Complex equipment assembly performance prediction method based on improved virus optimization algorithm

Abstract

The invention discloses a complex equipment assembly performance prediction method based on an improved virus optimization algorithm. The method comprises the following steps: establishing a quality feature model of a complex equipment assembly part and a core part; establishing a neural network model for predicting the assembly quality performance of the complex equipment and an initial assembly quality performance prediction model of the complex equipment; obtaining sample data for predicting the assembling quality performance of the complex equipment, and carrying out standardization processing on the sample data; obtaining an initial weight of the neural network model by using an improved virus optimization algorithm; and substituting the initial weight into the neural network model, updating the weight by using a back propagation principle according to a chain rule, and performing final training on the model. According to the method, the virus evolutionary algorithm is creatively endowed to the weight initialization process of the neural network model, and fast iteration of the training process is theoretically achieved. According to the invention, assistance can be provided for workshop quality management and control personnel to carry out complex equipment assembly quality performance prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of product assembly quality control, evolutionary algorithm, neural network algorithm design, in particular to a complex equipment assembly performance prediction method based on an improved virus optimization algorithm. BACKGROUND

[0002] Complex equipment production is a concentrated embodiment of national industrial manufacturing and scientific research strength. Under the constraint of core component-assembly cascade, the quality performance index of the core component is the key and core to determine the final performance index of the assembly. The quality performance of the core component-assembly under the constraint of assembly cascade has a complex nonlinear implicit correlation.

[0003] Assembly performance, as a key indicator to measure the assembly quality of precision products, has attracted more and more attention. Due to the limitations of current manufacturing, processing, measurement, assembly and other technical capabilities, the performance of the product after assembly is unstable, the number of invalid assemblies is too large, and the product delivery cycle is too long, and a series of problems. Therefore, before the assembly operation of the parts, learning prior knowledge from the historical data of the important features of the product assembly can accurately predict the assembly performance of the product, which can not only avoid invalid assembly and improve production efficiency, but also guarantee the assembly quality of the product. With the advent of artificial intelligence methods such as neural networks and support vector machines, researchers at home and abroad have carried out a lot of research work on this. For example, the microstructure assembly performance prediction model based on back propagation (BP) neural network proposed in the literature [C. Shao, X. Ye, L. Wang, and J. Qian, “An assembly process parameters optimization method for precision assembly performance,” Journal of Physics: Conference Series, vol. 1303, pp. 1-8, 2019.] establishes a mapping model between assembly process parameters and assembly performance, and optimizes the assembly process parameters through genetic algorithm.

[0004] The literature research shows that the BP neural network algorithm has excellent performance in solving the quality performance prediction problem. However, due to the large number of key quality performance parameters of assembly parts and core components of complex equipment, a large number of weight parameters need to be trained into the neural network model even for the simplest three-layer neural network model. The number of weight parameters directly determines the length of the chromosome in the evolutionary algorithm. The excessively long chromosome will cause the solution space to be too large, thereby reducing the effect of genetic and mutation operations, causing the algorithm to converge slowly, and even it is difficult to find a better solution. Therefore, it is necessary to improve the current algorithm to meet the requirements of complex equipment assembly performance prediction. SUMMARY

[0005] The present application aims to at least solve one of the technical problems existing in the related art to some extent.

[0006] One purpose of the present application is to provide a complex equipment assembly performance prediction method based on an improved virus optimization algorithm, which innovatively enables the virus evolution algorithm to the weight initialization process of the neural network model, realizes the rapid iteration of the model training process, and can provide assistance for the complex equipment assembly quality performance prediction for workshop quality control personnel.

