Real-time scheduling method and system for workshop production based on AI industrial Internet of Things

Abstract

The present disclosure relates to the technical field of workshop scheduling, and proposes a real-time scheduling method and system for workshop production based on AI industrial Internet of Things, including: constructing an industrial Internet of Things scheduling model; obtaining tasks to be processed, constructing a workshop state diagram and a workpiece state set, and performing vectorization processing to convert them into vectors of the same dimension; solving the industrial Internet of Things scheduling model based on the vectorized workshop state and workpiece state within the current moment and a set time period in the past using a bidirectional long short-term memory network, performing inverse quantization processing on the solved workpiece state, and obtaining a probabilistic generation scheduling plan for the machine where each workpiece is located. The present disclosure adopts vectorization processing and inverse quantization processing after inference, considers the state of the workshop within a certain period of time before in the scheduling process, and uses a long short-term memory network to evaluate the scheduling of the workpiece at the next moment, so as to avoid the extra consumption that may be generated by focusing only on the current moment.

Description

Technical Field

[0001] The present disclosure relates to the technical field related to workshop scheduling, and more specifically, to a real-time workshop production scheduling method and system based on AI industrial Internet of Things. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] The workshop scheduling problem is one of the most difficult problems in the manufacturing system. The complex combination of processing machines and workpieces leads to a high computational complexity and is a very strong NP-hard problem.

[0004] Existing traditional methods divide the total production time into multiple time windows T, making decisions at the beginning of each time window T to generate a production plan. These decisions consider minimizing the maximum completion time and the maximum load on a single machine to generate the final scheduling plan. Other approaches use graph neural networks for scheduling. By constructing a graphical model for job shop scheduling, acquiring job shop workpiece processing data, and establishing a Markov decision model, the method utilizes actor networks for scheduling decisions, uses critic networks to evaluate the effectiveness of actor decisions, and uses graph neural networks to update node states, ultimately generating a workpiece scheduling plan.

[0005] The aforementioned methods, whether employing traditional methods, heuristics, or deep learning approaches for shop floor scheduling (deep learning methods primarily include graph neural networks and deep reinforcement learning networks), all calculate the shop floor state at the next moment based solely on the current state. Time-based scheduling can lead to long-distance scheduling, resulting in the consumption of additional resources. Furthermore, simple data pre-processing, such as integer encoding, one-hot encoding, and normalization, prior to employing deep learning algorithms can lose the dependencies between shop floor machines and workpiece operations, impacting subsequent calculations. Furthermore, reinforcement learning approaches suffer from high computational complexity and difficulty in achieving convergence in model training. Summary of the Invention

[0006] In order to solve the above problems, the present disclosure proposes a real-time scheduling method and system for workshop production based on AI industrial Internet of Things. Taking into account the status of the workshop within a certain period of time before, the method ensures the scheduling of workpieces in the vicinity during scheduling, avoids or reduces scheduling across long distances, and reduces additional resource consumption.

[0007] In order to achieve the above objectives, the present disclosure adopts the following technical solutions:

[0008] One or more embodiments provide a real-time scheduling method for workshop production based on AI industrial Internet of Things, including the following steps:

[0009] Build an industrial IoT scheduling model to minimize the total processing completion time of all workpieces;

[0010] Obtain the tasks to be processed, construct the workshop state diagram and workpiece state set, and perform vectorization processing to convert them into vectors of the same dimension;

[0011] Based on the quantized workshop status and workpiece status at the current moment and within the past set time period, the bidirectional long short-term memory network is used to solve the industrial Internet of Things scheduling model to obtain the workpiece status at the next moment and predict the workpiece status within the future set time period;

[0012] The obtained workpiece status is subjected to inverse quantization processing to obtain the probability of each workpiece being located on a machine. The machine corresponding to the maximum probability is the processing machine of the workpiece, and a scheduling plan is generated based on the correspondence between the workpiece and the machine.

