Data-driven intelligent distribution method for transformer production materials

Abstract

The invention discloses a data-driven intelligent distribution method for transformer production materials, and relates to the technical field of workshop material distribution. The intelligent distribution method for production materials divides the distribution process into two stages: production material preparation and production material distribution. Firstly, process data is continuously collected and classified and stored in the workshop production process. Based on the workshop production history data and real-time monitoring data, KNN (K-Nearest Neighbor) regression prediction is used in the material preparation stage to realize intelligent preparation of workshop production materials; in the production material distribution stage, A* (A-Star) algorithm with improved time window is used to realize autonomous decision-making of AGV (Automated Guided Vehicle) material distribution path with good obstacle avoidance ability, smooth and efficient path. The invention can meet the actual use needs of transformer production material distribution, improve the distribution efficiency of transformer production and manufacturing process, and has certain reference significance and application value.

Description

Technical Field

[0001] The present invention relates to the technical field of workshop material distribution, and in particular to a data-driven intelligent distribution method for transformer production materials. Background Art

[0002] In recent years, with the practical application of digital factories in the transformer manufacturing industry, intelligent distribution of transformer production materials has become an inevitable development trend. With the introduction of the lean production concept, the operating time of each device in the factory and the actions of each worker are precisely required. In the distribution process of production materials, the control of time is equally important. According to the theory of lean modular supply JIS (just in sequence) and just-in-time production method JIT (just in time), workshop logistics needs to supply materials in a timely and appropriate amount according to the needs of each production line; at the same time, material distribution should adjust the overall distribution rhythm according to the production rhythm; ensure that the material distribution of each production line is replenished in time according to the expected type and quantity of material consumption, while meeting the production and processing rhythm requirements of the production line, so that the distribution resource consumption is minimized and the material inventory is minimized.

[0003] Production practice has proved that high-quality and efficient enterprise production cannot be separated from the support of the production logistics system. Production material distribution is a key link in the construction of digital factories. However, regarding the problem of workshop material distribution, traditional scheduling methods often ignore the historical operation data accumulated by the manufacturing system, resulting in the production laws hidden in the manufacturing data not being deeply mined. The application of new generation information and communication technology can quickly control, perceive and collect the supply and demand relationship of all resources in the production system, significantly improve the scheduling level of the production execution system, and realize the efficient operation of the entire manufacturing execution process.

[0004] In summary, it is of great significance to realize data-driven intelligent distribution of production materials based on mining and analysis based on big data and artificial intelligence algorithms on the basis of historical operation data of transformer manufacturing plants. At present, in the actual material distribution process, the production preparation problem before distribution is rarely considered. KNN (K-Nearest Neighbor) is one of the machine learning algorithms, which can be used for classification and regression. It is a supervised learning algorithm. In the classification decision, this method only determines the category of the sample to be classified based on the category of the nearest one or several samples. Due to the advantages of KNN (K-nearest neighbor) regression, such as high accuracy and less training time, the common production material preparation problem in the distribution process of transformer production materials is analyzed by using KNN regression to analyze the relevant factors in the production time series data, so as to accurately predict the production material preparation time. At the same time, in the aspect of AGV material distribution path planning, A* (A-Star) algorithm is the most effective direct search method for solving the shortest path in a static road network, and it is also an effective algorithm for solving many search problems. The closer the distance estimate in the algorithm is to the actual value, the faster the final search speed will be. During a critical period in the time cycle, if an influence or action is exerted on a specific thing, the chance of success will increase. The A* (A-Star) algorithm with time window constraints has good obstacle avoidance ability, smooth and efficient path, and autonomous decision-making ability for material distribution paths, which can meet the actual use needs of transformer production material distribution. Summary of the invention