[0007] In order to achieve the above purpose, the present application provides a complex equipment assembly performance prediction method based on an improved virus optimization algorithm, comprising the following steps:

[0008] S1, a quality characteristic model of complex equipment assembly parts and core components is established, including constructing an assembly part bill of materials model, an assembly part quality performance characteristic model and a core component quality performance characteristic model;

[0009] S2, a neural network framework including an input layer, a hidden layer and an output layer is constructed as an initial complex equipment assembly quality performance prediction model, and the number of nodes of each layer is determined, wherein the number of input layer nodes corresponds to the total number of key quality characteristics of core components, and the number of output layer nodes corresponds to the total number of assembly parts to be predicted quality characteristics;

[0010] S3, sample data for complex equipment assembly quality performance prediction is obtained, and the sample data is standardized;

[0011] S4, the initial weight of the neural network model is obtained by using an improved virus optimization algorithm, and the improved virus optimization algorithm realizes directional search through the bidirectional evolution mechanism of the main population and the virus population, specifically including:

[0012] The main population and the virus population are initialized, wherein each main individual in the main population is real number coded for all weight parameters of the neural network through a two-stage coding method, and each virus individual in the virus population is formed by copying the main individual coding and randomly inserting wildcards.

[0013] The fitness value of each individual in the main population is calculated based on the neural network forward propagation, the fitness value is determined by the loss of the predicted output under the weight parameter represented by the individual code and the sample label;

[0014] The excellent individual with a high ranking fitness value is selected from the main population, and the virus population is directionally replicated based on the code of the excellent individual through the infection mechanism;

[0015] The effective gene of the replicated virus individual is deleted to a wildcard with a preset probability, and local fine search of the virus population is performed;

[0016] The above process is iteratively executed until the termination condition is met, and the weight parameter corresponding to the optimal individual code is output as the initial weight of the neural network model;

[0017] S5, the initial weight is substituted into the initial complex equipment assembly quality performance prediction model, the model is trained by using the back propagation algorithm, and a trained complex equipment assembly quality performance prediction model is obtained, and the complex equipment assembly performance is predicted by using the complex equipment assembly quality performance prediction model.

[0018] The further preferred technical scheme of the present application is that the quality characteristic model of the assembly part and the core component of the complex equipment in step S1 is established, including constructing the assembly part bill of materials model, the assembly part quality performance characteristic model and the core component quality performance characteristic model; Specifically:

[0019] S11, an assembly part bill of materials model is established:

[0020] ;

[0021] The assembly part is represented by The core component is composed of ;

[0022] S12, an assembly part quality performance characteristic model is established:

[0023] ;

[0024] The assembly part is represented by The core component contains quality characteristics to be predicted;

[0025] S13, a core component quality performance characteristic model is established:

[0026] ;

[0027] The core component is represented by co-comprising key quality features .

[0028] As preferred, the neural network framework comprising input layer, hidden layer and output layer is constructed in step S2, specifically:

[0029] Firstly, the number of input layers of the neural network is determined and the number of nodes of the output layer , denoted as:

[0030] ;

[0031] ;

[0032] According to the empirical formula, the number of hidden layer nodes of the neural network model is determined :

[0033] ;

[0034] The activation function of each layer is determined, wherein the activation function of the hidden layer is ReLU function, and the output layer does not use activation function.

[0035] As preferred, the initial complex equipment assembly quality performance prediction model is represented as:

[0036] ;

[0037] ;

[0038] wherein, is the parameter of the neural network learned from the training data; represents the weight between the input layer and the hidden layer of the neural network, represents the weight between the hidden layer and the output layer of the neural network, represents the bias of the mapping between the input layer and the hidden layer of the neural network, represents the bias of the mapping between the hidden layer and the output layer of the neural network.

[0039] As preferred, the bias of the mapping between the input layer and the hidden layer of the neural network and the bias of the mapping between the hidden layer and the output layer of the neural network are both 0.

[0040] As preferred, the sample data is standardized in step S3, specifically:

[0041] The sample data is defined to consist of input data and output data, denoted as and For the first Group sample data and They are represented as follows:

[0042] ;

[0043] ;

[0044] The Robust Scaler standardization method was used to normalize the input data in the sample data. The processing is represented as follows:

[0045] ;

[0046] in, and They represent the first Core components in group sample data The The values ​​of key quality characteristics before and after standardization; express Sample data The median value, This represents the difference between the 75th percentile and the 25th percentile of the corresponding sample data.

