Process Optimization Method and System for Multilayer Flexible Printed Circuit Board Production Line

Abstract

The process optimization method and system for a multi-layer flexible printed circuit board production line provided by the present invention relate to the field of artificial intelligence technology. In the present invention, first, a lamination process vector corresponding to target lamination process data of lamination process parameters to be optimized is mined, and at least two data modalities included in the target lamination process data are determined; secondly, process optimization operations are respectively performed based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data to generate a first optimized lamination process parameter and a second optimized lamination process parameter; then, based on the first optimized lamination process parameter and the second optimized lamination process parameter, a target optimized lamination process parameter is determined to complete the optimization operation of the lamination process parameters to be optimized. Based on the above method, the problem of relatively low reliability of process optimization of multi-layer flexible printed circuit board production lines existing in the prior art can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology. Specifically, it relates to a process optimization method and system for a multi-layer flexible printed circuit board production line. Background Art

[0002] The lamination process is one of the key processes in the production of multi-layer flexible printed circuits (FPC). It plays a crucial role in stacking, pressing, and forming the final product structure of each layer of the circuit board. This process ensures that the layers of the circuit board can be firmly bonded and meet the requirements of electrical and mechanical properties. Therefore, the optimization of the lamination process is the quality guarantee of multi-layer flexible printed circuits. However, in the prior art, generally, relevant staff optimize the lamination process according to experience (such as the optimization of parameters such as temperature, pressure, and time). However, due to the limitations of relevant staff in terms of knowledge, experience, etc., it is difficult to effectively and reliably optimize the lamination process. Therefore, there is a problem that the reliability of the lamination process optimization is relatively low. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a process optimization method and system for a multi-layer flexible printed circuit board production line to improve the problem of relatively low reliability of the process optimization of the multi-layer flexible printed circuit board production line existing in the prior art.

[0004] To achieve the above purpose, the embodiments of the present invention adopt the following technical solutions:

[0005] A process optimization method for a multi-layer flexible printed circuit board production line includes:

[0006] Using the semantic mining model in the process optimization network, mining the lamination process vector corresponding to the target lamination process data of the lamination process parameters to be optimized, and determining at least two data modalities included in the target lamination process data, where the process optimization network belongs to a neural network and further includes a process optimization model, the lamination process parameters to be optimized are applied to a multi-layer flexible printed circuit board production line, the target lamination process data includes the first lamination process data and the second lamination process data obtained by performing lamination operations based on the lamination process parameters to be optimized respectively, the at least two data modalities include lamination process parameter text and lamination process monitoring images, the lamination process parameter text is at least used to describe temperature, pressure, and time, and the lamination process monitoring images are at least used to describe the lamination deformation of the multi-layer flexible printed circuit board;

[0007] Using the process optimization model, perform process optimization operations respectively based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data, and generate first optimized lamination process parameters and second optimized lamination process parameters;

[0008] Based on the first optimized lamination process parameters and the second optimized lamination process parameters, determine the target optimized lamination process parameters to complete the optimization operation of the lamination process parameters to be optimized, wherein the target optimized lamination process parameters are applied to a multi-layer flexible printed circuit board production line.

[0009] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the step of using the process optimization model to perform process optimization operations respectively based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data, and generating first optimized lamination process parameters and second optimized lamination process parameters includes:

[0010] Using the internal focusing unit included in the process optimization model, perform a focusing and mining operation within the data modality, output the internal focusing parameters corresponding to the at least two data modalities, and based on the internal focusing parameters and the lamination process vector, determine the lamination process internal focusing vector;

[0011] Using the external focusing unit included in the process optimization model, perform a focusing and mining operation between the data modalities, output the external focusing parameters corresponding to the lamination process internal focusing vector, and based on the external focusing parameters and the lamination process internal focusing vector, determine the lamination process external focusing vector;

[0012] Using the first decoding output unit included in the process optimization model, generate first optimized lamination process parameters based on the lamination process external focusing vector corresponding to the first lamination process data;

[0013] Using the second decoding output unit included in the process optimization model, generate second optimized lamination process parameters based on the lamination process external focusing vector corresponding to the second lamination process data.

[0014] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the at least two data modalities include a target data modality, wherein the step of using the internal focusing unit included in the process optimization model to perform a focusing and mining operation within the data modality, output the internal focusing parameters corresponding to the at least two data modalities, and based on the internal focusing parameters and the lamination process vector, determine the lamination process internal focusing vector includes:

[0015] Extract the lamination process local vector of the target lamination process data in the target data modality from the lamination process vector;

[0016] Perform a focus mining operation within the data modality to semantically focus the lamination process local vector in the target data modality, and output the internal focus sub-parameters corresponding to the target data modality;

[0017] After obtaining the internal focus sub-parameters corresponding to each data modality, combine them based on the internal focus sub-parameters corresponding to each data modality to form the internal focus parameters corresponding to the at least two data modalities;

[0018] Based on the internal focus parameters and the lamination process vector, determine the lamination process internal focus vector corresponding to the target lamination process data.

[0019] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the internal focus sub-parameters include a first internal focus sub-parameter and a second internal focus sub-parameter;

[0020] The step of performing a focus mining operation within the data modality to semantically focus the lamination process local vector in the target data modality and output the internal focus sub-parameters corresponding to the target data modality includes:

[0021] Perform a focus mining operation within the data modality to perform a vector compression operation on the lamination process local vector of the target data modality, and output the first internal focus sub-parameter corresponding to the target data modality, where the first internal focus sub-parameter includes one parameter value;

[0022] Determine the mean of each vector parameter in the lamination process local vector of the target data modality, and calculate the squared value of the difference between each vector parameter and the mean, and calculate the mean of each squared value to obtain the second internal focus sub-parameter corresponding to the target data modality, where the second internal focus sub-parameter includes one parameter value.

[0023] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the step of, after obtaining the internal focus sub-parameters corresponding to each data modality, combining them based on the internal focus sub-parameters corresponding to each data modality to form the internal focus parameters corresponding to the at least two data modalities includes:

[0024] Stitch the first internal focus sub-parameters corresponding to each data modality to form the corresponding first internal focus parameter, and stitch the second internal focus sub-parameters corresponding to each data modality to form the corresponding second internal focus parameter;

[0025] Multiply the first internal focusing parameter by a predetermined first weight parameter to output a corresponding first internally weighted focusing parameter, and multiply the second internal focusing parameter by a predetermined second weight parameter to output a corresponding second internally weighted focusing parameter;

[0026] Perform an addition operation on the first internally weighted focusing parameter and the second internally weighted focusing parameter, and perform a linear mapping operation on the result of the addition operation to form the internal focusing parameter corresponding to the at least two data modalities.

[0027] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the internal focusing parameter includes an internal focusing sub-parameter corresponding to each of the data modalities, wherein the step of determining the internal focusing vector of the lamination process corresponding to the target lamination process data based on the internal focusing parameter and the lamination process vector includes:

[0028] Multiply the internal focusing sub-parameter corresponding to the target data modality by the lamination process local vector in the target data modality to output the lamination process local focusing vector of the target data modality;

[0029] After outputting the lamination process local focusing vector corresponding to each of the data modalities, splice the lamination process local focusing vectors corresponding to each of the data modalities to form the lamination process internal focusing vector of the target lamination process data.