[0005] The purpose of the present invention is to provide a data-driven intelligent distribution method for transformer production materials. The distribution method divides the distribution process into two stages: production material preparation and production material distribution. By collecting and classifying the historical production data of the workshop, KNN regression prediction is used in the preparation stage to realize the intelligent preparation of workshop production materials; in the production material distribution stage, the A* (A-Star) algorithm with an improved time window is used to realize the autonomous decision-making of AGV material distribution path with good obstacle avoidance ability, smooth path and efficient operation.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A data-driven intelligent distribution method for transformer production materials, the distribution method steps are as follows:

[0008] Step 1: First, during the transformer production process, AGV, stereoscopic warehouse, manufacturing station and charging station constitute the workshop intelligent IIoT data acquisition sensor platform, which collects workshop production data information in real time, pre-processes the data, and extracts key fields related to production material distribution;

[0009] Step 2: Further call the KNN regression model to predict the delivery time of the next period of time for the materials to be delivered;

[0010] Step 3: Further calculate the material preparation and delivery time based on the predicted delivery time and issue the material preparation instruction. In the three-dimensional warehouse, the production materials are transferred to the material preparation area for AGV loading and transportation;

[0011] Step 4: After material preparation is completed, the workshop dispatcher issues a production material delivery instruction to the AGV, and the AGV starts to perform the delivery task;

[0012] Step 5: Grid the production workshop. By marking the obstacles in the production area of ​​the grid workshop and the corresponding labels, the algorithm can use the labels to quickly obtain the grid information when searching for the path;

[0013] Step 6: In the grid information network, the improved time window A* (A-Star) algorithm selects the location with the lowest current "cost" according to the extended node for the next search until the search end, thereby planning the path with the lowest cost;

[0014] Step 7: AGV implements the material delivery task according to the path planned by the improved time window A* (A-Star) algorithm;

[0015] Step 8: During the AGV delivery process, the improved time window A* (A-Star) algorithm performs real-time collision detection based on the AGV's location and the real-time status of the workshop, discards nodes that collide with each other, retains valid nodes, calculates its improved time window and adds it to the set, and backtracks to obtain a collision-free path after finding the target point, and updates the AGV's walking path in real time;

[0016] Step 9: After the AGV delivers the materials to the designated location, the improved time window A* (A-Star) algorithm plans the return path in real time and issues instructions.

[0017] As a further solution of the present invention: the parameters that need to be considered in the KNN prediction of the delivery time in step 2 are described as the production work order start time (planned), the production work order start time (actual), the time required to deliver production materials (planned), and the historical production material delivery time (actual), which are respectively recorded as All the hours, minutes, and seconds are converted into minutes, and the production speed of the product before the predicted time is recorded as The consumption rate of production materials before the prediction time is recorded as The production material number and quantity required by the production unit before the forecast time are recorded as Based on the above data description, the data required for discrete manufacturing workshop material prediction can be obtained: D = {{Xi, Yi}}, Yi=yi,i=1,2,…,n, Yi is the target variable, Xi is the associated variable, and n is the data length.

[0018] As a further solution of the present invention: the workflow steps of KNN prediction of delivery time in step 2 are:

[0019] S1: Input the production order start time (planned), production order start time (actual), time required to deliver production materials (planned), and historical production material delivery time (actual) related parameters into the KNN prediction model;

[0020] S2: Determine the numerical value between all the set points in the known data set and the current set point, and arrange them in descending order according to the numerical value;

[0021] S3: Use N points with similar numerical values ​​to the target set point and determine the mean coordinates of these N points;

[0022] S4: Return the coordinate output of the current node as the prediction of the material preparation time for the next period;

[0023] S5: Combine the material demand information related to MES and WMS to clarify the types and quantities of materials required.

[0024] As a further solution of the present invention: the A* (A-Star) algorithm in step six is ​​based on the A* (A-Star) algorithm, and the heuristic and time window of the A* (A-Star) algorithm are improved. The two methods are combined by using the heuristic time factor to perform spatiotemporal search, which can avoid quadratic programming while searching for the shortest time; at the same time, an improved focus search operator is used to improve the algorithm search efficiency while reducing the impact of negative feedback on the search efficiency.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The present invention can meet the actual use requirements of transformer production material distribution, improve the distribution efficiency of the transformer production process, and has certain reference significance and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The overall architecture diagram of the data-driven intelligent distribution method for transformer production materials.

