Collaborative manufacturing management platform and method based on industrial Internet

Through the industrial Internet collaborative manufacturing management platform, the generated scheduling solution solves the problem of workpiece quality and machine load not being considered in flexible workshop scheduling, and realizes workpiece processing quality and Optimization of machine load.

CN118798825BActive Publication Date: 2024-12-03FUJIAN YUNCHUANG XINAN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411272121.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-11
Publication Date
2024-12-03
Estimated Expiration
2044-09-11

AI Technical Summary

Technical Problem

The existing flexible workshop scheduling only considers the workpiece machining completion time and fails to fully consider the workpiece machining quality and machining machine load, resulting in the generated scheduling scheme that may not be the optimal solution.

Method used

The collaborative manufacturing management platform based on the industrial Internet generates a scheduling solution through workpiece information collection, machine information collection, heterogeneous graph data construction, prediction model construction and red-mouthed blue magpi optimization algorithm to ensure the optimal workpiece processing quality and the minimum machine load.

Benefits of technology

Through the information aggregation of heterogeneous graph data and prediction model, combined with the red-mouthed blue magpi optimization algorithm, the generated scheduling scheme can effectively improve the quality of workpiece processing and optimize the machine load, ensuring the optimal production scheme.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118798825B_ABST
    Figure CN118798825B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of collaborative manufacturing, and discloses a collaborative manufacturing management platform and method based on the industrial Internet, including: a heterogeneous graph data construction module, which is used to construct heterogeneous graph data according to workpiece information of a workpiece to be processed and machine information of a processing machine; a prediction model construction module, which is used to construct and train a prediction model; a scheduling plan generation module, which is used to generate a scheduling plan by combining a red-billed blue magpie optimization algorithm with a prediction model; the present invention constructs heterogeneous graph data according to information of the workpiece to be processed and the processing machine, aggregates information on the heterogeneous graph data through a prediction model to obtain the processing quality of the workpiece to be processed and the load of the processing machine, and generates a scheduling plan by combining a red-billed blue magpie optimization algorithm with the prediction model, so that the processing quality of the workpiece to be processed is optimized and the load of the processing machine is minimized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of collaborative manufacturing technology, and more specifically, to a collaborative manufacturing management platform and method based on industrial Internet. Background Art

[0002] With the development of industrial Internet technology, the manufacturing industry is undergoing profound changes. Industrial Internet has achieved comprehensive interconnection of production equipment and information systems through technologies such as the Internet of Things, big data and cloud computing, greatly improving production efficiency. Among them, collaborative manufacturing management is one of the key strategies to enhance the competitiveness of the manufacturing industry. Collaborative manufacturing management emphasizes cross-organizational collaboration and aims to achieve overall optimization of production through information sharing and optimal resource utilization. In collaborative manufacturing management, flexible workshop scheduling plays an important role. Compared with traditional workshop scheduling, flexible workshop scheduling allows the processing procedures of different workpieces to be processed on multiple machines, which is more in line with the actual workshop production scenario.

[0003] The existing flexible workshop scheduling constructs the allocation relationship between the processing procedures of workpieces and processing machines into a matrix code representation. The rows of the matrix code represent workpieces, the columns represent procedures, and the elements of the matrix code are represented by real number codes. The integer part of the real number code represents the serial number of the processing machine, and the decimal part represents the processing priority. The processing is performed in order from small to large. For example, there are 4 workpieces, and each workpiece has 3 processing procedures. Then the matrix code includes 4 rows and 3 columns. If the element values ​​of the 1st row to the 4th row of the 1st column are 2.1, 2.2, 2.3 and 3.1 respectively, it means that the 1st procedure of the 1st workpiece, the 2nd workpiece and the 3rd workpiece are all processed on the processing machine with serial number 2, and the processing order is the 1st workpiece, the 2nd workpiece and the 3rd workpiece. Then, with minimizing the workpiece processing completion time as the objective function, a scheduling plan is generated by a genetic algorithm to complete the allocation of the processing procedures and processing machines of the workpieces.

[0004] However, the above scheme only considers the completion time of workpiece processing, but does not consider the processing quality of the workpiece and the load of the processing machine, resulting in that the generated scheduling scheme may not be the optimal solution. Summary of the invention

[0005] The present invention provides a collaborative manufacturing management platform and method based on the Industrial Internet, which solves the technical problem that the existing flexible workshop scheduling in the related technology only considers the completion time of workpiece processing, but does not consider the processing quality of the workpiece and the load of the processing machine, resulting in that the generated scheduling plan may not be the optimal solution.