[0030] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the at least two data modalities include a first data modality and a second data modality, wherein the step of using the external focusing unit included in the process optimization model to perform a focusing mining operation between the data modalities, outputting the external focusing parameter corresponding to the lamination process internal focusing vector, and determining the lamination process external focusing vector based on the external focusing parameter and the lamination process internal focusing vector includes:

[0031] Extract, from the lamination process vector, a first lamination process local vector of the target lamination process data in the first data modality and a second lamination process local vector of the target lamination process data in the second data modality, respectively;

[0032] Using the first external focusing subunit in the external focusing unit included in the process optimization model, perform a focusing mining operation based on the first lamination process local vector and the second lamination process local vector, output the first external focusing parameter corresponding to the lamination process vector, and determine the first external focusing vector corresponding to the target lamination process data based on the first external focusing parameter and the lamination process vector;

[0033] Extract the lamination process local focusing vector of the target lamination process data in the first data modality and the lamination process local focusing vector of the target lamination process data in the second data modality respectively from the lamination process internal focusing vector;

[0034] Using the second external focusing subunit in the external focusing unit, perform a focusing mining operation based on the lamination process local focusing vector in the first data modality and the lamination process local focusing vector in the second data modality, output the second external focusing parameter corresponding to the lamination process internal focusing vector, and determine the second external focusing vector corresponding to the target lamination process data based on the second external focusing parameter and the lamination process internal focusing vector;

[0035] Concatenate the first external focusing vector and the second external focusing vector to obtain the lamination process external focusing vector corresponding to the target lamination process data.

[0036] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the step of using the first external focusing subunit in the external focusing unit included in the process optimization model to perform a focusing mining operation based on the first lamination process local vector and the second lamination process local vector, output the first external focusing parameter corresponding to the lamination process vector, and determine the first external focusing vector corresponding to the target lamination process data includes:

[0037] Based on the first lamination process local vector, perform a multi-level focusing mining operation on the second lamination process local vector, and output corresponding multi-level first focusing mining results;

[0038] Based on the second lamination process local vector, perform a multi-level focusing mining operation on the first lamination process local vector, and output corresponding multi-level second focusing mining results;

[0039] Aggregate the multi-level first focusing mining results and the multi-level second focusing mining results to obtain the first external focusing vector corresponding to the target lamination process data;

[0040] Among them, on the first focused mining path, in the focused mining operation of the first level, the dot product between the first lamination process local vector and the transposed vector of the second lamination process local vector is used as the corresponding first external focusing parameter, and the result of the weighted sum between this first external focusing parameter and the second lamination process local vector is used as the first focused mining result of the first level. In the focused mining operations of the second level and each subsequent level, dilated convolutions are respectively performed on the first lamination process local vector and the first focused mining result of the first level to form the corresponding first lamination process convolution vector and second lamination process convolution vector. Moreover, the dot product between this first lamination process convolution vector and the transposed vector of this second lamination process convolution vector is used as the corresponding first external focusing parameter, and the result of the weighted sum between this first external focusing parameter and this second lamination process convolution vector is used as the first focused mining result of the corresponding level. And, the dilation factor of the dilated convolution has a positive correlation with the corresponding level.

[0041] In some preferred embodiments, in the above process optimization method for a multi-layer flexible printed circuit board production line, the step of using the second external focusing subunit in the external focusing unit to perform a focused mining operation based on the lamination process local focusing vector in the first data modality and the lamination process local focusing vector in the second data modality, outputting the second external focusing parameter corresponding to the lamination process internal focusing vector, and determining the second external focusing vector corresponding to the target lamination process data based on the second external focusing parameter and the lamination process internal focusing vector includes:

[0042] Determine the dot product between the lamination process local focusing vector in the first data modality and the transposed vector of the lamination process local focusing vector in the second data modality, and use this dot product as the corresponding second external focusing parameter. Moreover, based on this second external focusing parameter, perform a weighted sum calculation on the lamination process local focusing vector in the second data modality to obtain the corresponding first second focused mining result;

[0043] Determine the dot product between the lamination process local focusing vector in the second data modality and the transposed vector of the lamination process local focusing vector in the first data modality, and use this dot product as the corresponding second external focusing parameter. Moreover, based on this second external focusing parameter, perform a weighted sum calculation on the lamination process local focusing vector in the first data modality to obtain the corresponding second second focused mining result;

[0044] Aggregate the first second focused mining result and the second second focused mining result to obtain the second external focusing vector corresponding to the target lamination process data.

[0045] An embodiment of the present invention further provides a process optimization system for a multi-layer flexible printed circuit board production line, including a processor and a memory. The memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-mentioned process optimization method for a multi-layer flexible printed circuit board production line.

[0046] For the process optimization method and system for a multi-layer flexible printed circuit board production line provided by the embodiments of the present invention, first, a lamination process vector corresponding to target lamination process data of the lamination process parameters to be optimized is mined, and at least two data modalities included in the target lamination process data are determined; secondly, process optimization operations are respectively performed based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data to generate a first optimized lamination process parameter and a second optimized lamination process parameter; then, based on the first optimized lamination process parameter and the second optimized lamination process parameter, a target optimized lamination process parameter is determined to complete the optimization operation of the lamination process parameters to be optimized. Based on the above method, on the one hand, the powerful learning ability of the neural network can be utilized to make the reliability of the optimization operation higher. On the other hand, during the optimization operation, optimization will be performed respectively based on the first lamination process data and the second lamination process data corresponding to the lamination process parameters to be optimized, so that the limitations and one-sidedness existing when optimizing based on a single lamination process data can be improved. In this way, the reliability of the optimization can also be further improved. In addition, since the lamination process data includes at least two data modalities, the aggregation of semantic information of multiple data modalities can be realized, thereby further improving the reliability of the optimization basis, and then improving the reliability of the optimization, so that the problem of relatively low reliability of the process optimization of the multi-layer flexible printed circuit board production line existing in the prior art can be fully improved to ensure the quality of multi-layer flexible printed circuit board production.

[0047] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, makes the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a structural block diagram of a process optimization system for a multi-layer flexible printed circuit board production line provided by an embodiment of the present invention.

[0049] Figure 2 It is a schematic flowchart of each step included in the process optimization method for a multi-layer flexible printed circuit board production line provided by an embodiment of the present invention.

[0050] Figure 3 It is a schematic diagram of the focused mining operations at multiple levels on the first focused mining path provided by an embodiment of the present invention.

[0051] Figure 4Schematic diagram of the multi-level focused mining operations on the second focused mining path provided by the embodiments of the present invention. Detailed implementation manners

[0052] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are only a part rather than all of the embodiments of the present invention. Components of the embodiments of the present invention described and illustrated herein usually can be arranged and designed in various different configurations.

[0053] Therefore, the detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0054] As Figure 1 shown, the embodiments of the present invention provide a process optimization system for a multi-layer flexible printed circuit board production line. Among them, the process optimization system for the multi-layer flexible printed circuit board production line may include a memory and a processor.

[0055] Specifically, the memory and the processor are directly or indirectly electrically connected to achieve data transmission or interaction. For example, they can be electrically connected through one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that can exist in the form of software or firmware. The processor can be used to execute the executable computer program stored in the memory, so as to implement the process optimization method for the multi-layer flexible printed circuit board production line provided by the embodiments of the present invention (as described later).