[0028] Figure 2 Diagram of the intelligent material preparation model for transformer production.

[0029] Figure 3 The overall flow chart of data-driven intelligent distribution of transformer production materials.

[0030] Figure 4 This is a schematic diagram of the prediction results of intelligent material preparation for transformer production based on KNN regression.

[0031] Figure 5a It is a schematic diagram of the autonomous decision-making results of the material distribution path for the first specific transformer production based on the improved A* (A-Star) algorithm;

[0032] Figure 5b This is a schematic diagram of the autonomous decision-making results of the material distribution path for the second specific transformer production based on the improved A* (A-Star) algorithm. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Intelligent material preparation for transformer production is realized based on KNN regression prediction, including:

[0035] Production material preparation is the process of determining the types and quantities of production materials required before the execution of material distribution tasks based on the work orders and warehouse management system (WMS) information for the next period. The present invention analyzes workshop production data based on KNN regression to predict the specific material demand time.

[0036] See also Figure 1 5. In an embodiment of the present invention, a data-driven intelligent distribution method for transformer production materials is provided, and the distribution method steps are as follows:

[0037] Step 1: First, during the transformer production process, AGV, stereoscopic warehouse, manufacturing station, charging station, etc. constitute the workshop intelligent IIoT data acquisition sensor platform, which collects workshop production data information in real time, pre-processes these data, and extracts key fields related to production material distribution. (Such as production work order start time (planned), production work order start time (actual), time required to distribute production materials (planned), historical production material distribution time (actual), etc.);

[0038] Step 2: Further call the KNN regression model to predict the delivery time of the next period of time for the materials to be delivered;

[0039] Step 3: Further calculate the material preparation and delivery time based on the predicted delivery time and issue the material preparation instruction. In the three-dimensional warehouse, the production materials are transferred to the material preparation area for AGV loading and transportation;

[0040] Step 4: After material preparation is completed, the workshop dispatcher issues a production material delivery instruction to the AGV, and the AGV starts to perform the delivery task;

[0041] Step 5: Grid the production workshop. By marking the obstacles in the production area of ​​the grid workshop and the corresponding labels, the algorithm can use the labels to quickly obtain the grid information when searching for the path;

[0042] Step 6: In the grid information network, the improved time window A* (A-Star) algorithm selects the location with the lowest current "cost" according to the extended node for the next search until the search end, thereby planning the path with the lowest cost;

[0043] Step 7: AGV implements the material delivery task according to the path planned by the improved time window A* (A-Star) algorithm;

[0044] Step 8: During the AGV delivery process, the improved time window A* (A-Star) algorithm performs real-time collision detection based on the AGV's location and the real-time status of the workshop, discards nodes that collide with each other, retains valid nodes, calculates its improved time window and adds it to the set, and backtracks to obtain a collision-free path after finding the target point, and updates the AGV's walking path in real time;

[0045] Step 9: After the AGV delivers the materials to the designated location, the improved time window A* (A-Star) algorithm plans the return path in real time and issues instructions.

[0046] The parameters that need to be considered in the KNN prediction of delivery time in step 2 are described as the production order start time (planned), the production order start time (actual), the time required to deliver production materials (planned), and the historical production material delivery time (actual), which are respectively denoted as All the hours, minutes, and seconds are converted into minutes. The production speed of the product before the prediction time is recorded as The consumption rate of production materials before the prediction time is recorded as The production material number and quantity required by the production unit before the forecast time are recorded as Based on the above data description, the data required for discrete manufacturing workshop material prediction can be obtained: D = {{Xi, Yi}}, Yi=yi,i=1,2,…,n,Yi is the target variable, Xi is the associated variable, and n is the data length;