[0006] The present invention provides a collaborative manufacturing management platform based on the Industrial Internet, including:

[0007] A workpiece information acquisition module, which is used to acquire workpiece information of the workpiece to be processed;

[0008] Workpiece information includes: unique identifier, weight, maximum length, maximum width, maximum height, processing completion time and number of processing steps;

[0009] A machine information collection module, which is used to collect machine information of the processing machine;

[0010] Machine information includes: unique identifier, purchase amount, energy consumption, rated power, historical maintenance times, historical maintenance amount, wear degree and processing accuracy;

[0011] A heterogeneous graph data construction module, which is used to construct heterogeneous graph data according to workpiece information of the workpiece to be processed and machine information of the processing machine;

[0012] Heterogeneous graph data includes: nodes, initial features, and edges between nodes;

[0013] The nodes include: a first node and a second node;

[0014] A mapping relationship is established between the first node and the workpiece to be processed, and an initial feature of the first node is represented by workpiece information of the workpiece to be processed that has a mapping relationship established with the first node;

[0015] A mapping relationship is established between the second node and the processing machine, and the initial feature of the second node is represented by the machine information of the processing machine that establishes a mapping relationship with the second node;

[0016] A prediction model building module, which is used to build and train a prediction model;

[0017] The prediction model inputs heterogeneous graph data, and the output values ​​include the processing quality of each workpiece to be processed and the load of each processing machine;

[0018] A scheduling plan generation module, which is used to generate a scheduling plan by combining the red-billed blue magpie optimization algorithm with the prediction model;

[0019] According to the scheduling plan, the workpieces to be processed are assigned to the processing machines for processing.

[0020] Furthermore, the unique identifier of the workpiece to be processed is represented by a positive integer code, the processing completion time is represented by the processing completion time minus the current time, and the number of processing steps for each workpiece to be processed is different.

[0021] Furthermore, the unique identifier of the processing machine is represented by a positive integer code, the energy consumption is represented by the power consumption per hour, the degree of wear is represented by the usage time divided by the maximum usage time, and the processing accuracy is represented by the yield rate of the processed workpiece per unit time.

[0022] Furthermore, the edges between nodes include: an edge constructed between a first node and a second node indicates that a workpiece to be processed that has a mapping relationship with the first node is assigned to a processing machine that has a mapping relationship with the second node; an edge constructed between second nodes indicates that the processing machines that have a mapping relationship with the second node are powered by the same power system.

[0023] Further, the prediction model includes: a first feature conversion layer, a second feature conversion layer, a graph update layer, a first extraction layer, a second extraction layer, a first classifier, and a second classifier;

[0024] The first feature conversion layer inputs the initial feature of the first node of the heterogeneous graph data and outputs the first updated feature;

[0025] The second feature conversion layer inputs the initial feature of the second node of the heterogeneous graph data and outputs the second updated feature;

[0026] The first updated feature and the second updated feature have the same size;

[0027] The graph update layer inputs heterogeneous graph data and outputs an update matrix;

[0028] A row vector of the update matrix corresponds to a third updated feature of the first node or the second node;

[0029] The first extraction layer is used to extract the third updated feature corresponding to the first node and input it into the first classifier, and the classification space of the first classifier represents the processing quality of the workpiece to be processed;

[0030] The second extraction layer is used to extract the third updated feature corresponding to the second node and input it into the second classifier. The classification space of the second classifier represents the load of the processing machine.

[0031] Furthermore, the calculation formula of the prediction model includes:

[0032] The calculation formula of the first feature conversion layer is as follows:

[0033] ;

[0034] in represents the first updated feature output by the first feature conversion layer, A represents the initial feature of the first node, and They represent the weight parameters and bias parameters corresponding to the first feature conversion layer respectively, and sigmoid represents the sigmoid activation function;

[0035] The calculation formula of the second feature conversion layer is as follows:

[0036] ;

[0037] in represents the second updated feature output by the second feature conversion layer, B represents the initial feature of the second node, and Respectively represent the weight parameters and bias parameters corresponding to the second feature conversion layer;

[0038] The calculation formula of the graph update layer is as follows:

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] Where 1≤i≤S, Matrix represents the update matrix output by the graph update layer, S represents the sum of the number of the first node and the second node, represents the i-th row vector of the update matrix and represents the third updated feature of the i-th first node or second node, and denote the first updated feature of the i-th and j-th first nodes or the second updated feature of the second nodes, respectively. represents the set of nodes that have edges connected to the i-th first node or second node. , , and Respectively represent the first coefficient, second coefficient, bias parameter and number of edges between the i-th node and the j-th node, , and represent the first weight parameter, the second weight parameter and the third weight parameter respectively, It indicates that the third updated features of S nodes are stacked, T indicates a transposition operation, and concat indicates a concatenation operation.

[0044] Furthermore, the sample labels corresponding to the training samples used to train the prediction model are obtained through manual labeling, and the processing quality of the workpiece to be processed and the load of the processing machine are both customized discrete values.

