Method, device and equipment for predicting manufacturing time of aeronautical composite material part and medium

Abstract

The invention provides an aviation composite material part manufacturing time prediction method and device, equipment and a medium, and relates to the technical field of intelligent manufacturing. The method comprises the following steps: firstly, determining a plurality of candidate influence factors which influence the manufacturing time of the aeronautical composite material part; secondly, mapping the plurality of candidate impact factors into a high-dimensional feature space to perform dimension reduction processing on the plurality of candidate impact factors to obtain at least one target impact factor; and then, based on the at least one target influence factor and the corresponding manufacturing man-hour label, training the candidate man-hour prediction network to form a target man-hour prediction network for predicting the manufacturing man-hour of the aeronautical composite part. On the basis of the content, the problem that in the prior art, the precision of manufacturing man-hour prediction is relatively low can be solved.

Description

Technical Field

[0001] The present application relates to the field of intelligent manufacturing technology, and more specifically, to a method, device, equipment, and medium for predicting the manufacturing time of aviation composite parts. Background Art

[0002] With the rapid development of smart manufacturing and the Internet of Things (IoT) technologies, the manufacturing industry is undergoing an unprecedented data revolution. In this context, manufacturers are able to collect unprecedented amounts of data from their machining equipment. This data, derived from the component manufacturing process, exhibits typical big data characteristics, including but not limited to high dimensionality, high velocity, high complexity, and high value density. Leveraging this data, companies can gain deeper insights into manufacturing processes, optimize production procedures, and improve product quality.

[0003] For example, in the manufacture of aviation composite parts, manufacturing time can be predicted, thereby promoting the accuracy and scientific nature of capacity assessment and quota management. However, in existing technologies, the accuracy of manufacturing time prediction is relatively low. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method, device, equipment and medium for predicting the manufacturing time of aviation composite parts, so as to improve the problem of relatively low accuracy of manufacturing time prediction in the prior art.

[0005] To achieve the above objectives, this application adopts the following technical solutions: A method for predicting the manufacturing time of aviation composite parts, comprising: Determining a plurality of candidate influencing factors that have an impact on the manufacturing time of aviation composite parts, wherein the plurality of candidate influencing factors include at least one nonlinear influencing factor; Mapping the multiple candidate impact factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate impact factors to obtain at least one target impact factor, wherein the number of the at least one target impact factor is less than the number of the multiple candidate impact factors; Based on the at least one target influencing factor and the corresponding manufacturing time label, a candidate man-hour prediction network is trained to form a target man-hour prediction network, wherein the target man-hour prediction network is used to predict the manufacturing time of aviation composite parts.

[0006] In a preferred embodiment of the present application, in the above-mentioned method for predicting the manufacturing time of aviation composite parts, the step of mapping the multiple candidate influencing factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate influencing factors to obtain at least one target influencing factor includes: Mapping the multiple candidate influencing factors into a high-dimensional feature space to form a high-dimensional feature space matrix, wherein each matrix parameter in the high-dimensional feature space matrix is ​​used to reflect the similarity between two candidate influencing factors among the multiple candidate influencing factors; Performing eigendecomposition on the high-dimensional feature space matrix to obtain an eigenvalue corresponding to each of the multiple candidate influencing factors; Based on the characteristic value corresponding to each of the candidate influencing factors, the multiple candidate influencing factors are screened to determine at least one target influencing factor, wherein the number of the at least one target influencing factor is less than the number of the multiple candidate influencing factors.

[0007] In a preferred embodiment of the present application, in the above-mentioned method for predicting the manufacturing time of aviation composite parts, the step of mapping the multiple candidate influencing factors into a high-dimensional feature space to form a high-dimensional feature space matrix includes: For each candidate influencing factor, similarity calculation is performed on the candidate influencing factor and each candidate influencing factor to obtain a corresponding column of similarity parameters. A column of similarity parameters corresponding to each of the multiple candidate influencing factors is combined to form a high-dimensional feature space matrix including multiple columns of matrix parameters, wherein a column of similarity parameters corresponds to a column of matrix parameters in the high-dimensional feature space matrix.

[0008] In a preferred embodiment of the present application, in the above-mentioned method for predicting the manufacturing time of aviation composite parts, the step of performing similarity calculation on each of the multiple candidate influencing factors and each of the multiple candidate influencing factors to obtain a corresponding column of similarity parameters includes: Determining the square value of the difference between the two candidate influencing factors to obtain a first square value, and determining the square value of a preset width parameter to obtain a second square value, wherein the width parameter is used to determine the influence range of each of the candidate influencing factors in the high-dimensional feature space; A ratio between the first square value and the second square value is determined, and a natural exponential function calculation is performed based on the ratio to obtain a similarity parameter between the two candidate influencing factors.

[0009] In a preferred embodiment of the present application, in the above-mentioned method for predicting the manufacturing time of aviation composite parts, the step of combining a column of similarity parameters corresponding to each of the multiple candidate influencing factors to form a high-dimensional feature space matrix including multiple columns of matrix parameters includes: Combining a column of similarity parameters corresponding to each of the multiple candidate influencing factors to form an initial feature space matrix including multiple columns of matrix parameters, wherein a column of similarity parameters serves as a column of matrix parameters in the high-dimensional feature space matrix; The initial feature space matrix is ​​column-centered and row-centered to form a high-dimensional feature space matrix including multiple columns of matrix parameters, wherein, in the high-dimensional feature space matrix, the mean of the matrix parameters of each column is equal to zero, and the mean of the matrix parameters of each row is equal to zero.

