection accuracy of the prepared lithium hexafluorophosphate, so as to improve the quality of the electrolyte and the battery performance used in lithium-ion batteries.
  • the present application provides an automated sampling and analysis system for the preparation of lithium hexafluorophosphate, which includes: a sampling gas data acquisition module, used to obtain a gas chromatogram of the sampling gas; a denoising module, used to pass the gas chromatogram through a denoising module based on an automatic codec to obtain a denoised gas chromatogram; a chromatographic feature extraction module, used to pass the denoised gas chromatogram through a convolutional neural network model including multiple mixed convolutional layers to obtain a denoised gas chromatogram feature graph; a coding compensation module, used to perform defocusing fuzzy optimization of feature clustering on the denoised gas chromatogram feature graph to obtain an optimized denoised gas chromatogram feature graph; a feature distribution enhancement module, used to pass the optimized denoised gas chromatogram through a convolutional neural network model including multiple mixed convolutional layers to obtain a denoised gas chromatogram feature graph;
  • the chromatographic characteristic graph is passed through a residual double attention mechanism module to obtain
  • FIG. 1 illustrates an application scenario diagram of an automated sampling and analysis system and method for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • a lithium hexafluorophosphate solution sample e.g., L as shown in FIG. 1
  • a sealed container e.g., A as shown in FIG. 1 , it should be understood that one-fifth to one-third of the volume is reserved in the sealed container to accommodate volatile gases
  • a gas sampling tube e.g., G as shown in FIG.
  • a gas chromatograph e.g., T as shown in FIG. 1
  • a gas chromatograph e.g., T as shown in FIG. 1
  • the collected gas chromatogram is input into a server (e.g., S as shown in FIG. 1 ) in which an automated sampling and analysis algorithm for preparing lithium hexafluorophosphate is deployed, wherein the server can process the gas chromatogram using the automated sampling and analysis algorithm for preparing lithium hexafluorophosphate to generate an analysis result for representing the content value of phosphorus trifluoride oxide in the sampled gas.
  • FIG2 illustrates a block diagram of an automated sampling and analysis system for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • the automated sampling and analysis system 100 for preparing lithium hexafluorophosphate according to the embodiment of the present application includes: a sampling gas data acquisition module 110, which is used to obtain a gas chromatogram of the sampling gas; a denoising module 120, which is used to pass the gas chromatogram through a denoising module based on an automatic codec to obtain a denoised gas chromatogram; a chromatographic feature extraction module 130, which is used to pass the denoised gas chromatogram through a convolutional neural network model including multiple mixed convolutional layers to obtain a denoised gas chromatogram feature graph; a coding compensation module 140, which is used to perform defocusing fuzzy optimization of feature clustering on the denoised gas chromatogram feature graph to obtain an optimized denoised gas chromatogram feature graph; a feature distribution enhancement module 150, which is used
  • the sampling gas data acquisition module 110 is used to obtain a gas chromatogram of the sampling gas.
  • the quality of lithium hexafluorophosphate directly affects the quality of the electrolyte and the battery performance used in lithium-ion batteries, the monitoring of the quality of lithium hexafluorophosphate is critical for the electrolyte.
  • most of the existing quality inspection schemes conduct experimental analysis on the prepared lithium hexafluorophosphate to compare it with the reaction results under the standard reference mass, so as to determine the quality of the prepared lithium hexafluorophosphate.
  • the present application stores the lithium hexafluorophosphate solution sample in a sealed container, and reserves one-fifth to one-third of the volume in the sealed container to accommodate volatile gases.
  • a gas sampling tube is used to sample the volatile gas above the lithium hexafluorophosphate solution in the container, and then a gas chromatograph is used to quantitatively analyze the phosphorus oxyfluoride in the sampled gas components, and the quality of the lithium hexafluorophosphate solution is determined by the phosphorus oxyfluoride content. Therefore, in order to ensure the accuracy of the quality inspection of the prepared lithium hexafluorophosphate, it is necessary to accurately detect the phosphorus oxyfluoride content.
  • the chromatogram refers to the image of the distribution of the detection signal of the separated components over time
  • the content of phosphorus oxyfluoride can be obtained from the gas chromatogram, but because there are other gases in the sampled gas, such as hydrogen fluoride, it is difficult to obtain an accurate content of phosphorus oxyfluoride through the gas chromatogram.
  • an artificial intelligence detection algorithm based on deep learning is used to extract the multi-scale implicit feature distribution information of different sizes in the gas chromatogram of the sampled gas after noise reduction, and the attention mechanism is used to further strengthen the detection of small-scale phosphorus oxyfluoride content in the gas chromatogram and filter out the influence of useless interference features to improve the accuracy of phosphorus oxyfluoride content detection.
  • the phosphorus oxyfluoride content can be intelligently and accurately detected, thereby improving the quality inspection accuracy of the prepared lithium hexafluorophosphate, so as to improve the quality of the electrolyte and the battery performance used in lithium-ion batteries.
  • the noise reduction module 120 is used to pass the gas chromatogram through a noise reduction module based on an automatic codec to obtain a gas chromatogram after noise reduction. It should be understood that, considering that in the process of collecting the gas chromatogram of the sampled gas by an actual gas chromatograph, the influence of environmental gas factors in the air and other gas factors generated by itself will result in a low accuracy in determining the content of phosphorus oxyfluoride in the gas chromatogram, thereby affecting the quality detection of lithium hexafluorophosphate. Therefore, in the technical solution of the present application, it is necessary to further use the noise reduction module of the automatic codec to perform noise reduction processing on the gas chromatogram to obtain a gas chromatogram after noise reduction.
  • the denoising module 120 includes: a gas chromatogram encoding unit 121, used to input the gas chromatogram into an encoder of the denoising module, wherein the encoder uses a convolution layer to perform explicit spatial encoding on the gas chromatogram to obtain a gas chromatogram characteristic graph; a gas chromatogram characteristic decoding unit 122, used to input the gas chromatogram characteristic graph into a decoder of the signal denoising module, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the gas chromatogram characteristic graph to obtain the denoised gas chromatogram.
  • the gas chromatography feature decoding unit includes: an image method unit, used to fill the gas chromatography feature map with zeros according to the step size of the convolution kernel of the encoder to obtain a zero-filled feature map; and a transposed convolution unit, used to convolve the zero-filled feature map using a transposed convolution kernel that is transposed of the convolution kernel of the encoder to generate the denoised gas chromatogram, wherein the denoised gas chromatogram is the same size as the gas chromatogram.
  • the chromatographic feature extraction module 130 is used to obtain the denoised gas chromatogram through a convolutional neural network model including multiple mixed convolutional layers.
  • Feature graph It should be understood that, considering that the traditional convolutional neural network uses a single convolution kernel, although it can extract local features, it cannot take into account both large-scale periodic features and small-scale timing features. Therefore, in the technical solution of the present application, the denoised gas chromatogram is processed through a mixed convolution layer to extract the multi-scale implicit correlation features of the denoised gas chromatogram, thereby obtaining a denoised gas chromatogram feature graph.
  • the chromatographic feature extraction module 130 is further used to: use each mixed convolution layer of the convolutional neural network model to respectively perform on the input data in the forward transfer of the layer: use a first convolution kernel with a first size to perform convolution encoding on the denoised gas chromatogram to obtain a first scale feature map; use a second convolution kernel with a first void ratio to perform convolution encoding on the denoised gas chromatogram to obtain a second scale feature map; use a third convolution kernel with a second void ratio to perform convolution encoding on the denoised gas chromatogram to obtain a third scale feature map; use a fourth convolution kernel with a third void ratio to perform convolution encoding on the denoised gas chromatogram to obtain A fourth-scale feature map, wherein the first convolution kernel, the second convolution kernel, the third convolution kernel and the fourth convolution kernel have the same size, and the second convolution kernel, the third convolution kernel and the fourth convolution kernel have
  • the design of this module includes four branches in parallel, which are composed of a normal convolution layer with a convolution kernel size of 3 ⁇ 3 and three dilated convolution layers with a convolution kernel size of 3 ⁇ 3.
  • the input feature map is operated respectively, and the expansion rates of the three branches of the dilated convolution are set to 2, 3, and 4 respectively.
  • image information of different receptive domains can be obtained, and feature maps of different scales can be obtained.
  • the downsampling loss information is avoided.
  • the four branch feature maps are fused to make the sampling more intensive, with both high-level features and no additional parameters.
  • the coding compensation module 140 is used to perform defocusing fuzzy optimization of feature clustering on the denoised gas chromatographic feature map to obtain an optimized denoised gas chromatographic feature map. It should be understood that in the technical solution of the present application, the hybrid convolution layer obtains multiple branch feature maps from the denoised gas chromatogram through branches with different expansion rates, and then fuses the branch feature maps by point addition to finally obtain the denoised gas chromatogram feature map.
  • the applicant of the present application takes into account that since each branch of the hybrid convolution layer has its own expansion rate, the fusion of the multiple branch feature maps by point addition will conform to the Gaussian distribution in the natural state, that is, the local area with average correlation between the feature values of each point addition and thus average fusion effect has the highest probability density, while the local area with relatively high and relatively low correlation and thus relatively high and relatively low fusion effect has a lower probability density.
  • the obtained denoised gas chromatogram feature map may have a poor clustering effect in the global correlation relationship between the feature values, and the spatial attention mechanism and channel attention mechanism of the residual dual attention mechanism module will strengthen the spatial attention mechanism and channel attention mechanism of the residual dual attention mechanism module.
  • the attention weights of the local distribution are used to further reduce the clustering effect on the global correlation relationship, which leads to poor representation of the key global feature distribution that affects the decoding result in the decoding feature graph, affecting the decoding accuracy of the decoding feature graph. Based on this, the defocused fuzzy optimization of feature clustering is performed on the denoised gas chromatographic feature graph.
  • the coding compensation module 140 is further used to: perform defocusing fuzzy optimization of feature clustering on the denoised gas chromatogram characteristic graph according to the following formula to obtain the optimized denoised gas chromatogram characteristic graph; wherein the formula is:
  • fi,j,k represents the characteristic value of the (i,j,k)th position of the gas chromatogram characteristic graph after denoising
  • ⁇ and ⁇ represent the mean and standard deviation of the characteristic value set of each position of the gas chromatogram characteristic graph after denoising, respectively.
  • the defocusing fuzzy optimization of the feature clustering is to compensate for the dependency similarity of the high-frequency distribution features following the Gaussian point distribution relative to the homogenized representation of the overall feature distribution by indexing the focus stack representation used to estimate the clustering metric value based on statistical information, thereby avoiding the focus fuzziness of the overall feature distribution caused by the low dependency similarity.
  • the feature clustering effect of the denoised gas chromatographic feature graph is improved, and the decoding accuracy of the decoding feature graph obtained by further passing the denoised gas chromatographic feature graph through the residual double attention mechanism module is optimized.
