Polycrystalline mullite fiber production line and method

Abstract

The present application relates to the field of intelligent control, and specifically discloses a polycrystalline mullite fiber production line and method, which uses artificial intelligence technology based on a deep learning model to extract and encode the characteristics of the spinning speed, the distance between the nozzle and the collector, and the spinning angle, so as to obtain a classification result indicating whether the spinning speed at the current time point should be increased or decreased. In this way, the control efficiency is improved and the labor cost is reduced by intelligently controlling the spinning speed.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically to a polycrystalline mullite fiber production line and method. Background Art

[0002] Polycrystalline mullite fiber is a high-performance fiber material with excellent properties such as high temperature resistance, chemical corrosion resistance and mechanical strength. The production process of polycrystalline mullite fiber includes: crushing and grinding of mullite ore, melting, fiberization, curing and post-treatment. In the process of fiberization, the rotary spinning method is usually used, that is, the molten mullite liquid is ejected from the nozzle and stretched into fibers by the centrifugal force of high-speed rotation while ejecting. In the traditional rotary spinning method, various parameters in the spinning process are often manually controlled to control the diameter and length distribution of the fiber to obtain the desired fiber properties. However, this control method requires high manual proficiency and is prone to errors.

[0003] Therefore, an optimized production scheme for polycrystalline mullite fibers is desired. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. The embodiment of the present application provides a polycrystalline mullite fiber production line and method, which uses artificial intelligence technology based on a deep learning model to extract and encode the features of the spinning speed, the distance between the nozzle and the collector, and the spinning angle, so as to obtain a classification result indicating whether the spinning speed at the current time point should be increased or decreased. In this way, the control efficiency is improved and the labor cost is reduced by intelligently controlling the spinning speed.

[0005] According to one aspect of the present application, a polycrystalline mullite fiber production line is provided, comprising:

[0006] A data acquisition module, used to acquire the values ​​of the spinning speed, the distance between the nozzle and the collector, and the spinning angle at a plurality of predetermined time points;

[0007] A vector construction module, used for arranging the spinning speed value, the distance value between the nozzle and the collector, and the spinning angle value into a first input vector, a second input vector, and a third input vector according to the time dimension respectively;

[0008] a fusion module, configured to fuse the first input vector, the second input vector and the third input vector to obtain a fused feature matrix;

[0009] A deep-shallow feature fusion module, used for passing the fused feature matrix through a first convolutional neural network model including the deep-shallow fusion module to obtain a deep-shallow feature matrix;

[0010] A first scale feature extraction module, used for inputting the deep and shallow feature matrix into a second convolutional neural network model with a first scale to obtain a first scale correlation feature matrix;

[0011] A second scale feature extraction module, used for inputting the deep and shallow feature matrix into a second convolutional neural network model with a second scale to obtain a second scale correlation feature matrix;

[0012] An optimization module, used for performing homography density domain uniformization on the first scale correlation feature matrix and the second scale correlation feature matrix to obtain an optimized first scale correlation feature matrix and an optimized second scale correlation feature matrix;

[0013] A feature fusion module, used for calculating the weighted sum of the optimized first-scale correlation feature matrix and the optimized second-scale correlation feature matrix according to position to obtain a classification feature matrix;

[0014] The spinning speed control module is used to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the spinning speed at the current time point should be increased or decreased.

[0015] In the above-mentioned polycrystalline mullite fiber production line, the deep and shallow feature fusion module includes:

[0016] A shallow feature extraction unit, used to extract a shallow feature matrix from the Mth layer of the first convolutional neural network model, where M is greater than or equal to 1 and less than or equal to 6;

[0017] A deep feature extraction unit, used to extract a deep feature matrix from the Nth layer of the first convolutional neural network model, wherein N / M is greater than or equal to 5 and less than or equal to 10;

[0018] A deep-shallow fusion unit is used to fuse the shallow feature matrix and the deep feature matrix using the deep-shallow feature fusion module of the first convolutional neural network model to obtain a deep-shallow feature matrix.

[0019] In the above-mentioned polycrystalline mullite fiber production line, the first scale feature extraction module is used to:

[0020] The deep and shallow feature matrix is ​​input into the first convolution layer of the second convolutional neural network model using a first scale to obtain a first-scale correlation feature matrix, wherein the first convolution layer has a first two-dimensional convolution kernel of a first length.