[0047] Preferably, step S4, which involves obtaining the initial weights of the neural network model using an improved virus optimization algorithm, specifically involves:

[0048] S41. Select one unselected data point from the sample data;

[0049] S42. Initialize the main population and virus population for the virus evolution algorithm, including:

[0050] (a) Determine the size of the main population and virus population size ;

[0051] (b) For each principal individual in the principal group, randomly generate the weight parameters of the neural network. and And using a two-stage coding method, a master individual code is formed;

[0052] (c) The virus individual is a substring of the main individual. Wildcards are randomly added to the main individual to form a virus individual code with the same length as the main individual code.

[0053] S43. Calculate the fitness value of the main population:

[0054] Utilize the decoding mechanism of the neural network forward transmission, take the current selected sample data as the input of the neural network And the output label , the weight represented by the encoding of each master individual in the master population Is substituted into the neural network model for forward calculation to obtain an estimated value , the estimated value And the loss value of the label As the fitness value of the individual;

[0055] S44, the directional evolution of the population of virus individuals, comprising:

[0056] (a) select individuals with a proportion of In the fitness value ranking from the master individual, and put them into the excellent individual set;

[0057] (b) select an unselected chromosome from the excellent individual set, and replicate all virus individuals;

[0058] (c) perform a deletion operation on the replicated virus chromosome, and replace the effective gene with a wildcard symbol with a set probability;

[0059] (d) determine whether the virus evolution algorithm meets the termination condition, if yes, execute step S45; otherwise, calculate the fitness value of the new population and re-perform the directional evolution of the population of virus individuals;

[0060] S45, determine whether the training termination condition is met, if yes, output the encoding of the optimal individual as the initial weight of the neural network model; otherwise, execute step S41 repeatedly.

[0061] Another aspect of the present application provides a non-transitory computer readable storage medium having computer instructions stored thereon, the computer instructions causing a computer to execute the improved virus optimization algorithm-based complex equipment assembly performance prediction method described above.

[0062] Still another aspect of the present application provides an electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus, and the processor invokes the logic instructions in the memory to execute the improved virus optimization algorithm-based complex equipment assembly performance prediction method described above.

[0063] Still another aspect of the present application provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program being executed by a processor to perform the improved virus optimization algorithm-based complex equipment assembly performance prediction method described above.

[0064] Beneficial effects: the complex equipment assembly performance prediction method based on the improved virus optimization algorithm of the application innovatively enables the virus evolution algorithm to the weight initialization process of the neural network model, realizes the rapid iteration of the training process, and uses the trained complex equipment assembly quality performance prediction model to predict the complex equipment assembly performance, which can provide assistance for the complex equipment assembly quality performance prediction of the workshop quality control personnel.

[0065] The application utilizes the infection and evolution mechanism of the virus evolution algorithm itself, and the interaction between the main group and the virus group, realizes the directional evolution of the group, makes the genes of the excellent individuals in the group act on the virus directionally, and makes the virus evolve, realizes the horizontal transmission of the genes in the same group through the infection operation of the virus, and performs local search. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 The flowchart for obtaining the initial weight of the neural network by using the improved virus optimization algorithm in embodiment 1;

[0067] Figure 2 The example diagram of the two-stage encoding of the main individual in embodiment 1;

[0068] Figure 3 The example diagram of the encoding of the virus individual in embodiment 1;

[0069] Figure 4 The example diagram of the replication of the virus individual in embodiment 1;

[0070] Figure 5 The example diagram of the virus individual deletion operation in embodiment 1. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical scheme and advantages of the application clearer, the technical scheme in the application will be described clearly and completely in combination with the drawings in the application. Obviously, the described embodiments are part of the embodiments of the application, not all the embodiments, and they should not be understood as limiting the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application. In the description of the application, it should be understood that the terms used are only for the purpose of description, and should not be understood as indicating or implying relative importance.

[0072] The following will be described in combination with Figures 1-5 The complex equipment assembly performance prediction method based on the improved virus optimization algorithm provided by the application is described.