[0056] Optionally, the memory may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electric Erasable Programmable Read-Only Memory (EEPROM), etc. The processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), a System on Chip (SoC), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0057] And, Figure 1 The structure shown is only schematic, and the process optimization system for a multi-layer flexible printed circuit board production line may further include more or fewer components than those shown in Figure 1 or have a different configuration from that shown in Figure 1 For example, it may include a communication unit for information interaction with other devices. Among them, in an alternative example, the process optimization system for a multi-layer flexible printed circuit board production line may be a server with data processing capabilities.

[0058] Combined with Figure 2 , an embodiment of the present invention further provides a process optimization method for a multi-layer flexible printed circuit board production line, which can be applied to the above-mentioned process optimization system for a multi-layer flexible printed circuit board production line. Among them, the method steps defined by the process related to the process optimization method for a multi-layer flexible printed circuit board production line can be implemented by the process optimization system for a multi-layer flexible printed circuit board production line (which can be abbreviated as the process optimization system hereinafter).

[0059] Next, the specific process shown in Figure 2 will be elaborated in detail.

[0060] Step S110: Use the semantic mining model in the process optimization network to mine the lamination process vector corresponding to the target lamination process data of the lamination process parameters to be optimized, and determine at least two data modalities included in the target lamination process data.

[0061] In an embodiment of the present invention, the process optimization system may utilize a semantic mining model in a process optimization network to mine a lamination process vector corresponding to target lamination process data of lamination process parameters to be optimized, and determine at least two data modalities included in the target lamination process data. Wherein, the process optimization network belongs to a neural network and further includes a process optimization model, the lamination process parameters to be optimized are applied to a multi-layer flexible printed circuit board production line, and the target lamination process data includes first lamination process data and second lamination process data obtained by performing lamination operations respectively based on the lamination process parameters to be optimized (which may be two tests based on the lamination process parameters to be optimized, or actual production based on the lamination process parameters to be optimized. It can be understood that in other embodiments, based on reliability requirements, third lamination process data, fourth lamination process data, fifth lamination process data, etc. may also be included), the at least two data modalities include lamination process parameter text and lamination process monitoring images, the lamination process parameter text is at least used to describe temperature, pressure and time (or may also include other parameters, such as adhesive film materials, such as polyimide film or epoxy resin film, which will bond each layer under high temperature and high pressure), and the lamination process monitoring images are at least used to describe the lamination deformation of the multi-layer flexible printed circuit board (such as obtained by performing an image acquisition operation on the multi-layer flexible printed circuit board after the lamination process is completed). Additionally, as an example, the lamination process parameter text may include the following content:

[0062] Lamination temperature: Set temperature: 180 °C; Heat preservation time: 20 minutes; Temperature control: Use a uniform heating system;

[0063] Lamination pressure: Pressure: 2.1 MPa; Pressure holding time: Hold pressure for 20 minutes;

[0064] Lamination time: Total lamination time: 50 minutes, including heating-up time, heat preservation time, and cooling time; Heating-up time: 10 minutes, heating from normal temperature (20 °C) to 180 °C; Heat preservation time: 20 minutes, maintaining a constant temperature at the set temperature;

[0065] Cooling method: Natural cooling until the temperature drops to normal temperature (20 °C);

[0066] Vacuum degree: Use a vacuum bag to remove air, and the vacuum degree is -0.9 bar (relative air pressure).

[0067] It should be noted that since the target lamination process data includes at least two data modalities, and the corresponding semantic mining methods are generally different for different data modalities. Based on this, the semantic mining model may include a first semantic mining unit and a second semantic mining unit. Among them, the first semantic mining unit can be a word embedding model, which can be used to perform word embedding processing on the lamination process parameter text to obtain corresponding word embedding vectors. Then, the word embedding vectors are concatenated or averaged to obtain corresponding local lamination process vectors. The second semantic mining unit can be a convolutional neural network, which can be used to perform convolutional processing on the lamination process monitoring image to obtain corresponding local lamination process vectors. Taking "using a vacuum bag to remove air, with a vacuum degree of -0.9 bar" as an example, the word embedding processing is illustrated, and the corresponding word embedding vectors can be obtained:

[0068] "Use": [0.2, -0.1, 0.3, 0.5, -0.2, 0.1, -0.1, 0.6, 0.4, -0.3, 0.3,......, 0.2];

[0069] "Vacuum bag": [0.1, -0.3, 0.4, 0.6, -0.1, 0.5, 0.2, 0.4, 0.1, -0.2, 0.4,......, 0.5];

[0070] "To": [0.05, 0.1, -0.2, 0.4, 0.3, 0.2, 0.6, -0.1, 0.5, -0.3, 0.2,......, 0.1];

[0071] "Remove": [0.3, -0.1, 0.5, 0.2, -0.3, 0.1, 0.3, 0.4, 0.2, 0.6, -0.1,......, 0.3];

[0072] "Air": [0.6, -0.1, 0.2, 0.4, 0.1, -0.3, 0.4, 0.2, 0.3, 0.1, -0.2,......, 0.5];

[0073] "Vacuum degree": [0.4, -0.2, 0.3, 0.6, 0.1, 0.5, 0.2, 0.7, -0.1, 0.4, 0.3,......, -0.2];

[0074] "Is": [0.2, 0.1, 0.3, 0.4, -0.1, 0.2, 0.5, -0.3, 0.4, 0.6, -0.2,......, 0.1];

[0075] "-0.9": [0.1, 0.3, -0.2, 0.5, 0.3, -0.1, 0.2, 0.4, -0.3, 0.1, 0.2,......, 0.0];

[0076] "bar": [0.3, 0.1, 0.5, -0.1, 0.4, 0.2, 0.4, 0.5, -0.3, 0.3, 0.2,......, -0.1].

[0077] Step S120: Using the process optimization model, perform process optimization operations respectively based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data, and generate the first optimized lamination process parameters and the second optimized lamination process parameters.

[0078] In the embodiment of the present invention, after obtaining the lamination process vector, the process optimization system may use the process optimization model to perform process optimization operations respectively based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data, and generate the first optimized lamination process parameters and the second optimized lamination process parameters. That is to say, a process optimization operation may be performed based on the lamination process vector corresponding to the first lamination process data to generate the first optimized lamination process parameters, and, a process optimization operation may be performed based on the lamination process vector corresponding to the second lamination process data to generate the second optimized lamination process parameters.

[0079] Step S130: Based on the first optimized lamination process parameters and the second optimized lamination process parameters, determine the target optimized lamination process parameters to complete the optimization operation of the lamination process parameters to be optimized.

[0080] In the embodiment of the present invention, after obtaining the first optimized lamination process parameters and the second optimized lamination process parameters, the process optimization system may determine the target optimized lamination process parameters based on the first optimized lamination process parameters and the second optimized lamination process parameters to complete the optimization operation of the lamination process parameters to be optimized; exemplarily, for parameters that can be quantified, such as time, temperature, pressure, etc., the first optimized lamination process parameters and the second optimized lamination process parameters may be averaged to obtain the target optimized lamination process parameters, and for parameters that cannot be quantified, such as the adhesive film material, any one of the first optimized lamination process parameters and the second optimized lamination process parameters may be used as the target optimized lamination process parameters. Among them, the target optimized lamination process parameters are applied to the multi-layer flexible printed circuit board production line.