  • the feature distribution enhancement module 150 is used to pass the optimized denoised gas chromatographic feature graph through the residual double attention mechanism module to obtain an enhanced gas chromatographic feature graph as a decoding feature graph.
  • the optimized noise-reduced gas chromatogram characteristic graph is further enhanced with characteristic data to obtain an enhanced gas chromatogram characteristic graph.
  • the optimized denoised gas chromatogram feature map is passed through a residual dual attention mechanism module to obtain an enhanced gas chromatogram feature map as a decoding feature map.
  • the network will obtain partial feature information, but will not automatically distinguish the detailed information between high and low frequencies and the differences between the features of each category.
  • the network's ability to selectively use features is limited. In view of the fact that the attention mechanism can select the focus position, a more discriminative feature representation is produced, and the features after the attention module is added will undergo adaptive changes as the network deepens.
  • a channel attention mechanism and a spatial attention mechanism are introduced, and by introducing a residual structure, it is combined with the proposed dual attention network to construct a residual dual attention model.
  • This model combines spatial attention and channel attention in parallel, so that different types of effective information in the gas chromatogram are captured in large quantities, which can effectively enhance the feature recognition learning ability.
  • the task processing system is more focused on finding useful information in the input data that is significantly related to the current output, thereby improving the quality of the output, and the increasing attention module will bring continuous performance improvement.
  • the force mechanism module uses a combination of spatial attention and channel attention in parallel, so that different types of effective information in the gas chromatographic characteristic graph after noise reduction are captured in large quantities, which can effectively enhance the feature recognition learning ability and thus improve the detection of the content value of phosphorus trifluoride in the sampled gas.
  • the feature distribution enhancement module 150 includes: a spatial attention unit 151, which is used to input the optimized denoised gas chromatography feature map into the spatial attention module of the residual dual attention mechanism module to obtain a spatial attention map; a channel attention unit 152, which is used to input the optimized denoised gas chromatography feature map into the channel attention module of the residual dual attention mechanism module to obtain a channel attention map; an attention fusion unit 153, which is used to fuse the spatial attention map and the channel attention map to obtain a fused attention map; an activation unit 154, which is used to input the fused attention map into a Sigmoid activation function for activation to obtain a fused attention feature map; an attention application unit 155, which is used to calculate the position point multiplication of the fused attention feature map and the optimized denoised gas chromatography feature map to obtain a weighted feature map; and a residual fusion unit 156, which is used to fuse the weighted feature map and the optimized denoised gas chromatography feature
  • the spatial attention unit 151 includes: a convolutional encoding subunit 1511, used to use the convolutional layer of the spatial attention module of the residual dual attention mechanism module to perform convolution encoding on the depth-enhanced gas chromatography feature map to obtain a convolutional feature map; a probabilistic subunit 1512, used to pass the spatial attention map through a Softmax function to obtain a spatial attention score map; and a spatial attention application subunit 1513, used to perform positional point multiplication of the spatial attention score map and the depth-enhanced gas chromatography feature map to obtain a spatial attention map.
  • a convolutional encoding subunit 1511 used to use the convolutional layer of the spatial attention module of the residual dual attention mechanism module to perform convolution encoding on the depth-enhanced gas chromatography feature map to obtain a convolutional feature map
  • a probabilistic subunit 1512 used to pass the spatial attention map through a Softmax function to obtain a spatial attention score map
  • the channel attention unit 152 includes: a global mean pooling subunit 1521, used to perform global mean pooling on the depth enhanced gas chromatography feature map along the channel dimension to obtain a channel feature vector; a normalization subunit 1522, used to pass the channel feature vector through a Softmax function to obtain a normalized channel feature vector; and a channel attention application subunit 1523, used to weight the feature matrix of the depth enhanced gas chromatography feature map along the channel dimension using the eigenvalues of each position in the normalized channel feature vector as weights to obtain a channel attention map.
  • the analysis result generation module 160 is used to pass the decoding characteristic graph through a decoder to obtain a decoding value, and the decoding value is used to represent the content value of phosphorus oxyfluoride in the sampled gas. That is, the enhanced gas chromatography characteristic graph is imported as the decoding characteristic graph for decoding regression to obtain a decoding value used to represent the content value of phosphorus oxyfluoride in the sampled gas, so that the content of phosphorus oxyfluoride in the sampled gas can be accurately detected, thereby improving the quality inspection accuracy for the preparation of lithium hexafluorophosphate.
  • the analysis result generation module is further used to: use the decoder to perform decoding regression on the decoding feature map to obtain the decoding value using the following formula, wherein the formula is Among them, X is the decoded feature map, Y is the decoded value, and W is the weight matrix. Represents matrix multiplication.
  • the automated sampling and analysis system for preparing lithium hexafluorophosphate uses an artificial intelligence detection algorithm based on deep learning to extract the gas chromatogram of the sampled gas after noise reduction.
  • the multi-scale implicit feature distribution information of different sizes in the spectrum is obtained, and the attention mechanism is used to further enhance the detection of small-scale phosphorus oxyfluoride content in the gas chromatogram and filter out the influence of useless interference features to improve the accuracy of phosphorus oxyfluoride content detection.
  • the phosphorus oxyfluoride content can be detected intelligently and accurately, thereby improving the quality inspection accuracy of the prepared lithium hexafluorophosphate, thereby improving the quality of the electrolyte and the battery performance used in lithium-ion batteries.
  • the automated sampling and analysis system 100 for the preparation of lithium hexafluorophosphate can be implemented in various terminal devices, such as a server deployed with a quality inspection algorithm for hexafluorobutadiene.
  • the automated sampling and analysis system 100 for the preparation of lithium hexafluorophosphate can be integrated into the terminal device as a software module and/or a hardware module.
  • the automated sampling and analysis system 100 for the preparation of lithium hexafluorophosphate can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the automated sampling and analysis system 100 for the preparation of lithium hexafluorophosphate can also be one of the many hardware modules of the terminal device.
  • the automated sampling and analysis system 100 for the preparation of lithium hexafluorophosphate and the terminal device may also be separate devices, and the automated sampling and analysis system 100 for the preparation of lithium hexafluorophosphate may be connected to the terminal device via a wired and/or wireless network and transmit interactive information in accordance with an agreed data format.
  • FIG7 illustrates a flow chart of an automated sampling and analysis method for preparing lithium hexafluorophosphate according to an embodiment of the present application.
  • the automated sampling and analysis method for preparing lithium hexafluorophosphate according to an embodiment of the present application includes: S110, obtaining a gas chromatogram of the sampled gas; S120, passing the gas chromatogram through a denoising module based on an automatic codec to obtain a denoised gas chromatogram; S130, passing the denoised gas chromatogram through a convolutional neural network model comprising multiple mixed convolutional layers to obtain a denoised gas chromatogram feature graph; S140, performing feature clustering defocusing fuzzy optimization on the denoised gas chromatogram feature graph to obtain an optimized denoised gas chromatogram feature graph; S150, passing the optimized denoised gas chromatogram feature graph through a residual dual attention mechanism module to obtain an enhanced gas chromatogram feature graph as a decoding feature graph; and, S160, passing the
  • FIG8 illustrates a schematic diagram of the system architecture of the automated sampling and analysis method for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • a gas chromatogram of the sampled gas is obtained, and the gas chromatogram is passed through a denoising module based on an automatic codec to obtain a denoised gas chromatogram.
  • the denoised gas chromatogram is passed through a convolutional neural network model comprising a plurality of mixed convolutional layers to obtain a denoised gas chromatogram feature graph, and the denoised gas chromatogram feature graph is subjected to feature clustering defocusing fuzzy optimization to obtain an optimized denoised gas chromatogram feature graph.
  • the optimized denoised gas chromatogram feature graph is passed through a residual dual attention mechanism module to obtain an enhanced gas chromatogram feature graph as a decoding feature graph.
  • the decoded feature graph is passed through a decoder to obtain a decoded value, which is used to represent the content value of phosphorus oxyfluoride in the sampled gas.
  • Fig. 9 illustrates a block diagram of an electronic device according to an embodiment of the present application.
  • the electronic device 10 includes one or more processors 11 and a memory 12 .
  • the processor 11 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and/or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.
  • CPU central processing unit
  • the memory 12 may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory.
  • the volatile memory may include, for example, random access memory (RAM) and/or cache memory (cache), etc.
  • the non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.
  • One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may run the program instructions to implement the automated sampling analysis and/or other desired functions for the preparation of lithium hexafluorophosphate in the various embodiments of the present application described above.
  • Various contents such as gas chromatograms of sampled gas may also be stored in the computer-readable storage medium.
  • the electronic device 10 may further include: an input device 13 and an output device 14, and these components are interconnected via a bus system and/or other forms of connection mechanisms (not shown).
  • the input device 13 may include, for example, a keyboard, a mouse, etc.
  • the output device 14 can output various information to the outside, including classification results, etc.
  • the output device 14 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.
  • an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the automated sampling and analysis method for the preparation of lithium hexafluorophosphate according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.
  • the computer program product may be written in any combination of one or more programming languages to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages.
  • the program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.
  • the embodiment of the present application may also be a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the processor executes the method for preparing hexafluorophosphorus according to various embodiments of the present application described in the above “Exemplary Methods” section of the present specification. Steps of automated sampling and analysis method for lithium ion preparation.
  • the computer readable storage medium can adopt any combination of one or more readable media.
  • the readable medium can be a readable signal medium or a readable storage medium.
  • the readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above.
  • readable storage media include: an electrical connection with one or more wires, a portable 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.
  • RAM random access memory
  • ROM read-only memory
  • EPROM or flash memory erasable programmable read-only memory
  • CD-ROM compact disk read-only memory
  • magnetic storage device or any suitable combination of the above.
  • each component or each step can be decomposed and/or recombined.
  • These decompositions and/or recombinations should be regarded as equivalent schemes of the present application.
  • the above description of the disclosed aspects is provided to enable any technician in the field to make or use the present application.
  • Various modifications to these aspects are very obvious to those skilled in the art, and the general principles defined here can be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown here, but according to the widest scope consistent with the principles disclosed here and novel features.