[0010] In a preferred embodiment of the present application, in the above-mentioned method for predicting the manufacturing time of aviation composite parts, the step of screening the plurality of candidate influencing factors based on the characteristic value corresponding to each of the candidate influencing factors to determine at least one target influencing factor includes: Determine the sum of the characteristic values ​​corresponding to each of the candidate influencing factors to obtain a target sum value; Traversing the plurality of candidate influencing factors in descending order of the characteristic values ​​corresponding to each candidate influencing factor to form a currently traversed candidate influencing factor; For the currently traversed candidate influencing factor, calculate the sum of the eigenvalues ​​corresponding to the currently traversed candidate influencing factor and each previously traversed candidate influencing factor to obtain a traversal sum value; calculating a ratio between the traversal sum value and the target sum value; If the ratio is greater than a predetermined ratio threshold, the traversal is stopped, and the currently traversed candidate influencing factor and each candidate influencing factor that has been traversed are used as target influencing factors. If the ratio is not greater than the ratio threshold, the traversal is continued until the ratio corresponding to the currently traversed candidate influencing factor is greater than the ratio threshold.

[0011] In a preferred embodiment of the present application, in the above-mentioned method for predicting the manufacturing time of aviation composite parts, the plurality of candidate influencing factors include: Part length, part width, part height, part surface area, part type, machine model, number of layers, table length, number of blanks, part weight, whether it is a functional part, number of part productions, scanning step length, part complexity, projection area, main material type, main material quota, adhesive type, surface type, number of auxiliary materials, maximum blank area, release agent type, sandwich type, and cutting type; The at least one target impact factor includes: Main material quota, part surface area, number of layers, projected area, part complexity, maximum blank area, surface type, number of blanks, part type and machine model.

[0012] The present application also provides a device for predicting the manufacturing time of aviation composite parts, comprising: A candidate factor determination module is used to determine a plurality of candidate influencing factors that have an impact on the manufacturing time of aviation composite parts, wherein the plurality of candidate influencing factors include at least one nonlinear influencing factor; a target factor determination module, configured to map the plurality of candidate influencing factors into a high-dimensional feature space to perform dimensionality reduction processing on the plurality of candidate influencing factors to obtain at least one target influencing factor, wherein the number of the at least one target influencing factor is less than the number of the plurality of candidate influencing factors; A prediction network training module is used to train a candidate manufacturing time prediction network based on the at least one target influencing factor and the corresponding manufacturing time label to form a target manufacturing time prediction network, wherein the target manufacturing time prediction network is used to predict the manufacturing time of aviation composite parts.

[0013] Based on the above, the present application further provides an electronic device, including: memory for storing computer programs; The processor connected to the memory is used to execute the computer program stored in the memory to implement the above-mentioned method for predicting the manufacturing time of aviation composite parts.

[0014] On the basis of the above, the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is run, it executes each step of the above-mentioned method for predicting the manufacturing time of aviation composite parts.

[0015] The present application provides a method, apparatus, device, and medium for predicting the manufacturing time of aviation composite parts. First, multiple candidate influencing factors that have an impact on the manufacturing time of aviation composite parts are determined; second, the multiple candidate influencing factors are mapped into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate influencing factors to obtain at least one target influencing factor; then, based on the at least one target influencing factor and the corresponding manufacturing time label, the candidate time prediction network is trained to form a target time prediction network for predicting the manufacturing time of aviation composite parts. Based on the above content, on the one hand, by performing dimensionality reduction processing, the influence of irrelevant interference information in some candidate influencing factors with little effect on network training can be reduced, thereby improving the reliability of network training, so that when the network formed by training is used to predict the working time, a more accurate working time prediction result can be obtained. On the other hand, since the candidate influencing factors are mapped to a high-dimensional feature space during dimensionality reduction processing, the nonlinear influencing factors in this high-dimensional space may also show a linear relationship (or a near-linear relationship), thereby achieving effective processing of nonlinear influencing factors (compared to the conventional technical solution of identifying the main components in the data by calculating the covariance matrix of the data and finding its eigenvalues ​​and eigenvectors, since the nonlinear influencing factors can be processed, some nonlinear influencing factors with larger effects can be utilized, and the reliability is higher), thereby improving the reliability of the dimensionality reduction processing, obtaining reliable target influencing factors, further improving the accuracy of training, and further improving the accuracy of manufacturing working time prediction based on the trained network. Therefore, the problem of relatively low accuracy of manufacturing working time prediction in the existing technology can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

[0017] Figure 1 This is a structural block diagram of an electronic device provided in an embodiment of the present application.

[0018] Figure 2 A schematic flow chart of a method for predicting the manufacturing time of aviation composite parts provided in an embodiment of the present application.

[0019] Figure 3 This is a logical relationship diagram of the MLP algorithm provided in the embodiment of this application.

[0020] Figure 4 A block diagram of a device for predicting the manufacturing time of aviation composite parts provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in the present application without creative work are within the scope of protection of the present application.

[0023] like Figure 1 As shown, an embodiment of the present application provides an electronic device, wherein the electronic device may include a memory, a processor, and a device for predicting the manufacturing time of aviation composite parts.

[0024] In detail, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, the memory and the processor can be electrically connected via one or more communication buses or signal lines. The aviation composite parts manufacturing time prediction device includes at least one software function module stored in the memory in the form of software or firmware. The processor is used to execute an executable computer program stored in the memory, for example, the software function modules and computer programs included in the aviation composite parts manufacturing time prediction device, to implement the aviation composite parts manufacturing time prediction method provided in the embodiment of the present application.

[0025] Optionally, the memory may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0026] Optionally, 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.