[0073] Embodiment 1: This embodiment is directed to the complex equipment assembly quality performance prediction problem, and proposes a complex equipment assembly performance prediction method based on an improved virus optimization algorithm. The method focuses on building and training a complex equipment assembly quality performance prediction model. Since the number of key quality performance parameters of the assembly parts and core components of the complex equipment is large, even the simplest three-layer neural network model needs to train a large number of weight parameters. The number of weight parameters directly determines the length of the chromosome in the evolutionary algorithm. Too long chromosome will cause the solution space to be too large, thereby reducing the effect of genetic and mutation operations, causing the algorithm to converge slowly, and even it is difficult to find a better solution. The model of the present application introduces an improved virus evolution algorithm to generate initial weight factors in the traditional neural network algorithm.

[0074] The complex equipment assembly performance prediction method based on the improved virus optimization algorithm of the embodiment has the following specific steps:

[0075] S1, a quality characteristic model of complex equipment assembly parts and core components is established, including establishing an assembly part bill of materials model, an assembly part quality performance characteristic model and a core component quality performance characteristic model.

[0076] S11, an assembly part bill of materials model is established:

[0077] ;

[0078] The assembly part is represented by The assembly part is composed of core components .

[0079] S12, an assembly part quality performance characteristic model is established:

[0080] ;

[0081] The assembly part is represented by The assembly part contains quality characteristics to be predicted .

[0082] S13, a core component quality performance characteristic model is established:

[0083] ;

[0084] The core component is represented by The core component contains key quality characteristics .

[0085] S2, a neural network framework comprising an input layer, a hidden layer and an output layer is constructed as an initial complex equipment assembly quality performance prediction model, and the number of nodes of each layer is determined, wherein the number of nodes of the input layer corresponds to the total number of key quality characteristics of the core components, and the number of nodes of the output layer corresponds to the total number of quality characteristics to be predicted of the assembly.

[0086] S21, the constructed neural network framework comprises an input layer, a hidden layer and an output layer.

[0087] First, the number of input layers of the neural network is determined and the number of nodes of the output layer , denoted as:

[0088] ;

[0089] ;

[0090] According to the empirical formula, the number of hidden layer nodes of the neural network model is determined :

[0091] ;

[0092] Determine the activation function of each layer, wherein the activation function of the hidden layer is ReLU function, and the output layer does not use the activation function.

[0093] S22, according to the established neural network framework, an initial complex equipment assembly quality performance prediction model is established.

[0094] The initial complex equipment assembly quality performance prediction model is denoted as:

[0095] ;

[0096] ;

[0097] wherein, is the parameter of the neural network learned from the training data; denotes the weight between the input layer and the hidden layer of the neural network, denotes the weight between the hidden layer and the output layer of the neural network, denotes the bias of the mapping between the input layer and the hidden layer of the neural network, denotes the bias of the mapping between the hidden layer and the output layer of the neural network. In the embodiment, the bias and are both set to 0.

[0098] S3, sample data for complex equipment assembly quality performance prediction is obtained, and the sample data is standardized.

[0099] Define sample data by The input and output data consist of two groups, denoted as follows: and For the first Group sample data and They are represented as follows:

[0100] ;

[0101] ;

[0102] Normalization ensures that input features and target variables have similar dimensions, preventing certain features from dominating the loss function due to excessively large numerical ranges. This embodiment utilizes Robust Scaler standardization, which scales based on the median and interquartile range. It is insensitive to outliers and suitable for datasets that may contain noise or outliers. The Robust Scaler standardization method is used to normalize the input data in the sample data. The processing is represented as follows:

[0103] ;

[0104] in, and They represent the first Core components in group sample data The The values ​​of key quality characteristics before and after standardization; express Sample data The median value, This represents the difference between the 75th percentile and the 25th percentile of the corresponding sample data.

[0105] S4. Obtain the initial weights of the neural network model using an improved virus optimization algorithm. The improved virus optimization algorithm achieves targeted search through a bidirectional evolutionary mechanism between the main population and the virus population. The specific process is as follows: Figure 1 As shown, it includes:

[0106] S41. Select one unselected data point from the sample data;

[0107] S42. Initialize the main population and virus population for the virus evolution algorithm, including:

[0108] (a) Determine the size of the main population and virus population size ;

[0109] (b) For each principal individual in the principal group, randomly generate the weight parameters of the neural network. and And using a two-stage coding method, a master individual code is formed, such as Figure 2 As shown;

