Twin workshop production-logistics resource real-time cooperative scheduling method

Abstract

The invention discloses a twinning workshop production-logistics resource real-time collaborative scheduling method, which belongs to the technical field of workshop scheduling and comprises the following steps: S1, in a digital twinning environment, establishing a twinning workshop production-logistics resource real-time collaborative scheduling model; s2, when a task triggers a scheduling mechanism, obtaining production information of a physical workshop in real time, completing simulation and prediction in a virtual space, and optimizing a scheduling method according to a prediction result; s3, evaluating the service capability of the production-logistics resource group according to the optimized scheduling method, and selecting an optimal resource service group for the task; and S4, feeding back the task allocation scheme to the physical workshop, and waiting for the next task to trigger the scheduling mechanism. According to the twinning workshop production-logistics resource real-time collaborative scheduling method, production-logistics resources are distributed for tasks in real time, and production scheduling of workshops in a high-disturbance environment is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of workshop scheduling, and in particular to a real-time collaborative scheduling method for production and logistics resources of twin workshops. Background Art

[0002] Workshop scheduling is the core of intelligent manufacturing and is constrained by manufacturing tasks, resource types, resource status, and resource coordination modes. In actual production workshops, some uncertain dynamic interference events often occur, such as machine failures, random order arrivals, ambiguous workpiece processing time, ambiguous AGV transportation time, etc. These dynamic interferences will cause the production process to deviate from the plan and affect production execution efficiency. In a high-disturbance environment, the operating status and coordination relationship of production-logistics resources cannot be accurately predicted, and the scheduling system needs to optimize and regulate based on real-time information.

[0003] Digital Twin (DT) can simulate, predict, monitor and optimize physical objects through interactive feedback between virtual and real objects, data fusion analysis, iterative decision optimization and other means. It has become a core method for countries around the world to seize the commanding heights of a new round of global industrial competition. Combining digital twins with workshop scheduling can achieve real-time integration of the physical world and the information world through multi-factor mapping of multi-modal real-time data, obtain disturbances in the workshop processing process, and predict and correct the impact of disturbances on subsequent processes in real time, and then conduct precise hyper-temporal and spatial deduction in the twin space to meet the needs of a dynamic production environment. Summary of the invention

[0004] The purpose of the present invention is to provide a real-time collaborative scheduling method for production and logistics resources in a twin workshop. Through the information fusion and virtual-reality interaction characteristics of digital twin technology, a deep learning network is used to predict the completion time of workpieces in a production workshop under a high-disturbance environment, and to achieve iterative optimization of the algorithm, thereby obtaining the optimal service resource group for the task and completing the scheduling process.

[0005] To achieve the above object, the present invention provides a twin workshop production-logistics resource real-time collaborative scheduling method, comprising the following steps:

[0006] S1. In the digital twin environment, establish a real-time collaborative scheduling model for production and logistics resources in the twin workshop;

[0007] S2. When a task triggers the scheduling mechanism, it obtains scheduling-related information in the physical workshop production in real time, simulates multiple times in the virtual space, and uses the LSTM model to predict the completion time of the workpiece based on the real-time simulation data. The task urgency model is dynamically updated according to the prediction results;

[0008] S3. Evaluate the service capabilities of the production-logistics resource group in real time based on the updated urgency model, select the optimal resource service group for the task, and realize real-time collaborative scheduling of production-logistics resources;

[0009] S4: Feedback the task allocation plan to the physical workshop and wait for the next task to trigger the scheduling mechanism.

[0010] Preferably, step S1 specifically comprises:

[0011] S101. Establish a twin workshop production-logistics resource real-time collaborative scheduling problem model with completion time, total energy consumption and average delay as optimization objectives: including problem description and mathematical model;

[0012] S102, establishing a task urgency model based on real-time information;

[0013] S103, establishing a production-logistics resource service combination and a corresponding evaluation function;

[0014] The problem is described as follows:

[0015] There are d independent workpieces {n = 1, 2, ..., d}, each workpiece goes through s processes {k = 1, 2, ..., s} in sequence, there are i machines {m = 1, 2, ..., i} and j AGVs {v = 1, 2, ..., j} to choose from, and each job has two time attributes, namely arrival time and deadline;

[0016] The assumptions of the problem are as follows:

[0017] Workpieces arrive randomly, and each has a different deadline;

[0018] Each operation of the workpiece can be performed on a set of selectable machines;

[0019] The arrival time and deadline of the workpiece are not known until the workpiece arrives;

[0020] Each machine can only perform one task at a time;

[0021] Each AGV can only transport one task at a time;

[0022] Consider the AGV transportation time, and AGV resources are limited;

[0023] Once a process is processed on the machine, it should be completed before the subsequent process begins.