[0013] One or more embodiments provide a real-time workshop production scheduling system based on AI industrial IoT, including:

[0014] Model building module: configured to build an industrial Internet of Things scheduling model so that the total processing completion time of all workpieces is minimized;

[0015] Pre-processing module: It is configured to obtain the tasks to be processed, build the workshop state diagram and workpiece state set, and perform vectorization processing to convert them into vectors of the same dimension;

[0016] Solution module: Based on the quantized workshop status and workpiece status at the current moment and within a set time period in the past, the bidirectional long short-term memory network is used to solve the industrial Internet of Things scheduling model to obtain the workpiece status at the next moment and predict the workpiece status within a set time period in the future;

[0017] Scheduling plan generation module: It is configured to perform inverse quantization processing on the obtained workpiece status to obtain the probability of the machine where each workpiece is located. The machine corresponding to the maximum probability is the processing machine of the workpiece, and a scheduling plan is generated based on the correspondence between the workpiece and the machine.

[0018] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of the above method are completed.

[0019] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the steps of the above method are completed.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] In this disclosure, in order to speed up the reasoning, the data is vectorized before reasoning and inversely quantized after reasoning. During the scheduling process, the status of the workshop in a certain period of time before is considered, and the long short-term memory network (BiLSTM) is used to evaluate the scheduling of the workpiece at the next moment to avoid the extra consumption that may be caused by focusing only on the current moment.

[0022] In the present disclosure, matrix vectorization pre-processing operations are used to retain the dependency between the workpiece state and the machine state, and vector data is transmitted to the model for prediction, thereby improving the accuracy of workshop scheduling and reducing the amount of calculation.

[0023] The advantages of the present disclosure and additional advantages will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which constitute a part of the present disclosure, are used to provide a further understanding of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure but do not constitute a limitation of the present disclosure.

[0025] Figure 1 is a flowchart of a real-time scheduling method for workshop production according to embodiment 1 of the present disclosure;

[0026] Figure 2 This is a workshop status diagram at an example moment of embodiment 1 of the present disclosure;

[0027] Figure 3 Schematic diagram of the vectorization process of embodiment 1 of the present disclosure;

[0028] Figure 4 This is a schematic diagram of calculating the workpiece state by BiLSTM in Example 1 of the present disclosure;

[0029] Figure 5 It is a schematic diagram of the inverse quantization processing process of embodiment 1 of the present disclosure. DETAILED DESCRIPTION

[0030] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0031] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present disclosure belongs.

[0032] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments in the present disclosure and the features in the embodiments can be combined with each other. The embodiments will be described in detail below with reference to the accompanying drawings.

[0033] Example 1

[0034] In the technical solutions disclosed in one or more embodiments, Figures 1 to 5 As shown in FIG, a real-time scheduling method for workshop production based on AI industrial Internet of Things may include the following steps:

[0035] Step 1: Build an industrial Internet of Things scheduling model to minimize the total processing completion time of all workpieces;

[0036] Step 2: Get the tasks to be processed and build the workshop status graph G t With artifact state set A t , and vectorize it into a vector representation of the same dimension and

[0037] Step 3: Based on the quantized workshop status and workpiece status at the current moment and within the past set time period, a bidirectional long short-term memory network is used to solve the industrial Internet of Things scheduling model to obtain the workpiece status at the next moment and predict the workpiece status within the future set time period;

[0038] Step 4: Perform inverse quantization processing on the obtained workpiece status to obtain the probability of each workpiece being located on the machine. The machine corresponding to the maximum probability is the processing machine of the workpiece, and a scheduling plan is generated based on the correspondence between the workpiece and the machine.

[0039] In this embodiment, to speed up training and inference, the data is vectorized before inference and inversely quantized after inference. During the scheduling process, the state of the workshop within a certain period of time is taken into consideration, and the long short-term memory network (BiLSTM) is used to evaluate the scheduling of the workpiece at the next moment to avoid the extra consumption that may be caused by focusing only on the current moment.

[0040] In this embodiment, matrix vectorization pre-processing operations are used to retain the dependency between the workpiece state and the machine state, and the vector data is transmitted to the model for prediction, thereby improving the accuracy of workshop scheduling and reducing the amount of calculation.