[0110] (c) The virus individual is a substring of the main individual. Wildcards are randomly added to the main individual to form a virus individual code with the same length as the main individual code. Figure 3 The following is from Figure 2 Virus individuals generated by the primary individual;

[0111] S43. Calculate the fitness value of the main population:

[0112] By utilizing the forward propagation decoding mechanism of a neural network, the currently selected sample data is used as the input to the neural network. and output tags The weight represented by the encoding of each principal individual in the principal group. Substitute the values ​​into the neural network model for forward computation to obtain the estimated values. , estimate With tags The loss value is used as the fitness value of the individual;

[0113] S44. Population-oriented evolution based on individual viruses, including:

[0114] (a) The percentage of individuals with better fitness values ​​selected from the primary individuals is: Individuals that are not good are placed into the set of excellent individuals;

[0115] (b) Select one unselected chromosome from the set of excellent individuals, and then... Figure 4 The method shown replicates all individual viruses;

[0116] (c) To reduce the effective genes on the viral chromosome, deletion operations are performed on the replicated viral chromosome. This is done by replacing effective genes with wildcards with a certain probability, as described in the following procedure. Figure 5 As shown;

[0117] (d) Determine whether the virus evolution algorithm has reached the termination condition. If so, proceed to step S45; otherwise, calculate the fitness value of the new population and start the population-oriented evolution of individual viruses again.

[0118] S45. Determine whether the training termination condition has been met. If so, output the code of the optimal individual as the initial weight of the neural network model; otherwise, repeat step S41.

[0119] S5. Substitute the initial weights into the initial complex equipment assembly quality performance prediction model, train the model using the backpropagation algorithm, obtain the trained complex equipment assembly quality performance prediction model, and use the complex equipment assembly quality performance prediction model to predict the assembly performance of complex equipment.

[0120] Example 2: This example provides a non-transitory computer-readable storage medium storing computer instructions that cause a computer to execute a method for predicting the performance of complex equipment assembly based on an improved virus optimization algorithm. The method includes the following steps:

[0121] S1. Establish quality characteristic models for complex equipment assemblies and core components, including constructing a bill of materials model for assemblies, a quality performance characteristic model for assemblies, and a quality performance characteristic model for core components.

[0122] S2. Construct a neural network framework containing an input layer, a hidden layer, and an output layer as an initial complex equipment assembly quality performance prediction model, and determine the number of nodes in each layer, where the number of nodes in the input layer corresponds to the total number of key quality features of the core components, and the number of nodes in the output layer corresponds to the total number of quality features of the assembly to be predicted.

[0123] S3. Obtain sample data for predicting the assembly quality and performance of complex equipment, and standardize the sample data.

[0124] S4. The initial weights of the neural network model are obtained using an improved virus optimization algorithm, which achieves directional search through a bidirectional evolutionary mechanism between the main population and the virus population.

[0125] S5. Substitute the initial weights into the initial complex equipment assembly quality performance prediction model, train the model using the backpropagation algorithm, obtain the trained complex equipment assembly quality performance prediction model, and use the complex equipment assembly quality performance prediction model to predict the assembly performance of complex equipment.

[0126] Example 3: This example provides an electronic device that may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The processor can call logical instructions from the memory to execute a method for predicting the performance of complex equipment assembly based on an improved virus optimization algorithm. This method includes the following steps:

[0127] S1. Establish quality characteristic models for complex equipment assemblies and core components, including constructing a bill of materials model for assemblies, a quality performance characteristic model for assemblies, and a quality performance characteristic model for core components.

[0128] S2. Construct a neural network framework containing an input layer, a hidden layer, and an output layer as an initial complex equipment assembly quality performance prediction model, and determine the number of nodes in each layer, where the number of nodes in the input layer corresponds to the total number of key quality features of the core components, and the number of nodes in the output layer corresponds to the total number of quality features of the assembly to be predicted.

[0129] S3. Obtain sample data for predicting the assembly quality and performance of complex equipment, and standardize the sample data.

[0130] S4. The initial weights of the neural network model are obtained using an improved virus optimization algorithm, which achieves directional search through a bidirectional evolutionary mechanism between the main population and the virus population.