Landscapes

  • Sampling And Sample Adjustment (AREA)
  • Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)

Abstract

一种用于六氟磷酸锂制备的自动化采样分析系统及其方法，涉及六氟磷酸锂的采样分析领域。其采用基于深度学习的人工智能检测算法来提取出降噪后的抽样气体气相色谱图中不同尺寸大小的多尺度隐含特征分布信息，并利用注意力机制来进一步加强对于气相色谱图中的小尺度三氟氧化磷含量检测而滤除无用的干扰特征的影响，以提高三氟氧化磷含量检测的精准度，通过这样的方式，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。

Description

用于六氟磷酸锂制备的自动化采样分析系统及其方法 技术领域
本申请涉及六氟磷酸锂的采样分析领域，且更为具体地，涉及一种用于六氟磷酸锂制备的自动化采样分析系统及其方法。
背景技术
锂离子电池具有能量密度高、比功率大、循环性能好、无记忆效应、无污染等特点，目前已广泛应用于电子数码产品中，同时也是未来电动汽车能源的理想选择。
电池电解液由有机溶剂和锂盐电解质组成，目前常用的锂盐电解质有高氯酸锂、六氟磷酸锂、四氟硼酸锂等，其中六氟磷酸锂具有良好的导电性和电化学稳定性，是目前应用范围最广的锂盐电解质。
由于六氟磷酸锂的质量直接影响到电解液的品质以及应用于锂离子电池中的电池性能，因此，六氟磷酸锂品质的监测对于电解液十分关键。然而，现有的质检方案大多数都是将制备后的六氟磷酸锂进行实验分析来与标准参考质量下的反应结果进行比对，以此来确定制备后的六氟磷酸锂质量。这样不仅会浪费大量的原材料和资源，而且在实验过程中会存在较多的变量因素，难以控制反应的结果，进而对于六氟磷酸锂质检的准确度难以保证。
因此，期待一种优化的用于六氟磷酸锂制备的自动化采样分析系统。
发明内容
为了解决上述技术问题，提出了本申请。本申请的实施例提供了一种用于六氟磷酸锂制备的自动化采样分析系统及其方法，采用基于深度学习的人工智能检测算法来提取出降噪后的抽样气体气相色谱图中不同尺寸大小的多尺度隐含特征分布信息，并利用注意力机制来进一步加强对于气相色谱图中的小尺度三氟氧化磷含量检测而滤除无用的干扰特征的影响，以提高三氟氧化磷含量检测的精准度。这样，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。
根据本申请的一个方面，提供了一种用于六氟磷酸锂制备的自动化采样分析系统，其包括：抽样气体数据采集模块，用于获取抽样气体的气相色谱图；降噪模块，用于将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图；色谱特征提取模块，用于将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图；编码补偿模块，用于对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图；特征分布增强模块，用于将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图；以及分析结果生成模块，用于将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
根据本申请的另一方面，提供了一种用于六氟磷酸锂制备的自动化采样分析方法，其包括：获取抽样气体的气相色谱图；将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图；将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图；对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图；将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图；以及将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
根据本申请的再一方面，提供了一种电子设备，包括：处理器；以及，存储器，在所述存储器中存储有计算机程序指令，所述计算机程序指令在被所述处理器运行时使得所述处理器执行如上所述的用于六氟磷酸锂制备的自动化采样分析方法。
根据本申请的又一方面，提供了一种计算机可读介质，其上存储有计算机程序指令，所述计算机程序指令在被处理器运行时使得所述处理器执行如上所述的用于六氟磷酸锂制备的自动化采样分析方法。
与现有技术相比，本申请提供的一种用于六氟磷酸锂制备的自动化采样分析系统及其方法，采用基于深度学习的人工智能检测算法来提取出降噪后的抽样气体气相色谱图中不同尺寸大小的多尺度隐含特征分布信息，并利用注意力机制来进一步加强对于气相色谱图中的小尺度三氟氧化磷含量检测而滤除无用的干扰特征的影响，以提高三氟氧化磷含量检测的精准度。这样，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。
附图说明
通过结合附图对本申请实施例进行更详细的描述，本申请的上述以及其他目的、特征和优势将变得更加明显。附图用来提供对本申请实施例的进一步理解，并且构成说明书的一部分，与本申请实施例一起用于解释本申请，并不构成对本申请的限制。在附图中，相同的参考标号通常代表相同部件或步骤。
图1图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统及其方法的应用场景图。
图2图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统的框图示意图。
图3图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统中降噪模块的框图。
图4图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统中特征增强模块的框图。
图5图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统中空间注意力单元的框图。
图6图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统 中通道注意力单元的框图。
图7图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统的流程图。
图8图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统的系统架构的示意图。
图9图示了根据本申请实施例的电子设备的框图。
具体实施方式
下面，将参考附图详细地描述根据本申请的示例实施例。显然，所描述的实施例仅仅是本申请的一部分实施例，而不是本申请的全部实施例，应理解，本申请不受这里描述的示例实施例的限制。
场景概述：如上所述，电池电解液由有机溶剂和锂盐电解质组成，目前常用的锂盐电解质有高氯酸锂、六氟磷酸锂、四氟硼酸锂等，其中六氟磷酸锂具有良好的导电性和电化学稳定性，是目前应用范围最广的锂盐电解质。
由于六氟磷酸锂的质量直接影响到电解液的品质以及应用于锂离子电池中的电池性能，因此，六氟磷酸锂品质的监测对于电解液十分关键。然而，现有的质检方案大多数都是将制备后的六氟磷酸锂进行实验分析来与标准参考质量下的反应结果进行比对，以此来确定制备后的六氟磷酸锂质量。这样不仅会浪费大量的原材料和资源，而且在实验过程中会存在较多的变量因素，难以控制反应的结果，进而对于六氟磷酸锂质检的准确度难以保证。因此，期待一种优化的用于六氟磷酸锂制备的自动化采样分析系统。
针对上述技术问题，本申请通过将六氟磷酸锂溶液样品存储于密封的容器中，并使密封容器中预留五分之一到三分之一的体积以容纳挥发性气体，放置一段时间后，对容器内六氟磷酸锂溶液上方的挥发性气体进行抽样，再采用气相色谱仪对于抽样气体成份中的三氟氧化磷进行定量分析，以三氟氧化磷含量来判定六氟磷酸锂溶液的品质。因此，为了能够保证制备后的六氟磷酸锂的质检准确度，需要对于三氟氧化磷含量进行准确地检测。
目前，深度学习以及神经网络已经广泛应用于计算机视觉、自然语言处理、语音信号处理等领域。此外，深度学习以及神经网络在图像分类、物体检测、语义分割、文本翻译等领域，也展现出了接近甚至超越人类的水平。
近年来，深度学习以及神经网络的发展为三氟氧化磷含量的智能检测提供了新的解决思路和方案。
具体地，在本申请的技术方案中，采用基于深度学习的人工智能检测算法来提取出降噪后的抽样气体气相色谱图中不同尺寸大小的多尺度隐含特征分布信息，并利用注意力机制来进一步加强对于气相色谱图中的小尺度三氟氧化磷含量检测而滤除无用的干扰特征的影响，以提高三氟氧化磷含量检测的精准度。这样，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。
具体地，在本申请的技术方案中，首先，获取抽样气体的气相色谱图，并将 所述气相色谱图通过基于自动编解码器的降噪模块中进行降噪处理以得到降噪后气相色谱图。应可以理解，考虑到在实际的气相色谱仪对于抽样气体的气相色谱图进行采集的过程中，会由于空气中的环境气体因素以及自身产生的其他气体因素的影响导致对于气相色谱图中三氟氧化磷含量的确定准确性较低，进而影响六氟磷酸锂的品质检测。因此，在本申请的技术方案中，需要进一步使用自动编解码器的降噪模块对于所述气相色谱图进行降噪处理以得到降噪后气相色谱图。