[0045] Furthermore, a scheduling plan is generated by combining the red-billed blue magpie optimization algorithm with the prediction model, including the following steps:

[0046] Step S301, randomly generate an initialization population that meets the constraint conditions, the number of individuals in the initialization population is U, where U is a custom parameter;

[0047] Constraints include: the same processing step of each workpiece to be processed can only be assigned to one processing machine; the processing order of each workpiece to be processed must be strictly processed in sequence according to the process flow, and the next processing step of the same workpiece to be processed can only be started after the previous processing step is completed;

[0048] The individual is represented by a matrix code, which includes N rows and M columns, where N represents the number of workpieces to be processed, M represents the maximum number of processing steps, and the element value of the matrix code is represented by a real number, the integer part of the real number represents the unique identifier of the processing machine, and the decimal part represents the processing priority, and the processing is performed in order from small to large;

[0049] The calculation formula of the element value of the matrix encoding is as follows:

[0050] ;

[0051] Where 1≤m≤M, 1≤n≤N, ub and lb represent the upper and lower bounds respectively. Indicates a random number ranging from 0 to 1;

[0052] Step S302, calculating the fitness values ​​of all individuals in the initialized population through an objective function, wherein the objective function is constructed based on a prediction model;

[0053] The objective function is calculated as follows:

[0054] ;

[0055] Where Fit represents the fitness value of an individual, N represents the number of workpieces to be processed, and R represents the number of processing machines. represents the machining quality of the nth workpiece to be machined output by the prediction model, represents the load of the rth processing machine output by the prediction model, and Represent the first attention and the second attention respectively, and are custom parameters, and and The sum of is 1, and represent the first adjustment factor and the second adjustment factor respectively, and All are custom parameters;

[0056] Step S303, generating a first random number ranging from 0 to 1. When the first random number is less than the balance coefficient, the initialized population is divided into multiple subgroups, each of which has Q individuals, and the matrix code of the individuals in each subgroup is updated through the food search strategy. Otherwise, the matrix code of the individuals in the initialized population is directly updated through the food search strategy, where Q is a custom parameter.

[0057] Step S304, generating a second random number in the range of 0 to 1. When the second random number is less than the balance coefficient, the initialized population is divided into multiple subgroups, each of which has Q individuals, and the matrix code of the individuals in each subgroup is updated through the food search strategy, otherwise the matrix code of the individuals in the initialized population is directly updated through the prey attack strategy;

[0058] Step S305, updating the matrix codes of individuals of the initialized population by storing food strategies;

[0059] Step S306, the current number of iterations is increased by 1, and if it is determined that the iteration termination condition is met, the individual with the largest fitness value in the initialized population is used as the scheduling plan, otherwise it returns to step S302 to continue execution, where the iteration termination condition is that the current number of iterations reaches the maximum number of iterations or the maximum value of the fitness value within K consecutive iterations does not exceed the set threshold, where the maximum number of iterations, K and the set threshold are all custom parameters.

[0060] Furthermore, the calculation formula of the red-billed blue magpie optimization algorithm includes:

[0061] The calculation formula for finding food strategy is as follows:

[0062] ;

[0063] Where P is assigned to Q or U, 1≤p≤Q or 1≤p≤U, and Respectively represent the matrix code of the i-th individual with the current iteration number t+1 and t, and They represent the matrix codes of the pth individual and a random individual at the current iteration number t, respectively. Indicates a random number ranging from 0 to 1;

[0064] The calculation formula for attacking prey strategy includes:

[0065] ;

[0066] ;

[0067] in It represents the matrix encoding of the individual with the largest fitness value at the current iteration number t, Max represents the maximum number of iterations, CF represents the flight factor, Represents random numbers that conform to the standard normal distribution;

[0068] The calculation formula for the food storage strategy is as follows:

[0069] ;

[0070] in Represents the maximum fitness value of the i-th individual in the historical iteration number, Represents the fitness value of the i-th individual with the current iteration number t.

[0071] The present invention provides a collaborative manufacturing management method based on the industrial Internet, comprising the following steps:

[0072] Step S401, collecting workpiece information of the workpiece to be processed;

[0073] Step S402, collecting machine information of the processing machine;

[0074] Step S403, constructing heterogeneous graph data according to workpiece information of the workpiece to be processed and machine information of the processing machine;

[0075] Step S404, constructing and training a prediction model;

[0076] Step S405, generating a scheduling plan by combining the red-billed blue magpie optimization algorithm with the prediction model.

[0077] The beneficial effects of the present invention are as follows: the present invention constructs heterogeneous graph data according to the information of the workpiece to be processed and the processing machine, obtains the processing quality of the workpiece to be processed and the load of the processing machine by performing information aggregation on the heterogeneous graph data through a prediction model, and generates a scheduling plan through a red-billed blue magpie optimization algorithm combined with the prediction model, so that the processing quality of the workpiece to be processed is optimized and the load of the processing machine is minimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 is a schematic diagram of a collaborative manufacturing management platform based on the industrial Internet of the present invention;

[0079] Figure 2 is a schematic diagram of a prediction model of the present invention;

[0080] Figure 3 It is a flow chart of generating a scheduling plan by combining a red-billed blue magpie optimization algorithm with a prediction model according to the present invention;

[0081] Figure 4 It is a flow chart of the collaborative manufacturing management method based on industrial Internet of the present invention.