[0041] Optionally, in step 1, the constraints of the IIoT scheduling model constructed include:

[0042] (1) The processing task includes n workpieces to be processed;

[0043] (2) Each workpiece i has m processes, and the processing order of each process needs to follow the set order, where i∈{1, 2, ..., n};

[0044] (3) There are m processing machines, each of which is responsible for a separate process. The processing time of each machine is different, and the same machine can only process one workpiece at a time.

[0045] The goal of the industrial Internet of Things scheduling model is to determine the workpiece processing sequence of each machine and the specific start time of each process, so that the total processing completion time of all workpieces is minimized, which is the objective function of the industrial Internet of Things scheduling model. The objective function L of the entire optimization can be expressed as:

[0046] L=min(max Ci)i∈{1, 2,...,n}

[0047] Where: Ci is the total time taken for workpiece i from the beginning to the processing to the final completion, including processing time and waiting time.

[0048] Optionally, in step 2, construct the workshop state diagram G t Specifically: In the workshop state graph G = (V, E), V represents the vertex set, which is the state of each machine, mainly including information such as the current machine's processing status; E is the edge set, which is the dependency relationship between each process, used to connect the machines that may be used in the previous process with the machines that may be used in the next process. Figure 2 As shown in the figure, Vi,j represents the j-th machine that completes the i-th process.

[0049] A further technical solution is to improve the processing efficiency of BiLSTM by converting the workshop state graph G t Convert it into a vector representation and then perform reasoning. In this embodiment, the shop floor state diagram can be vectorized to include machine dependencies to solve the problem of one-hot encoding.

[0050] Optional, for the workshop status diagram G t Vectorization processing includes the following steps:

[0051] Step 21: Based on the workshop status diagram, the occupancy of the machines in the workshop and the dependency relationship between the machines are expressed using a two-dimensional matrix. To express it, we get the two-dimensional matrix of workshop status;

[0052] First, the occupancy of the machines in the workshop at time t and the dependency relationship between the machines are expressed as a two-dimensional matrix To express it, where m is the number of processing machines, each row and column represents a machine, and the row and column with the same index is a machine.

[0053] Two-dimensional matrix The value on the diagonal line represents the percentage of resource usage of machine i, and the value range is G t [i, i]∈[0, 1], where 0 represents no load and 1 represents full load; the values ​​on the off-diagonal lines represent the dependencies between machines agreed upon by the process, which only take the values ​​0 and 1 and are forward dependencies.

[0054] For example, in the workshop state example: suppose there are four processing machines M1, M2, M3, and M4 in the workshop, where M1 is the machine required for the first process, M2 and M3 are the machines required for the second process, and M4 is the machine required for the third process. The workpieces need to be produced in ascending order of the process numbers. Then the workshop state diagram matrix is:

[0055]

[0056] The parameters on the main diagonal have the following meanings: 0.1 means that the load of machine M1 is 10%, 0.4 means that the load of machine M2 is 40%, 0.2 means that the load of machine M3 is 20%, and 0.5 means that the load of machine M4 is 50%.

[0057] The meanings of the parameters on the off-diagonal lines are as follows: the first row 1 indicates that the subsequent process machines of machine M1 include M2 ​​and M3;

[0058] The second line 1 indicates that the successor machine of machine M2 is M4, and the third line 1 indicates that the successor machine of machine M3 is M4;

[0059] The absence of 1 in the fourth row indicates that machine M4 is the last processing machine in the process and there are no other subsequent machines.

[0060] Step 22: To further integrate and process the workshop status diagram matrix information, convert the workshop status two-dimensional matrix into a vector form as the input of BiLSTM.