[0047] The workflow steps of KNN prediction of delivery time in step 2 are as follows:

[0048] S1: Input relevant parameters such as production order start time (planned), production order start time (actual), time required to deliver production materials (planned), and historical production material delivery time (actual) into the KNN prediction model;

[0049] S2: Determine the numerical value between all the set points in the known data set and the current set point, and arrange them in descending order according to the numerical value;

[0050] S3: Use N points with similar numerical values ​​to the target set point and determine the mean coordinates of these N points;

[0051] S4: Return the coordinate output of the current node as the prediction of the material preparation time for the next period;

[0052] S5: Combine the material demand information related to MES and WMS to clarify the types and quantities of materials required;

[0053] The A*(A-Star) algorithm in step 6 is based on the A*(A-Star) algorithm, and the A*(A-Star) algorithm heuristic and time window are improved. The two methods are combined to perform spatiotemporal search using an instructive time factor, which can avoid quadratic programming while searching for the shortest time; at the same time, an improved focus search operator is used to improve the algorithm search efficiency while reducing the impact of negative feedback on the search efficiency;

[0054] Autonomous decision-making of transformer production material distribution path based on improved A* (A-Star) algorithm. It should be noted that:

[0055] 1. When the AGV is performing the task of material distribution, it needs to move from the current position to the material preparation area to load, and then move to the material demand station. Material distribution plans the AGV driving path with the decision-making goals of maximizing resource utilization and shortening the path mileage. The algorithm is used to plan the path with the production material demand station as the end point and the AGV's current position as the starting point;

[0056] 2. The improved time window A*(A-Star) algorithm after optimizing the A*(A-Star) algorithm has better performance;

[0057] The working principle of the present invention is:

[0058] The invention divides the distribution process into two stages: production material preparation and production material distribution. First, process data is continuously collected and classified and stored during the workshop production process. Based on the workshop production history data and real-time monitoring data, KNN regression prediction is used in the material preparation stage to realize intelligent material preparation of workshop production materials; in the production material distribution stage, the A* (A-Star) algorithm with improved time window is used to realize autonomous decision-making of AGV material distribution path with good obstacle avoidance ability, smooth and efficient path;

[0059] The present invention can meet the actual use requirements of transformer production material distribution, improve the distribution efficiency of the transformer production process, and has certain reference significance and application.

[0060] The original technical solution of the present invention:

[0061] The present invention is a data-driven intelligent distribution method for transformer production materials: the distribution process is divided into two stages: production material preparation and production material distribution. First, process data is continuously collected and classified and stored during the workshop production process. Based on the workshop production history data and real-time monitoring data, KNN (K-Nearest Neighbor) regression prediction is used in the material preparation stage to realize intelligent preparation of workshop production materials; in the production material distribution stage, the A* (A-Star) algorithm with improved time window is used to realize autonomous decision-making of AGV (Automated Guided Vehicle) material distribution path with good obstacle avoidance ability, smooth and efficient path. The present invention can meet the actual use needs of transformer production material distribution, improve the distribution efficiency of transformer production and manufacturing process, and has certain reference significance and application value.

[0062] The present invention relates to the technical field of workshop material distribution, and in particular to a data-driven intelligent distribution method for transformer production materials.

[0063] In recent years, with the practical application of digital factories in the transformer manufacturing industry, intelligent distribution of transformer production materials has become an inevitable development trend. With the introduction of the lean production concept, the operating time of each device in the factory and the actions of each worker are precisely required. In the distribution process of production materials, the control of time is equally important. According to the theory of lean modular supply JIS (just in sequence) and just-in-time production method JIT (just in time), workshop logistics needs to supply materials in a timely and appropriate amount according to the needs of each production line; at the same time, material distribution should adjust the overall distribution rhythm according to the production rhythm; ensure that the material distribution of each production line is replenished in time according to the expected type and quantity of material consumption, while meeting the production and processing rhythm requirements of the production line, so that the distribution resource consumption is minimized and the material inventory is minimized.