[0082] In the figure: workpiece information acquisition module 101, machine information acquisition module 102, heterogeneous graph data construction module 103, prediction model construction module 104, scheduling plan generation module 105, first feature conversion layer 201, second feature conversion layer 202, graph update layer 203, first extraction layer 204, second extraction layer 205, first classifier 206 and second classifier 207. DETAILED DESCRIPTION

[0083] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0084] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0085] like Figure 1 to Figure 4 As shown in the figure, the collaborative manufacturing management platform based on the Industrial Internet includes:

[0086] A workpiece information acquisition module 101 is used to acquire workpiece information of a workpiece to be processed;

[0087] Workpiece information includes: unique identifier, weight, maximum length, maximum width, maximum height, processing completion time and number of processing steps;

[0088] A machine information collection module 102, which is used to collect machine information of a processing machine;

[0089] Machine information includes: unique identifier, purchase amount, energy consumption, rated power, historical maintenance times, historical maintenance amount, wear degree and processing accuracy;

[0090] A heterogeneous graph data construction module 103, which is used to construct heterogeneous graph data according to workpiece information of the workpiece to be processed and machine information of the processing machine;

[0091] Heterogeneous graph data includes: nodes, initial features, and edges between nodes;

[0092] The nodes include: a first node and a second node;

[0093] A mapping relationship is established between the first node and the workpiece to be processed, and an initial feature of the first node is represented by workpiece information of the workpiece to be processed that has a mapping relationship established with the first node;

[0094] A mapping relationship is established between the second node and the processing machine, and the initial feature of the second node is represented by the machine information of the processing machine that establishes a mapping relationship with the second node;

[0095] A prediction model building module 104, which is used to build and train a prediction model;

[0096] The prediction model inputs heterogeneous graph data, and the output values ​​include the processing quality of each workpiece to be processed and the load of each processing machine;

[0097] A scheduling scheme generating module 105, which is used to generate a scheduling scheme by combining a red-billed blue magpie optimization algorithm with a prediction model;

[0098] According to the scheduling plan, the workpieces to be processed are assigned to the processing machines for processing.

[0099] In one embodiment of the present invention, the unique identifier of the workpiece to be processed is represented by a positive integer code, the processing completion time is represented by the processing completion time minus the current time, and the number of processing steps for each workpiece to be processed is different.

[0100] In one embodiment of the present invention, the unique identifier of the workpiece to be processed can also be represented by binary code or hash code, and the number of processing steps for each workpiece to be processed can be the same or different, and is set according to the number of processing steps actually required for each workpiece to be processed.

[0101] It should be noted that the initial feature of the first node is expressed as:

[0102] ,in They respectively represent the unique identifier, weight, maximum length, maximum width, maximum height, processing completion time and number of processing steps of the workpiece to be processed that establishes a mapping relationship with the first node, that is, the initial feature of the first node is a vector with 7 dimensions.

[0103] In one embodiment of the present invention, the unique identifier of the processing machine is represented by a positive integer code, the energy consumption is represented by the power consumption per hour, the degree of wear is represented by the usage time divided by the maximum usage time, and the processing accuracy is represented by the yield rate of the processed workpiece per unit time.

[0104] In one embodiment of the present invention, the unique identifier of the processing machine can also be represented by a binary code or a hash code, and the energy consumption situation can also be represented by an energy consumption level. For example, the energy consumption levels include: level one energy consumption, level two energy consumption and level three energy consumption. The energy consumption level can be represented by a real number code. For example, a real number code of 1.1 indicates that the energy consumption level is level one energy consumption. The units of the maximum usage time and the usage time of the processing machine are both days, and the maximum usage time is provided by the manufacturer of the processing machine. The processing accuracy is represented by the yield of the workpiece processed in unit time. The unit time can be one day or one month, which will not be elaborated here.

[0105] It should be noted that the initial feature of the second node is expressed as:

[0106] ,in They respectively represent the unique identifier, purchase amount, energy consumption, rated power, historical maintenance times, historical maintenance amount, wear degree and processing accuracy of the processing machine that establishes a mapping relationship with the second node, that is, the initial feature of the second node is a vector with 8 dimensions.

[0107] In one embodiment of the present invention, the edges between nodes include: the edge constructed between the first node and the second node indicates that the workpiece to be processed that establishes a mapping relationship with the first node is assigned to the processing machine that establishes a mapping relationship with the second node; the edge constructed between the second nodes indicates that the processing machines that establish a mapping relationship with the second node are powered by the same power system.

[0108] It should be noted that the edge constructed between the first node and the second node represents the allocation relationship between the workpiece to be processed and the processing machine, and there may be multiple edges between the first node and the second node, because different processing procedures of the same workpiece to be processed may be assigned to the same processing machine. In addition, the edge constructed between the second nodes indicates that different processing machines share the same resources, such as being powered by the same power system, cooled by the same cooling system, etc., and the processing machines sharing the same resources have a mutual influence relationship. The edge constructed between the second nodes can also represent the production line coordination relationship of the processing machines. For example, each processing machine can only process one processing procedure of the workpiece. The workpiece to be processed needs to be preliminarily processed on the processing machine with a unique identifier of 1, and then transferred to the processing machine with a unique identifier of 2 for further processing. Therefore, the edge constructed between the second nodes can also represent this processing order constraint.