[0081] Based on the above method, on the one hand, the powerful learning ability of the neural network can be utilized to make the reliability of the optimization operation higher. On the other hand, during the optimization operation, the optimization is carried out respectively based on the first lamination process data and the second lamination process data corresponding to the lamination process parameters to be optimized, so that the limitations and one-sidedness existing in the optimization based on a single lamination process data can be improved. In this way, the reliability of the optimization can be further improved. In addition, since the lamination process data includes at least two data modalities, the aggregation of semantic information of multiple data modalities can be realized, thereby further improving the reliability of the optimization basis, and further improving the reliability of the optimization, so that the problem of relatively low reliability of the process optimization of the multi-layer flexible printed circuit board production line existing in the prior art can be fully improved to ensure the quality of the multi-layer flexible printed circuit board production.

[0082] Based on the above embodiments, it should be noted that for step S120, the specific manner of performing the process optimization operation based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data is not limited.

[0083] For example, in a specific application scenario, in order to improve the reliability of the process optimization operation, the above step S120 may further include step S121, step S122, step S123, and step S124. The implementation content of each step is described as follows.

[0084] Step S121, using the internal focusing unit included in the process optimization model, perform a focusing and mining operation within the data modality, output the internal focusing parameters corresponding to the at least two data modalities, and determine the lamination process internal focusing vector based on the internal focusing parameters and the lamination process vector.

[0085] In the embodiment of the present invention, the internal focusing unit included in the process optimization model can be used to perform a focusing and mining operation within the data modality, output the internal focusing parameters corresponding to the at least two data modalities, and determine the lamination process internal focusing vector based on the internal focusing parameters and the lamination process vector. That is to say, for each data modality, a focusing and mining operation can be performed within the data modality to extract the key information within the data modality.

[0086] Step S122, using the external focusing unit included in the process optimization model, perform a focusing and mining operation between the data modalities, output the external focusing parameters corresponding to the lamination process internal focusing vector, and determine the lamination process external focusing vector based on the external focusing parameters and the lamination process internal focusing vector.

[0087] In an embodiment of the present invention, the external focusing unit included in the process optimization model may also be used to perform a focusing mining operation between data modalities, output an external focusing parameter corresponding to the internal focusing vector of the lamination process, and determine an external focusing vector of the lamination process based on the external focusing parameter and the internal focusing vector of the lamination process. That is, a focusing mining operation may be performed between data modalities so that the information associated between data modalities is mined out.

[0088] Step S123: Use the first decoding output unit included in the process optimization model to generate a first optimized lamination process parameter based on the external focusing vector of the lamination process corresponding to the first lamination process data.

[0089] In an embodiment of the present invention, after obtaining the external focusing vector of the lamination process corresponding to the first lamination process data, the first decoding output unit included in the process optimization model may be used to generate a first optimized lamination process parameter based on the external focusing vector of the lamination process corresponding to the first lamination process data.

[0090] Step S124: Use the second decoding output unit included in the process optimization model to generate a second optimized lamination process parameter based on the external focusing vector of the lamination process corresponding to the second lamination process data.

[0091] In an embodiment of the present invention, after obtaining the lamination process external focus vector corresponding to the second lamination process data, the second decoding output unit included in the process optimization model can be used to generate a second optimized lamination process parameter based on the lamination process external focus vector corresponding to the second lamination process data. It should be noted that the first decoding output unit and the second decoding output unit can be the same decoding output unit or different decoding output units. For example, they can have the same network architecture, but the network parameters can be different. The specific network parameters can be formed during the training process. Among them, the decoding output unit is a decoder, and its working principle can be that at the first time step, the input vector of the decoder (such as the lamination process external focus vector) will be mapped to the size of the vocabulary (i.e., the number of words in the vocabulary) through a linear transformation (usually a fully connected layer), and then, the result of the full connection is converted into a probability distribution through the Softmax function, that is, the probability of each word in the vocabulary, and then, the word with the largest probability is used as the first output word, and then, at the second time step, the first output word can be embedded to obtain the corresponding word embedding vector, and the word embedding vector and the input vector of the decoder at the first time step are cross-attention processed or spliced ​​+ convolution processed to obtain the input vector of the decoder at the second time step, and then, based on the input vector, the second output word is determined, and in this way, the corresponding optimized lamination process parameters can be generated by cyclic processing in sequence. Among them, the specific working principle of the decoder can refer to the relevant prior art, and no specific limitation and description are made here. The focus of the present invention is how to dig out a reliable lamination process external focus vector.

[0092] It can be selected that, in the above step S121, the specific manner of performing the focus mining operation within the data modality is not limited. For example, in a specific application scenario, the at least two data modalities include a target data modality (the target data modality can be any data modality). In order to be able to perform a focus mining operation within the data modality to mine the internal focus vector of the lamination process with higher semantic representation accuracy, the above step S121 can further include step S121a, step S121b, step S121c and step S121d. The implementation content of each step is as follows.

[0093] Step S121a: extracting the lamination process local vector of the target lamination process data in the target data mode from the lamination process vector.

[0094] In an embodiment of the present invention, the lamination process local vector of the target lamination process data in the target data modality can be extracted from the lamination process vector, such as the lamination process local vector corresponding to the lamination process parameter text (i.e., the semantic mining result of the semantic mining model), or the lamination process local vector corresponding to the lamination process monitoring image.

[0095] Step S121b, perform a focusing mining operation within the data modality to perform a semantic focusing operation on the lamination process local vector in the target data modality, and output the internal focusing sub-parameters corresponding to the target data modality.

[0096] In an embodiment of the present invention, a focusing mining operation can be performed within the data modality to perform a semantic focusing operation on the lamination process local vector in the target data modality, and output the internal focusing sub-parameters corresponding to the target data modality, that is, the key semantic information in the lamination process local vector in the target data modality is mined and represented by the internal focusing sub-parameters.

[0097] Step S121c, after obtaining the internal focusing sub-parameters corresponding to each data modality, based on the internal focusing sub-parameters corresponding to each data modality, combine to form the internal focusing parameters corresponding to the at least two data modalities.

[0098] In an embodiment of the present invention, after obtaining the internal focusing sub-parameters corresponding to each data modality (i.e., each data modality is respectively used as the target data modality), based on the internal focusing sub-parameters corresponding to each data modality, combine to form the internal focusing parameters corresponding to the at least two data modalities. For example, when including two data modalities, the internal focusing parameters may include two internal focusing sub-parameters; when including three data modalities, the internal focusing parameters may include three internal focusing sub-parameters; when including four data modalities, the internal focusing parameters may include four internal focusing sub-parameters.

[0099] Step S121d, based on the internal focusing parameters and the lamination process vector, determine the lamination process internal focusing vector corresponding to the target lamination process data.

[0100] In an embodiment of the present invention, the lamination process internal focusing vector corresponding to the target lamination process data can be determined based on the internal focusing parameters and the lamination process vector. In this way, the important information characterized by the internal focusing parameters can be fused into the lamination process vector, so that the semantic representation ability of the obtained lamination process internal focusing vector can be better.