[0027] I understand. Figure 1 The structure shown is only for illustration, and the electronic device may also include Figure 1 More or fewer components than shown, or with Figure 1 The different configurations shown may, for example, further include a communication unit for exchanging information with other devices.

[0028] Combine Figure 2 The present application also provides a method for predicting the manufacturing time of aviation composite parts applicable to the above-mentioned electronic device. The method steps defined in the process related to the method for predicting the manufacturing time of aviation composite parts can be implemented by the electronic device.

[0029] The following will Figure 2 The specific process shown is explained in detail.

[0030] Step S110 : determining a plurality of candidate influencing factors that have an impact on the manufacturing time of aviation composite parts.

[0031] In an embodiment of the present application, the electronic device can determine a plurality of candidate influencing factors that have an impact on the manufacturing time of aviation composite parts. The plurality of candidate influencing factors include at least one nonlinear influencing factor. Exemplarily, the plurality of candidate influencing factors can be all influencing factors that have an impact on the manufacturing time of aviation composite parts, or can be a part of all influencing factors that have an impact on the manufacturing time of aviation composite parts. For example, the production of aviation composite parts will involve multiple processes, and the manufacturing time of the parts will adopt the process addition method, that is, by predicting the part manufacturing time of a single process, and finally adding up to obtain the processing time data of the entire part. The relevant factors (i.e., influencing factors) in the part production process can be divided into two categories: basic part information and process processing information.

[0032] Step S120 : Mapping the multiple candidate impact factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate impact factors to obtain at least one target impact factor.

[0033] In an embodiment of the present application, after determining the multiple candidate influencing factors, the electronic device may map the multiple candidate influencing factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate influencing factors to obtain at least one target influencing factor. The number of the at least one target influencing factor is smaller than the number of the multiple candidate influencing factors. That is, some major candidate influencing factors may be screened out from the multiple candidate influencing factors to reduce the dimensionality from the number of the multiple candidate influencing factors to the number of the at least one target influencing factor.

[0034] Step S130 : Based on the at least one target influencing factor and the corresponding manufacturing man-hour label, the candidate man-hour prediction network is trained to form a target man-hour prediction network.

[0035] In an embodiment of the present application, after obtaining the at least one target influencing factor, the electronic device can train a candidate man-hour prediction network based on the at least one target influencing factor and the corresponding manufacturing man-hour label to form a target man-hour prediction network. The target man-hour prediction network is used to predict the manufacturing man-hour of aviation composite parts. That is, for a sample data item obtained, the at least one target influencing factor (corresponding specific parameter) and the corresponding manufacturing man-hour label (i.e., actual manufacturing man-hour) can then be used to predict the at least one target influencing factor in the sample data to obtain a corresponding man-hour prediction result. Subsequently, the network parameters of the candidate man-hour prediction network can be updated based on the error between the man-hour prediction result and the manufacturing man-hour label until the error converges, thereby obtaining a man-hour prediction network. Based on this, in practical applications, after determining a target data item, the at least one target influencing factor (corresponding specific parameter) included in the target data item can be loaded into the target man-hour prediction network, thereby predicting and outputting a man-hour prediction result corresponding to the target data item.

[0036] Based on the above content, on the one hand, by performing dimensionality reduction processing, the influence of irrelevant interference information in some candidate influencing factors with little effect on network training can be reduced, thereby improving the reliability of network training, so that when the network formed by training is used to predict the working time, a more accurate working time prediction result can be obtained. On the other hand, since the candidate influencing factors are mapped to a high-dimensional feature space during dimensionality reduction processing, the nonlinear influencing factors in this high-dimensional space may also show a linear relationship (or a near-linear relationship), thereby achieving effective processing of nonlinear influencing factors (compared to the conventional technical solution of identifying the main components in the data by calculating the covariance matrix of the data and finding its eigenvalues ​​and eigenvectors, since the nonlinear influencing factors can be processed, some nonlinear influencing factors with larger effects can be utilized, and the reliability is higher), thereby improving the reliability of the dimensionality reduction processing, obtaining reliable target influencing factors, further improving the accuracy of training, and further improving the accuracy of manufacturing working time prediction based on the trained network. Therefore, the problem of relatively low accuracy of manufacturing working time prediction in the existing technology can be improved.

[0037] First, regarding step S110, it should be noted that the specific content of the multiple candidate influencing factors is not limited and can be selected accordingly based on actual needs. For example, in an alternative embodiment, the multiple candidate influencing factors may include at least one of the following influencing factors: part length, part width, part height, part surface area, part type, machine model, number of layers, table length, number of blanks, part weight, whether it is a functional part, number of part productions, scanning step length, part complexity, projected area, main material type, main material quota, glue type, curved surface type, number of auxiliary materials, maximum blank area, release agent type, sandwich type, and cutting type. Alternatively, in other embodiments, other influencing factors may also be included.

[0038] Secondly, it should be noted that for step S120 , the specific method of obtaining at least one target impact factor is not limited and can be selected accordingly according to actual needs.

[0039] For example, in an alternative embodiment, in order to avoid the high complexity problem caused by directly performing high-dimensional calculations, the above-mentioned step S120 can further include step S121, step S122 and step S123, and the specific content of each step is as follows.

[0040] Step S121 : Mapping the multiple candidate influencing factors into a high-dimensional feature space to form a high-dimensional feature space matrix.

[0041] In an embodiment of the present application, the multiple candidate influencing factors can be mapped into a high-dimensional feature space to form a high-dimensional feature space matrix. In the high-dimensional feature space matrix, each matrix parameter is used to reflect the similarity between two candidate influencing factors among the multiple candidate influencing factors. That is, by calculating the similarity, the calculation of the high-dimensional mapping is avoided. Therefore, when dealing with nonlinear problems, the efficiency is improved.