[0064] Production practice has proved that high-quality and efficient enterprise production cannot be separated from the support of production logistics systems. Production material distribution is a key link in the construction of digital factories. However, regarding the problem of workshop material distribution, traditional scheduling methods often ignore the historical operation data accumulated by the manufacturing system, resulting in the production laws hidden in the manufacturing data not being deeply mined. The application of new generation information and communication technology can quickly control, perceive and collect the supply and demand relationship of all resources in the production system, significantly improve the scheduling level of the production execution system, and realize the efficient operation of the entire manufacturing execution process. In summary, based on the historical operation data of the transformer manufacturing plant, it is of great significance to mine and analyze based on big data and artificial intelligence algorithms to realize data-driven intelligent distribution of production materials. At present, in the actual material distribution process, the problem of production preparation before distribution is rarely considered. KNN (K-Nearest Neighbor) is one of the machine learning algorithms, which can be used for classification and regression and is a supervised learning algorithm. In the classification decision, this method only determines the category of the sample to be classified based on the category of the nearest one or several samples. Since KNN (K-nearest neighbor) regression has many advantages such as high accuracy and less training time, KNN regression is used to analyze the relevant factors in the production time series data for the common production material preparation problems in the transformer production material distribution process, so that the production material preparation time can be accurately predicted. At the same time, in terms of AGV material distribution path planning, the A* (A-Star) algorithm is the most effective direct search method for solving the shortest path in a static road network, and it is also an effective algorithm for solving many search problems. The closer the distance estimate in the algorithm is to the actual value, the faster the final search speed. In a critical period in the time cycle, if you exert influence or take action on a specific thing, the chance of success will increase. The A* (A-Star) algorithm with time window constraints has good obstacle avoidance ability, smooth and efficient material distribution path autonomous decision-making ability, and can meet the actual use needs of transformer production material distribution.

[0065] In view of this, it is necessary to provide a data-driven intelligent distribution method for transformer production materials.

[0066] The present invention provides a data-driven intelligent distribution method for transformer production materials. The distribution method divides the distribution process into two stages: production material preparation and production material distribution. Through the collection and classified storage of workshop production history data, KNN regression prediction is used in the material preparation stage to realize intelligent preparation of workshop production materials; in the production material distribution stage, the A* (A-Star) algorithm with improved time window is used to realize autonomous decision-making of AGV material distribution path with good obstacle avoidance capability, smooth path and efficient operation.

[0067] The present invention realizes intelligent material preparation for transformer production based on KNN regression prediction. Material preparation for production is the process of determining the type and quantity of production materials required before the material distribution task is executed based on the work order and warehouse management system (WMS) information for the next period. The present invention analyzes workshop production data based on KNN regression to predict the specific material demand time.

[0068] Step 1: During the transformer production process, AGV, stereoscopic warehouse, manufacturing station, charging station, etc. constitute the workshop intelligent industrial Internet (IIoT) data acquisition sensor platform, which collects workshop production data information in real time, pre-processes the data, and extracts key fields related to production material distribution. (Such as production work order start time (planned), production work order start time (actual), time required to distribute production materials (planned), historical production material distribution time (actual), etc.)

[0069] Step 2: Call the KNN regression model to predict the delivery time of the next period of time for the materials to be delivered.

[0070] Step 3: Calculate the material preparation and delivery time based on the predicted delivery time and issue preparation instructions. In the three-dimensional warehouse, production materials are transferred to the preparation area for AGV loading and transportation.

[0071] Step 4: After material preparation is completed, the workshop dispatcher issues a production material delivery instruction to the AGV, and the AGV starts to perform the delivery task.