[0131] S5. Substitute the initial weights into the initial complex equipment assembly quality performance prediction model, train the model using the backpropagation algorithm, obtain the trained complex equipment assembly quality performance prediction model, and use the complex equipment assembly quality performance prediction model to predict the assembly performance of complex equipment.

[0132] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part 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 to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0133] Example 4: This example provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a method for predicting the performance of complex equipment assembly based on an improved virus optimization algorithm. This method includes the following steps:

[0134] S1. Establish quality characteristic models for complex equipment assemblies and core components, including constructing a bill of materials model for assemblies, a quality performance characteristic model for assemblies, and a quality performance characteristic model for core components.

[0135] S2. Construct a neural network framework containing an input layer, a hidden layer, and an output layer as an initial complex equipment assembly quality performance prediction model, and determine the number of nodes in each layer, where the number of nodes in the input layer corresponds to the total number of key quality features of the core components, and the number of nodes in the output layer corresponds to the total number of quality features of the assembly to be predicted.

[0136] S3. Obtain sample data for predicting the assembly quality and performance of complex equipment, and standardize the sample data.

[0137] S4. The initial weights of the neural network model are obtained using an improved virus optimization algorithm, which achieves directional search through a bidirectional evolutionary mechanism between the main population and the virus population.

[0138] S5. Substitute the initial weights into the initial complex equipment assembly quality performance prediction model, train the model using the backpropagation algorithm, obtain the trained complex equipment assembly quality performance prediction model, and use the complex equipment assembly quality performance prediction model to predict the assembly performance of complex equipment.

[0139] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the assembly performance of complex equipment based on an improved virus optimization algorithm, characterized in that, Includes the following steps: S1. Establish quality characteristic models for complex equipment assemblies and core components, including constructing a bill of materials model for assemblies, a quality performance characteristic model for assemblies, and a quality performance characteristic model for core components. S2. Construct a neural network framework containing an input layer, a hidden layer, and an output layer as an initial complex equipment assembly quality performance prediction model, and determine the number of nodes in each layer, where the number of nodes in the input layer corresponds to the total number of key quality features of the core components, and the number of nodes in the output layer corresponds to the total number of quality features of the assembly to be predicted. S3. Obtain sample data for predicting the assembly quality and performance of complex equipment, and standardize the sample data. S4. Obtain the initial weights of the neural network model using an improved virus optimization algorithm. The improved virus optimization algorithm achieves targeted search through a bidirectional evolutionary mechanism between the main population and the virus population, specifically including: Initialize the main population and the virus population. In the main population, each master individual encodes all weight parameters of the neural network with real numbers through a two-stage encoding method. In the virus population, each virus individual is formed by copying the encoding of the master individual and randomly inserting wildcards. The fitness value of each individual in the main population is calculated based on the forward propagation of the neural network. The fitness value is determined by the predicted output under the weight parameters represented by the individual encoding and the loss of the sample label. Select the best individuals with the highest fitness values ​​from the main population, and use the infection mechanism to perform targeted replication on the virus population based on the encoding of the best individuals; After replication, the effective genes of individual viruses are deleted into wildcards with a preset probability, and a local fine search is performed on the virus population. The above process is executed iteratively until the termination condition is met, and the weight parameters corresponding to the optimal individual code are output as the initial weights of the neural network model. S5. Substitute the initial weights into the initial complex equipment assembly quality performance prediction model, train the model using the backpropagation algorithm, obtain the trained complex equipment assembly quality performance prediction model, and use the complex equipment assembly quality performance prediction model to predict the assembly performance of complex equipment.

2. The method for predicting the assembly performance of complex equipment based on an improved virus optimization algorithm according to claim 1, characterized in that, Step S1 involves establishing quality characteristic models for complex equipment assemblies and core components, including constructing a bill of materials model for assemblies, a quality and performance characteristic model for assemblies, and a quality and performance characteristic model for core components; specifically: S11. Establish a bill of materials model for complex equipment components: ； Indicates assembly parts Depend on Core components Assembly composition; S12. Establish a quality performance characteristic model for the assembly: ； Indicates assembly parts Total of One quality characteristic to be predicted ; S13. Establish a quality performance characteristic model for core components: ； Indicates core components Total of Key quality characteristics .