[0101] Optionally, in the above step S121b, the specific manner of performing semantic focusing operation on the lamination process local vector in the target data modality is not limited. For example, in a specific application scenario, in order to mine more important information, the internal focusing sub-parameters include a first internal focusing sub-parameter and a second internal focusing sub-parameter. Based on this, the above step S121b may further include the following implementable content:

[0102] First, a focusing and mining operation can be performed inside the data modality to perform vector compression operation on the lamination process local vector of the target data modality, and output the first internal focusing sub-parameter corresponding to the target data modality, where the first internal focusing sub-parameter includes a parameter value; Exemplarily, the mean value of the lamination process local vector can be calculated to obtain the corresponding first internal focusing sub-parameter. In this way, while realizing data compression, the lamination process local vector can also be effectively characterized by the first internal focusing sub-parameter;

[0103] Second, the mean value of each vector parameter in the lamination process local vector of the target data modality can be determined, and the square value of the difference between each vector parameter and the mean value is calculated, and the mean value of each square value is calculated to obtain the second internal focusing sub-parameter corresponding to the target data modality, where the second internal focusing sub-parameter includes a parameter value. In this way, while realizing data compression, the volatility between the parameters in the lamination process local vector can also be reflected by the second internal focusing sub-parameter, realizing the characterization of important semantic information.

[0104] Optionally, in the above step S121c, the specific manner of combining to form the internal focusing parameters corresponding to the at least two data modalities is not limited. For example, in a specific application scenario, in order to further improve the characterization ability of the internal focusing parameters, the above step S121c may further include the following implementable content:

[0105] First, the first internal focusing sub-parameters corresponding to each data modality can be spliced to form the corresponding first internal focusing parameter (such as {the first internal focusing sub-parameter corresponding to data modality 1, the first internal focusing sub-parameter corresponding to data modality 2}), and the second internal focusing sub-parameters corresponding to each data modality are spliced to form the corresponding second internal focusing parameter (such as {the second internal focusing sub-parameter corresponding to data modality 1, the second internal focusing sub-parameter corresponding to data modality 2});

[0106] Secondly, the first internal focusing parameter can be multiplied by a predetermined first weight parameter to output a corresponding first internally weighted focusing parameter, and the second internal focusing parameter can be multiplied by a predetermined second weight parameter to output a corresponding second internally weighted focusing parameter; Exemplarily, the first weight parameter and the second weight parameter can be used as network parameters of the internal focusing unit and formed during corresponding training processes. Thus, due to different focusing methods, the importance of the information concerned is different. Therefore, it can be characterized by corresponding weight parameters;

[0107] Then, an addition operation can be performed on the first internally weighted focusing parameter and the second internally weighted focusing parameter (for example, the two sub-parameters corresponding to data modality 1 can be added, and the two sub-parameters corresponding to data modality 2 can be added), and a linear mapping operation can be performed on the result of the addition operation (such as implemented through a corresponding linear activation function, and the specific linear activation function can be selected according to actual needs) to form the internal focusing parameters corresponding to the at least two data modalities.

[0108] Optionally, in step S121d above, the specific manner of determining the lamination process internal focusing vector corresponding to the target lamination process data is not limited. For example, in a specific application scenario, the internal focusing parameter includes an internal focusing sub-parameter corresponding to each of the data modalities. Thus, step S121d above can include the following implementable content:

[0109] First, the internal focusing sub-parameter corresponding to the target data modality can be multiplied by the lamination process local vector in the target data modality to output the lamination process local focusing vector of the target data modality, that is, the internal focusing sub-parameter is used as the weight of the corresponding data modality to weight the corresponding lamination process local vector. Thus, due to the different importance of different data modalities, through weighting, the representation accuracy of the corresponding vector can be made higher;

[0110] Secondly, after the lamination process local focusing vector corresponding to each of the data modalities is output, the lamination process local focusing vectors corresponding to each of the data modalities can be concatenated to form the lamination process internal focusing vector of the target lamination process data, such as {the lamination process local focusing vector corresponding to data modality 1, the lamination process local focusing vector corresponding to data modality 2}. For example, when the size of the lamination process local focusing vector is A*B, if the data modality is 2, the size of the lamination process internal focusing vector is 2*A*B; if the data modality is 3, the size of the lamination process internal focusing vector is 3*A*B; if the data modality is 4, the size of the lamination process internal focusing vector is 4*A*B.

[0111] Optionally, in the above step S122, the specific manner of performing the focusing and mining operation between data modalities is not limited. For example, in a specific application scenario, the at least two data modalities include a first data modality and a second data modality. Based on this, in order to achieve the full integration of semantic information between different data modalities, the above step S122 may further include step S122a, step S122b, step S122c, step S122d, and step S122e. The implementation content of each step is described as follows.

[0112] Step S122a: respectively extract the first lamination process partial vector of the target lamination process data in the first data modality and the second lamination process partial vector of the target lamination process data in the second data modality from the lamination process vector.

[0113] In the embodiment of the present invention, respectively extract the first lamination process partial vector of the target lamination process data in the first data modality and the second lamination process partial vector of the target lamination process data in the second data modality from the lamination process vector. Exemplarily, the first lamination process partial vector may be the lamination process partial vector corresponding to the lamination process parameter text, and the second lamination process partial vector may be the lamination process partial vector corresponding to the lamination process monitoring image.

[0114] Step S122b: use the first external focusing subunit in the external focusing unit included in the process optimization model to perform a focusing and mining operation based on the first lamination process partial vector and the second lamination process partial vector, output the first external focusing parameter corresponding to the lamination process vector, and determine the first external focusing vector corresponding to the target lamination process data based on the first external focusing parameter and the lamination process vector.

[0115] In an embodiment of the present invention, after obtaining the first lamination process local vector and the second lamination process local vector, the first external focusing subunit in the external focusing unit included in the process optimization model may be used to perform a focusing and mining operation based on the first lamination process local vector and the second lamination process local vector, output the first external focusing parameter corresponding to the lamination process vector, and determine the first external focusing vector corresponding to the target lamination process data based on the first external focusing parameter and the lamination process vector. That is to say, the first lamination process local vector and the second lamination process local vector may be associated and fused, so that the semantic information of different data modalities can be fused. In addition, directly processing the original first lamination process local vector and the second lamination process local vector can capture the associated information in the original semantic information and avoid the problem of semantic loss during the deep processing process.

[0116] Step S122c: respectively extract the lamination process local focusing vector of the target lamination process data in the first data modality and the lamination process local focusing vector of the target lamination process data in the second data modality from the lamination process internal focusing vector.

[0117] In an embodiment of the present invention, the lamination process local focusing vector of the target lamination process data in the first data modality and the lamination process local focusing vector of the target lamination process data in the second data modality may also be respectively extracted from the lamination process internal focusing vector. Exemplarily, the first data modality may refer to the modality corresponding to the lamination process parameter text, and the second data modality may refer to the modality corresponding to the lamination process monitoring image.

[0118] Step S122d: use the second external focusing subunit in the external focusing unit to perform a focusing and mining operation based on the lamination process local focusing vector in the first data modality and the lamination process local focusing vector in the second data modality, output the second external focusing parameter corresponding to the lamination process internal focusing vector, and determine the second external focusing vector corresponding to the target lamination process data based on the second external focusing parameter and the lamination process internal focusing vector.