[0021] In the above-mentioned polycrystalline mullite fiber production line, the second scale feature extraction module is used to:

[0022] The deep and shallow feature matrix is ​​input into the second convolutional layer of the second convolutional neural network model using a second scale to obtain a second-scale correlation feature matrix, wherein the second convolutional layer has a second two-dimensional convolution kernel of a second length, and the first length is different from the second length.

[0023] In the above-mentioned polycrystalline mullite fiber production line, the optimization module includes:

[0024] A semantic encoding unit, configured to perform semantic encoding on the first scale correlation feature matrix and the second scale correlation feature matrix to obtain a first semantic feature vector and a second semantic feature vector;

[0025] an association coding unit, configured to perform semantic association coding on the first semantic feature vector and the second semantic feature vector to obtain a common semantic association feature matrix;

[0026] A homography density domain probabilization unit, used for obtaining a common semantic association homography density domain probabilization feature matrix by passing the common semantic association feature matrix through a Softmax activation function;

[0027] The position point multiplication unit is used to perform position point multiplication on the first scale association feature matrix and the second scale association feature matrix with the common semantic association homography density domain probabilistic feature matrix respectively to obtain the optimized first scale association feature matrix and the optimized second scale association feature matrix.

[0028] In the above-mentioned polycrystalline mullite fiber production line, the spinning speed control module is used to:

[0029] Using the classifier to process the classification feature matrix according to the following classification formula to obtain the classification result;

[0030] Wherein, the classification formula is: O = softmax{(W c ,B c )|Project(F)}, where Project(F) represents projecting the classification feature matrix into a vector, W c is the weight matrix, B c represents a bias vector, softmax represents a normalized exponential function, and O represents the classification result.

[0031] According to another aspect of the present application, a control method for a polycrystalline mullite fiber production line is also provided, comprising:

[0032] Obtaining values ​​of the spinning speed, the distance between the nozzle and the collector, and the spinning angle at a plurality of predetermined time points;

[0033] Arrange the spinning speed value, the distance value between the nozzle and the collector, and the spinning angle value into a first input vector, a second input vector, and a third input vector respectively according to the time dimension;

[0034] Fusing the first input vector, the second input vector and the third input vector to obtain a fused feature matrix;

[0035] Passing the fused feature matrix through a first convolutional neural network model including a deep and shallow fusion module to obtain a deep and shallow feature matrix;

[0036] Inputting the deep and shallow feature matrix into a second convolutional neural network model with a first scale to obtain a first scale correlation feature matrix;

[0037] Inputting the deep and shallow feature matrix into a second convolutional neural network model with a second scale to obtain a second scale correlation feature matrix;

[0038] Performing homography density domain uniformity on the first-scale correlation feature matrix and the second-scale correlation feature matrix to obtain an optimized first-scale correlation feature matrix and an optimized second-scale correlation feature matrix;

[0039] Calculating a weighted sum by position of the optimized first-scale correlation feature matrix and the optimized second-scale correlation feature matrix to obtain a classification feature matrix;

[0040] The classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the spinning speed at the current time point should be increased or decreased.

[0041] Compared with the prior art, the polycrystalline mullite fiber production line and method provided by the present application use artificial intelligence technology based on a deep learning model to extract and encode the characteristics of the spinning speed, the distance between the nozzle and the collector, and the spinning angle, so as to obtain a classification result indicating whether the spinning speed at the current time point should be increased or decreased. In this way, the control efficiency is improved and the labor cost is reduced by intelligently controlling the spinning speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0043] Figure 1 A block diagram of a polycrystalline mullite fiber production line according to an embodiment of the present application is illustrated.

[0044] Figure 2 The diagram shows a system architecture diagram of a polycrystalline mullite fiber production line according to an embodiment of the present application.

[0045] Figure 3 The diagram illustrates a block diagram of a deep and shallow feature fusion module in a polycrystalline mullite fiber production line according to an embodiment of the present application.

[0046] Figure 4 The flowchart of the control method of the polycrystalline mullite fiber production line according to the embodiment of the present application is illustrated.