[0042] Step S122: performing eigendecomposition on the high-dimensional feature space matrix to obtain an eigenvalue corresponding to each of the multiple candidate influencing factors.

[0043] In an embodiment of the present application, after forming the high-dimensional feature space matrix, the high-dimensional feature space matrix can be subjected to eigendecomposition to obtain the eigenvalue corresponding to each of the multiple candidate influencing factors. For example, the corresponding eigenvalue can be calculated based on the following formula: K centered v i =λ i v i , where K centered Represents the high-dimensional feature space matrix, v i represents the eigenvector, λ i represents the eigenvalue, and i is an integer greater than or equal to 1 and less than or equal to the number of the plurality of candidate influencing factors. In addition, the specific process of eigendecomposition can refer to the relevant prior art and will not be described in detail here.

[0044] Step S123 : Screening the plurality of candidate influencing factors based on the characteristic value corresponding to each candidate influencing factor to determine at least one target influencing factor.

[0045] In an embodiment of the present application, after obtaining the characteristic value corresponding to each candidate influencing factor, the plurality of candidate influencing factors may be screened based on the characteristic value corresponding to each candidate influencing factor to determine at least one target influencing factor, wherein the number of the at least one target influencing factor is less than the number of the plurality of candidate influencing factors.

[0046] It can be understood that, in the above-mentioned step S121, the specific method of mapping the multiple candidate influencing factors into the high-dimensional feature space is not limited. For example, in an alternative embodiment, in order to fully utilize the similarity between the candidate influencing factors to characterize each candidate influencing factor, the above-mentioned step S121 can further include step S121a and step S121b. The specific content of each step is described below.

[0047] Step S121a: For each candidate influencing factor, perform similarity calculation on the candidate influencing factor with each candidate influencing factor to obtain a corresponding column of similarity parameters.

[0048] In an embodiment of the present application, for each of the multiple candidate influencing factors, similarity calculation is performed on the candidate influencing factor and each of the multiple candidate influencing factors to obtain a corresponding column of similarity parameters. Exemplarily, the multiple candidate influencing factors can form a column vector, and then the column vector can be transposed to form a row vector, and then the column vector and the row vector can be subjected to a dot product operation (or other operations can be performed as long as the corresponding similarity can be represented) to obtain a corresponding dot product matrix, in which a column of matrix parameters can be a column of similarity parameters.

[0049] Step S121b: Combining a column of similarity parameters corresponding to each of the multiple candidate influencing factors to form a high-dimensional feature space matrix including multiple columns of matrix parameters.

[0050] In an embodiment of the present application, after forming a column of similarity parameters corresponding to each candidate influencing factor, the column of similarity parameters corresponding to each candidate influencing factor among the multiple candidate influencing factors may be combined to form a high-dimensional feature space matrix including multiple columns of matrix parameters. A column of similarity parameters corresponds to a column of matrix parameters in the high-dimensional feature space matrix.

[0051] It is understood that, in the above step S121a, the specific method of obtaining the corresponding column of similarity parameters is not limited. For example, in an alternative embodiment, in order to fully capture the complex nonlinear relationship in the data, the above step S121a may further include: First, the square value of the difference between the two candidate influencing factors can be determined to obtain a first square value, and the square value of a preset width parameter can be determined to obtain a second square value, wherein the width parameter is used to determine the influence range of each candidate influencing factor in the high-dimensional feature space, and its specific value can be set based on experience (wherein, when the width parameter is large, it means that more distant points also have greater similarity, and the global characteristics of the data can be captured; when the width parameter is small, it means that only points with closer distances have significant similarity, and more attention is paid to the local structure of the data). It can also be calculated. For example, first, the Euclidean distance between each of the multiple candidate influencing factors can be calculated; then, a distance statistic (such as the average distance or the median distance) can be selected; finally, the width parameter can be set to a multiple of the statistic (such as half of the average distance, or a constant multiple, etc.); Secondly, the ratio between the first square value and the second square value can be determined, and a natural exponential function calculation can be performed based on the ratio to obtain the similarity parameter between the two candidate influencing factors (based on this, the corresponding similarity parameter can be calculated for every two candidate influencing factors, which can be combined to form a corresponding column of similarity parameters and a high-dimensional feature space matrix).

[0052] For the above calculation process, in one application, it can be: ; in, Represents x and x i The similarity parameter between these two candidate impact factors, () represents the natural exponential function, Indicates the width parameter.

[0053] It is understood that, in the above step S121b, the specific manner of combining and processing a column of similarity parameters corresponding to each of the multiple candidate influencing factors is not limited. For example, in an alternative embodiment, in order to remove the global bias in the data and ensure that the analysis is based on the essential structure of the data rather than the influence of the mean, the above step S121b may further include: First, a column of similarity parameters corresponding to each of the multiple candidate influencing factors can be combined to form an initial feature space matrix including multiple columns of matrix parameters, wherein a column of similarity parameters serves as a column of matrix parameters in the high-dimensional feature space matrix; Secondly, the initial feature space matrix can be column-centered and row-centered to form a high-dimensional feature space matrix including multiple columns of matrix parameters, wherein, in the high-dimensional feature space matrix, the mean of the matrix parameters of each column is equal to zero, and the mean of the matrix parameters of each row is equal to zero. In addition, column-centered processing can be performed first (that is, the mean of the matrix parameters of each column is equal to zero, for example, the mean of the matrix parameters of a column is calculated, and then the difference between the matrix parameters of this column and the mean is taken to obtain the parameters after column-centered processing) and then row-centered processing can be performed (that is, the mean of the matrix parameters of each row is equal to zero, for example, the mean of the matrix parameters of a row is calculated, and then the difference between the matrix parameters of this row and the mean is taken to obtain the parameters after row-centered processing), or row-centered processing can be performed first and then column-centered processing can be performed.