[0109] In one embodiment of the present invention, Figure 2 As shown, the prediction model includes: a first feature conversion layer 201, a second feature conversion layer 202, a graph update layer 203, a first extraction layer 204, a second extraction layer 205, a first classifier 206 and a second classifier 207;

[0110] The first feature conversion layer inputs the initial feature of the first node of the heterogeneous graph data and outputs the first updated feature;

[0111] The second feature conversion layer inputs the initial feature of the second node of the heterogeneous graph data and outputs the second updated feature;

[0112] The first updated feature and the second updated feature have the same size;

[0113] The graph update layer inputs heterogeneous graph data and outputs an update matrix;

[0114] A row vector of the update matrix corresponds to a third updated feature of the first node or the second node;

[0115] The first extraction layer is used to extract the third updated feature corresponding to the first node and input it into the first classifier, and the classification space of the first classifier represents the processing quality of the workpiece to be processed;

[0116] The second extraction layer is used to extract the third updated feature corresponding to the second node and input it into the second classifier. The classification space of the second classifier represents the load of the processing machine.

[0117] In one embodiment of the present invention, the calculation formula of the prediction model includes:

[0118] The calculation formula of the first feature conversion layer is as follows:

[0119] ;

[0120] in represents the first updated feature output by the first feature conversion layer, A represents the initial feature of the first node, and They represent the weight parameters and bias parameters corresponding to the first feature conversion layer respectively, and sigmoid represents the sigmoid activation function;

[0121] The calculation formula of the second feature conversion layer is as follows:

[0122] ;

[0123] in represents the second updated feature output by the second feature conversion layer, B represents the initial feature of the second node, and Respectively represent the weight parameters and bias parameters corresponding to the second feature conversion layer;

[0124] The calculation formula of the graph update layer is as follows:

[0125] ;

[0126] ;

[0127] ;

[0128] ;

[0129] Where 1≤i≤S, Matrix represents the update matrix output by the graph update layer, S represents the sum of the number of the first node and the second node, represents the i-th row vector of the update matrix and represents the third updated feature of the i-th first node or second node, and denote the first updated feature of the i-th and j-th first nodes or the second updated feature of the second nodes, respectively. represents the set of nodes that have edges connected to the i-th first node or second node. , , and Respectively represent the first coefficient, second coefficient, bias parameter and number of edges between the i-th node and the j-th node, , and represent the first weight parameter, the second weight parameter and the third weight parameter respectively, It indicates that the third updated features of S nodes are stacked, T indicates a transposition operation, and concat indicates a concatenation operation.

[0130] It should be noted that the initial feature of the first node is a vector of size 1×7, and the weight parameter corresponding to the first feature conversion layer can be a matrix of size 7×64. The second updated feature is a vector of size 1×64, and the initial feature of the second node is a vector of size 1×8. Therefore, the weight parameter corresponding to the second feature conversion layer can be a matrix of size 8×64. In addition, the first feature conversion layer and the second feature conversion layer can be designed as multiple nonlinear units. For example, the first feature conversion layer can include 3 nonlinear units, and the corresponding weight parameters are matrices of size 7×16, 16×32 and 32×64, respectively. The second feature conversion layer can also include 3 nonlinear units, and the corresponding weight parameters are matrices of size 8×16, 16×32 and 32×64, respectively.

[0131] It should be noted that, assuming that the first updated feature of the first node and the second updated feature of the second node are both vectors of size 1×64, then the first updated feature of the first node or the second updated feature of the second node can be concatenated to obtain a vector of size 1×128, the third weight parameter can be designed as a matrix of size 128×16, the first coefficient is a vector of size 1×16, and the first coefficient is transposed to obtain a vector of size 16×1, the second weight parameter can be designed as a vector of size 1×64, and the first updated feature of the first node or the second updated feature of the second node is transposed to obtain a vector of size 64×1, then the vectors of size 16×1, 1×64 and 64×1 are multiplied to obtain a vector of size 16×1, the first weight parameter can be designed as a matrix of size 16×64, and then multiplied by the transpose of the first update feature of the first node or the second update feature of the second node, and a vector of size 16×1 can also be obtained, that is, the third update feature of the first node or the second node is a vector of size 16×1.

[0132] In one embodiment of the present invention, both the first classifier and the second classifier are softmax classifiers.

[0133] In one embodiment of the present invention, sample labels corresponding to training samples used to train the prediction model are obtained through manual labeling, and the processing quality of the workpiece to be processed and the load of the processing machine are both user-defined discrete values.

[0134] For example, the processing quality of the workpiece to be processed includes excellent, medium and low, and the load of the processing machine includes idle, general and busy, all of which are represented by positive integer codes. For example, the processing quality of the workpiece to be processed is excellent, which is represented by the positive integer 1, and the load of the processing machine is idle, which can also be represented by the positive integer 1. In addition, the processing quality of the workpiece to be processed and the load of the processing machine can also be discrete values ​​ranging from 1 to 100.