[0024] The mathematical model is defined as follows:

[0025] F = min{f1,f2,f3};

[0026] f1=max{C n};

[0027]

[0028] St

[0029] L n =max(0,C n -d n );

[0030] n={1,2,…,d};

[0031] k={1,2,…,s};

[0032] m={1,2,…,i};

[0033] v = {1, 2, ..., j};

[0034] Where F represents the multi-objective optimization function; f1 represents the completion time; f2 represents the total energy consumption, including machine energy consumption and AGV energy consumption; f3 represents the average delay of the operation; C n represents the actual completion time of workpiece n; represents the kth process of workpiece n Processed on machine m, otherwise Indicates the processing step of machine tool m The power at Indicates the processing step of machine tool m The duration of the hour; Indicates the process Transported by AGVv, otherwise Indicates AGVv transportation process The power at Indicates AGVv transportation process Total duration of I m represents the total standby time of machine tool m; e m represents the standby energy consumption of machine tool m; I v Indicates the total standby time of AGVv; e v Indicates the standby energy consumption of AGVv; L n represents the delay of job n; d n Represents the specified completion time of job n.

[0035] Preferably, step S102 is specifically as follows:

[0036] At time t, the kth process of workpiece n The remaining processing time is:

[0037]

[0038] Process The urgency expression is:

[0039]

[0040] Define the urgency of the task It represents the preference factor of task execution, indicating the current urgency of the task, which is related to the remaining workload and the specified deadline of the task artifact. Indicates the process The sum of the average processing time of all machine tools and the average transportation time of AGV in the subsequent process.

[0041] Preferably, step S103 is specifically as follows:

[0042] Create a selection of machine tools based on the type of task at hand and optional AGV sets Let G U A collection of service groups representing the service requirements of the task:

[0043]

[0044] Assume that g z It's a task Service group set G U A service combination of G U = {g z |z=1,2,…,U},g z =(m x ,v y ), U = X × Y;

[0045] Service Group z The total service time is:

[0046]

[0047] Service Group z The total energy consumption is:

[0048]

[0049] in, Indicates machine tool m x Processing The duration of the hour; Indicates AGVv y Take process time spent; Indicates v y Delivery process time spent; Indicates v y Take process Waiting time; Indicates mx Waiting for processing free time; Indicates v y Transportation process The power at Indicates m x Processing procedures The power at the time; e y Indicates v y Standby power consumption; e x Indicates m x Standby power consumption;

[0050] Establish the f0 function as a real-time evaluation function:

[0051] f0=ω e ×T nor +(1-ω e )×W nor ;

[0052]

[0053] Among them, T min represents the minimum service time of the optional combined resources; T max represents the maximum service time of the optional combined resources; W min represents the minimum energy consumption of the optional combined resources; W max It represents the maximum energy consumption of the optional combination of resources. The evaluation function balances time and energy consumption based on task characteristics, and selects the service combination with the smallest evaluation function f0 for the task, ultimately achieving the optimization of the three indicators of time, energy consumption and delay.

[0054] Preferably, step S2 specifically includes:

[0055] The real-time production information of the physical workshop includes production order information, processing task information and equipment operation information, including:

[0056] The production order information is whether there are new orders arriving at the current moment;

[0057] The processing task information is whether there is a task triggering scheduling mechanism at the current moment;

[0058] The equipment operation information is the real-time status of processing resources and logistics resources at the current moment. The real-time status model of processing resources is The real-time status attributes of processing equipment have eight characteristics: equipment number, service options, resource location, manufacturing energy consumption, standby energy consumption, manufacturing time, current service queue, and completion time of the current service queue:

[0059]