[0054] It is understandable that, in the above step S123, the specific method of screening the multiple candidate influencing factors is not limited. For example, in an alternative embodiment, in order to fully screen out candidate influencing factors with greater effects, the above step S123 may include: First, the sum of the characteristic values ​​corresponding to each candidate influencing factor can be determined to obtain the target sum value, such as ,in, refers to the eigenvalue corresponding to the kth candidate influencing factor, and m refers to the number of candidate influencing factors; Secondly, the plurality of candidate influencing factors may be traversed in descending order of the eigenvalues ​​corresponding to each candidate influencing factor to form the currently traversed candidate influencing factor, that is, the first traversed candidate influencing factor has the largest eigenvalue. Then, for the candidate influencing factor currently traversed, the sum of the eigenvalues ​​corresponding to the candidate influencing factor currently traversed and each candidate influencing factor that has been traversed is calculated to obtain the traversal sum value, such as , where s refers to the number of the candidate influencing factors currently traversed and each candidate influencing factor that has been traversed; Afterwards, the ratio between the traversal sum value and the target sum value can be calculated, such as ; Finally, if the ratio is greater than a predetermined ratio threshold (which can be selected according to actual needs, such as 80%, 85%, 90%, etc.), the traversal is stopped, and the currently traversed candidate influencing factor and each candidate influencing factor that has been traversed are used as target influencing factors. If the ratio is not greater than the ratio threshold, the traversal is continued until the ratio corresponding to the currently traversed candidate influencing factor is greater than the ratio threshold.

[0055] For example, in a specific application, the characteristic values ​​and ratios shown in the following table can be obtained:

[0056] Analyzing the data in the table above reveals that the eigenvalues ​​of the top 10 candidate influencing factors account for over 85% of the total. Therefore, in subsequent research, these top 10 candidate influencing factors can be used as target influencing factors (i.e., main material quota, part surface area, number of plies, projected area, part complexity, maximum sheet area, surface type, number of sheets, part type, and machine model) to be loaded into the candidate man-hour prediction network for training. This maximizes the preservation of the basic information of the original sample, enabling dimensionality reduction of the influencing factors, reducing the original 24 influencing factors to 10. This reduces redundancy between the influencing factors and lays the foundation for further prediction research using the network.

[0057] Thirdly, it should be noted that the candidate man-hour prediction network can be a multilayer perceptron (MLP), which is a traditional supervised learning method that simulates human neurons. When dealing with multi-classification problems, it can infinitely approximate the actual mapping relationship between input features and output labels by learning nonlinear functions. The multilayer perceptron adds hidden layers to the perceptron structure, which increases the complexity of the model while enhancing the expressive power of the model. The output layer neurons can have multiple outputs and can be flexibly applied to classification and regression problems. Multilayer perceptrons such as Figure 3 As shown in Figure 1, it consists of an input layer, a hidden layer, and an output layer, and is a fully connected neural network. The neurons in the same layer of each layer are independent of each other and not connected to each other, while the neurons between two adjacent layers are fully connected. Each neuron has a corresponding input weight (such as V ij 、W jk ), bias, and activation function. The connection strength is determined by the weights between neurons. The number of neurons in the input layer and the output layer can be one or more, and the number of hidden layers can be single or multi-layer. The transmission of data between neurons is directional. During forward calculation, the data is calculated layer by layer from input to output. During reverse calculation, the error is back-propagated to correct the connection weights. Let X be the label of the input layer neurons, and the number of neurons in the input layer is n; H is the label of the hidden layer neurons, and the number of neurons in the hidden layer is d. The activation function is ; The output layer neurons are labeled Y, the number of neurons in the output layer is m, and the corresponding activation function is .

[0058] In addition, common activation functions include the Sigmoid function, the Tanh function, and the Relu function. The Sigmoid function is relatively simple and has good nonlinear mapping. The mean of the Tanh function's output is 0, and it converges faster than the Sigmoid function. The Relu function's gradient converges quickly, and the amount of gradient calculation is less than that of the first two functions. To solve the problem of gradient calculation, stochastic gradient descent (SGD) can be used (a portion of samples (batch) is selected each time, and the gradient value is calculated as the overall gradient value. By adjusting the number of batches, both computational efficiency and value accuracy can be achieved).

[0059] In addition, for each of the above-mentioned influencing factors (candidate influencing factors, target influencing factors), corresponding collection methods, data correction and normalization processing logic can be respectively used. For example: For the main material quota, the collection method can be MBOM extraction, the data correction method is historical material usage data, and the normalization processing logic is in square meters; for the part surface area, the collection method can be MBOM extraction, the data correction method is digital-analog data verification, and the normalization processing logic is in square meters; for the number of layers, the collection method can be digital-analog data, the data correction method is FO data verification, and the normalization processing logic is the number of layers; for the projection area, the collection method can be MBOM extraction, the data correction method is digital-analog data verification, and the normalization processing logic is in square meters; for the part complexity, the collection method can be defined by the technician association, the data correction method is sample survey data, and the normalization processing logic is defined according to the complexity, with 1, 2, 3, 4, etc.; for the maximum sheet area, the collection method can be NC program extraction, the data correction method is FO data verification, and the normalization processing logic is in square meters; for the surface type, the collection method can be defined by the technicians association, the data correction method is digital-analog data verification, and the normalization processing logic is defined according to the surface type, expressed as 1, 2, 3, etc.; for the number of sheets, the collection method can be NC program extraction, the data correction method is FO data verification, and the normalization processing logic is the number of sheets; for the part type, the collection method can be the process-defined value, and the normalization processing logic is defined according to the part type, expressed as 1, 2, 3, etc.; for the machine model, the collection method can be the actual value, and the normalization processing logic is to establish a machine model correspondence table, expressed as 1-20.