[0072] Furthermore, the parameters that need to be considered in the KNN prediction of the delivery time in step 2 are described as: the production order start time (planned), the production order start time (actual), the time required to deliver production materials (planned), and the historical production material delivery time (actual), which are respectively denoted as All the hours, minutes, and seconds are converted into minutes. The production speed of the product before the prediction time is recorded as The consumption rate of production materials before the prediction time is recorded as The production material number and quantity required by the production unit before the forecast time are recorded as Based on the above data description, the data required for discrete manufacturing workshop material prediction can be obtained: D = {{Xi, Yi}}, Yi=yi,i=1,2,…,n, Yi is the target variable, Xi is the associated variable, and n is the data length.

[0073] Furthermore, the workflow of KNN prediction of delivery time in step 2 is described as follows:

[0074] Step 2.1: Input relevant parameters such as production order start time (planned), production order start time (actual), time required to deliver production materials (planned), and historical production material delivery time (actual) into the KNN prediction model.

[0075] Step 2.2: Determine the numerical value between all the set points in the known data set and the current set point, and arrange them in order from large to small according to the numerical value.

[0076] Step 2.3: Use N points with similar numerical values ​​to the target set point and determine the mean coordinates of these N points.

[0077] Step 2.4: Return the coordinate output of the current node as the predicted material preparation time for the next period.

[0078] Step 2.5: Combine the Manufacturing Execution System (MES) to optimize the entire production process from order placement to product completion through information transmission and the Warehouse Management System (WMS) to more perfectly manage information, resources, behaviors, inventory and distribution operations according to the business rules and algorithms of operation, so as to clarify the types and quantities of materials required.

[0079] Autonomous decision-making of transformer production material distribution paths based on the improved A* (A-Star) algorithm, including:

[0080] It should be noted that when AGV performs production material distribution tasks, it needs to move from its current location to the material preparation area to load, and then move to the material demand station. Material distribution is based on the decision-making objectives of maximizing resource utilization and shortening the path mileage.

[0081] The AGV driving path is planned. The algorithm is used to plan the path, taking the production material demand station as the end point and the AGV current position as the starting point.

[0082] It should be noted that the improved time window A*(A-Star) algorithm optimized by the A*(A-Star) algorithm has better performance, using the heuristic time factor for space-time search, ensuring the shortest search time while avoiding quadratic programming. At the same time, the improved focus search operator is used to improve the algorithm search efficiency while reducing the impact of negative feedback on the search efficiency.

[0083] Step 1: Grid the production workshop. By marking the obstacles in the production area of ​​the gridded workshop and the corresponding labels, the algorithm can use the labels to quickly obtain the grid information when searching for the path.

[0084] Step 2: In the grid information network, the improved time window A* (A-Star) algorithm selects the location with the lowest current "cost" according to the extended node for the next search until the search end, thereby planning the path with the lowest cost.

[0085] Step 3: AGV implements the material delivery task according to the path planned by the improved time window A* (A-Star) algorithm.

[0086] Step 4: During the AGV delivery process, the improved time window A* (A-Star) algorithm performs real-time collision detection based on the AGV's location and the real-time status of the workshop, discards nodes that have opposite collisions, retains valid nodes, calculates its improved time window and adds it to the set, and when the target point is found, backtracks to obtain a collision-free path and updates the AGV's walking path in real time.

[0087] Step 5: After the AGV delivers the material to the designated location, the improved time window A* (A-Star) algorithm plans the return path in real time and issues instructions.

[0088] Figure 1 The overall architecture diagram of the data-driven intelligent distribution method for transformer production materials.

[0089] Figure 2 Diagram of the intelligent material preparation model for transformer production.

[0090] Figure 3 The overall flow chart of data-driven intelligent distribution of transformer production materials.

[0091] Figure 4 This is a schematic diagram of the prediction results of intelligent material preparation for transformer production based on KNN regression.

[0092] Figure 5a and Figure 5b This is a schematic diagram of the autonomous decision-making results of transformer production material distribution path based on the improved A* (A-Star) algorithm.