[0119] In an embodiment of the present invention, after obtaining the local focusing vectors of the lamination process in two data modalities, the second external focusing subunit in the external focusing unit may be used to perform a focusing mining operation based on the local focusing vector of the lamination process in the first data modality and the local focusing vector of the lamination process in the second data modality, output the second external focusing parameter corresponding to the internal focusing vector of the lamination process, and determine the second external focusing vector corresponding to the target lamination process data based on the second external focusing parameter and the internal focusing vector of the lamination process. That is, the local focusing vectors of the lamination process in two data modalities can be associated and fused, so that the semantic information of different data modalities can be fused.

[0120] Step S122e, splice the first external focusing vector and the second external focusing vector to obtain the external focusing vector of the lamination process corresponding to the target lamination process data.

[0121] In an embodiment of the present invention, after obtaining the first external focusing vector and the second external focusing vector, the first external focusing vector and the second external focusing vector may be spliced to obtain the external focusing vector of the lamination process corresponding to the target lamination process data. Exemplarily, in some applications, for the spliced vector, convolution, pooling, etc. may also be performed to obtain the corresponding external focusing vector of the lamination process. Based on this, since the first external focusing vector is obtained by focusing on the original semantic information, more detailed information can be obtained. In addition, since the second external focusing vector is obtained by focusing on the internally focused semantic information and has a certain depth, more abstract information can be obtained. Thus, after splicing the two external focusing vectors, the formed external focusing vector of the lamination process can take into account both detailed information and abstract information, thereby having better semantic representation ability. In addition, due to the existence of the first external focusing vector, problems such as overfitting caused by using only the second external focusing vector can be improved, and the problem of distorted semantic information extraction can be avoided.

[0122] Optionally, in the above step S122b, the specific manner of performing the focusing mining operation based on the first local lamination process vector and the second local lamination process vector is not limited. For example, in a specific application scenario, in order to achieve the full fusion of the first local lamination process vector and the second local lamination process vector, the above step S122b may further include the following implementable contents:

[0123] First, multiple levels of focused mining operations can be performed on the second lamination process local vector based on the first lamination process local vector, and the corresponding first focused mining results of multiple levels are output, where the semantic information represented by the first focused mining result of each level is different;

[0124] Second, multiple levels of focused mining operations can be performed on the first lamination process local vector based on the second lamination process local vector, and the corresponding second focused mining results of multiple levels are output, where the semantic information represented by the second focused mining result of each level is different;

[0125] Then, aggregate the first focused mining results of multiple levels and the second focused mining results of multiple levels to obtain the first external focused vector corresponding to the target lamination process data; Exemplarily, after adjusting the sizes of the first focused mining results of multiple levels and the second focused mining results of multiple levels to be the same, the adjusted first focused mining results of multiple levels and the second focused mining results of multiple levels are averaged to obtain the corresponding first external focused vector. For example, the focused mining results with a large size can be pooled so that the pooled result can have the same size as the focused mining results with a small size.

[0126] Among them, combined with Figure 3, on the first focused mining path, in the focused mining operation of the first level, the dot product between the first lamination process local vector and the transposed vector of the second lamination process local vector is used as the corresponding first external focus parameter, and the result of the weighted sum between this first external focus parameter and the second lamination process local vector is used as the first focused mining result of the first level. In the focused mining operations of the second level and each subsequent level, dilated convolutions are respectively performed on the first lamination process local vector and the first focused mining result of the first level to form the corresponding first lamination process convolution vector and second lamination process convolution vector. Also, the dot product between this first lamination process convolution vector and the transposed vector of this second lamination process convolution vector is used as the corresponding first external focus parameter, and the result of the weighted sum between this first external focus parameter and this second lamination process convolution vector is used as the first focused mining result of the corresponding level. Moreover, the dilation factor of the dilated convolution has a positive correlation with the corresponding level. That is to say, as the level increases, the dilation factor of the dilated convolution can gradually increase. For example, in the focused mining operation of the second level, the dilation factor of the dilated convolution can be equal to 1, which belongs to standard convolution and the convolution kernel has no holes. In the focused mining operation of the third level, the dilation factor of the dilated convolution can be equal to 2, and one hole is inserted between every two adjacent elements of the convolution kernel. In this way, by increasing the holes, the receptive field of the convolution operation (i.e., the range that can perceive the input data) can be increased, enabling the network to capture more context information. In this way, for the focused mining operations of different levels, different semantic information can be fused, so that more semantic information can be represented by the first focused mining results obtained at multiple levels.

[0127] In some other application scenarios, on the first focused mining path, in the focused mining operation of the first level, the dot product between the first lamination process local vector and the transposed vector of the second lamination process local vector is used as the corresponding first external focus parameter, and the result of the weighted sum between this first external focus parameter and the second lamination process local vector is used as the first focused mining result of the first level; in the focused mining operation of the second level, dilated convolutions are respectively performed on the first lamination process local vector and the first focused mining result of the first level to obtain the first lamination process convolution vector and the second lamination process convolution vector of the second level, and the dot product between this first lamination process convolution vector and the transposed vector of this second lamination process convolution vector is used as the corresponding first external focus parameter, and the result of the weighted sum between this first external focus parameter and this second lamination process convolution vector is used as the first focused mining result of the second level; in the focused mining operation of each level from the third level onwards, dilated convolutions are respectively performed on the first lamination process convolution vector of the previous level and the first focused mining result of the previous level to form the first lamination process convolution vector and the second lamination process convolution vector of the current level, and the dot product between this first lamination process convolution vector and the transposed vector of this second lamination process convolution vector is used as the corresponding first external focus parameter, and the result of the weighted sum between this first external focus parameter and this second lamination process convolution vector is used as the first focused mining result of the current level.

[0128] Similarly, in combination with Figure 4 , on the second focused mining path, in the focused mining operation of the first level, the dot product between the second lamination process local vector and the transposed vector of the first lamination process local vector is used as the corresponding first external focus parameter, and the result of the weighted sum between this first external focus parameter and the first lamination process local vector is used as the second focused mining result of the first level. In the focused mining operation of each level from the second level onwards, dilated convolutions are respectively performed on the second lamination process local vector and the second focused mining result of the first level to form the corresponding second lamination process convolution vector and the first lamination process convolution vector, and the dot product between this second lamination process convolution vector and the transposed vector of this first lamination process convolution vector is used as the corresponding first external focus parameter, and the result of the weighted sum between this first external focus parameter and this first lamination process convolution vector is used as the second focused mining result of the corresponding level. Moreover, the dilation factor of the dilated convolution has a positive correlation with the corresponding level, and the specific content is as described in the previous related description.