[0061] Optionally, convert the two-dimensional matrix of the workshop status into a vector form using Figure 3 The processing module shown performs the conversion, including the following process:

[0062] 221. Perform a first convolution operation on the two-dimensional matrix of the workshop status using a first convolution kernel;

[0063] Optionally, in this embodiment, the size of the first convolution kernel is set to 3x3; after the data is input, it is implemented in the first 3x3 convolution layer;

[0064] 222. For the result of the first convolution operation, perform activation and batch normalization operations in sequence, and perform a second convolution operation according to the first convolution kernel;

[0065] 223. Perform a third convolution operation on the workshop status two-dimensional matrix using the second convolution kernel;

[0066] Optionally, in this embodiment, the size of the second convolution kernel is 1x1;

[0067] 224. After weighted fusion of the third convolution operation result and the second convolution operation result, perform a flattening operation to flatten them into a one-dimensional vector;

[0068] 225. Perform dimensionality transformation to transform the one-dimensional vector into a vector of the same dimension as the input of the bidirectional long short-term memory network.

[0069] Figure 3 The 3x3 convolution layer (Conv-3x3-d) at the middle input is a convolution operation with a convolution kernel size of 3x3 and an output channel dimension of d; the 1x1 convolution layer (Conv-1x1-d) on the far right is a convolution operation with a convolution kernel size of 1x1 and an output channel dimension of d; the activation and batch normalization layer uses ReLU as the activation function to enhance the nonlinear relationship between data and facilitate reasoning; BN batch normalization is used to adjust the data distribution and facilitate network convergence; the dimensional flattening layer (Flatten) is a flattening operation that flattens the two-dimensional matrix into a one-dimensional vector by row; the fully connected layer (FC-d) is a fully connected operation with an output dimension of d, which is used to transform the vector dimension to meet the BiLSTM input dimension requirements.

[0070] If the order Figure 3 The operation shown is a V operation, then in

[0071] In step 2, the artifact state set A t It includes the start time of workpiece processing, the current process of the workpiece and the processing machine.

[0072] Optionally, for the artifact state set A t Vectorization processing includes the following steps:

[0073] 2.1) Based on the workpiece state set, the workpiece state at a certain moment is represented by a two-dimensional matrix to obtain a two-dimensional matrix of workpiece state;

[0074] and workshop status Figure 1 Similarly, the workpiece state at a certain moment is also represented by a two-dimensional matrix Where n is the total number of workpieces and m is the total number of machines.

[0075] A tEach row of the matrix represents the processing progress of the current workpiece, and the value range is A t [i, j]∈[0, 1], that is, the percentage of workpiece i that has been processed on machine j at time t.

[0076] For example, assume that the workshop machine conditions are the same as those described in the previous workshop state example, and there are three workpieces C1, C2, and C3 being processed. The workpieces are represented by the rows of the matrix, with three rows representing three workpieces, each column representing one machine, and four columns representing four machines. Then:

[0077]

[0078] Among them, 0.1 in the first row means that workpiece C1 has been processed 10% in machine M1; 0.6 in the second row means that workpiece C2 has been processed 60% in machine M1; and 0.4 in the third row means that workpiece C3 has been processed 40% in machine M4.

[0079] In order to further process the concentrated information of the workpiece status, it is converted into a vector form as the input of BiLSTM. Figure 3 The processing module shown is vectorized. Figure 3 The operation shown is a V operation, then in

[0080] 2.2) Convert the workpiece state two-dimensional matrix into vector form, using Figure 3 The processing module shown performs the conversion, including the following process:

[0081] 2.21. Perform the first convolution operation on the workpiece state two-dimensional matrix using the first convolution kernel;

[0082] Optionally, in this embodiment, the size of the first convolution kernel is set to 3x3;

[0083] 2.22. Based on the result of the first convolution operation, perform activation and batch normalization operations in sequence, and perform a second convolution operation according to the first convolution kernel;

[0084] 2.23. Perform the third convolution operation on the workpiece state two-dimensional matrix using the second convolution kernel;

[0085] Optionally, in this embodiment, the size of the second convolution kernel is 1x1;

[0086] 2.24. After weighted fusion of the third convolution operation result and the second convolution operation result, flatten the result into a one-dimensional vector.

[0087] 2.25. Perform dimensionality transformation to transform the one-dimensional vector into a vector of the same dimension as the input of the bidirectional long short-term memory network.