[0060] In addition, it should be noted that the entire manufacturing big data is divided into two categories: data that affects the manufacturing time of parts and data on the actual processing time of each process of parts. The actual processing time t of a process can be obtained by the start time of the process. , end time , processing quantity n, etc., the specific working hours can be expressed as: ; Part processing status information, such as process start and end times, processing quantity, and number of people involved, can be collected through start and completion time recording. The actual processing time collected and statistically analyzed represents the entire time consumed by the process, including equipment commissioning, tool preparation, and worker rest periods. Therefore, the actual processing time must be corrected before it can be used as manufacturing hours in the overall manufacturing big data.

[0061] In addition, it should be noted that the process of training the network can be divided into a testing phase and a training phase. Therefore, the sample data needs to be divided into a test set and a training set. Among them, in order to ensure the quality of the constructed test set and training set, it is crucial to perform refined data preprocessing on the manufacturing big data obtained from multiple source systems such as MBOM (Bill of Materials), NC program (Numerical Control Program), Technician Association, Digital Model (Digital Model) and MES (Manufacturing Execution System). This process aims to improve the integrity and accuracy of the data, specifically including removing duplicate data, filling in missing data, and correcting erroneous data, thereby laying a solid foundation for the subsequent man-hour prediction network training. The following are the main steps of data preprocessing: a) Duplicate Data Processing: During data collection, identical data records may appear due to data synchronization or recording errors between systems. This duplicate data not only takes up storage space but can also interfere with model training, reducing model learning efficiency. Therefore, a data deduplication algorithm is needed to identify and remove duplicate data records to ensure the uniqueness of each piece of data. b) Missing data processing: In the acquired manufacturing big data, some data elements may be missing, such as key information such as the complexity of the part and processing parameters. Different strategies should be adopted for handling missing data based on the degree of missingness and the importance of the data, such as: Ignore: When multiple key data elements are missing from a data record, the data record may not be used for model training. Therefore, choose to ignore and delete these data records. Filling: For data records with only a small number of missing elements, data filling methods such as the mean method, median method, or prediction filling based on other relevant data can be used to maintain data integrity; c) Error data correction: During the data collection process, erroneous data may be introduced due to various reasons, such as sensor failure and human input errors. The presence of erroneous data can seriously affect the training effect and prediction accuracy of the model. Therefore, it is necessary to set reasonable data verification rules to define the reasonable value range of each type of data. Outliers outside the normal range should be regarded as erroneous data and corrected or eliminated to purify the data set and improve the training quality of the model. The above data preprocessing steps significantly improve the quality of manufacturing big data, providing a more reliable and accurate data foundation for subsequent man-hour prediction network training, thereby improving prediction accuracy and stability and promoting the intelligent upgrade of manufacturing enterprises. Furthermore, in addition to cleaning and processing the data, it is also normalized. From this normalized data set, 90% of the data is extracted as a training set, and 10% of the data is used as a test set.

[0062] In addition, it should be noted that the iterative optimization of the MLP algorithm mainly relies on the backpropagation algorithm and the gradient descent algorithm to adjust the weights and biases in the network to minimize the prediction error. The following is the iterative optimization process of the MLP algorithm: a) Initialize weights and biases: Use the weight optimizer lbfgs, also known as the quasi-Newton method optimizer, which converges well for small datasets. Set up two hidden layers with 100 neurons each. b) Forward propagation: During forward propagation, the input data is passed from the input layer to the output layer after being processed by each layer, including weighted summation, bias addition, and application of activation functions. The logistic function, f(x) = 1 / (1 + exp(-x)), is selected as the activation function. c) Calculate loss: The error between the model's predicted value and the true value is calculated through the loss function, and the mean squared error (MSE) function is used to calculate the error; d) Backpropagation: Based on the gradient of the loss function, the gradient of each weight is calculated layer by layer from the output layer to the input layer through the chain input rule to determine the contribution of each weight to the final error; e) Update weights and biases: Use the calculated gradients to update the weights and biases in the network to reduce the loss function. This is an unconstrained problem in this network, so gradient descent (SGD) is used to solve the gradient calculation problem. Stochastic gradient descent (SGD) selects a batch of samples each time and calculates the gradient as the overall gradient. By adjusting the number of batches, both computational efficiency and value accuracy can be balanced. Iterative optimization: Repeat the steps of forward propagation, loss calculation, backpropagation, and weight update until the stopping condition is met. The maximum number of iterations in this round is set to 1000, which is the stopping condition of the iteration.