[0129] Optionally, in the above step S122d, the specific manner of performing the focus mining operation based on the lamination process local focus vector in the first data modality and the lamination process local focus vector in the second data modality is not limited. For example, in a specific application scenario, in order to avoid overfitting and semantic distortion while achieving sufficient fusion of semantic information in different data modalities, the above step S122d may include the following implementable content:

[0130] First, the dot product between the lamination process local focus vector in the first data modality and the transposed vector of the lamination process local focus vector in the second data modality can be determined, and this dot product can be used as the corresponding second external focus parameter. Then, based on this second external focus parameter, a weighted sum calculation is performed on the lamination process local focus vector in the second data modality to obtain the corresponding first second focus mining result;

[0131] Second, the dot product between the lamination process local focus vector in the second data modality and the transposed vector of the lamination process local focus vector in the first data modality can be determined, and this dot product can be used as the corresponding second external focus parameter. Then, based on this second external focus parameter, a weighted sum calculation is performed on the lamination process local focus vector in the first data modality to obtain the corresponding second second focus mining result. It should be noted that in some other application scenarios, before performing the dot product and weighted sum calculations, a vector mapping operation can also be performed first. Taking the second external focus sub-unit as an example, the second external focus sub-unit may carry a first mapping vector, a second mapping vector, and a third mapping vector (which can be formed during network training). The lamination process local focus vector in the second data modality can be multiplied by the first mapping vector in matrix form to obtain a first focus mapping vector. The lamination process local focus vector in the first data modality can be multiplied by the second mapping vector and the third mapping vector respectively in matrix form to obtain the corresponding second focus mapping vector and third focus mapping vector. Then, the dot product between the first focus mapping vector and the transposed vector of the second focus mapping vector can be used as the second external focus parameter. Then, based on this second external focus parameter, a weighted sum calculation is performed on the third focus mapping vector to obtain the corresponding second second focus mining result;

[0132] Finally, the first second focus mining result and the second second focus mining result can be aggregated to obtain the second external focus vector corresponding to the target lamination process data. Exemplarily, the first second focus mining result and the second second focus mining result can be averaged, or they can also be concatenated, convolved, pooled, etc. to obtain the corresponding second external focus vector.

[0133] Finally, it should be noted that, in order to ensure the reliability of data processing of the process optimization network, before performing the above step S110, the process optimization network can also be trained. Specifically, the first lamination process sample and the second lamination process sample corresponding to the lamination process sample to be optimized can be obtained, and then, semantic mining is performed based on the semantic mining model to obtain the corresponding first sample lamination process vector and second sample lamination process vector. Then, the process optimization model can be used to process the first sample lamination process vector and the second sample lamination process vector respectively to obtain the corresponding first optimized lamination process sample and second optimized lamination process sample. After that, the target optimized lamination process sample is determined based on the first optimized lamination process sample and the second optimized lamination process sample. Finally, the network parameters of the process optimization network can be updated along the direction of reducing the error between the target optimized lamination process sample and the optimized lamination process label corresponding to the lamination process sample to be optimized (which can be determined based on other reliable neural networks or corresponding experts), until the error is reduced to a preset value, etc., thereby completing the training of the network.

[0134] In summary, for the process optimization method and system for a multi-layer flexible printed circuit board production line provided by the present invention, first, the lamination process vector corresponding to the target lamination process data of the lamination process parameters to be optimized is mined, and at least two data modalities included in the target lamination process data are determined; secondly, process optimization operations are respectively performed based on the lamination process vector corresponding to the first lamination process data and the lamination process vector corresponding to the second lamination process data to generate the first optimized lamination process parameters and the second optimized lamination process parameters; then, based on the first optimized lamination process parameters and the second optimized lamination process parameters, the target optimized lamination process parameters are determined to complete the optimization operation of the lamination process parameters to be optimized. Based on the above method, on the one hand, the powerful learning ability of the neural network can be utilized to make the reliability of the optimization operation higher. On the other hand, during the optimization operation, the optimization is performed respectively based on the first lamination process data and the second lamination process data corresponding to the lamination process parameters to be optimized, so that the limitations and one-sidedness existing in the optimization based on a single lamination process data can be improved. In this way, the reliability of the optimization can also be further improved. In addition, since the lamination process data includes at least two data modalities, the aggregation of semantic information of multiple data modalities can be realized, thereby further improving the reliability of the optimization basis and further improving the reliability of the optimization, so that the problem of relatively low reliability of the process optimization of the multi-layer flexible printed circuit board production line existing in the prior art can be fully improved to ensure the quality of the multi-layer flexible printed circuit board production.

[0135] In several embodiments provided by the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0136] In addition, each functional module in various embodiments of the present invention can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part.

[0137] If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, an electronic device, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes. It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to this process, method, article, or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article, or device including the said element.

[0138] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A process optimization method for a multi-layer flexible circuit board production line, characterized in that: The process optimization method for a multi-layer flexible circuit board production line comprises: Using a semantic mining model in a process optimization network, a lamination process vector corresponding to target lamination process data of a lamination process parameter to be optimized is mined, and at least two data modes included in the target lamination process data are determined, wherein the process optimization network belongs to a neural network and also includes a process optimization model, the lamination process parameters to be optimized are applied to a multi-layer flexible circuit board production line, the target lamination process data include first lamination process data and second lamination process data obtained by lamination operations based on the lamination process parameters to be optimized, respectively, the first lamination process data and the second lamination process data are formed by performing two tests or actual production, the at least two data modes include a lamination process parameter text and a lamination process monitoring image, the lamination process parameter text is at least used to describe temperature, pressure and time, and the lamination process monitoring image is at least used to describe the lamination deformation of the multi-layer flexible circuit board; Using the internal focusing unit included in the process optimization model, a focusing mining operation is performed inside the data modality, and the internal focusing parameters corresponding to the at least two data modalities are output, and based on the internal focusing parameters and the lamination process vector, the internal focusing vector of the lamination process is determined; using the external focusing unit included in the process optimization model, a focusing mining operation is performed between data modalities, and the external focusing parameters corresponding to the internal focusing vector of the lamination process are output, and based on the external focusing parameters and the internal focusing vector of the lamination process, the external focusing vector of the lamination process is determined; using the first decoding output unit included in the process optimization model, based on the external focusing vector of the lamination process corresponding to the first lamination process data, a first optimized lamination process parameter is generated; using the second decoding output unit included in the process optimization model, based on the external focusing vector of the lamination process corresponding to the second lamination process data, a second optimized lamination process parameter is generated; wherein, the at least two data modalities include a target Data modality, the step of using the internal focusing unit included in the process optimization model to perform focusing mining operations inside the data modality, outputting internal focusing parameters corresponding to the at least two data modalities, and determining the internal focusing vector of the lamination process based on the internal focusing parameters and the lamination process vector, includes: extracting the lamination process local vector of the target lamination process data in the target data modality from the lamination process vector; performing focusing mining operations inside the data modality to perform semantic focusing operations on the lamination process local vector in the target data modality, and outputting the internal focusing sub-parameters corresponding to the target data modality; after obtaining the internal focusing sub-parameters corresponding to each data modality, based on the internal focusing sub-parameters corresponding to each data modality, combining to form the internal focusing parameters corresponding to the at least two data modalities; determining the internal focusing vector of the lamination process corresponding to the target lamination process data based on the internal focusing parameters and the lamination process vector; Based on the first optimized lamination process parameter and the second optimized lamination process parameter, a target optimized lamination process parameter is determined to complete the optimization operation of the lamination process parameter to be optimized, wherein the target optimized lamination process parameter is applied to a multi-layer flexible circuit board production line.