[0088] In some embodiments, the scheduling process considers the state of the workshop within a certain period of time. Long short-term memory (BiLSTM) networks are used to evaluate the scheduling of the next workpiece, thus avoiding the extra overhead that may be incurred by focusing only on the current moment. Compared to reinforcement learning, BiLSTM has a lower computational load and is easier to converge.

[0089] A further technical solution is to use the vectorized workshop status and workpiece status as the input of BiLSTM to solve the industrial Internet of Things scheduling model.

[0090] Through the pre-processing stage of vectorization, the workshop state graph G t With artifact state set A t , converted to a vector representation of the same dimension and Because it is necessary to consider the workshop status over the past period of time to calculate the workpiece status at the next moment and predict the workpiece status within a certain period of time in the future, in order to avoid the extra consumption caused by long-distance scheduling, this solution uses BiLSTM to infer the workpiece status. B is abbreviated as the state of each layer in BiLSTM and the additional parameters. The workpiece status at the next moment i can be calculated as follows:

[0091]

[0092] Specifically, the current workpiece processing state set A t , the current workshop state diagram, and the workshop state diagram of the past i moments are fed into the BiLSTM model, and the workpiece state set A at the next moment in the future can be calculated. t+1 , and estimate the workpiece processing status set at the next i moments. By obtaining the workpiece processing status set at the next i moments, pre-scheduling can be performed to improve the scheduling plan.

[0093] like Figure 4 As shown, G t Represents the current workshop state diagram, in vector form:

[0094] G t-1 Represents the workshop status diagram at the previous moment, in vector form:

[0095] G t-i Represents the state diagram of the workshop at the previous i moments, in vector form:

[0096] A t Represents the processing status set of each workpiece at the current moment, which mainly includes the start time of workpiece processing, the current process of the workpiece and the processing machine. The vector form is

[0097] A t+1 is the workpiece processing state set at the next moment, in vector form: That is, the scheduling of each workpiece at the next moment is carried out based on this;

[0098] A t+i is the estimated workpiece processing status set at the next i moments Based on the estimated workpiece status at a future moment, a certain degree of pre-scheduling can be performed to improve scheduling efficiency.

[0099] After obtaining the BiLSTM results, the workpiece status set is in vectorized form, and the status information of each workpiece cannot be directly obtained. It needs to be restored to the original two-dimensional matrix representation to obtain the machine used to process the current workpiece at a certain moment in the future.

[0100] Optionally, the inverse quantization process can be performed using Figure 5 The processing module shown performs the conversion from a vector to a two-dimensional matrix, including the following steps:

[0101] Step 41: Perform a fully connected layer operation on the workpiece status output by the bidirectional long short-term memory network, and output the workpiece status output dimension as n×m, where n is the number of workpieces and m is the number of machines, to obtain the row matrix

[0102] Step 42: In the shape deformation layer, the data after the full connection layer operation is deformed, that is, the Reshape transformation operation is performed, and the data is transformed into an n-row and m-column matrix. Transformed into

[0103] Step 43: Calculate the probability of each element of the matrix along the rows of the matrix. Optionally, the probability calculation can use the Softmax function, that is:

[0104]

[0105] Where i is the row index of the matrix after Reshape transformation, j is the column index of the matrix, and x ij is the element in row i and column j in the matrix.

[0106] If the order Figure 5 The inverse quantization operation shown is V -1 Operation, then in,

[0107] The state probability matrix [PA t+1 , P.A. t+2 ,…,PA t+i], the machine with the maximum probability in each row is the target machine where the current workpiece is located at the target time. The scheduling plan for all workpieces can be obtained by traversing all future times.

[0108] In this embodiment, the proposed workshop production scheduling scheme can be described by the following formula:

[0109] [PA t+1 , P.A. t+2 ,…,PA t+i ]=V -1 (BiLSTM(V(G t-i ), V(G t-1 ), V(G t ), V(A t ), B))

[0110] Among them, PA t+i is the probability matrix of the workpiece state at the future i-th moment, G t is the workshop status diagram at the current moment, and At is the workpiece status set at the current moment.