然后，考虑到所述三氟氧化磷对于所述抽样气体来说是小尺寸的，因此所述三氟氧化磷在所述气相色谱图中也属于小尺度特征，因此，在本申请的技术方案中，选择将所述降噪后气相色谱图通过混合卷积层中进行处理，以提取出所述降噪后气相色谱图的多尺度隐含关联特征，从而得到降噪后气相色谱特征图。也就是，在本申请的一个具体示例中，在所述混合卷积层(mixed convolution layer，MCL)中，此模块的设计包括并联的四个分支，由一个卷积核大小为3×3的普通卷积层以及三个卷积核大小为3×3的空洞卷积层构成，分别对输入特征图进行操作，将空洞卷积三个分支的扩张率分别设置为2、3、4，通过不同扩张率的设置可获得不同感受域的图像信息，即可得到不同尺度的特征图，在扩大感受野的同时，又避免了下采样损失信息，接着将4个分支特征图进行融合，使得采样更为密集，既拥有了高层特征，也没有增加额外的参数量。
进一步地，考虑到所述三氟氧化磷属于小尺寸对象(即，其在整个抽样气体的气相色谱图中所占比例较小)，其次，所述抽样气体具有各种气体类型，这会对所述三氟氧化磷的目标检测造成干扰。因此，在本申请的技术方案中，进一步对所述降噪后气相色谱特征图进行特征数据增强以得到增强气相色谱特征图。
具体地，将所述降噪后气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图。应可以理解，网络经过一系列卷积之后，会得到部分特征信息，但不会自动区分高低频间的详细信息与各个类别特征间的差异性，网络选择性地使用特征的能力有限，鉴于注意力机制能够选择聚焦位置，产生更具分辨性的特征表示，且加入注意力模块后的特征会随着网络的加深产生适应性的改变。因此，在本申请的技术方案中，引入通道注意力机制和空间注意力机制，并通过引入残差结构，将其与提出的双注意力网络相结合来构造残差双注意力模型，此模型将空间注意力和通道注意力并行组合，使得所述气相色谱图中的不同类型的有效信息被大量捕捉到，可有效增强特征辨别学习能力，在网络训练过程中，任务处理系统更专注于找到输入数据中显著的与当前输出相关的有用信息，从而提高输出的质量，且渐增的注意力模块将带来持续的性能提升。应可以理解，所述残差双注意力机制模块通过使用空间注意力和通道注意力并行的组合，使得所述降噪后气相色谱特征图中不同类型的有效信息被大量捕捉到，可有效增强特征辨别学习能力，进而提高对于所述抽样气体中三氟氧化磷的含量值检测。
然后，再将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。也就是，将所述增强气相色谱特征图作为所述解码特征图来进行解码回归，以得到用于表示所述抽样气体中三氟氧化磷的含量值的解码值。这样，能够对于所述抽样气体中的三氟氧化磷的含量进行准确检测，进而提高对于制备六氟磷酸锂的质检准确度。
特别地，在本申请的技术方案中，所述混合卷积层通过具有不同扩张率的各个分支从所述降噪后气相色谱图得到多个分支特征图，再通过点加的方式融合各个分支特征图来最终得到降噪后气相色谱特征图。这里，本申请的申请人考虑到由于所述混合卷积层的各个分支具有各自的扩张率，因此所述多个分支特征图通过点加的方式进行的融合将符合自然状态下的高斯分布，即各点加的特征值之间具有平均关联度从而具有平均融合效果的局部具有最高的概率密度，而具有相对高和相对低的关联度从而具有相对高和相对低的融合效果的局部具有较低的概率密度。由此，所获得的所述降噪后气相色谱特征图可能存在各个特征值之间在全局关联关系上的聚类效果不良的情况，而残差双注意力机制模块的空间注意力机制和通道注意力机制又会通过强化对局部分布的注意力权值来进一步降低全局关联关系上的聚类效果，这就导致所述解码特征图中影响解码结果的关键全局特征分布的表示性差，影响所述解码特征图的解码准确性。
基于此，对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化，表示为：
μ和δ分别是特征集合fi,j,k∈F'的均值和标准差，且fi,j,k是所述降噪后气相色谱特征图的第(i,j,k)位置的特征值。
这里，所述特征聚类的去聚焦模糊优化通过将用于估计聚类度量值的聚焦堆栈表示进行基于统计信息的特征聚类索引，来补偿遵循高斯点分布的高频分布特征相对于整体特征分布的均一化表示的依赖相似度，从而避免由于该依赖相似度低而引起整体特征分布的聚焦模糊，这样，就提升了所述降噪后气相色谱特征图的特征聚类效果，优化了所述降噪后气相色谱特征图进一步通过所述残差双注意力机制模块所得到的解码特征图的解码准确性。这样，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。基于此，本申请提供了一种用于六氟磷酸锂制备的自动化采样分析系统，其包括：抽样气体数据采集模块，用于获取抽样气体的气相色谱图；降噪模块，用于将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图；色谱特征提取模块，用于将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图；编码补偿模块，用于对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图；特征分布增强模块，用于将所述优化降噪气相 色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图；以及，分析结果生成模块，用于将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
图1图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统及其方法的应用场景图。如图1所示，在该应用场景中，将六氟磷酸锂溶液样品(例如，图1中所示意的L)存储于密封容器(例如，图1中所示意的A，应可以理解，密封容器中预留五分之一到三分之一的体积以容纳挥发性气体)中，放置一段时间后，使用采气管(例如，如图1中所示意的G)对容器内六氟磷酸锂溶液上方的挥发性气体进行抽样，再使用气相色谱仪(例如，如图1中所示意的T)采集抽样气体(例如，如图1中所示意的P)的气相色谱图。然后，将采集的所述气相色谱图输入至部署有用于六氟磷酸锂制备的自动化采样分析算法的服务器中(例如，图1中所示意的S)，其中，所述服务器能够使用所述用于六氟磷酸锂制备的自动化采样分析算法对所述气相色谱图进行处理以生成用于表示所述抽样气体中三氟氧化磷的含量值的分析结果。
在介绍了本申请的基本原理之后，下面将参考附图来具体介绍本申请的各种非限制性实施例。
示例性系统：图2图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析系统的框图示意图。如图2所示，根据本申请实施例的所述用于六氟磷酸锂制备的自动化采样分析系统100，包括：抽样气体数据采集模块110，用于获取抽样气体的气相色谱图；降噪模块120，用于将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图；色谱特征提取模块130，用于将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图；编码补偿模块140，用于对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图；特征分布增强模块150，用于将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图；以及，分析结果生成模块160，用于将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
在本申请实施例中，所述抽样气体数据采集模块110，用于获取抽样气体的气相色谱图。如上所述，由于六氟磷酸锂的质量直接影响到电解液的品质以及应用于锂离子电池中的电池性能，因此，六氟磷酸锂品质的监测对于电解液十分关键。然而，现有的质检方案大多数都是将制备后的六氟磷酸锂进行实验分析来与标准参考质量下的反应结果进行比对，以此来确定制备后的六氟磷酸锂质量。这样不仅会浪费大量的原材料和资源，而且在实验过程中会存在较多的变量因素，难以控制反应的结果，进而对于六氟磷酸锂质检的准确度难以保证。同时，考虑到六氟磷酸锂溶液中存在水时，会产生挥发性的杂质三氟氧化磷，通过对六氟磷酸锂溶液挥发气体中的三氟氧化磷含量检测， 就能反向推测出六氟磷酸锂溶液生成过程中杂质水分的介入程度，从而可以对六氟磷酸锂进行质检。
因此，本申请通过将六氟磷酸锂溶液样品存储于密封的容器中，并使密封容器中预留五分之一到三分之一的体积以容纳挥发性气体，放置一段时间后，使用采气管对容器内六氟磷酸锂溶液上方的挥发性气体进行抽样，再使用气相色谱仪对于抽样气体成份中的三氟氧化磷进行定量分析，以三氟氧化磷含量来判定六氟磷酸锂溶液的品质。因此，为了能够保证制备后的六氟磷酸锂的质检准确度，需要对于三氟氧化磷含量进行准确地检测。同时，考虑到由于色谱图是指被分离组分的检测信号随时间分布的图象，因此，从气相色谱图可以得出三氟氧化磷的含量，但是因为在抽样气体中还存在其他的气体，如氟化氢，因此，很难通过气相色谱图得到精准的三氟氧化磷的含量。