[0060] In the formula, Indicates the type of service that the machine tool can provide; L x Indicates the machine tool position; Indicates the processing energy consumption of the machine tool for the task; Indicates the service time of the machine tool for the task; sq x Indicates the current service queue of the machine tool; p tx Indicates the completion time of the current service queue;

[0061] The real-time status model of logistics resources is The real-time status attributes of logistics equipment have seven characteristics: equipment number, service options, resource location, handling energy consumption, standby energy consumption, current service queue, and completion time of the current service queue:

[0062]

[0063] In the formula, Indicates handling equipment v y Type of service provided; y Indicates the location of handling equipment; Indicates the energy consumption of the handling equipment for the task; e y Indicates idle energy consumption; sq y Indicates the current service queue of the handling equipment; p ty Indicates the completion time of the current service queue.

[0064] Preferably, the specific steps of multiple simulations in the virtual space are:

[0065] When a task reaches the trigger time, the scheduling mechanism is triggered. The virtual space obtains the workshop processing information and the processing status of all current tasks. Based on the real-time collaborative scheduling method of production and logistics resources in the twin workshop, the current task and subsequent tasks to be triggered are simulated multiple times to obtain the workpiece completion time series.

[0066] Preferably, predicting the workpiece completion time using an LSTM model based on real-time simulation data includes:

[0067] By obtaining the completion time series set of different workpieces, the completion time series of different workpieces are established. For example, the completion time series of workpiece n is:

[0068] T n =(X1,X2,X3,…,X k );

[0069] Among them, X k is the completion time of workpiece n simulated k times in the virtual workshop;

[0070] The collected completion time series set is subjected to standardization preprocessing. The formula for standardization preprocessing is:

[0071]

[0072] The collected data is preprocessed by standardization so that the mapping interval of the collected data values ​​is [-1, 1]. In the formula of standardization preprocessing, x is the original data; x mean 、x std are the mean and standard deviation of each group of original data respectively; x now The result data after standardization has a mean of 0 and a standard deviation of 1;

[0073] Construct the input-output sequence pair S through the preprocessed data set n , each input sequence contains the first 5 data points, corresponding to an output value, and the l-th input-output sequence pair of workpiece n Where X l ,X l+1 ,X l+2 ,X l+3 ,X l+4 , which is the input sequence that will be passed to the model to generate the predicted value Y for the next time step l , and calculate the predicted value X l With the true value X l+5 Losses between

[0074] The number of neurons in the input layer is n, the number of LSTM structural units in the hidden layer is 50, the number of neurons in the output layer is n, the maximum number of iterations is 200, and the loop is exited when the error is less than 0.005; the mean square error loss function is used to measure the difference between the predicted value and the true value, and the Adam optimizer is used in the back propagation process to adjust the weights of the model to minimize the loss function;

[0075] Use the trained LSTM model to output the prediction data, and the output result is the completion time yt(O n ).

[0076] Preferably, dynamically updating the task urgency model according to the prediction results is specifically as follows:

[0077] In the virtual space, according to the prediction results, the workpiece The urgency expression is:

[0078]

[0079] Dynamic updates to:

[0080]

[0081] in, It is a workpiece The sum of the average processing time of all machine tools and the average transportation time of AGV in the subsequent process; yt(O n ) is based on the LSTM model to predict the completion time of job n.

[0082] Preferably, the specific steps of selecting the optimal resource service group for the task in step S3 are:

[0083] To trigger the scheduling mechanism Building optional machine tool sets and optional AGV sets Calculate all its service groups g z The total service time and total energy consumption of production are calculated, and the service combination with the minimum evaluation function f0 is found as the task Providing services while serving this task Next process Add the scheduling trigger time and continue to wait for scheduling.

[0084] Preferably, in step S4, all tasks are gradually allocated to the most suitable machine tools and AGVs in real time according to the pairing results of the service groups and tasks, so as to realize real-time collaborative scheduling of production and logistics resources of twin workshops based on deep learning.

[0085] The present invention provides a real-time collaborative scheduling method for production and logistics resources in a twin workshop. Through the information fusion and virtual-reality interaction characteristics of digital twin technology, a deep learning network is used to predict the completion time of workpieces in a production workshop under a high-disturbance environment, and iterative optimization of the algorithm is achieved, thereby obtaining the optimal service resource group for the task and completing the scheduling process.