[0111] The data processing during the training of the bidirectional long short-term memory network is the same as steps 1 to 4 above. The training set construction includes the workshop status diagram and workpiece status set of each time period. The status at the previous moment is used to predict the status at the next moment. Based on the comparison between the predicted results and the actual status, the network parameters are adjusted to obtain the trained bidirectional long short-term memory network.

[0112] Example 2

[0113] Based on Example 1, this embodiment provides a real-time workshop production scheduling system based on AI industrial Internet of Things, including:

[0114] Model building module: configured to build an industrial Internet of Things scheduling model so that the total processing completion time of all workpieces is minimized;

[0115] Pre-processing module: It is configured to obtain the tasks to be processed, build the workshop state diagram and workpiece state set, and perform vectorization processing to convert them into vectors of the same dimension;

[0116] Solution module: Based on the quantized workshop status and workpiece status at the current moment and within a set time period in the past, the bidirectional long short-term memory network is used to solve the industrial Internet of Things scheduling model to obtain the workpiece status at the next moment and predict the workpiece status within a set time period in the future;

[0117] Scheduling plan generation module: It is configured to perform inverse quantization processing on the obtained workpiece status to obtain the probability of the machine where each workpiece is located. The machine corresponding to the maximum probability is the processing machine of the workpiece, and a scheduling plan is generated based on the correspondence between the workpiece and the machine.

[0118] It should be noted here that the various modules in this embodiment correspond one-to-one to the various steps in Example 1, and the specific implementation processes are the same, which will not be repeated here.

[0119] Example 3

[0120] This embodiment provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are run by the processor, the steps described in the method of embodiment 1 are completed.

[0121] Example 4

[0122] This embodiment provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps described in the method of embodiment 1 are completed.

[0123] The foregoing description is merely a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present disclosure shall be included within the scope of protection of the present disclosure.

[0124] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Those skilled in the art should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A real-time scheduling method for workshop production based on AI industrial Internet of Things, characterized by: The steps include: Build an industrial IoT scheduling model to minimize the total processing completion time of all workpieces; Obtain the tasks to be processed, construct a workshop state graph and a workpiece state set, and perform vectorization processing to convert them into vectors of the same dimension; the workshop state graph is constructed as follows: the workshop state graph includes a vertex set and an edge set consisting of edges between vertices. The vertices represent the processing status of each machine; the edge set is the dependency relationship between each process, which is used to connect the machines that may be used in the previous process with the machines that may be used in the subsequent process; The shop state diagram is vectorized, comprising the following steps: based on the shop state diagram, the occupancy status of machines in the shop and the dependency relationship between the machines are represented by a two-dimensional matrix to obtain a two-dimensional shop state matrix; and the two-dimensional shop state matrix is ​​converted into a vector form; Vectorizing the workpiece state set includes the following steps: representing the workpiece state at a certain moment by a two-dimensional matrix according to the workpiece state set to obtain a two-dimensional matrix of the workpiece state; converting the two-dimensional matrix of the workpiece state into a vector form; Based on the quantized workshop status and workpiece status at the current moment and within the past set time period, the bidirectional long short-term memory network is used to solve the industrial Internet of Things scheduling model to obtain the workpiece status at the next moment and predict the workpiece status within the future set time period; The obtained workpiece status is subjected to inverse quantization processing to obtain the probability of each workpiece being located on a machine. The machine corresponding to the maximum probability is the processing machine of the workpiece, and a scheduling plan is generated based on the correspondence between the workpiece and the machine.

2. The real-time scheduling method for workshop production based on AI industrial Internet of Things according to claim 1 is characterized in that: The constraints of the IIoT scheduling model include: The task to be processed includes a plurality of workpieces to be processed; Each workpiece has multiple processes, and the processing order of each process needs to follow a set sequence; The workshop includes multiple processing machines, each of which is responsible for a separate process. The processing time of each machine is different, and the same machine processes a workpiece at the same time.