[0135] It should be noted that the weight parameters and bias parameters in the prediction model are all learnable parameters. In the process of prediction model training, the difference between the predicted value output by the prediction model at each iteration and the true value corresponding to the sample label of the training sample is specified as the loss function. The chain rule is used to calculate the gradient of the weight parameters and bias parameters in the prediction model, and then the weight parameters and bias parameters in the prediction model are updated through back propagation and gradient descent algorithm to minimize the loss function until the maximum number of iterations is reached and the training is terminated.

[0136] In one embodiment of the present invention, Figure 3 As shown, the scheduling plan is generated by combining the red-billed blue magpie optimization algorithm with the prediction model, including the following steps:

[0137] Step S301, randomly generate an initialization population that meets the constraint conditions, the number of individuals in the initialization population is U, where U is a custom parameter, preferably, U is set to 20;

[0138] Constraints include: the same processing step of each workpiece to be processed can only be assigned to one processing machine; the processing order of each workpiece to be processed must be strictly processed in sequence according to the process flow, and the next processing step of the same workpiece to be processed can only be started after the previous processing step is completed;

[0139] The individual is represented by a matrix code, which includes N rows and M columns, where N represents the number of workpieces to be processed, M represents the maximum number of processing steps, and the element value of the matrix code is represented by a real number, the integer part of the real number represents the unique identifier of the processing machine, and the decimal part represents the processing priority, and the processing is performed in order from small to large;

[0140] The calculation formula of the element value of the matrix encoding is as follows:

[0141] ;

[0142] Where 1≤m≤M, 1≤n≤N, ub and lb represent the upper and lower bounds respectively. Indicates a random number ranging from 0 to 1;

[0143] For example, if the unique identifier of a processing machine ranges from 1 to 10, and the maximum number of processing steps is 5, then ub can be assigned a value of 10.5 and lb can be assigned a value of 1.1;

[0144] Step S302, calculating the fitness values ​​of all individuals in the initialized population through an objective function, wherein the objective function is constructed based on a prediction model;

[0145] The objective function is calculated as follows:

[0146] ;

[0147] Where Fit represents the fitness value of an individual, N represents the number of workpieces to be processed, and R represents the number of processing machines. represents the machining quality of the nth workpiece to be machined output by the prediction model, represents the load of the rth processing machine output by the prediction model, and Represent the first attention and the second attention respectively, and are custom parameters, and and The sum of is 1, and represent the first adjustment factor and the second adjustment factor respectively, and All are custom parameters, preferably, Set to 1 / 10, Set to 10;

[0148] Step S303, generating a first random number ranging from 0 to 1. When the first random number is less than the balance coefficient, the initialized population is divided into multiple subgroups, each of which has Q individuals, and the matrix code of the individuals in each subgroup is updated by the food search strategy. Otherwise, the matrix code of the individuals in the initialized population is directly updated by the food search strategy, where Q is a custom parameter, preferably, Q is set to 5;

[0149] Step S304, generating a second random number in the range of 0 to 1. When the second random number is less than the balance coefficient, the initialized population is divided into multiple subgroups, each of which has Q individuals, and the matrix code of the individuals in each subgroup is updated through the food search strategy, otherwise the matrix code of the individuals in the initialized population is directly updated through the prey attack strategy;

[0150] Step S305, updating the matrix codes of individuals of the initialized population by storing food strategies;

[0151] Step S306, the current number of iterations is increased by 1, and if it is determined that the iteration termination condition is met, the individual with the largest fitness value in the initialized population is used as the scheduling plan, otherwise, it returns to step S302 to continue execution, where the iteration termination condition is that the current number of iterations reaches the maximum number of iterations or the maximum value of the fitness value within K consecutive iterations does not change by more than the set threshold, where the maximum number of iterations, K and the set threshold are all custom parameters. Preferably, the maximum number of iterations is set to 200, K is set to 3, and the set threshold is set to 0.5.

[0152] It should be noted that the prediction model can also be used to predict the processing time of the workpiece to be processed, and the processing time of the workpiece to be processed can be integrated into the objective function to achieve optimization in multiple objective dimensions, which will not be elaborated here.

[0153] In one embodiment of the present invention, the calculation formula of the red-billed blue magpie optimization algorithm includes:

[0154] The calculation formula for finding food strategy is as follows:

[0155] ;

[0156] Where P is assigned to Q or U, 1≤p≤Q or 1≤p≤U, and Respectively represent the matrix code of the i-th individual with the current iteration number t+1 and t, and They represent the matrix codes of the pth individual and a random individual at the current iteration number t, respectively. Indicates a random number ranging from 0 to 1;

[0157] The calculation formula for attacking prey strategy includes:

[0158] ;

[0159] ;

[0160] in It represents the matrix encoding of the individual with the largest fitness value at the current iteration number t, Max represents the maximum number of iterations, CF represents the flight factor, Represents random numbers that conform to the standard normal distribution;

[0161] The calculation formula for the food storage strategy is as follows:

[0162] ;

[0163] in Represents the maximum fitness value of the i-th individual in the historical iteration number, Represents the fitness value of the i-th individual with the current iteration number t.