[0047] Figure 5 A block diagram of an electronic device according to an embodiment of the present application is illustrated. DETAILED DESCRIPTION

[0048] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0049] Exemplary Systems

[0050] Figure 1 FIG. 1 is a block diagram of a polycrystalline mullite fiber production line according to an embodiment of the present application. Figure 1As shown, according to the embodiment of the present application, the polycrystalline mullite fiber production line 100 includes: a data acquisition module 110, which is used to obtain the spinning speed value, the distance value between the nozzle and the collector, and the spinning angle value at multiple predetermined time points; a vector construction module 120, which is used to arrange the spinning speed value, the distance value between the nozzle and the collector, and the spinning angle value into a first input vector, a second input vector, and a third input vector according to the time dimension; a fusion module 130, which is used to fuse the first input vector, the second input vector, and the third input vector to obtain a fusion feature matrix; a depth feature fusion module 140, which is used to pass the fusion feature matrix through a first convolutional neural network model including a depth fusion module to obtain a depth feature matrix; a first scale feature extraction module 150, which is used to input the depth feature matrix into a first convolutional neural network model having a first scale feature extraction module. a second convolutional neural network model with a second scale to obtain a first-scale correlation feature matrix; a second-scale feature extraction module 160, used to input the depth feature matrix into a second convolutional neural network model with a second scale to obtain a second-scale correlation feature matrix; an optimization module 170, used to perform homography density domain uniformization on the first-scale correlation feature matrix and the second-scale correlation feature matrix to obtain an optimized first-scale correlation feature matrix and an optimized second-scale correlation feature matrix; a feature fusion module 180, used to calculate the position-weighted sum of the optimized first-scale correlation feature matrix and the optimized second-scale correlation feature matrix to obtain a classification feature matrix; and a spinning speed control module 190, used to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the spinning speed at the current time point should be increased or decreased.

[0051] Figure 2 The system architecture diagram of the polycrystalline mullite fiber production line according to the embodiment of the present application is shown in FIG. Figure 2As shown, in the system architecture, first, the spin speed values, the distance values ​​between the nozzle and the collector, and the spin angle values ​​at multiple predetermined time points are obtained. Then, the spin speed values, the distance values ​​between the nozzle and the collector, and the spin angle values ​​are arranged as a first input vector, a second input vector, and a third input vector according to the time dimension. Then, the first input vector, the second input vector, and the third input vector are fused to obtain a fusion feature matrix. Then, the fusion feature matrix is ​​passed through a first convolutional neural network model including a deep and shallow fusion module to obtain a deep and shallow feature matrix. Then, the deep and shallow feature matrix is ​​input into a second convolutional neural network model with a first scale to obtain a first scale association feature matrix. Then, the deep and shallow feature matrix is ​​input into a second convolutional neural network model with a second scale to obtain a second scale association feature matrix. Then, the first scale association feature matrix and the second scale association feature matrix are fused to obtain a classification feature matrix. Furthermore, the classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the spin speed at the current time point should be increased or decreased.

[0052] In the above-mentioned polycrystalline mullite fiber production line 100, the data acquisition module 110 is used to obtain the values ​​of the spinning speed, the distance between the nozzle and the collector, and the spinning angle at a plurality of predetermined time points. As mentioned in the above-mentioned background technology, in the traditional rotary spinning method, various parameters in the spinning process are often manually controlled to control the diameter and length distribution of the fiber to obtain the required fiber properties. However, this control method has high requirements on the proficiency of the manual labor and is prone to errors. Therefore, an optimized polycrystalline mullite fiber production scheme is expected.

[0053] In response to the above technical problems, an optimized polycrystalline mullite fiber production line is proposed, which uses artificial intelligence technology based on a deep learning model to extract and encode the features of the spinning speed, the distance between the nozzle and the collector, and the spinning angle, so as to obtain a classification result indicating whether the spinning speed at the current time point should be increased or decreased. In this way, the control efficiency is improved and the labor cost is reduced by intelligently controlling the spinning speed.

[0054] At present, deep learning and neural networks have been widely used in computer vision, natural language processing, speech signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields.

[0055] In recent years, the development of deep learning and neural networks has provided new solutions and solutions for polycrystalline mullite fiber production lines.

[0056] Specifically, first, the spinneret speed values, the distance values ​​between the nozzle and the collector, and the spinneret angle values ​​at multiple predetermined time points are obtained. The spinneret speed values, the distance values ​​between the nozzle and the collector, and the spinneret angle values ​​obtained at multiple predetermined time points can be used to establish an accurate model, and these data can be used to learn and predict the change trend of the spinneret speed.