3. The real-time scheduling method for workshop production based on AI industrial Internet of Things according to claim 1 is characterized in that: Convert the two-dimensional matrix of workshop status into vector form, including: Perform a first convolution operation on the two-dimensional matrix of the workshop status using the first convolution kernel; For the result of the first convolution operation, activation, batch normalization operations are performed in sequence, and a second convolution operation is performed according to the first convolution kernel; The workshop status two-dimensional matrix is ​​subjected to a third convolution operation using the second convolution kernel; After weighted fusion of the third convolution operation result and the second convolution operation result, a flattening operation is performed to flatten them into a one-dimensional vector; Perform dimensionality transformation to transform the one-dimensional vector into a vector of the same dimension as the bidirectional long short-term memory network input.

4. The method for real-time workshop production scheduling based on AI industrial Internet of Things according to claim 1 is characterized in that: The workpiece status set includes the start time of workpiece processing, the current process of the workpiece, and the processing machine.

5. The real-time scheduling method for workshop production based on AI industrial Internet of Things according to claim 4 is characterized in that: Convert the workpiece state two-dimensional matrix into vector form, including: Perform a first convolution operation on the workpiece state two-dimensional matrix using a first convolution kernel; For the result of the first convolution operation, activation and batch normalization operations are performed in sequence, and a second convolution operation is performed according to the first convolution kernel; The workpiece state two-dimensional matrix is ​​subjected to a third convolution operation using the second convolution kernel; After weighted fusion of the third convolution operation result and the second convolution operation result, a flattening operation is performed to flatten them into a one-dimensional vector; Perform dimensionality transformation to transform the one-dimensional vector into a vector of the same dimension as the bidirectional long short-term memory network input.

6. The real-time scheduling method for workshop production based on AI industrial Internet of Things according to claim 1 is characterized in that: The inverse quantization process includes: Perform a fully connected layer operation on the workpiece status output by the bidirectional long short-term memory network, and output the workpiece status output dimension to n×m, where n is the number of workpieces and m is the number of machines, to obtain a row matrix; Perform a deformation operation on the matrix after the fully connected layer operation and transform it into an n-row and m-column matrix; Compute the probability of belonging to each matrix element along the matrix rows.

7. The real-time scheduling system for workshop production based on AI industrial Internet of Things is characterized by: include: Model building module: configured to build an industrial Internet of Things scheduling model so that the total processing completion time of all workpieces is minimized; Pre-processing module: This module is configured to obtain tasks to be processed, construct a workshop state graph and a workpiece state set, and perform vectorization processing to convert them into vectors of the same dimension. The workshop state graph is constructed as follows: the workshop state graph includes a vertex set and an edge set consisting of edges between vertices. The vertices represent the processing status of each machine. The edge set is the dependency relationship between each process, which is used to connect the machines that may be used in the previous process with the machines that may be used in the subsequent process. The shop state diagram is vectorized, comprising the following steps: based on the shop state diagram, the occupancy status of machines in the shop and the dependency relationship between the machines are represented by a two-dimensional matrix to obtain a two-dimensional shop state matrix; and the two-dimensional shop state matrix is ​​converted into a vector form; Vectorizing the workpiece state set includes the following steps: representing the workpiece state at a certain moment by a two-dimensional matrix according to the workpiece state set to obtain a two-dimensional matrix of the workpiece state; converting the two-dimensional matrix of the workpiece state into a vector form; Solution module: Based on the quantized workshop status and workpiece status at the current moment and within a set time period in the past, the bidirectional long short-term memory network is used to solve the industrial Internet of Things scheduling model to obtain the workpiece status at the next moment and predict the workpiece status within a set time period in the future; Scheduling plan generation module: It is configured to perform inverse quantization processing on the obtained workpiece status to obtain the probability of the machine where each workpiece is located. The machine corresponding to the maximum probability is the processing machine of the workpiece, and a scheduling plan is generated based on the correspondence between the workpiece and the machine.

8. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps of any one of the methods of claims 1 to 6 are completed.

9. A computer-readable storage medium, characterized in that Used to store computer instructions, which, when executed by a processor, complete the steps of any one of the methods of claims 1 to 6.

Images