基于此，在本申请的技术方案中，采用基于深度学习的人工智能检测算法来提取出降噪后的抽样气体气相色谱图中不同尺寸大小的多尺度隐含特征分布信息，并利用注意力机制来进一步加强对于气相色谱图中的小尺度三氟氧化磷含量检测而滤除无用的干扰特征的影响，以提高三氟氧化磷含量检测的精准度。这样，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。
在本申请实施例中，所述降噪模块120，用于将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图。应可以理解，考虑到在实际的气相色谱仪对于抽样气体的气相色谱图进行采集的过程中，会由于空气中的环境气体因素以及自身产生的其他气体因素的影响导致对于气相色谱图中三氟氧化磷含量的确定准确性较低，进而影响六氟磷酸锂的品质检测。因此，在本申请的技术方案中，需要进一步使用自动编解码器的降噪模块对于所述气相色谱图进行降噪处理以得到降噪后气相色谱图。
在本申请一个具体的实施例中，所述降噪模块120，包括：气相色谱图编码单元121，用于将所述气相色谱图输入所述降噪模块的编码器，其中，所述编码器使用卷积层对所述气相色谱图进行显式空间编码以得到气相色谱特征图；气相色谱特征解码单元122，用于将所述气相色谱特征图输入所述信号降噪模块的解码器，其中，所述解码器使用反卷积层对所述气相色谱特征图进行反卷积处理以得到所述降噪后气相色谱图。
更为具体的，在本申请的一个实施例中，所述气相色谱特征解码单元，包括：图像方法单元，用于根据所述编码器的卷积核的步长对所述气相色谱特征图进行补零以得到补零特征图；以及，转置卷积单元，用于使用与所述编码器的卷积核互为转置的转置卷积核对所述补零特征图进行卷积处理以生成所述降噪后气相色谱图，其中，所述降噪后气相色谱图与所述气相色谱图尺寸相同。
在本申请实施例中，所述色谱特征提取模块130，用于将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱 特征图。应可以理解，考虑传统的卷积神经网络使用单一卷积核虽然能提取到局部特征，但无法无法兼顾大尺度的周期特征与小尺度的时序特征，因此，在本申请的技术方案中，选择将所述降噪后气相色谱图通过混合卷积层中进行处理，以提取出所述降噪后气相色谱图的多尺度隐含关联特征，从而得到降噪后气相色谱特征图。
在本申请一个具体的实施例中，所述色谱特征提取模块130，进一步用于：使用所述卷积神经网络模型的各个混合卷积层在层的正向传递中分别对输入数据进行：使用具有第一尺寸的第一卷积核对所述降噪后气相色谱图进行卷积编码以得到第一尺度特征图；使用具有第一空洞率的第二卷积核对所述降噪后气相色谱图进行卷积编码以得到第二尺度特征图；使用具有第二空洞率的第三卷积核对所述降噪后气相色谱图进行卷积编码以得到第三尺度特征图；使用具有第三空洞率的第四卷积核对所述降噪后气相色谱图进行卷积编码以得到第四尺度特征图，其中，所述第一卷积核、所述第二卷积核、所述第三卷积核和所述第四卷积核具有相同的尺寸，且所述第二卷积核、所述第三卷积核和所述第四卷积核具有不同的空洞率；将所述第一尺度特征图、所述第二尺度特征图、所述第三尺度特征图和所述第四尺度特征图进行沿通道维度的聚合以得到聚合特征图；对所述聚合特征图进行池化处理以生成池化特征图；以及，对所述池化特征图进行激活处理以生成激活特征图；其中，所述包含多个混合卷积层的卷积神经网络模型的最后一层的输出为所述降噪后气相色谱特征图。
在本申请的一个具体示例中，在所述混合卷积层(mixed convolution layer，MCL)中，此模块的设计包括并联的四个分支，由一个卷积核大小为3×3的普通卷积层以及三个卷积核大小为3×3的空洞卷积层构成，分别对输入特征图进行操作，将空洞卷积三个分支的扩张率分别设置为2、3、4，通过不同扩张率的设置可获得不同感受域的图像信息，即可得到不同尺度的特征图，在扩大感受野的同时，又避免了下采样损失信息，接着将4个分支特征图进行融合，使得采样更为密集，既拥有了高层特征，也没有增加额外的参数量。在本申请实施例中，所述编码补偿模块140，用于对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图。应可以理解，在本申请的技术方案中，所述混合卷积层通过具有不同扩张率的各个分支从所述降噪后气相色谱图得到多个分支特征图，再通过点加的方式融合各个分支特征图来最终得到降噪后气相色谱特征图。这里，本申请的申请人考虑到由于所述混合卷积层的各个分支具有各自的扩张率，因此所述多个分支特征图通过点加的方式进行的融合将符合自然状态下的高斯分布，即各点加的特征值之间具有平均关联度从而具有平均融合效果的局部具有最高的概率密度，而具有相对高和相对低的关联度从而具有相对高和相对低的融合效果的局部具有较低的概率密度。由此，所获得的所述降噪后气相色谱特征图可能存在各个特征值之间在全局关联关系上的聚类效果不良的情况，而残差双注意力机制模块的空间注意力机制和通道注意力机制又会通过强化 对局部分布的注意力权值来进一步降低全局关联关系上的聚类效果，这就导致所述解码特征图中影响解码结果的关键全局特征分布的表示性差，影响所述解码特征图的解码准确性。基于此，对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化。
在本申请一个具体的实施例中，所述编码补偿模块140，进一步用于：以如下公式对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到所述优化降噪气相色谱特征图；其中，所述公式为：
其中fi,j,k表示所述降噪后气相色谱特征图的第(i,j,k)位置的特征值，μ和δ分别表示所述降噪后气相色谱特征图的各个位置的特征值集合的均值和标准差。
这里，所述特征聚类的去聚焦模糊优化通过将用于估计聚类度量值的聚焦堆栈表示进行基于统计信息的特征聚类索引，来补偿遵循高斯点分布的高频分布特征相对于整体特征分布的均一化表示的依赖相似度，从而避免由于该依赖相似度低而引起整体特征分布的聚焦模糊，这样，就提升了所述降噪后气相色谱特征图的特征聚类效果，优化了所述降噪后气相色谱特征图进一步通过所述残差双注意力机制模块所得到的解码特征图的解码准确性。这样，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。在本申请实施例中，所述特征分布增强模块150，用于将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图。应可以理解，考虑到所述三氟氧化磷属于小尺寸对象(即，其在整个抽样气体的气相色谱图中所占比例较小)，其次，所述抽样气体具有各种气体类型，这会对所述三氟氧化磷的目标检测造成干扰。因此，在本申请的技术方案中，进一步对所述优化降噪气相色谱特征图进行特征数据增强以得到增强气相色谱特征图。
具体地，将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图。应可以理解，网络经过一系列卷积之后，会得到部分特征信息，但不会自动区分高低频间的详细信息与各个类别特征间的差异性，网络选择性地使用特征的能力有限，鉴于注意力机制能够选择聚焦位置，产生更具分辨性的特征表示，且加入注意力模块后的特征会随着网络的加深产生适应性的改变。因此，在本申请的技术方案中，引入通道注意力机制和空间注意力机制，并通过引入残差结构，将其与提出的双注意力网络相结合来构造残差双注意力模型，此模型将空间注意力和通道注意力并行组合，使得所述气相色谱图中的不同类型的有效信息被大量捕捉到，可有效增强特征辨别学习能力，在网络训练过程中，任务处理系统更专注于找到输入数据中显著的与当前输出相关的有用信息，从而提高输出的质量，且渐增的注意力模块将带来持续的性能提升。应可以理解，所述残差双注意 力机制模块通过使用空间注意力和通道注意力并行的组合，使得所述降噪后气相色谱特征图中不同类型的有效信息被大量捕捉到，可有效增强特征辨别学习能力，进而提高对于所述抽样气体中三氟氧化磷的含量值检测。
在本申请一个具体的实施例中，所述特征分布增强模块150，包括：空间注意力单元151，用于将所述优化降噪气相色谱特征图输入所述残差双注意力机制模块的空间注意力模块以得到空间注意力图；通道注意力单元152，用于将所述优化降噪气相色谱特征图输入所述残差双注意力机制模块的通道注意力模块以得到通道注意力图；注意力融合单元153，用于融合所述空间注意力图和所述通道注意力图以得到融合注意力图；激活单元154，用于将所述融合注意力图输入Sigmoid激活函数进行激活以得到融合注意力特征图；注意力施加单元155，用于计算所述融合注意力特征图和所述优化降噪气相色谱特征图的按位置点乘以得到加权特征图；以及，残差融合单元156，用于融合所述加权特征图和所述优化降噪气相色谱特征图以得到所述增强气相色谱特征图。
在本申请一个具体的实施例中，所述空间注意力单元151，包括：卷积编码子单元1511，用于使用所述残差双注意力机制模块的空间注意力模块的卷积层对所述深度增强气相色谱特征图进行卷积编码以得到卷积特征图；概率化子单元1512，用于将所述空间注意力图通过Softmax函数以得到空间注意力得分图；以及，空间注意力施加子单元1513，用于将所述空间注意力得分图与所述深度增强气相色谱特征图进行按位置点乘以得到空间注意力图。