[0057] In the above-mentioned polycrystalline mullite fiber production line 100, the vector construction module 120 is used to arrange the spin speed value, the distance value between the nozzle and the collector, and the spin angle value as a first input vector, a second input vector, and a third input vector according to the time dimension. In the polycrystalline mullite fiber production line, the changes in the spin speed, distance, and angle are usually related to time. By arranging according to the time dimension, this time dependency can be retained, so that the model can better capture the dynamic changes between parameters. Using the parameter values ​​at different time points as different input vectors helps the model extract features from data at different time points. This can more comprehensively describe the relationship between parameters and provide more information for model learning and prediction. Arranging the parameters as different input vectors according to the time dimension allows the model to choose which time points to use as needed. In this way, the input of the model can be flexibly adjusted to adapt to different production needs and scenarios.

[0058] In the above-mentioned polycrystalline mullite fiber production line 100, the fusion module 130 is used to fuse the first input vector, the second input vector and the third input vector to obtain a fused feature matrix. The first input vector, the second input vector and the third input vector respectively contain different parameter information. By fusing them together, the influence of these parameters can be comprehensively considered. This helps the model to more comprehensively understand and predict the changing trend of the spinning speed and improve the accuracy and robustness of the model. The fused feature matrix can promote the feature interaction between different parameters. By fusing different input vectors together, the model can learn the interaction and association between parameters and further extract richer feature representations. At the same time, the fused feature matrix aligns the dimensions of different input vectors so that the model can more conveniently process and analyze these features. This can simplify the structure and calculation of the model and improve the efficiency and interpretability of the model. Further, by fusing the feature matrix, the loss of information can be reduced. Different input vectors may contain complementary information. By fusing, this information can be comprehensively utilized to provide a more complete and accurate feature representation.

[0059] In the above-mentioned polycrystalline mullite fiber production line 100, the deep and shallow feature fusion module 140 is used to pass the fused feature matrix through the first convolutional neural network model including the deep and shallow fusion module to obtain the deep and shallow feature matrix. The deep and shallow fusion module can extract features from the fused feature matrix through the convolutional neural network. Convolutional neural networks have excellent feature extraction capabilities in processing image and sequence data. By inputting the fused feature matrix into the convolutional neural network, the model can learn more abstract and advanced feature representations, thereby better capturing the relationship and pattern between parameters. The deep and shallow fusion module can extract deep features and shallow features at the same time. Deep features refer to high-level abstract features learned through multi-layer convolutional neural networks, which have stronger expression capabilities and semantic information. Shallow features refer to low-level features learned through a few layers of convolutional neural networks, which are closer to the original input data. By extracting deep features and shallow features at the same time, feature representations at different levels can be comprehensively considered to improve the robustness and generalization ability of the model.

[0060] Figure 3 The block diagram of the deep and shallow feature fusion module in the polycrystalline mullite fiber production line according to the embodiment of the present application is shown. Figure 3 As shown, the deep and shallow feature fusion module 140 includes: a shallow feature extraction unit 141, used to extract a shallow feature matrix from the Mth layer of the first convolutional neural network model, wherein M is greater than or equal to 1 and less than or equal to 6; a deep feature extraction unit 142, used to extract a deep feature matrix from the Nth layer of the first convolutional neural network model, wherein N / M is greater than or equal to 5 and less than or equal to 10; and a deep and shallow fusion unit 143, used to use the deep and shallow feature fusion module of the first convolutional neural network model to fuse the shallow feature matrix and the deep feature matrix to obtain a deep and shallow feature matrix.

[0061] In the above-mentioned polycrystalline mullite fiber production line 100, the first scale feature extraction module 150 is used to input the depth feature matrix into the second convolutional neural network model with a first scale to obtain a first scale correlation feature matrix. By using a convolutional neural network model with a first scale, the model can be made more sensitive to scale changes in the input data. In a polycrystalline mullite fiber production line, features of different scales have different effects on the spinning speed. Therefore, using a model with a first scale can better capture the scale information related to the spinning speed. The first scale correlation feature matrix can help the model learn the correlation between parameters. In a fiber production line, there are complex correlations between different parameters, such as the spinning speed and the distance between the nozzle and the collector, the spinning angle, etc. By using the first scale correlation feature matrix, the model can better capture these correlation features, thereby improving the prediction accuracy of the spinning speed.