[0086] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 It is a flowchart of the steps of an embodiment of a method for real-time collaborative scheduling of production and logistics resources of a twin workshop of the present invention;

[0088] Figure 2 It is a scheduling trigger diagram of an embodiment of a twin workshop production-logistics resource real-time collaborative scheduling method of the present invention;

[0089] Figure 3 This is a network architecture diagram of an embodiment of a twin workshop production-logistics resource real-time collaborative scheduling method of the present invention;

[0090] Figure 4 It is a Gantt chart of an embodiment of a twin workshop production-logistics resource real-time collaborative scheduling method of the present invention. DETAILED DESCRIPTION

[0091] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0092] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0093] Embodiment 1

[0094] like Figure 1-3 As shown, the present invention provides a twin workshop production-logistics resource real-time collaborative scheduling method, comprising the following steps:

[0095] S1. In the digital twin environment, establish a real-time collaborative scheduling model for production and logistics resources in the twin workshop;

[0096] S2. When a task triggers the scheduling mechanism, it obtains scheduling-related information in the physical workshop production in real time, simulates multiple times in the virtual space, and uses the LSTM model to predict the completion time of the workpiece based on the real-time simulation data. The task urgency model is dynamically updated according to the prediction results;

[0097] S3. Evaluate the service capabilities of the production-logistics resource group in real time based on the updated urgency model, select the optimal resource service group for the task, and realize real-time collaborative scheduling of production-logistics resources;

[0098] S4: Feedback the task allocation plan to the physical workshop and wait for the next task to trigger the scheduling mechanism.

[0099] S1.1. The problem description of the real-time collaborative scheduling method of production and logistics resources in twin workshops is:

[0100] There are d independent workpieces {n = 1, 2, ..., d}, each workpiece goes through s processes {k = 1, 2, ..., s} in sequence, there are i machines {m = 1, 2, ..., i} and j AGVs {v = 1, 2, ..., j} to choose from, and each job has two time attributes, namely arrival time and deadline;

[0101] The assumptions of the problem are as follows:

[0102] Workpieces arrive randomly, and each has a different deadline;

[0103] Each operation of the workpiece can be performed on a set of selectable machines;

[0104] The arrival time and deadline of the workpiece are not known until the workpiece arrives;

[0105] Each machine can only perform one task at a time;

[0106] Each AGV can only transport one task at a time;

[0107] Consider the AGV transportation time, and AGV resources are limited;

[0108] Once a process is processed on the machine, it should be completed before the subsequent process begins. S1.2 The mathematical model of the real-time collaborative scheduling method of production and logistics resources in twin workshops is:

[0109] F = min{f1,f2,f3};

[0110] f1=max{C n};

[0111]

[0112] St

[0113] L n =max(0,C n -d n );

[0114] n={1,2,…,d};

[0115] k={1,2,…,s};

[0116] m={1,2,…,i};

[0117] v = {1, 2, ..., j};

[0118] Where F represents the multi-objective optimization function; f1 represents the completion time; f2 represents the total energy consumption, including machine energy consumption and AGV energy consumption; f3 represents the average delay of the operation; C n represents the actual completion time of workpiece n; represents the kth process of workpiece n Processed on machine m, otherwise Indicates the processing step of machine tool m The power at Indicates the processing step of machine tool m The duration of the hour; Indicates the process Transported by AGVv, otherwise Indicates AGVv transportation process The power at Indicates AGVv transportation process Total duration of I m represents the total standby time of machine tool m; e m represents the standby energy consumption of machine tool m; I v Indicates the total standby time of AGVv; e v Indicates the standby energy consumption of AGVv; L nrepresents the delay of job n; d n Represents the specified completion time of job n.

[0119] S1.3. Establish a task urgency model based on real-time information:

[0120] At time t, the kth process of workpiece n The remaining processing time is:

[0121]

[0122] Process The urgency expression is:

[0123]

[0124] Define the urgency of the task is the preference factor for task execution, indicating the current urgency of the task, which is related to the remaining workload and the specified deadline of the task artifact. It is a process The sum of the average processing time of all machine tools and the average transportation time of AGV in the subsequent process.