在本申请一个具体的实施例中，所述通道注意力单元152，包括：全局均值池化子单元1521，用于对所述深度增强气相色谱特征图进行沿通道维度的全局均值池化以得到通道特征向量；归一化子单元1522，用于将所述通道特征向量通过Softmax函数以得到归一化通道特征向量；以及，通道注意力施加子单元1523，用于以所述归一化通道特征向量中各个位置的特征值作为权重对所述深度增强气相色谱特征图的沿通道维度的特征矩阵进行加权以得到通道注意力图。
在本申请实施例中，所述分析结果生成模块160，用于将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。也就是，将所述增强气相色谱特征图作为所述解码特征图导入进行解码回归，以获得用于表示所述抽样气体中三氟氧化磷的含量值的解码值，这样，能够对于所述抽样气体中的三氟氧化磷的含量进行准确检测，进而提高对于制备六氟磷酸锂的质检准确度。
在本申请一个具体的实施例中，所述分析结果生成模块，进一步用于：使用所述解码器以如下公式对所述解码特征图进行解码回归以得到所述解码值，其中，所述公式为其中，X是解码特征图，Y是解码值，W是权重矩阵，表示矩阵乘法。
综上，基于本申请实施例的所述用于六氟磷酸锂制备的自动化采样分析系统，采用基于深度学习的人工智能检测算法来提取出降噪后的抽样气体气相色 谱图中不同尺寸大小的多尺度隐含特征分布信息，并利用注意力机制来进一步加强对于气相色谱图中的小尺度三氟氧化磷含量检测而滤除无用的干扰特征的影响，以提高三氟氧化磷含量检测的精准度。这样，能够对于三氟氧化磷含量进行智能准确地检测，进而提高对于制备后的六氟磷酸锂的质检准确度，以提高电解液的品质和应用于锂离子电池中的电池性能。
如上所述，根据本申请实施例的所述用于六氟磷酸锂制备的自动化采样分析系统100可以实现在各种终端设备中，例如部署有六氟丁二烯的质检算法的服务器等。在一个示例中，用于六氟磷酸锂制备的自动化采样分析系统100可以作为一个软件模块和/或硬件模块而集成到终端设备中。例如，该用于六氟磷酸锂制备的自动化采样分析系统100可以是该终端设备的操作系统中的一个软件模块，或者可以是针对于该终端设备所开发的一个应用程序；当然，该用于六氟磷酸锂制备的自动化采样分析系统100同样可以是该终端设备的众多硬件模块之一。
替换地，在另一示例中，该用于六氟磷酸锂制备的自动化采样分析系统100与该终端设备也可以是分立的设备，并且用于六氟磷酸锂制备的自动化采样分析系统100可以通过有线和/或无线网络连接到该终端设备，并且按照约定的数据格式来传输交互信息。
示例性方法：图7图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析方法的流程图。如图7所示，根据本申请实施例的所述用于六氟磷酸锂制备的自动化采样分析方法，包括：S110，获取抽样气体的气相色谱图；S120，将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图；S130，将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图；S140，对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图；S150，将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图；以及，S160，将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
图8图示了根据本申请实施例的用于六氟磷酸锂制备的自动化采样分析方法的系统架构的示意图。如图8所示，在本申请实施例的所述用于六氟磷酸锂制备的自动化采样分析方法的系统架构中，首先，获取抽样气体的气相色谱图，并将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图。然后，将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图，并对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图。接着，将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图。最后，将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
这里，本领域技术人员可以理解，上述用于六氟磷酸锂制备的自动化采样分 析方法中的各个步骤的具体操作已经在上面参考图1到图6的用于六氟磷酸锂制备的自动化采样分析系统的描述中得到了详细介绍，并因此，将省略其重复描述。
示例性电子设备：下面，参考图9来描述根据本申请实施例的电子设备。图9图示了根据本申请实施例的电子设备的框图。
如图9所示，电子设备10包括一个或多个处理器11和存储器12。
处理器11可以是中央处理单元(CPU)或者具有数据处理能力和/或指令执行能力的其他形式的处理单元，并且可以控制电子设备10中的其他组件以执行期望的功能。
存储器12可以包括一个或多个计算机程序产品，所述计算机程序产品可以包括各种形式的计算机可读存储介质，例如易失性存储器和/或非易失性存储器。所述易失性存储器例如可以包括随机存取存储器(RAM)和/或高速缓冲存储器(cache)等。所述非易失性存储器例如可以包括只读存储器(ROM)、硬盘、闪存等。在所述计算机可读存储介质上可以存储一个或多个计算机程序指令，处理器11可以运行所述程序指令，以实现上文所述的本申请的各个实施例的用于六氟磷酸锂制备的自动化采样分析以及/或者其他期望的功能。在所述计算机可读存储介质中还可以存储诸如抽样气体的气相色谱图等各种内容。
在一个示例中，电子设备10还可以包括：输入装置13和输出装置14，这些组件通过总线系统和/或其他形式的连接机构(未示出)互连。
该输入装置13可以包括例如键盘、鼠标等等。
该输出装置14可以向外部输出各种信息，包括分类结果等。该输出装置14可以包括例如显示器、扬声器、打印机、以及通信网络及其所连接的远程输出设备等等。
示例性计算机程序产品和计算机可读存储介质：除了上述方法和设备以外，本申请的实施例还可以是计算机程序产品，其包括计算机程序指令，所述计算机程序指令在被处理器运行时使得所述处理器执行本说明书上述“示例性方法”部分中描述的根据本申请各种实施例的用于六氟磷酸锂制备的自动化采样分析方法的步骤。
所述计算机程序产品可以以一种或多种程序设计语言的任意组合来编写用于执行本申请实施例操作的程序代码，所述程序设计语言包括面向对象的程序设计语言，诸如Java、C++等，还包括常规的过程式程序设计语言，诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户计算设备上部分在远程计算设备上执行、或者完全在远程计算设备或服务器上执行。
此外，本申请的实施例还可以是计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令在被处理器运行时使得所述处理器执行本说明书上述“示例性方法”部分中描述的根据本申请各种实施例的用于六氟磷 酸锂制备的自动化采样分析方法的步骤。
所述计算机可读存储介质可以采用一个或多个可读介质的任意组合。可读介质可以是可读信号介质或者可读存储介质。可读存储介质例如可以包括但不限于电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。可读存储介质的更具体的例子(非穷举的列表)包括：具有一个或多个导线的电连接、便携式盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。
以上结合具体实施例描述了本申请的基本原理，但是，需要指出的是，在本申请中提及的优点、优势、效果等仅是示例而非限制，不能认为这些优点、优势、效果等是本申请的各个实施例必须具备的。另外，上述公开的具体细节仅是为了示例的作用和便于理解的作用，而非限制，上述细节并不限制本申请为必须采用上述具体的细节来实现。
本申请中涉及的器件、装置、设备、系统的方框图仅作为例示性的例子并且不意图要求或暗示必须按照方框图示出的方式进行连接、布置、配置。如本领域技术人员将认识到的，可以按任意方式连接、布置、配置这些器件、装置、设备、系统。诸如“包括”、“包含”、“具有”等等的词语是开放性词汇，指“包括但不限于”，且可与其互换使用。这里所使用的词汇“或”和“和”指词汇“和/或”，且可与其互换使用，除非上下文明确指示不是如此。这里所使用的词汇“诸如”指词组“诸如但不限于”，且可与其互换使用。
还需要指出的是，在本申请的装置、设备和方法中，各部件或各步骤是可以分解和/或重新组合的。这些分解和/或重新组合应视为本申请的等效方案。提供所公开的方面的以上描述以使本领域的任何技术人员能够做出或者使用本申请。对这些方面的各种修改对于本领域技术人员而言是非常显而易见的，并且在此定义的一般原理可以应用于其他方面而不脱离本申请的范围。因此，本申请不意图被限制到在此示出的方面，而是按照与在此公开的原理和新颖的特征一致的最宽范围。
为了例示和描述的目的已经给出了以上描述。此外，此描述不意图将本申请的实施例限制到在此公开的形式。尽管以上已经讨论了多个示例方面和实施例，但是本领域技术人员将认识到其某些变型、修改、改变、添加和子组合。