[0062] Specifically, in an embodiment of the present application, the first-scale feature extraction module 150 is used to: input the deep and shallow feature matrix into the first convolutional layer of the second convolutional neural network model using a first scale to obtain a first-scale correlation feature matrix, wherein the first convolutional layer has a first two-dimensional convolution kernel of a first length.

[0063] In the above-mentioned polycrystalline mullite fiber production line 100, the second scale feature extraction module 160 is used to input the depth feature matrix into the second convolutional neural network model with a second scale to obtain a second scale correlation feature matrix. The second convolutional neural network model has a stronger feature expression capability. By inputting the depth feature matrix into the model with the first scale, more advanced and abstract feature representations can be further extracted. This helps the model to better understand and represent the relationship between parameters, thereby improving the accuracy of prediction. The second scale correlation feature matrix helps the model learn more complex parameter correlations. In a fiber production line, the correlation between different parameters involves features on multiple scales. By using a model with a second scale, the correlation features between these scales can be better learned, thereby improving the accuracy of prediction of the spinning speed.

[0064] Specifically, in an embodiment of the present application, the second-scale feature extraction module 160 is used to: input the deep and shallow feature matrix into the second convolutional layer of the second convolutional neural network model using a second scale to obtain a second-scale correlation feature matrix, wherein the second convolutional layer has a second two-dimensional convolution kernel of a second length, and the first length is different from the second length.

[0065] In the above-mentioned polycrystalline mullite fiber production line 100, the optimization module 170 is used to perform homography density domain uniformization on the first scale correlation feature matrix and the second scale correlation feature matrix to obtain an optimized first scale correlation feature matrix and an optimized second scale correlation feature matrix. In the technical solution of the present application, the first scale correlation feature matrix and the second scale correlation feature matrix are data manifolds distributed in different high-dimensional feature spaces. Therefore, if the first scale correlation feature matrix and the second scale correlation feature matrix can be mapped to a common feature space, it is beneficial to reduce the feature domain differences between the first scale correlation feature matrix and the second scale correlation feature matrix, so as to solve the problems of domain adaptation and domain generation of data in different domains, and improve the generalization ability and migration ability of the model. Specifically, in the technical solution of the present application, the first scale correlation feature matrix and the second scale correlation feature matrix are firstly semantically encoded, for example, the first scale correlation feature matrix and the second scale correlation feature matrix are matrix expanded to obtain a first semantic feature vector and a second semantic feature vector. Next, semantic association encoding is performed on the first semantic feature vector and the second semantic feature vector to obtain a common semantic association feature matrix, which is used to represent the collaborative high-dimensional manifold surface of the first scale association feature matrix and the second scale association feature matrix in the semantic space, and the common semantic association feature matrix is ​​activated by a Softmax function to obtain a common semantic association homography density domain probabilistic feature matrix, that is, the common semantic association feature matrix is ​​homography-densified activated using the Softmax activation function to map the eigenvalues ​​of each position in the common semantic association feature matrix to the probability density domain space, and then, the first scale association feature matrix and the second scale association feature matrix are respectively multiplied by the common semantic association homography density domain probabilistic feature matrix according to the position to obtain the optimized first scale association feature matrix and the optimized second scale association feature matrix, so that the common semantic association homography density domain probabilistic feature matrix is ​​equivalent to a transformation function so that in the transformed feature space, data from different domains have the same or approximately the same density function, so that the model can be trained and tested in a unified feature space without considering the differences between domains.

[0066] Specifically, in an embodiment of the present application, the optimization module 170 includes: a semantic encoding unit, which is used to perform semantic encoding on the first-scale association feature matrix and the second-scale association feature matrix to obtain a first semantic feature vector and a second semantic feature vector; an association encoding unit, which is used to perform semantic association encoding on the first semantic feature vector and the second semantic feature vector to obtain a common semantic association feature matrix; a homography density domain probabilistic unit, which is used to pass the common semantic association feature matrix through a Softmax activation function to obtain a common semantic association homography density domain probabilistic feature matrix; and a position point multiplication unit, which is used to perform position point multiplication on the first-scale association feature matrix and the second-scale association feature matrix with the common semantic association homography density domain probabilistic feature matrix respectively to obtain the optimized first-scale association feature matrix and the optimized second-scale association feature matrix.