[0125] S1.4. Establishing production-logistics resource service combination and corresponding evaluation function:

[0126] Create a selection of machine tools based on the type of task at hand and optional AGV sets Let G U A collection of service groups representing the service requirements of the task:

[0127]

[0128] Assume that g z It's a task Service group set G U A service combination of G U = {g z |z=1,2,…,U},g z =(m x ,v y ), U = X × Y;

[0129] Service Group z The total service time is:

[0130]

[0131] Service Group z The total energy consumption is:

[0132]

[0133] in, Indicates machine tool m x Processing The duration of the hour; Indicates AGVv y Take process time spent; Indicates v y Delivery process time spent; Indicates v y Take process Waiting time; Indicates m x Waiting for processing free time; Indicates v y Transportation process The power at Indicates m x Processing procedures The power at the time; e y Indicates v y Standby power consumption; e x Indicates m x Standby power consumption;

[0134] Establish the f0 function as a real-time evaluation function:

[0135] f0=ω e ×T nor +(1-ω e )×W nor ;

[0136]

[0137] Among them, T min represents the minimum service time of the optional combined resources; T max represents the maximum service time of the optional combined resources; W min represents the minimum energy consumption of the optional combined resources; W max It represents the maximum energy consumption of the optional combination of resources. The evaluation function is to balance time and energy consumption based on task characteristics. The service combination with the smallest evaluation function f0 is selected for the task, and finally the three indicators of time, energy consumption and delay are optimized.

[0138] S2.1. Obtain production order information, processing task information, and equipment operation information, and establish a real-time status model for processing equipment and handling equipment, including:

[0139] The production order information is whether there are new orders arriving at the current moment;

[0140] The processing task information is whether there is a task triggering scheduling mechanism at the current moment;

[0141] The equipment operation information is the real-time status of processing resources and logistics resources at the current moment. The real-time status model of processing resources is The real-time status attributes of processing equipment have eight characteristics: equipment number, service options, resource location, manufacturing energy consumption, standby energy consumption, manufacturing time, current service queue, and completion time of the current service queue:

[0142]

[0143] In the formula, Indicates the type of service that the machine tool can provide; L x Indicates the machine tool position; Indicates the processing energy consumption of the machine tool for the task; Indicates the service time of the machine tool for the task; sq x Indicates the current service queue of the machine tool; p tx Indicates the completion time of the current service queue;

[0144] The real-time status model of logistics resources is The real-time status attributes of logistics equipment have seven characteristics: equipment number, service options, resource location, handling energy consumption, standby energy consumption, current service queue, and completion time of the current service queue:

[0145]

[0146] In the formula, Indicates handling equipment v y Type of service provided; y Indicates the location of handling equipment; Indicates the energy consumption of the handling equipment for the task; e y Indicates idle energy consumption; sq y Indicates the current service queue of the handling equipment; p ty Indicates the completion time of the current service queue.

[0147] S2.2. When a task reaches the trigger time, the scheduling mechanism is triggered. The virtual space obtains the workshop processing information and the processing status of all current tasks. Based on the real-time collaborative scheduling method of production and logistics resources in the twin workshop, the current task and subsequent tasks to be triggered are simulated multiple times to obtain the workpiece completion time series.

[0148] S2.3. Obtain the completion time series set of different workpieces according to the simulation data, and establish the completion time series of different workpieces. For example, the completion time series of workpiece n is:

[0149] T n =(X1,X2,X3,…,X k );

[0150] Among them, X k is the completion time of workpiece n simulated k times in the virtual workshop.

[0151] S2.4. The collected completion time series set is subjected to standardization preprocessing. The formula for standardization preprocessing is:

[0152]

[0153] The collected data is preprocessed by standardization so that the mapping interval of the collected data values ​​is [-1, 1]. In the formula of standardization preprocessing, x is the original data; x mean 、x std are the mean and standard deviation of each group of original data respectively; x now The result data is standardized, with a mean of 0 and a standard deviation of 1.