[0164] In one embodiment of the present invention, Figure 4 As shown, the collaborative manufacturing management method based on the industrial Internet includes the following steps:

[0165] Step S401, collecting workpiece information of the workpiece to be processed;

[0166] Step S402, collecting machine information of the processing machine;

[0167] Step S403, constructing heterogeneous graph data according to workpiece information of the workpiece to be processed and machine information of the processing machine;

[0168] Step S404, constructing and training a prediction model;

[0169] Step S405, generating a scheduling plan by combining the red-billed blue magpie optimization algorithm with the prediction model.

[0170] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are protected by the present embodiment.

Claims

1. Collaborative manufacturing management platform based on industrial Internet, It is characterized in that include: A workpiece information acquisition module, which is used to acquire workpiece information of the workpiece to be processed; Workpiece information includes: unique identifier, weight, maximum length, maximum width, maximum height, processing completion time and number of processing steps; A machine information collection module, which is used to collect machine information of the processing machine; Machine information includes: unique identifier, purchase amount, energy consumption, rated power, historical maintenance times, historical maintenance amount, wear degree and processing accuracy; A heterogeneous graph data construction module, which is used to construct heterogeneous graph data according to workpiece information of the workpiece to be processed and machine information of the processing machine; Heterogeneous graph data includes: nodes, initial features, and edges between nodes; The nodes include: a first node and a second node; A mapping relationship is established between the first node and the workpiece to be processed, and an initial feature of the first node is represented by workpiece information of the workpiece to be processed that has a mapping relationship established with the first node; A mapping relationship is established between the second node and the processing machine, and the initial feature of the second node is represented by the machine information of the processing machine that establishes a mapping relationship with the second node; A prediction model building module, which is used to build and train a prediction model; The prediction model inputs heterogeneous graph data, and the output values ​​include the processing quality of each workpiece to be processed and the load of each processing machine; A scheduling plan generation module, which is used to generate a scheduling plan by combining the red-billed blue magpie optimization algorithm with the prediction model; Allocate the workpieces to be processed to the processing machines for processing according to the scheduling plan; The prediction model includes: a first feature conversion layer, a second feature conversion layer, a graph update layer, a first extraction layer, a second extraction layer, a first classifier and a second classifier; The first feature conversion layer inputs the initial feature of the first node of the heterogeneous graph data and outputs the first updated feature; The second feature conversion layer inputs the initial feature of the second node of the heterogeneous graph data and outputs the second updated feature; The first updated feature and the second updated feature have the same size; The graph update layer inputs heterogeneous graph data and outputs an update matrix; A row vector of the update matrix corresponds to a third updated feature of the first node or the second node; The first extraction layer is used to extract the third updated feature corresponding to the first node and input it into the first classifier, and the classification space of the first classifier represents the processing quality of the workpiece to be processed; The second extraction layer is used to extract the third updated feature corresponding to the second node and input it into the second classifier, and the classification space of the second classifier represents the load of the processing machine; The calculation formula of the prediction model includes: The calculation formula of the first feature conversion layer is as follows: ; in represents the first updated feature output by the first feature conversion layer, A represents the initial feature of the first node, and They represent the weight parameters and bias parameters corresponding to the first feature conversion layer respectively, and sigmoid represents the sigmoid activation function; The calculation formula of the second feature conversion layer is as follows: ; in represents the second updated feature output by the second feature conversion layer, B represents the initial feature of the second node, and Respectively represent the weight parameters and bias parameters corresponding to the second feature conversion layer; The calculation formula of the graph update layer is as follows: ; ; ; ; Where 1≤i≤S, Matrix represents the update matrix output by the graph update layer, S represents the sum of the number of the first node and the second node, represents the i-th row vector of the update matrix and represents the third updated feature of the i-th first node or second node, and denote the first updated feature of the i-th and j-th first nodes or the second updated feature of the second nodes, respectively. represents the set of nodes that have edges connected to the i-th first node or second node. , , and Respectively represent the first coefficient, second coefficient, bias parameter and number of edges between the i-th node and the j-th node, , and represent the first weight parameter, the second weight parameter and the third weight parameter respectively, It indicates that the third updated features of S nodes are stacked, T indicates a transposition operation, and concat indicates a concatenation operation.

2. According to the collaborative manufacturing management platform based on the industrial Internet according to claim 1, It is characterized in that The unique identifier of the workpiece to be processed is represented by a positive integer code, and the processing completion time is represented by the processing completion time minus the current time. The number of processing steps for each workpiece to be processed is different.

3. According to the collaborative manufacturing management platform based on the industrial Internet according to claim 1, It is characterized in that The unique identifier of the processing machine is represented by a positive integer code, the energy consumption is represented by the power consumption per hour, the degree of wear is represented by the usage time divided by the maximum usage time, and the processing accuracy is represented by the yield rate of the processed workpiece per unit time.