2. The process optimization method for a multi-layer flexible circuit board production line according to claim 1, characterized in that: The internal focus sub-parameters include a first internal focus sub-parameter and a second internal focus sub-parameter; The step of performing a focus mining operation within the data modality to perform a semantic focus operation on the lamination process local vector in the target data modality and outputting the internal focus sub-parameters corresponding to the target data modality includes: Performing a focus mining operation inside the data modality to perform a vector compression operation on a lamination process local vector of the target data modality, and outputting a first internal focus sub-parameter corresponding to the target data modality, wherein the first internal focus sub-parameter includes a parameter value; Determine the mean of each vector parameter in the local vector of the lamination process of the target data modality, calculate the square value of the difference between each vector parameter and the mean, and perform mean calculation on each square value to obtain a second internal focusing sub-parameter corresponding to the target data modality, wherein the second internal focusing sub-parameter includes a parameter value.

3. The process optimization method for a multi-layer flexible circuit board production line according to claim 2, characterized in that: After obtaining the internal focus sub-parameter corresponding to each data modality, the step of combining and forming the internal focus parameters corresponding to the at least two data modalities based on the internal focus sub-parameter corresponding to each data modality comprises: splicing the first internal focus sub-parameter corresponding to each of the data modalities to form a corresponding first internal focus parameter, and splicing the second internal focus sub-parameter corresponding to each of the data modalities to form a corresponding second internal focus parameter; Multiplying the first internal focusing parameter by a predetermined first weight parameter to output a corresponding first internal weighted focusing parameter, and multiplying the second internal focusing parameter by a predetermined second weight parameter to output a corresponding second internal weighted focusing parameter; An addition operation is performed on the first internal weight focusing parameter and the second internal weight focusing parameter, and a linear mapping operation is performed on the result of the addition operation to form internal focusing parameters corresponding to the at least two data modalities.

4. The process optimization method for a multi-layer flexible circuit board production line according to claim 1, characterized in that: The internal focus parameter includes an internal focus sub-parameter corresponding to each of the data modalities, wherein the step of determining the lamination process internal focus vector corresponding to the target lamination process data based on the internal focus parameter and the lamination process vector comprises: Performing a multiplication operation on the internal focusing sub-parameter corresponding to the target data modality and the lamination process local vector in the target data modality, and outputting the lamination process local focusing vector of the target data modality; After outputting the lamination process local focusing vector corresponding to each of the data modes, the lamination process local focusing vectors corresponding to each of the data modes are spliced ​​to form the lamination process internal focusing vector of the target lamination process data.

5. The process optimization method for a multi-layer flexible circuit board production line according to any one of claims 1 to 4, characterized in that: The at least two data modalities include a first data modality and a second data modality, the first data modality refers to the modality corresponding to the lamination process parameter text, and the second data modality refers to the modality corresponding to the lamination process monitoring image, wherein the step of using the external focusing unit included in the process optimization model to perform a focus mining operation between the data modalities, outputting the external focusing parameters corresponding to the internal focusing vector of the lamination process, and determining the external focusing vector of the lamination process based on the external focusing parameters and the internal focusing vector of the lamination process includes: extracting respectively from the lamination process vectors a first lamination process local vector of the target lamination process data in the first data modality and a second lamination process local vector of the target lamination process data in the second data modality; Using a first external focusing subunit in an external focusing unit included in the process optimization model, a focusing mining operation is performed based on the first lamination process local vector and the second lamination process local vector, a first external focusing parameter corresponding to the lamination process vector is output, and based on the first external focusing parameter and the lamination process vector, a first external focusing vector corresponding to the target lamination process data is determined; extracting, from the lamination process internal focus vector, a lamination process local focus vector of the target lamination process data in the first data modality and a lamination process local focus vector of the target lamination process data in the second data modality respectively; Using a second external focusing subunit in the external focusing unit, a focusing mining operation is performed based on a lamination process local focusing vector in the first data modality and a lamination process local focusing vector in the second data modality, a second external focusing parameter corresponding to the lamination process internal focusing vector is output, and based on the second external focusing parameter and the lamination process internal focusing vector, a second external focusing vector corresponding to the target lamination process data is determined; The first external focusing vector and the second external focusing vector are spliced ​​to obtain a lamination process external focusing vector corresponding to the target lamination process data.

6. The process optimization method for a multi-layer flexible circuit board production line according to claim 5, characterized in that: The step of using the first external focusing subunit in the external focusing unit included in the process optimization model to perform a focusing mining operation based on the first lamination process local vector and the second lamination process local vector, outputting a first external focusing parameter corresponding to the lamination process vector, and determining a first external focusing vector corresponding to the target lamination process data based on the first external focusing parameter and the lamination process vector, comprises: Based on the first lamination process local vector, performing a focused mining operation at multiple levels on the second lamination process local vector, and outputting corresponding first focused mining results at multiple levels; Based on the second lamination process local vector, performing a plurality of levels of focused mining operations on the first lamination process local vector, and outputting a plurality of levels of corresponding second focused mining results; Aggregating the first focused mining results of the multiple levels and the second focused mining results of the multiple levels to obtain a first external focused vector corresponding to the target lamination process data; Among them, on the first focused mining path, in the focused mining operation of the first level, the dot product between the first lamination process local vector and the transposed vector of the second lamination process local vector is used as the first external focused parameter corresponding to the first level, and the first external focused parameter and the second lamination process local vector are weighted summed, and the result of the weighted summation is used as the first focused mining result of the first level. In the focused mining operation of the second level and each subsequent level, the first lamination process local vector and the first focused mining result of the first level are respectively subjected to hole convolution to form the first lamination process convolution vector and the second lamination process convolution vector corresponding to the current level, and the dot product between the first lamination process convolution vector and the transposed vector of the second lamination process convolution vector is used as the first external focused parameter corresponding to the current level, and the first external focused parameter and the second lamination process convolution vector are weighted summed, and the result of the weighted summation is used as the first focused mining result of the current level, and the expansion factor of the hole convolution has a positive correlation with the corresponding level.

7. The process optimization method for a multi-layer flexible circuit board production line according to claim 5, characterized in that: The step of using the second external focusing subunit in the external focusing unit to perform a focusing mining operation based on the lamination process local focusing vector in the first data modality and the lamination process local focusing vector in the second data modality, outputting a second external focusing parameter corresponding to the lamination process internal focusing vector, and determining a second external focusing vector corresponding to the target lamination process data based on the second external focusing parameter and the lamination process internal focusing vector, comprises: Determine a dot product between a lamination process local focus vector in the first data modality and a transposed vector of the lamination process local focus vector in the second data modality, and use the dot product as a second external focus parameter corresponding to the lamination process local focus vector, and, based on the second external focus parameter, perform a weighted sum calculation on the lamination process local focus vector in the second data modality to obtain a first second focus mining result; Determine a dot product between a lamination process local focus vector in the second data modality and a transposed vector of the lamination process local focus vector in the first data modality, and use the dot product as a second external focus parameter corresponding to the lamination process local focus vector, and, based on the second external focus parameter, perform a weighted sum calculation on the lamination process local focus vector in the first data modality to obtain a second second focus mining result; The first second focused mining result and the second second focused mining result are aggregated to obtain a second external focused vector corresponding to the target lamination process data.

8. A process optimization system for a multi-layer flexible circuit board production line, characterized in that: It comprises a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the process optimization method for a multi-layer flexible circuit board production line as described in any one of claims 1 to 7.

Images