[0154] S2.5. Construct input and output sequence pairs S from the preprocessed data set n , each input sequence contains the first 5 data points, corresponding to an output value, and the l-th input-output sequence pair of workpiece n Where X l ,X l+1 ,X l+2 ,X l+3 ,X l+4 , which is the input sequence that will be passed to the model to generate the predicted value Y for the next time step l , and calculate the predicted value Y l With the true value X l+5 Losses between

[0155] The LSTM model is set as follows: the number of neurons in the input layer is the number of workpieces n, the number of LSTM structural units in the hidden layer is 50, the number of neurons in the output layer is the number of workpieces n, the maximum number of iterations is 200, and the loop is exited when the error is less than 0.005. The mean square error loss function is used to measure the difference between the predicted value and the true value. The Adam optimizer is used in the back propagation process to adjust the weights of the model to minimize the loss function.

[0156] S2.6, use the input and output sequences to train the LSTM model, adjust the weight parameters through back propagation, reduce the error between the predicted value and the true value, and finally output the completion time yt(O n ).

[0157] S3.1. According to the prediction results, the workpiece The urgency expression is:

[0158]

[0159] Dynamic updates to:

[0160]

[0161] in, Representation of workpiece The sum of the average processing time of all machine tools and the average transportation time of AGV in the subsequent process; yt(O n ) indicates the completion time of job n is predicted based on the LSTM model.

[0162] S3.2. Tasks that trigger the scheduling mechanism Building optional machine tool sets and optional AGV sets Calculate all its service groups g z The total service time and total energy consumption of production are calculated, and the service combination with the minimum evaluation function f0 is found as the task Providing services while serving this task Next process Add the scheduling trigger time and continue to wait for scheduling.

[0163] S4: Feedback the task allocation plan to the physical workshop, while continuing to collect current workshop information and waiting for the next task to trigger the scheduling mechanism. Finally, according to the matching results of the service group and the task, all tasks are gradually allocated to the most suitable machine tools and AGVs in real time, realizing the real-time collaborative scheduling of production and logistics resources in the twin workshop based on deep learning.

[0164] In order to fully verify the performance of the proposed method, a flexible workshop production scenario is constructed by taking the processing of ten workpieces in the workshop as an example. The specific parameters are shown in the following table:

[0165] Table 1 Experimental conditions

[0166] feature describe scale 6 machine tools, 1 automated warehouse and 5 AGVs Order Quantity 10 Order random arrival quantity 6 Order process quantity 5 Order deadline [400,400,400,450,450,500,550,580,650,700]

[0167] This experiment was conducted on a personal PC using Python 3.9 and an Intel(R) Core(TM) i5-12500 CPU @ 3.0 GHz. The scheduling method was run 200 times according to the experimental conditions in Table 1, and the experimental results are shown in Table 2. In large-scale flexible workshop production, the method proposed in the present invention can respond to changes in the production site in real time, provide timely and accurate scheduling solutions for the physical workshop, and meet the technical characteristics of digital twins.

[0168] Table 2 Experimental results

[0169] feature describe Number of runs 200 Average total running time 25.06s Average order dispatch time 2.506s Average process scheduling time 0.501s

[0170] The Gantt chart of the method proposed in the present invention is as follows: Figure 4As shown, the order arrival time is [0,0,0,0,23,45,80,110,135,198]. The machine tool in the figure has two states, "ST" means it is currently in standby state, and the second box "o41" of machine tool M5 indicates the process Processing on machine tool M5. The AGV in the figure has three states. "ST" means it is currently in standby state. The fourth and fifth boxes of AGV1 mean that AGV1 will go to pick up the process first. Then the process Sent to machine tool M3 to wait for processing.

[0171] The real-time scheduling method is constructed based on the real workshop. The real-time scheduling method takes into account the dynamic disturbances of the real workshop (these disturbances include random arrival of orders, fuzzy workpiece processing time, and fuzzy AGV transportation time). These fuzzy factors will affect the production scheduling process.