[0067] In the above-mentioned polycrystalline mullite fiber production line 100, the feature fusion module 180 is used to calculate the position-weighted sum of the optimized first-scale correlation feature matrix and the optimized second-scale correlation feature matrix to obtain a classification feature matrix.

[0068] In the above-mentioned polycrystalline mullite fiber production line 100, the spinning speed control module 190 is used to pass the classification feature matrix through the classifier to obtain a classification result, and the classification result is used to indicate whether the spinning speed at the current time point should be increased or decreased. The classifier can establish a model to predict whether the spinning speed at the current time point should be increased or decreased by learning the correlation between the classification feature matrix and the spinning speed. The classifier can learn the patterns and rules between the features and the spinning speed based on the existing training data, and then apply these patterns and rules to the new input data to obtain the corresponding classification results. By matching the classification results with the spinning speed at the current time point, real-time speed control and adjustment can be achieved. If the classification result indicates that the spinning speed should be increased, the control system can increase the spinning speed accordingly; if the classification result indicates that the spinning speed should be reduced, the control system can reduce the spinning speed accordingly. In this way, the spinning speed can be dynamically adjusted according to real-time needs to improve product quality and production efficiency.

[0069] Specifically, in the embodiment of the present application, the spinning speed control module 190 is used to: use the classifier to process the classification feature matrix according to the following classification formula to obtain the classification result; wherein the classification formula is: O = softmax{(W c ,B c )|Project(F)}, where Project(F) represents projecting the classification feature matrix into a vector, W c is the weight matrix, Bc represents a bias vector, softmax represents a normalized exponential function, and O represents the classification result.

[0070] In summary, the polycrystalline mullite fiber production line according to the embodiment of the present application has been explained, which uses artificial intelligence technology based on a deep learning model to extract and encode the features of the spinning speed, the distance between the nozzle and the collector, and the spinning angle to obtain a classification result indicating whether the spinning speed at the current time point should be increased or decreased. In this way, the control efficiency is improved and the labor cost is reduced by intelligently controlling the spinning speed.

[0071] As described above, the polycrystalline mullite fiber production line 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server of the polycrystalline mullite fiber production line. In one example, the polycrystalline mullite fiber production line 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the polycrystalline mullite fiber production line 100 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 polycrystalline mullite fiber production line 100 can also be one of the many hardware modules of the terminal device.

[0072] Alternatively, in another example, the polycrystalline mullite fiber production line 100 and the terminal device may also be separate devices, and the polycrystalline mullite fiber production line 100 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.

[0073] Exemplary Methods

[0074] Figure 4 The flowchart of the control method of the polycrystalline mullite fiber production line according to the embodiment of the present application is shown. Figure 4As shown, the control method of the polycrystalline mullite fiber production line according to the embodiment of the present application includes the following steps: S110, obtaining the spin speed values, the distance values ​​between the nozzle and the collector, and the spin angle values ​​at a plurality of predetermined time points; S120, arranging the spin speed values, the distance values ​​between the nozzle and the collector, and the spin angle values ​​into a first input vector, a second input vector, and a third input vector according to the time dimension; S130, fusing the first input vector, the second input vector, and the third input vector to obtain a fused feature matrix; S140, fusing the fused feature matrix through a first convolution including a deep and shallow fusion module; Convolution neural network model to obtain a depth feature matrix; S150, the depth feature matrix is ​​input into a second convolution neural network model with a first scale to obtain a first scale correlation feature matrix; S160, the depth feature matrix is ​​input into a second convolution neural network model with a second scale to obtain a second scale correlation feature matrix; S170, the first scale correlation feature matrix and the second scale correlation feature matrix are fused to obtain a classification feature matrix; and S180, the classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the spinning speed at the current time point should be increased or decreased.

[0075] Here, those skilled in the art will appreciate that the specific operations of each step in the control method of the polycrystalline mullite fiber production line have been described in detail above. Figures 1 to 3 The polycrystalline mullite fiber production line is described in detail in the description of the polycrystalline mullite fiber production line, and therefore, its repeated description will be omitted.