[0063] Regarding the above-mentioned method for predicting the manufacturing time of aviation composite parts, it is also necessary to explain: Traditional manufacturing time prediction models are often based on statistical methods. However, this application innovatively introduces deep learning technology, leveraging the powerful nonlinear fitting and feature learning capabilities of neural network models to conduct in-depth mining of composite material manufacturing big data. This can capture the complex relationships and implicit patterns that affect manufacturing time, which is difficult to achieve with traditional methods. In this application, not only was big data on composite manufacturing collected from multiple systems, including MBOM (Bill of Materials), NC programs (Numerical Control Programs), technician associations, digital models (digital models), and MES (Manufacturing Execution Systems), but the data was also carefully pre-processed, including duplicate data processing, missing data filling, and erroneous data correction, to ensure the high quality of model training data, which is the key to achieving high-precision predictions; A specialized prediction network was constructed to address the specific problem of predicting the manufacturing time of aviation composite parts. Compared to general manufacturing time prediction models, this model more accurately considers the particularities of the aviation composite parts manufacturing process, such as material properties, process flow, and equipment performance, thereby improving the pertinence and accuracy of predictions. The prediction network was rigorously tested using real production data to verify its effectiveness and practicality. This process not only verified the model's prediction accuracy but also demonstrated its application value in actual production, providing strong data support for subsequent capacity assessment and quota management. By constructing a manufacturing man-hour prediction and evaluation system based on the improved MLP model, the research provides a new path for the intelligent upgrade and data-driven decision-making of manufacturing enterprises. This achievement not only helps enterprises optimize resource allocation and improve production efficiency, but also provides a powerful demonstration for data-driven decision-making in the era of intelligent manufacturing.

[0064] Combine Figure 4 The present application also provides a device for predicting the manufacturing time of composite aviation parts applicable to the aforementioned electronic equipment. The device can include a candidate factor determination module, a target factor determination module, and a prediction network training module.

[0065] In detail, the candidate factor determination module can be used to determine multiple candidate influencing factors that have an impact on the manufacturing time of aviation composite parts, wherein the multiple candidate influencing factors include at least one nonlinear influencing factor. In the embodiment of the present application, the candidate factor determination module can be used to perform Figure 2 Regarding step S110 shown, the relevant contents of the candidate factor determination module may refer to the above description of step S110.

[0066] In detail, the target factor determination module can be used to map the multiple candidate influencing factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate influencing factors to obtain at least one target influencing factor, wherein the number of the at least one target influencing factor is less than the number of the multiple candidate influencing factors. In the embodiment of the present application, the target factor determination module can be used to perform Figure 2 As shown in step S120, for the relevant content of the target factor determination module, reference can be made to the above description of step S120.

[0067] In detail, the prediction network training module can be used to train the candidate man-hour prediction network based on the at least one target influencing factor and the corresponding manufacturing man-hour label to form a target man-hour prediction network, wherein the target man-hour prediction network is used to predict the manufacturing man-hour of aviation composite parts. In the embodiment of the present application, the prediction network training module can be used to perform Figure 2 As shown in step S130, for the relevant content of the prediction network training module, reference can be made to the above description of step S130.

[0068] In an embodiment of the present application, corresponding to the above-mentioned method for predicting the manufacturing time of aviation composite parts applied to the electronic equipment, a computer-readable storage medium is also provided, in which a computer program is stored. When the computer program is run, each step of the method for predicting the manufacturing time of aviation composite parts is executed.

[0069] The steps executed when the aforementioned computer program is running will not be described in detail here. Please refer to the above explanation of the method for predicting the manufacturing time of aviation composite parts.

[0070] In summary, the method, apparatus, equipment, and medium for predicting the manufacturing time of aviation composite parts provided in the present application first determine multiple candidate influencing factors that have an impact on the manufacturing time of aviation composite parts; secondly, map the multiple candidate influencing factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate influencing factors to obtain at least one target influencing factor; then, based on the at least one target influencing factor and the corresponding manufacturing time label, train the candidate time prediction network to form a target time prediction network for predicting the manufacturing time of aviation composite parts. Based on the above content, on the one hand, by performing dimensionality reduction processing, the influence of irrelevant interference information in some candidate influencing factors with little effect on network training can be reduced, thereby improving the reliability of network training, so that when the network formed by training is used to predict the working time, a more accurate working time prediction result can be obtained. On the other hand, since the candidate influencing factors are mapped to a high-dimensional feature space during dimensionality reduction processing, the nonlinear influencing factors in this high-dimensional space may also show a linear relationship (or a near-linear relationship), thereby achieving effective processing of nonlinear influencing factors (compared to the conventional technical solution of identifying the main components in the data by calculating the covariance matrix of the data and finding its eigenvalues ​​and eigenvectors, since the nonlinear influencing factors can be processed, some nonlinear influencing factors with larger effects can be utilized, and the reliability is higher), thereby improving the reliability of the dimensionality reduction processing, obtaining reliable target influencing factors, further improving the accuracy of training, and further improving the accuracy of manufacturing working time prediction based on the trained network. Therefore, the problem of relatively low accuracy of manufacturing working time prediction in the existing technology can be improved.

[0071] In the several embodiments provided in the embodiments of the present application, 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 schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the 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 box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0072] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0073] If the functions are implemented in the form of software modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the portion that contributes to the prior art, or the portion of the 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, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. It should be noted that, in this document, the terms "comprise," "include," or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. Without further constraints, an element defined by the phrase "comprises a..." does not preclude the existence of additional identical elements in the process, method, article or apparatus that includes the element.

[0074] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for predicting the manufacturing time of aviation composite parts, characterized in that: include: Determining a plurality of candidate influencing factors that have an impact on the manufacturing time of aviation composite parts, wherein the plurality of candidate influencing factors include at least one nonlinear influencing factor; Mapping the multiple candidate impact factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate impact factors to obtain at least one target impact factor, wherein the number of the at least one target impact factor is less than the number of the multiple candidate impact factors; Based on the at least one target influencing factor and the corresponding manufacturing time label, a candidate man-hour prediction network is trained to form a target man-hour prediction network, wherein the target man-hour prediction network is used to predict the manufacturing time of aviation composite parts.