Claims (10)

  1. 一种用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，包括：抽样气体数据采集模块，用于获取抽样气体的气相色谱图；降噪模块，用于将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图；色谱特征提取模块，用于将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图；编码补偿模块，用于对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图；特征分布增强模块，用于将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图；以及分析结果生成模块，用于将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
  2. 根据权利要求1所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述降噪模块，包括：气相色谱图编码单元，用于将所述气相色谱图输入所述降噪模块的编码器，其中，所述编码器使用卷积层对所述气相色谱图进行显式空间编码以得到气相色谱特征图；以及气相色谱特征解码单元，用于将所述气相色谱特征图输入所述信号降噪模块的解码器，其中，所述解码器使用反卷积层对所述气相色谱特征图进行反卷积处理以得到所述降噪后气相色谱图。
  3. 根据权利要求2所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述色谱特征提取模块，进一步用于使用所述卷积神经网络模型的各个混合卷积层在层的正向传递中分别对输入数据进行：使用具有第一尺寸的第一卷积核对所述降噪后气相色谱图进行卷积编码以得到第一尺度特征图；使用具有第一空洞率的第二卷积核对所述降噪后气相色谱图进行卷积编码以得到第二尺度特征图；使用具有第二空洞率的第三卷积核对所述降噪后气相色谱图进行卷积编码以得到第三尺度特征图；使用具有第三空洞率的第四卷积核对所述降噪后气相色谱图进行卷积编码以得到第四尺度特征图，其中，所述第一卷积核、所述第二卷积核、所述第三卷积核和所述第四卷积核具有相同的尺寸，且所述第二卷积核、所述第三卷积核和所述第四卷积核具有不同的空洞率；将所述第一尺度特征图、所述第二尺度特征图、所述第三尺度特征图和所述第四尺度特征图进行沿通道维度的聚合以得到聚合特征图；对所述聚合特征图进行池化处理以生成池化特征图；以及对所述池化特征图进行激活处理以生成激活特征图；其中，所述包含多个混合卷积 层的卷积神经网络模型的最后一层的输出为所述降噪后气相色谱特征图。
  4. 根据权利要求3所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述编码补偿模块，进一步用于：以如下公式对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到所述优化降噪气相色谱特征图；其中，所述公式为：
    其中fi,j,k表示所述降噪后气相色谱特征图的第(i,j,k)位置的特征值，μ和δ分别表示所述降噪后气相色谱特征图的各个位置的特征值集合的均值和标准差。
  5. 根据权利要求4所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述特征分布增强模块，包括：空间注意力单元，用于将所述优化降噪气相色谱特征图输入所述残差双注意力机制模块的空间注意力模块以得到空间注意力图；通道注意力单元，用于将所述优化降噪气相色谱特征图输入所述残差双注意力机制模块的通道注意力模块以得到通道注意力图；注意力融合单元，用于融合所述空间注意力图和所述通道注意力图以得到融合注意力图；激活单元，用于将所述融合注意力图输入Sigmoid激活函数进行激活以得到融合注意力特征图；注意力施加单元，用于计算所述融合注意力特征图和所述优化降噪气相色谱特征图的按位置点乘以得到加权特征图；以及残差融合单元，用于融合所述加权特征图和所述优化降噪气相色谱特征图以得到所述增强气相色谱特征图。
  6. 根据权利要求5所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述空间注意力单元，包括：卷积编码子单元，用于使用所述残差双注意力机制模块的空间注意力模块的卷积层对所述深度增强气相色谱特征图进行卷积编码以得到卷积特征图；概率化子单元，用于将所述空间注意力图通过Softmax函数以得到空间注意力得分图；以及空间注意力施加子单元，用于将所述空间注意力得分图与所述深度增强气相色谱特征图进行按位置点乘以得到空间注意力图。
  7. 根据权利要求6所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述通道注意力单元，包括：全局均值池化子单元，用于对所述深度增强气相色谱特征图进行沿通道维度的全局均值池化以得到通道特 征向量；归一化子单元，用于将所述通道特征向量通过Softmax函数以得到归一化通道特征向量；以及通道注意力施加子单元，用于以所述归一化通道特征向量中各个位置的特征值作为权重对所述深度增强气相色谱特征图的沿通道维度的特征矩阵进行加权以得到通道注意力图。
  8. 根据权利要求7所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述分析结果生成模块，进一步用于：使用所述解码器以如下公式对所述解码特征图进行解码回归以得到所述解码值，其中，所述公式为其中，X是解码特征图，Y是解码值，W是权重矩阵，表示矩阵乘法。
  9. 一种用于六氟磷酸锂制备的自动化采样分析方法，其特征在于，包括：获取抽样气体的气相色谱图；将所述气相色谱图通过基于自动编解码器的降噪模块以得到降噪后气相色谱图；将所述降噪后气相色谱图通过包含多个混合卷积层的卷积神经网络模型以得到降噪后气相色谱特征图；对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图；将所述优化降噪气相色谱特征图通过残差双注意力机制模块以得到增强气相色谱特征图作为解码特征图；以及将所述解码特征图通过解码器以得到解码值，所述解码值用于表示所述抽样气体中三氟氧化磷的含量值。
  10. 根据权利要求9所述的用于六氟磷酸锂制备的自动化采样分析系统，其特征在于，所述对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到优化降噪气相色谱特征图，包括：以如下公式对所述降噪后气相色谱特征图进行特征聚类的去聚焦模糊优化以得到所述优化降噪气相色谱特征图；其中，所述公式为：
    其中fi,j,k表示所述降噪后气相色谱特征图的第(i,j,k)位置的特征值，μ和δ分别表示所述降噪后气相色谱特征图的各个位置的特征值集合的均值和标准差。
PCT/CN2023/086450 2022-11-08 2023-04-06 用于六氟磷酸锂制备的自动化采样分析系统及其方法 WO2024098653A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211391469.3A CN115754107B (zh) 2022-11-08 2022-11-08 用于六氟磷酸锂制备的自动化采样分析系统及其方法
CN202211391469.3 2022-11-09