[0076] In summary, the control method of the polycrystalline mullite fiber production line according to the embodiment of the present application has been explained, which uses artificial intelligence technology based on a deep learning model to extract and encode the features of the spinning speed, the distance between the nozzle and the collector, and the spinning angle to obtain a classification result indicating whether the spinning speed at the current time point should be increased or decreased. In this way, the control efficiency is improved and the labor cost is reduced by intelligently controlling the spinning speed.

[0077] Exemplary Electronic Devices

[0078] Below, reference Figure 5 To describe an electronic device according to an embodiment of the present application. Figure 5 FIG. 1 is a block diagram of an electronic device according to an embodiment of the present application. Figure 5 As shown, the electronic device 10 includes one or more processors 11 and a memory 12 .

[0079] 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.

[0080] 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 functions of the control method of the polycrystalline mullite fiber production line of each embodiment of the present application described above and / or other desired functions. Various contents such as the spinneret speed value at multiple predetermined time points, the distance value between the nozzle and the collector, and the spinneret angle value may also be stored in the computer-readable storage medium.

[0081] In one example, 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).

[0082] The input device 13 may include, for example, a keyboard, a mouse, etc.

[0083] The output device 14 can output various information to the outside, including whether the spinning speed at the current time point should be increased or decreased, 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.

[0084] Of course, to simplify, Figure 5 Only some of the components related to the present application in the electronic device 10 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application situations, the electronic device 10 may also include any other appropriate components.

[0085] Exemplary computer program products and computer-readable storage media

[0086] In addition to the above-mentioned methods and devices, 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 in the functions of the control method of the polycrystalline mullite fiber production line according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0087] 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.

[0088] In addition, an embodiment of the present application may also be a computer-readable storage medium on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the processor executes the steps in the functions of the control method of the polycrystalline mullite fiber production line according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0089] 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. More specific examples (non-exhaustive list) of 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.

[0090] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0091] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0092] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0093] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0094] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A polycrystalline mullite fiber production line, characterized in that: include: A data acquisition module, used to acquire the values ​​of the spinning speed, the distance between the nozzle and the collector, and the spinning angle at a plurality of predetermined time points; A vector construction module, used for arranging the spinning speed value, the distance value between the nozzle and the collector, and the spinning angle value into a first input vector, a second input vector, and a third input vector according to the time dimension respectively; a fusion module, configured to fuse the first input vector, the second input vector and the third input vector to obtain a fused feature matrix; A deep-shallow feature fusion module, used for passing the fused feature matrix through a first convolutional neural network model including the deep-shallow fusion module to obtain a deep-shallow feature matrix; A first scale feature extraction module, used for inputting the deep and shallow feature matrix into a second convolutional neural network model with a first scale to obtain a first scale correlation feature matrix; A second scale feature extraction module, used for inputting the deep and shallow feature matrix into a second convolutional neural network model with a second scale to obtain a second scale correlation feature matrix; An optimization module, used for performing homography density domain uniformization on the first scale correlation feature matrix and the second scale correlation feature matrix to obtain an optimized first scale correlation feature matrix and an optimized second scale correlation feature matrix; A feature fusion module, used for calculating the weighted sum of the optimized first-scale correlation feature matrix and the optimized second-scale correlation feature matrix according to position to obtain a classification feature matrix; A spinning speed control module, used for passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the spinning speed at a current time point should be increased or decreased; Wherein, the deep and shallow feature fusion module includes: A shallow feature extraction unit, used to extract a shallow feature matrix from the Mth layer of the first convolutional neural network model, where M is greater than or equal to 1 and less than or equal to 6; A deep feature extraction unit, used to extract a deep feature matrix from the Nth layer of the first convolutional neural network model, wherein N / M is greater than or equal to 5 and less than or equal to 10; A deep-shallow fusion unit, used to fuse the shallow feature matrix and the deep feature matrix using the deep-shallow feature fusion module of the first convolutional neural network model to obtain a deep-shallow feature matrix; Wherein, the spinning speed control module is used for: Using the classifier to process the classification feature matrix according to the following classification formula to obtain the classification result; Wherein, the classification formula is: O = softmax{(W c ,B c )|Project(F)}, where Project(F) represents projecting the classification feature matrix into a vector, W c is the weight matrix, B c represents a bias vector, softmax represents a normalized exponential function, and O represents the classification result.