4. According to the collaborative manufacturing management platform based on the industrial Internet according to claim 1, It is characterized in that The edges between nodes include: the edge between the first node and the second node indicates that the workpiece to be processed that establishes a mapping relationship with the first node is assigned to the processing machine that establishes a mapping relationship with the second node; the edge between the second nodes indicates that the processing machines that establish a mapping relationship with the second node are powered by the same power system.

5. According to the collaborative manufacturing management platform based on the industrial Internet according to claim 1, It is characterized in that The sample labels corresponding to the training samples used to train the prediction model are obtained through manual annotation, and the processing quality of the workpiece to be processed and the load of the processing machine are both customized discrete values.

6. According to the collaborative manufacturing management platform based on the industrial Internet according to claim 1, It is characterized in that The scheduling plan is generated by combining the red-billed blue magpie optimization algorithm with the prediction model, including the following steps: Step S301, randomly generate an initialization population that meets the constraint conditions, the number of individuals in the initialization population is U, where U is a custom parameter; Constraints include: the same processing step of each workpiece to be processed can only be assigned to one processing machine; the processing order of each workpiece to be processed must be strictly processed in sequence according to the process flow, and the next processing step of the same workpiece to be processed can only be started after the previous processing step is completed; The individual is represented by a matrix code, which includes N rows and M columns, where N represents the number of workpieces to be processed, M represents the maximum number of processing steps, and the element value of the matrix code is represented by a real number, the integer part of the real number represents the unique identifier of the processing machine, and the decimal part represents the processing priority, and the processing is performed in order from small to large; The calculation formula of the element value of the matrix encoding is as follows: ; Where 1≤m≤M, 1≤n≤N, ub and lb represent the upper and lower bounds respectively. Indicates a random number ranging from 0 to 1; Step S302, calculating the fitness values ​​of all individuals in the initialized population through an objective function, wherein the objective function is constructed based on a prediction model; The objective function is calculated as follows: ; Where Fit represents the fitness value of an individual, N represents the number of workpieces to be processed, and R represents the number of processing machines. represents the machining quality of the nth workpiece to be machined output by the prediction model, represents the load of the rth processing machine output by the prediction model, and Represent the first attention and the second attention respectively, and are custom parameters, and and The sum of is 1, and represent the first adjustment factor and the second adjustment factor respectively, and All are custom parameters; Step S303, generating a first random number ranging from 0 to 1. When the first random number is less than the balance coefficient, the initialized population is divided into multiple subgroups, each of which has Q individuals, and the matrix code of the individuals in each subgroup is updated through the food search strategy. Otherwise, the matrix code of the individuals in the initialized population is directly updated through the food search strategy, where Q is a custom parameter. Step S304, generating a second random number in the range of 0 to 1. When the second random number is less than the balance coefficient, the initialized population is divided into multiple subgroups, each of which has Q individuals, and the matrix code of the individuals in each subgroup is updated through the food search strategy, otherwise the matrix code of the individuals in the initialized population is directly updated through the prey attack strategy; Step S305, updating the matrix codes of individuals of the initialized population by storing food strategies; Step S306, the current number of iterations is increased by 1, and if it is determined that the iteration termination condition is met, the individual with the largest fitness value in the initialized population is used as the scheduling plan, otherwise it returns to step S302 to continue execution, where the iteration termination condition is that the current number of iterations reaches the maximum number of iterations or the maximum value of the fitness value within K consecutive iterations does not exceed the set threshold, where the maximum number of iterations, K and the set threshold are all custom parameters.

7. According to the collaborative manufacturing management platform based on the industrial Internet according to claim 6, It is characterized in that The calculation formula of the red-billed blue magpie optimization algorithm includes: The calculation formula for finding food strategy is as follows: ; Where P is assigned to Q or U, 1≤p≤Q or 1≤p≤U, and Respectively represent the matrix code of the i-th individual with the current iteration number t+1 and t, and They represent the matrix codes of the pth individual and a random individual at the current iteration number t, respectively. Indicates a random number ranging from 0 to 1; The calculation formula for attacking prey strategy includes: ; ; in It represents the matrix encoding of the individual with the largest fitness value at the current iteration number t, Max represents the maximum number of iterations, CF represents the flight factor, Represents random numbers that conform to the standard normal distribution; The calculation formula for the food storage strategy is as follows: ; in Represents the maximum fitness value of the i-th individual in the historical iteration number, Represents the fitness value of the i-th individual with the current iteration number t.

8. Collaborative manufacturing management method based on industrial Internet, It is characterized in that Executing the collaborative manufacturing management platform based on the industrial Internet as described in any one of claims 1 to 7 comprises the following steps: Step S401, collecting workpiece information of the workpiece to be processed; Step S402, collecting machine information of the processing machine; Step S403, constructing heterogeneous graph data according to workpiece information of the workpiece to be processed and machine information of the processing machine; Step S404, constructing and training a prediction model; Step S405, generating a scheduling plan by combining the red-billed blue magpie optimization algorithm with the prediction model.