Publications (1)

Publication Number Publication Date
WO2024098653A1 true WO2024098653A1 (zh) 2024-05-16

Family

ID=85367948

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/086450 WO2024098653A1 (zh) 2022-11-08 2023-04-06 用于六氟磷酸锂制备的自动化采样分析系统及其方法

Country Status (2)

Country Link
CN (1) CN115754107B (zh)
WO (1) WO2024098653A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115754107B (zh) * 2022-11-08 2023-06-23 福建省龙德新能源有限公司 用于六氟磷酸锂制备的自动化采样分析系统及其方法
CN115950993B (zh) * 2023-03-15 2023-07-25 福建德尔科技股份有限公司 氟氮混合气中氟含量的测试方法
CN116106457B (zh) * 2023-04-13 2023-06-27 天津海河标测技术检测有限公司 空气采样检测一体化装置

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070211928A1 (en) * 2005-11-10 2007-09-13 Rosetta Inpharmatics Llc Discover biological features using composite images
CN103235049A (zh) * 2013-03-29 2013-08-07 东莞市杉杉电池材料有限公司 一种六氟磷酸锂溶液的检测方法
JP2019070547A (ja) * 2017-10-06 2019-05-09 日本電子株式会社 質量分析データ処理装置及び質量分析データ処理方法
CN114065820A (zh) * 2021-11-29 2022-02-18 北京唐智科技发展有限公司 一种多维数据故障决策方法、装置及存储介质
KR102380684B1 (ko) * 2021-11-09 2022-04-01 주식회사 셀키 Ai 기반으로 질량 스펙트럼에 따른 당펩타이드 분석을 통해 암 특이적 바이오마커를 결정하는 방법 및 장치
CN114913156A (zh) * 2022-05-17 2022-08-16 国网安徽省电力有限公司铜陵供电公司 变压器故障诊断系统及其诊断方法
CN114912533A (zh) * 2022-05-23 2022-08-16 国网安徽省电力有限公司铜陵供电公司 一种应用于变压器的状态监测系统及监测方法
CN115100456A (zh) * 2022-05-27 2022-09-23 福建天甫电子材料有限公司 用于电子级双氧水制备的智能化有毒有害气体报警系统
CN115754107A (zh) * 2022-11-08 2023-03-07 福建省龙德新能源有限公司 用于六氟磷酸锂制备的自动化采样分析系统及其方法

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020014533A1 (en) * 1995-12-18 2002-02-07 Xiaxun Zhu Automated object dimensioning system employing contour tracing, vertice detection, and forner point detection and reduction methods on 2-d range data maps
CN107646089B (zh) * 2015-03-06 2020-12-08 英国质谱公司 光谱分析
CN107330475A (zh) * 2017-07-19 2017-11-07 北京化工大学 一种新型无模型贝叶斯分类预测模型软测量方法
WO2021113977A1 (en) * 2019-12-13 2021-06-17 Etrace Medical Diagnostics Inc. Anomaly detection and diagnosis in chromatography applications
CN112634137B (zh) * 2020-12-28 2024-03-05 西安电子科技大学 基于ae提取多尺度空谱特征的高光谱和全色图像融合方法
CN113516626B (zh) * 2021-05-07 2023-09-26 中国海洋大学 基于多尺度卷积与池化策略的侧扫声纳海底沙波检测方法
CN113433206A (zh) * 2021-05-10 2021-09-24 上海交通大学 一种基于深度学习技术提高质谱谱图分组稳定性的方法
CN113158980A (zh) * 2021-05-17 2021-07-23 四川农业大学 一种基于高光谱图像和深度学习的茶叶分类方法
CN113705526B (zh) * 2021-09-07 2022-03-04 安徽大学 一种高光谱遥感影像分类方法
CN113588847B (zh) * 2021-09-26 2022-04-08 萱闱(北京)生物科技有限公司 一种生物代谢组学数据处理方法、分析方法及装置和应用
CN113902903A (zh) * 2021-09-30 2022-01-07 北京工业大学 一种基于下采样的双注意力多尺度融合方法
CN114184710A (zh) * 2021-12-24 2022-03-15 合肥国轩高科动力能源有限公司 一种锂离子电池电解液中六氟磷酸锂含量的检测方法
CN114841319A (zh) * 2022-04-29 2022-08-02 哈尔滨工程大学 一种基于多尺度自适应卷积核的多光谱图像变化检测方法
CN115063329A (zh) * 2022-06-10 2022-09-16 中国人民解放军国防科技大学 低光照环境下的可见光和红外图像融合增强方法及系统
CN115079572B (zh) * 2022-06-30 2023-02-03 福建省龙德新能源有限公司 用于六氟磷酸锂制备的能源管理控制系统及其控制方法

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070211928A1 (en) * 2005-11-10 2007-09-13 Rosetta Inpharmatics Llc Discover biological features using composite images
CN103235049A (zh) * 2013-03-29 2013-08-07 东莞市杉杉电池材料有限公司 一种六氟磷酸锂溶液的检测方法
JP2019070547A (ja) * 2017-10-06 2019-05-09 日本電子株式会社 質量分析データ処理装置及び質量分析データ処理方法
KR102380684B1 (ko) * 2021-11-09 2022-04-01 주식회사 셀키 Ai 기반으로 질량 스펙트럼에 따른 당펩타이드 분석을 통해 암 특이적 바이오마커를 결정하는 방법 및 장치
CN114065820A (zh) * 2021-11-29 2022-02-18 北京唐智科技发展有限公司 一种多维数据故障决策方法、装置及存储介质
CN114913156A (zh) * 2022-05-17 2022-08-16 国网安徽省电力有限公司铜陵供电公司 变压器故障诊断系统及其诊断方法
CN114912533A (zh) * 2022-05-23 2022-08-16 国网安徽省电力有限公司铜陵供电公司 一种应用于变压器的状态监测系统及监测方法
CN115100456A (zh) * 2022-05-27 2022-09-23 福建天甫电子材料有限公司 用于电子级双氧水制备的智能化有毒有害气体报警系统
CN115754107A (zh) * 2022-11-08 2023-03-07 福建省龙德新能源有限公司 用于六氟磷酸锂制备的自动化采样分析系统及其方法

Also Published As

Publication number Publication date
CN115754107A (zh) 2023-03-07
CN115754107B (zh) 2023-06-23

Similar Documents

Publication Publication Date Title
WO2024098653A1 (zh) 用于六氟磷酸锂制备的自动化采样分析系统及其方法
CN111652348B (zh) 一种基于改进cnn的动力电池组故障融合诊断方法及系统
Park et al. Probabilistic representations for video contrastive learning
CN112800774B (zh) 基于注意力机制的实体关系抽取方法、装置、介质和设备
CN116699297B (zh) 充电桩检测系统及其方法
CN113221687A (zh) 一种压板状态识别模型的训练方法及压板状态识别方法
CN114897889B (zh) 电池组点焊自动化全检方法及其系统
CN114882021B (zh) 电池锂膜的高效加工方法及其系统
CN115712740B (zh) 多模态蕴含增强图像文本检索的方法和系统
CN116030018A (zh) 用于门加工的来料合格检验系统及方法
CN111738319A (zh) 一种基于大规模样本的聚类结果评价方法及装置
WO2024103613A1 (zh) 一种电子级六氟丁二烯的质检系统及其方法
CN114580262A (zh) 一种锂离子电池健康状态估计方法
WO2024061141A1 (zh) 共性知识约束的遥感样本迁移方法
CN116381424A (zh) 一种基于多源数据融合的岸电箱绝缘性能异常检测方法
CN116520154A (zh) 锂电池极片检测的性能评估方法及其系统
CN114757097B (zh) 一种线路故障诊断方法及装置
CN113553844B (zh) 一种基于前缀树特征与卷积神经网络的领域识别方法
CN115143128A (zh) 小型潜水电泵的故障诊断方法及其系统
CN115600511A (zh) 电解质材料预测模型训练方法、装置和计算机设备
CN114418111A (zh) 标签预测模型训练及样本筛选方法、装置、存储介质
CN114077663A (zh) 应用日志的分析方法及装置
CN117250521B (zh) 充电桩电池容量监测系统及方法
CN112200254A (zh) 一种网络入侵检测模型生成方法、检测方法及电子设备
Li et al. Focus on local: transmission line defect detection via feature refinement