2. The method for predicting the manufacturing time of aviation composite parts according to claim 1, characterized in that: The step of mapping the multiple candidate impact factors into a high-dimensional feature space to perform dimensionality reduction processing on the multiple candidate impact factors to obtain at least one target impact factor includes: Mapping the multiple candidate influencing factors into a high-dimensional feature space to form a high-dimensional feature space matrix, wherein each matrix parameter in the high-dimensional feature space matrix is ​​used to reflect the similarity between two candidate influencing factors among the multiple candidate influencing factors; Performing eigendecomposition on the high-dimensional feature space matrix to obtain an eigenvalue corresponding to each of the multiple candidate influencing factors; Based on the characteristic value corresponding to each of the candidate influencing factors, the multiple candidate influencing factors are screened to determine at least one target influencing factor, wherein the number of the at least one target influencing factor is less than the number of the multiple candidate influencing factors.

3. The method for predicting the manufacturing time of aviation composite parts according to claim 2, characterized in that: The step of mapping the multiple candidate influencing factors into a high-dimensional feature space to form a high-dimensional feature space matrix includes: For each candidate influencing factor, similarity calculation is performed on the candidate influencing factor and each candidate influencing factor to obtain a corresponding column of similarity parameters. A column of similarity parameters corresponding to each of the multiple candidate influencing factors is combined to form a high-dimensional feature space matrix including multiple columns of matrix parameters, wherein a column of similarity parameters corresponds to a column of matrix parameters in the high-dimensional feature space matrix.

4. The method for predicting the manufacturing time of aviation composite parts according to claim 3, characterized in that: The step of performing similarity calculation on each candidate influencing factor among the multiple candidate influencing factors and obtaining a corresponding column of similarity parameters includes: Determining the square value of the difference between the two candidate influencing factors to obtain a first square value, and determining the square value of a preset width parameter to obtain a second square value, wherein the width parameter is used to determine the influence range of each of the candidate influencing factors in the high-dimensional feature space; A ratio between the first square value and the second square value is determined, and a natural exponential function calculation is performed based on the ratio to obtain a similarity parameter between the two candidate influencing factors.

5. The method for predicting the manufacturing time of aviation composite parts according to claim 3, characterized in that: The step of combining a column of similarity parameters corresponding to each candidate influencing factor among the multiple candidate influencing factors to form a high-dimensional feature space matrix including multiple columns of matrix parameters includes: Combining a column of similarity parameters corresponding to each of the multiple candidate influencing factors to form an initial feature space matrix including multiple columns of matrix parameters, wherein a column of similarity parameters serves as a column of matrix parameters in the high-dimensional feature space matrix; The initial feature space matrix is ​​column-centered and row-centered to form a high-dimensional feature space matrix including multiple columns of matrix parameters, wherein, in the high-dimensional feature space matrix, the mean of the matrix parameters of each column is equal to zero, and the mean of the matrix parameters of each row is equal to zero.

6. The method for predicting the manufacturing time of aviation composite parts according to claim 2, characterized in that: The step of screening the plurality of candidate impact factors based on the characteristic value corresponding to each candidate impact factor to determine at least one target impact factor includes: Determine the sum of the characteristic values ​​corresponding to each of the candidate influencing factors to obtain a target sum value; Traversing the plurality of candidate influencing factors in descending order of the characteristic values ​​corresponding to each candidate influencing factor to form a currently traversed candidate influencing factor; For the currently traversed candidate influencing factor, calculate the sum of the eigenvalues ​​corresponding to the currently traversed candidate influencing factor and each previously traversed candidate influencing factor to obtain a traversal sum value; calculating a ratio between the traversal sum value and the target sum value; If the ratio is greater than a predetermined ratio threshold, the traversal is stopped, and the currently traversed candidate influencing factor and each candidate influencing factor that has been traversed are used as target influencing factors. If the ratio is not greater than the ratio threshold, the traversal is continued until the ratio corresponding to the currently traversed candidate influencing factor is greater than the ratio threshold.

7. The method for predicting the manufacturing time of aviation composite parts according to any one of claims 1 to 6, characterized in that: The multiple candidate impact factors include: Part length, part width, part height, part surface area, part type, machine model, number of layers, table length, number of blanks, part weight, whether it is a functional part, number of part productions, scanning step length, part complexity, projection area, main material type, main material quota, adhesive type, surface type, number of auxiliary materials, maximum blank area, release agent type, sandwich type, and cutting type; The at least one target impact factor includes: Main material quota, part surface area, number of layers, projected area, part complexity, maximum blank area, surface type, number of blanks, part type and machine model.

8. A device for predicting the manufacturing time of aviation composite parts, characterized in that: include: A candidate factor determination module is used to determine a plurality of candidate influencing factors that have an impact on the manufacturing time of aviation composite parts, wherein the plurality of candidate influencing factors include at least one nonlinear influencing factor; a target factor determination module, configured to map the plurality of candidate influencing factors into a high-dimensional feature space to perform dimensionality reduction processing on the plurality of candidate influencing factors to obtain at least one target influencing factor, wherein the number of the at least one target influencing factor is less than the number of the plurality of candidate influencing factors; A prediction network training module is used to train a candidate manufacturing time prediction network based on the at least one target influencing factor and the corresponding manufacturing time label to form a target manufacturing time prediction network, wherein the target manufacturing time prediction network is used to predict the manufacturing time of aviation composite parts.

9. An electronic device, characterized in that: include: memory for storing computer programs; A processor connected to the memory is used to execute the computer program stored in the memory to implement the method for predicting the manufacturing time of aviation composite parts according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when running, executes the method for predicting the manufacturing time of aviation composite parts according to any one of claims 1 to 7.

Images