[0093] The present invention provides an excellent data-driven intelligent distribution method for transformer production materials, comprising the following steps:

[0094] Step 1: During the transformer production process, AGV, stereoscopic warehouse, manufacturing station, charging station, etc. constitute the workshop intelligent IIoT data acquisition sensor platform, which collects workshop production data information in real time, pre-processes the data, and extracts key fields related to production material distribution. (Such as production work order start time (planned), production work order start time (actual), time required to distribute production materials (planned), historical production material distribution time (actual), etc.).

[0095] Step 2: Call the KNN regression model to predict the delivery time of the next period of time for the materials to be delivered.

[0096] Furthermore, the parameters that need to be considered in the KNN prediction of the delivery time in step 2 are described as: the production order start time (planned), the production order start time (actual), the time required to deliver production materials (planned), and the historical production material delivery time (actual), which are respectively denoted as All the hours, minutes, and seconds are converted into minutes. The production speed of the product before the prediction time is recorded as The consumption rate of production materials before the prediction time is recorded as The production material number and quantity required by the production unit before the forecast time are recorded as Based on the above data description, the data required for discrete manufacturing workshop material prediction can be obtained: D = {{Xi, Yi}}, Yi=yi,i=1,2,…,n, Yi is the target variable, Xi is the associated variable, and n is the data length.

[0097] Furthermore, the workflow of KNN prediction of delivery time in step 2 is described as follows:

[0098] Step 2.1: Input relevant parameters such as production order start time (planned), production order start time (actual), time required to deliver production materials (planned), and historical production material delivery time (actual) into the KNN prediction model.

[0099] Step 2.2: Determine the numerical value between all the set points in the known data set and the current set point, and arrange them in order from large to small according to the numerical value.

[0100] Step 2.3: Use N points with similar numerical values ​​to the target set point and determine the mean coordinates of these N points.

[0101] Step 2.4: Return the coordinate output of the current node as the predicted material preparation time for the next period.

[0102] Step 2.5: Combine the material demand information from MES and WMS to clarify the types and quantities of materials required.

[0103] Step 3: Calculate the material preparation and delivery time based on the predicted delivery time and issue preparation instructions. In the three-dimensional warehouse, production materials are transferred to the preparation area for AGV loading and transportation.

[0104] Step 4: After material preparation is completed, the workshop dispatcher issues a production material delivery instruction to the AGV, and the AGV starts to perform the delivery task.

[0105] Step 5: Grid the production workshop. By marking the obstacles in the production area of ​​the gridded workshop and the corresponding labels, the algorithm can use the labels to quickly obtain the grid information when searching for the path.

[0106] Step 6: In the grid information network, the improved time window A* (A-Star) algorithm selects the location with the lowest current "cost" according to the extended node for the next search until the search end, thereby planning the path with the lowest cost.

[0107] Furthermore, the A*(A-Star) algorithm in step 7 is based on the A*(A-Star) algorithm, and improves the heuristic and time window of the A*(A-Star) algorithm. The two methods are combined to perform spatiotemporal search using an instructive time factor, which can avoid quadratic programming while searching in the shortest time; at the same time, an improved focus search operator is used to improve the algorithm search efficiency while reducing the impact of negative feedback on the search efficiency.

[0108] Step 7: AGV implements the material delivery task according to the path planned by the improved time window A* (A-Star) algorithm.

[0109] Step 8: During the AGV delivery process, the improved time window A* (A-Star) algorithm performs real-time collision detection based on the AGV's location and the real-time status of the workshop, discards nodes that have opposite collisions, retains valid nodes, calculates its improved time window and adds it to the set, and when the target point is found, backtracks to obtain a collision-free path and updates the AGV's walking path in real time.

[0110] Step 9: After the AGV delivers the materials to the designated location, the improved time window A* (A-Star) algorithm plans the return path in real time and issues instructions.