3. The method for predicting the assembly performance of complex equipment based on an improved virus optimization algorithm according to claim 2, characterized in that, Step S2 involves constructing a neural network framework comprising an input layer, hidden layers, and an output layer, specifically as follows: First, determine the number of input layers in the neural network. and the number of output layer nodes , is represented as: ； ； The number of hidden layer nodes in a neural network model is determined using empirical formulas. : ； Determine the activation function for each layer, where the activation function for the hidden layer is the ReLU function, and no activation function is used for the output layer.

4. The method for predicting the assembly performance of complex equipment based on an improved virus optimization algorithm according to claim 3, characterized in that, The initial complex equipment assembly quality performance prediction model is expressed as follows: ； ； in, These are the parameters of the neural network learned from the training data; This represents the weights between the input layer and the hidden layer of a neural network. This represents the weights between the hidden layers and the output layer of a neural network. This represents the bias in the mapping between the input layer and the hidden layer of a neural network. This represents the bias in the mapping between the hidden layers and the output layer of a neural network.

5. The method for predicting the assembly performance of complex equipment based on an improved virus optimization algorithm according to claim 4, characterized in that, The bias of the mapping between the input layer and the hidden layer of the neural network And the bias of the mapping between the hidden layers and the output layer of the neural network. All are 0.

6. The method for predicting the assembly performance of complex equipment based on an improved virus optimization algorithm according to claim 1, characterized in that, Step S3, which involves standardizing the sample data, specifically includes: Define sample data by The input and output data consist of two groups, denoted as follows: and For the first Group sample data and They are represented as follows: ； ； The Robust Scaler standardization method was used to normalize the input data in the sample data. The processing is represented as follows: ； in, and They represent the first Core components in group sample data The The values ​​of key quality characteristics before and after standardization; express Sample data The median value, This represents the difference between the 75th percentile and the 25th percentile of the corresponding sample data.

7. The method for predicting the assembly performance of complex equipment based on an improved virus optimization algorithm according to claim 4, characterized in that, Step S4, which involves obtaining the initial weights of the neural network model using an improved virus optimization algorithm, specifically includes: S41. Select one unselected data point from the sample data; S42. Initialize the main population and virus population for the virus evolution algorithm, including: (a) Determine the size of the main population and virus population size ; (b) For each principal individual in the principal group, randomly generate the weight parameters of the neural network. and And using a two-stage coding method, a master individual code is formed; (c) The virus individual is a substring of the main individual. Wildcards are randomly added to the main individual to form a virus individual code with the same length as the main individual code. S43. Calculate the fitness value of the main population: By utilizing the forward propagation decoding mechanism of a neural network, the currently selected sample data is used as the input to the neural network. and output tags The weight represented by the encoding of each principal individual in the principal group. Substitute the values ​​into the neural network model for forward computation to obtain the estimated values. , estimate With tags The loss value is used as the fitness value of the individual; S44. Population-oriented evolution based on individual viruses, including: (a) The percentage of individuals with better fitness values ​​selected from the primary individuals is: Individuals that are not good are placed into the set of excellent individuals; (b) Select one unselected chromosome from the set of excellent individuals and replicate it on all viral individuals; (c) Perform deletion operations on the replicated viral chromosome, replacing its effective genes with wildcards with a set probability; (d) Determine whether the virus evolution algorithm has reached the termination condition. If so, proceed to step S45; otherwise, calculate the fitness value of the new population and start the population-oriented evolution of individual viruses again. S45. Determine whether the training termination condition has been met. If so, output the code of the optimal individual as the initial weight of the neural network model; otherwise, repeat step S41.

8. A non-transitory computer-readable storage medium, characterized in that, It stores computer instructions that cause the computer to execute the complex equipment assembly performance prediction method based on the improved virus optimization algorithm as described in any one of claims 1-7.

9. An electronic device, characterized in that, include: The system includes a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The processor calls logical instructions from the memory to execute the complex equipment assembly performance prediction method based on the improved virus optimization algorithm as described in any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer executes the complex equipment assembly performance prediction method based on the improved virus optimization algorithm as described in any one of claims 1-7.

Images