2. The polycrystalline mullite fiber production line according to claim 1, characterized in that: The first scale feature extraction module is used to: The deep and shallow feature matrix is ​​input into the first convolution layer of the second convolutional neural network model using a first scale to obtain a first-scale correlation feature matrix, wherein the first convolution layer has a first two-dimensional convolution kernel of a first length.

3. The polycrystalline mullite fiber production line according to claim 2, characterized in that: The second scale feature extraction module is used to: The deep and shallow feature matrix is ​​input into the second convolutional layer of the second convolutional neural network model using a second scale to obtain a second-scale correlation feature matrix, wherein the second convolutional layer has a second two-dimensional convolution kernel of a second length, and the first length is different from the second length.

4. The polycrystalline mullite fiber production line according to claim 3, characterized in that: The optimization module comprises: A semantic encoding unit, configured to perform semantic encoding on the first scale correlation feature matrix and the second scale correlation feature matrix to obtain a first semantic feature vector and a second semantic feature vector; an association coding unit, configured to perform semantic association coding on the first semantic feature vector and the second semantic feature vector to obtain a common semantic association feature matrix; A homography density domain probabilization unit, used for obtaining a common semantic association homography density domain probabilization feature matrix by passing the common semantic association feature matrix through a Softmax activation function; The position point multiplication unit is used to perform position point multiplication on the first scale association feature matrix and the second scale association feature matrix with the common semantic association homography density domain probabilistic feature matrix respectively to obtain the optimized first scale association feature matrix and the optimized second scale association feature matrix.

5. A control method for a polycrystalline mullite fiber production line, characterized in that: include: Obtaining values ​​of the spinning speed, the distance between the nozzle and the collector, and the spinning angle at a plurality of predetermined time points; Arrange the spinning speed value, the distance value between the nozzle and the collector, and the spinning angle value into a first input vector, a second input vector, and a third input vector according to the time dimension respectively; Fusing the first input vector, the second input vector and the third input vector to obtain a fused feature matrix; Passing the fused feature matrix through a first convolutional neural network model including a deep and shallow fusion module to obtain a deep and shallow feature matrix; Inputting the deep and shallow feature matrix into a second convolutional neural network model with a first scale to obtain a first scale correlation feature matrix; Inputting the deep and shallow feature matrix into a second convolutional neural network model with a second scale to obtain a second scale correlation feature matrix; Performing homography density domain unification on the first-scale correlation feature matrix and the second-scale correlation feature matrix to obtain an optimized first-scale correlation feature matrix and an optimized second-scale correlation feature matrix; Calculating a weighted sum of the optimized first-scale correlation feature matrix and the optimized second-scale correlation feature matrix by position to obtain a classification feature matrix; Passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the spinning speed at the current time point should be increased or decreased; The fusion feature matrix is ​​passed through a first convolutional neural network model including a deep and shallow fusion module to obtain a deep and shallow feature matrix, including: Extracting a shallow feature matrix from the Mth layer of the first convolutional neural network model, where M is greater than or equal to 1 and less than or equal to 6; Extracting a deep feature matrix from the Nth layer of the first convolutional neural network model, where N / M is greater than or equal to 5 and less than or equal to 10; Using the deep-shallow feature fusion module of the first convolutional neural network model to fuse the shallow feature matrix and the deep feature matrix to obtain a deep-shallow feature matrix; Wherein, the classification feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the spinning speed at the current time point should be increased or decreased, including: Using the classifier to process the classification feature matrix according to the following classification formula to obtain the classification result; Wherein, the classification formula is: O = softmax{(W c ,B c )|Project(F)}, where Project(F) represents projecting the classification feature matrix into a vector, W c is the weight matrix, B c represents a bias vector, softmax represents a normalized exponential function, and O represents the classification result.

6. The control method of the polycrystalline mullite fiber production line according to claim 5, characterized in that: The deep and shallow feature matrix is ​​input into a second convolutional neural network model with a first scale to obtain a first scale correlation feature matrix, which is used for: The deep and shallow feature matrix is ​​input into the first convolution layer of the second convolutional neural network model using a first scale to obtain a first-scale correlation feature matrix, wherein the first convolution layer has a first two-dimensional convolution kernel of a first length.

Images