[0172] Therefore, the present invention adopts the above-mentioned twin workshop production-logistics resource real-time collaborative scheduling method. Through the information fusion and virtual-reality interaction characteristics of digital twin technology, a deep learning network is used to predict the completion time of workpieces in the production workshop under a high disturbance environment, and to achieve iterative optimization of the algorithm, thereby obtaining the optimal service resource group for the task and completing the scheduling process.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A twin workshop production-logistics resource real-time collaborative scheduling method, characterized in that: The following steps are involved: S1. In the digital twin environment, establish a real-time collaborative scheduling model for production and logistics resources in the twin workshop; S2. When a task triggers the scheduling mechanism, it obtains scheduling-related information in the physical workshop production in real time, simulates multiple times in the virtual space, and uses the LSTM model to predict the completion time of the workpiece based on the real-time simulation data. The task urgency model is dynamically updated according to the prediction results; S3. Evaluate the service capabilities of the production-logistics resource group in real time based on the updated urgency model, select the optimal resource service group for the task, and realize real-time collaborative scheduling of production-logistics resources; S4: Feedback the task allocation plan to the physical workshop and wait for the next task to trigger the scheduling mechanism.

2. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 1, characterized in that: Step S1 is specifically as follows: S101. Establish a twin workshop production-logistics resource real-time collaborative scheduling problem model with completion time, total energy consumption and average delay as optimization objectives: including problem description and mathematical model; S102, establishing a task urgency model based on real-time information; S103, establishing a production-logistics resource service combination and a corresponding evaluation function; The problem is described as follows: There are d independent workpieces {n = 1, 2, ..., d}, each workpiece goes through s processes {k = 1, 2, ..., s} in sequence, there are i machines {m = 1, 2, ..., i} and j AGVs {v = 1, 2, ..., j} to choose from, and each job has two time attributes, namely arrival time and deadline; The mathematical model is defined as follows: F = min{f1,f2,f3}; f1=max{C n }; St L n =max(0,C n -d n ); n={1,2,…,d}; k={1,2,…,s}; m={1,2,…,i}; v = {1, 2, ..., j}; Where F represents the multi-objective optimization function; f1 represents the completion time; f2 represents the total energy consumption, including machine energy consumption and AGV energy consumption; f3 represents the average delay of the operation; C n represents the actual completion time of workpiece n; represents the kth process of workpiece n Processed on machine m, otherwise Indicates the processing step of machine tool m The power at Indicates the processing step of machine tool m The duration of the hour; Indicates the process Transported by AGVv, otherwise Indicates AGVv transportation process The power at Indicates the AGVv transportation process Total duration of I m represents the total standby time of machine tool m; e m represents the standby energy consumption of machine tool m; I v Indicates the total standby time of AGVv; e v Indicates the standby energy consumption of AGVv; L n represents the delay of job n; d n Represents the specified completion time of job n.

3. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 2, characterized in that: Step S102 is specifically as follows: At time t, the kth process of workpiece n The remaining processing time is: Process The urgency expression is: Define the urgency of the task is the preference factor for task execution, which is the current urgency of the task and is related to the remaining workload and the specified deadline of the task artifact. It is a process The sum of the average processing time of all machine tools and the average transportation time of AGV in the subsequent process.

4. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 2, characterized in that: Step S103 is specifically as follows: Establish a set of optional machines based on the current task type and optional AGV set Let G U A collection of service groups representing the service requirements of the task: Assume that g z It's a task Service group set G U A service combination of G U = {g z |z=1,2,…,U},g z =(m x ,v y ), U = X × Y; Service Group z The total service time is: Service Group z The total energy consumption is: in, Indicates machine tool m x Processing The duration of the hour; Indicates AGVv y Take process time spent; Indicates v y Delivery process time spent; Indicates v y Take process Waiting time; Indicates m x Waiting for processing free time; Indicates v y Transportation process The power at Indicates m x Processing procedures The power at the time; e y Indicates v y Standby power consumption; e x Indicates m x Standby power consumption; Establish the f0 function as a real-time evaluation function: f0=ω e ×T nor +(1-ω e )×W nor ; Among them, T min represents the minimum service time of the optional combined resources; T max represents the maximum service time of the optional combined resources; W min represents the minimum energy consumption of the optional combined resources; W max It represents the maximum energy consumption of the optional combination of resources. The evaluation function balances time and energy consumption based on task characteristics, and selects the service combination with the smallest evaluation function f0 for the task, ultimately achieving the optimization of the three indicators of time, energy consumption and delay.

5. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 1, characterized in that: Step S2 specifically includes: The real-time production information of the physical workshop includes production order information, processing task information and equipment operation information, including: The production order information is whether there are new orders arriving at the current moment; The processing task information is whether there is a task triggering scheduling mechanism at the current moment; The equipment operation information is the real-time status of processing resources and logistics resources at the current moment. The real-time status model of processing resources is The real-time status attributes of processing equipment have eight characteristics: equipment number, service options, resource location, manufacturing energy consumption, standby energy consumption, manufacturing time, current service queue, and completion time of the current service queue: In the formula, Indicates the type of service that the machine tool can provide; L x Indicates the machine tool position; Indicates the processing energy consumption of the machine tool for the task; Indicates the service time of the machine tool for the task; sq x Indicates the current service queue of the machine tool; p tx Indicates the completion time of the current service queue; The real-time status model of logistics resources is The real-time status attributes of logistics equipment have seven characteristics: equipment number, service options, resource location, handling energy consumption, standby energy consumption, current service queue, and completion time of the current service queue: In the formula, Indicates handling equipment v y Type of service provided; y Indicates the location of handling equipment; Indicates the energy consumption of the handling equipment for the task; e y Indicates idle energy consumption; sq y Indicates the current service queue of the handling equipment; p ty Indicates the completion time of the current service queue.

6. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 5, characterized in that: The specific steps of multiple simulations in virtual space are as follows: When a task reaches the trigger time, the scheduling mechanism is triggered. The virtual space obtains the workshop processing information and the processing status of all current tasks. Based on the real-time collaborative scheduling method of production and logistics resources in the twin workshop, the current task and subsequent tasks to be triggered are simulated multiple times to obtain the workpiece completion time series.

7. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 6, characterized in that: The prediction of workpiece completion time using LSTM model based on real-time simulation data includes: By obtaining the completion time series set of different workpieces, the completion time series of different workpieces are established. The completion time series of workpiece n is: T n =(X1,X2,X3,…,X k ); Among them, X k is the completion time of workpiece n simulated k times in the virtual workshop; The collected completion time series set is subjected to standardization preprocessing. The formula for standardization preprocessing is: The collected data is preprocessed by standardization so that the mapping interval of the collected data values ​​is [-1, 1]. In the formula of standardization preprocessing, x is the original data; x mean 、x std are the mean and standard deviation of each group of original data respectively; x now The result data after standardization has a mean of 0 and a standard deviation of 1; Construct the input-output sequence pair S through the preprocessed data set n , each input sequence contains the first 5 data points, corresponding to an output value, and the l-th input-output sequence pair of artifact n Where X l ,X l+1 ,X l+2 ,X l+3 ,X l+4 , which is the input sequence that will be passed to the model to generate the predicted value Y for the next time step l , and calculate the predicted value Y l With the true value X l+5 Losses between The number of neurons in the input layer is n, the number of LSTM structural units in the hidden layer is 50, the number of neurons in the output layer is n, the maximum number of iterations is 200, and the loop is exited when the error is less than 0.005; the mean square error loss function is used to measure the difference between the predicted value and the true value, and the Adam optimizer is used in the back propagation process to adjust the weights of the model to minimize the loss function; Use the trained LSTM model to output the prediction data. The output result is the completion time yt(O n ).

8. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 7, characterized in that: The task urgency model is dynamically updated according to the prediction results as follows: In the virtual space, according to the prediction results, the workpiece The urgency expression is: Dynamic updates to: in, Representation of workpiece The sum of the average processing time of all machine tools and the average transportation time of AGV in the subsequent process; yt(O n ) indicates the completion time of job n is predicted based on the LSTM model.

9. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 1, characterized in that: The specific steps of selecting the optimal resource service group for the task in step S3 are as follows: To trigger the scheduling mechanism Building optional machine tool sets and optional AGV set Calculate all its service groups g z The total service time and total energy consumption of production are calculated, and the service combination with the minimum evaluation function f0 is found as the task Providing services while serving this task Next process Add the scheduling trigger time and continue to wait for scheduling.

10. A twin workshop production-logistics resource real-time collaborative scheduling method according to claim 1, characterized in that: In step S4, according to the pairing results of the service group and the task, all tasks are gradually allocated to the most suitable machine tools and AGVs in real time, realizing real-time collaborative scheduling of production and logistics resources in the twin workshop based on deep learning.

Images