[0111] The present invention provides a data-driven intelligent distribution method for transformer production materials: the distribution method divides the distribution process into two stages: production material preparation and production material distribution. First, process data is continuously collected and classified and stored during the workshop production process. Based on the workshop production history data and real-time monitoring data, KNN regression prediction is used in the material preparation stage to realize intelligent preparation of workshop production materials; in the production material distribution stage, the A* (A-Star) algorithm with improved time window is used to realize autonomous decision-making of AGV material distribution path with good obstacle avoidance ability, smooth and efficient path. The present invention can meet the actual use needs of transformer production material distribution, improve the distribution efficiency of transformer production and manufacturing process, and has certain reference significance and application.

[0112] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A data-driven intelligent distribution method for transformer production materials, characterized in that: The delivery method steps are as follows: Step 1: First, during the transformer production process, AGV, stereoscopic warehouse, manufacturing station and charging station constitute the workshop intelligent IIoT data acquisition sensor platform, which collects workshop production data information in real time, pre-processes the data, and extracts key fields related to production material distribution; Step 2: Further call the KNN regression model to predict the delivery time of the next period of time for the materials to be delivered; Step 3: Further calculate the material preparation and delivery time based on the predicted delivery time and issue preparation instructions, and transfer the production materials to the preparation area for AGV loading and transportation; Step 4: After material preparation is completed, the workshop dispatcher issues a production material delivery instruction to the AGV, and the AGV starts to perform the delivery task; Step 5: Grid the production workshop. By marking the obstacles in the production area of ​​the grid workshop and the corresponding labels, the algorithm can use the labels to quickly obtain the grid information when searching for the path; Step 6: In the grid information network, the improved time window A* (A-Star) algorithm selects the location with the lowest current "cost" according to the extended node for the next search until the search end, thereby planning the path with the lowest cost; Step 7: AGV implements the material delivery task according to the path planned by the improved time window A* (A-Star) algorithm; Step 8: During the AGV delivery process, the improved time window A* (A-Star) algorithm performs real-time collision detection based on the AGV's location and the real-time status of the workshop, discards nodes that collide with each other, retains valid nodes, calculates its improved time window and adds it to the set, and backtracks to obtain a collision-free path after finding the target point, and updates the AGV's walking path in real time; Step 9: After the AGV delivers the materials to the designated location, the improved time window A* (A-Star) algorithm plans the return path in real time and issues instructions.

2. The data-driven intelligent distribution method for transformer production materials according to claim 1 is characterized in that: The parameters that need to be considered in the KNN prediction of delivery time in step 2 are described as the production order start time (planned), the production order start time (actual), the time required to deliver production materials (planned), and the historical production material delivery time (actual), which are respectively denoted as All the hours, minutes, and seconds are converted into minutes, and the production speed of the product before the predicted time is recorded as The consumption rate of production materials before the prediction time is recorded as The production material number and quantity required by the production unit before the forecast time are recorded as Based on the above data description, the data required for discrete manufacturing workshop material forecasting can be obtained Yi=yi,i=1,2,…,n, Yi is the target variable, Xi is the associated variable, and n is the data length.

3. The data-driven intelligent distribution method for transformer production materials according to claim 1 is characterized in that: The workflow steps of KNN prediction of delivery time in step 2 are as follows: S1: Input the production order start time (planned), production order start time (actual), time required to deliver production materials (planned), and historical production material delivery time (actual) related parameters into the KNN prediction model; S2: Determine the numerical value between all the set points in the known data set and the current set point, and arrange them in descending order according to the numerical value; S3: Use N points with similar numerical values ​​to the target set point and determine the mean coordinates of these N points; S4: Return the coordinate output of the current node as the prediction of the material preparation time for the next period; S5: Combine the material demand information related to MES and WMS to clarify the types and quantities of materials required.

4. The data-driven intelligent distribution method for transformer production materials according to claim 1 is characterized in that: The A*(A-Star) algorithm in step 6 is based on the A*(A-Star) algorithm, and improves the heuristic and time window of the A*(A-Star) algorithm.

Images