A cutter motion trajectory optimization method and system

Abstract

The present application relates to the technical field of gypsum board production, and particularly relates to a cutting knife motion trajectory optimization method and system, comprising the following steps: a cutting knife digital twin platform for optimizing the cutting motion trajectory of a gypsum board cutting knife through virtual simulation is constructed by using a three-dimensional model of a gypsum board cutting motion area and a three-dimensional model of a gypsum board cutting knife through digital twin technology; on the cutting knife digital twin platform, a plurality of optimization tasks are established for the cutting motion trajectory of the gypsum board cutting knife, and then simulation learning is performed on the plurality of optimization tasks to obtain the optimized trajectory of the cutting motion of the gypsum board cutting knife. The present application uses a multi-task learning mechanism to perform optimized learning of the cutting motion trajectory of the gypsum board cutting knife, reduces the training and reasoning burden of a plurality of independent tasks, improves the calculation efficiency, and ultimately improves the generalization ability and efficiency of trajectory optimization.

Description

Technical Field

[0001] This invention relates to the field of gypsum board production technology, specifically to a method and system for optimizing the motion trajectory of a cutting tool. Background Technology

[0002] The cutting machine is a key piece of equipment in a gypsum board production line. Its function is to cut continuously formed wet gypsum board into groups. The cutting accuracy is closely related to subsequent production. Therefore, improving the cutting accuracy of the cutting machine is a guarantee for continuous production of gypsum board and a crucial factor in improving the input-output ratio of production. Improving the cutting accuracy of the cutting machine depends on optimizing the movement trajectory of the gypsum board cutter.

[0003] In existing technologies, the optimization of the movement trajectory of gypsum board cutters usually relies on the subjective settings of operators. This is highly subjective and has limited optimization effects. Moreover, the subjective optimization of the cutter's movement trajectory requires real-time data acquisition, transmission, calculation, and storage during the cutter's operation. The computing and storage resources of the gypsum board cutter control end are limited, making it difficult to maintain a hardware environment with high timeliness for data transmission and storage to optimize the cutter trajectory. This affects the optimization effect of the cutter trajectory from the hardware side. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing the motion trajectory of a cutting tool, so as to solve the technical problems of the prior art, which relies on the subjective settings of the operator, is highly subjective, and is difficult to maintain a hardware environment for data transmission and storage with high timeliness.

[0005] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution:

[0006] A method for optimizing the motion trajectory of a cutting tool includes the following steps:

[0007] Acquire real-world spatial data of the cutting motion area of ​​the plasterboard cutter, as well as the physical spatial data of the plasterboard cutter;

[0008] A three-dimensional model of the gypsum board cutter's cutting motion area was constructed based on the real-world spatial data of the gypsum board cutter's cutting motion area, and a three-dimensional model of the gypsum board cutter was constructed based on the solid spatial data of the gypsum board cutter.

[0009] By using digital twin technology, a digital twin platform for the gypsum board cutter is constructed using a 3D model of the gypsum board cutting motion area and a 3D model of the gypsum board cutter. This platform is used to optimize the cutting motion trajectory of the gypsum board cutter through virtual simulation.

[0010] On the digital twin platform for the gypsum board cutter, multiple optimization tasks are established for the cutting motion trajectory of the gypsum board cutter. Then, simulation learning is performed on the multiple optimization tasks to obtain the optimized trajectory of the gypsum board cutter's cutting motion.

[0011] The digital twin platform of the gypsum board cutter feeds back the optimized trajectory to the physical control terminal of the gypsum board cutter, so as to control the physical gypsum board cutter to perform optimal cutting motion along the optimized trajectory in the real space of the gypsum board cutter cutting motion area.

[0012] As a preferred embodiment of the present invention, the three-dimensional model of the cutting motion area of ​​the gypsum board cutter and the three-dimensional model of the gypsum board cutter are both obtained by three-dimensional modeling based on the real-scene space data and the entity space data using CAD software.

[0013] As a preferred embodiment of the present invention, the digital twin platform for the gypsum board cutter adopts a projection method to establish a positional mapping between the physical gypsum board cutter and the real-world space of the gypsum board cutter's cutting motion area, and to establish a data channel for data transmission with the physical gypsum board cutter.

[0014] As a preferred embodiment of the present invention, the position mapping includes:

[0015] Select several ground control points from the 3D model of the gypsum board cutting movement area and record their 3D coordinates. Then, find the positions of the observation points corresponding to the selected ground control points in the real space and record their latitude and longitude coordinates. Finally, project the latitude and longitude coordinates containing height information of each observation point into 3D coordinates in the Cartesian coordinate system using the projection forward formula.

[0016] Based on the 3D coordinates of several control points and the 3D coordinates of the corresponding observation points after projection transformation, the coordinate system deviation of the center point of the 3D model of the gypsum board cutting motion area from the latitude and longitude coordinates to the origin of the Cartesian coordinate system is estimated using the least squares method. The 3D coordinates of each point in the real space after projection transformation using the forward projection formula are added to the coordinate system deviation to obtain the 3D coordinates of the corresponding point in the cutting knife digital twin platform in the Cartesian coordinate system. The coordinate system deviation is subtracted from the 3D coordinates of each point in the cutting knife digital twin platform in the Cartesian coordinate system, and the 3D coordinates after projection transformation using the inverse projection formula are used as the 3D coordinates of the corresponding point in the real space.

[0017] As a preferred embodiment of the present invention, the step of establishing multiple optimization tasks for the cutting motion trajectory of the plasterboard cutter on the digital twin platform of the cutter includes:

[0018] Taking the object accuracy of the plasterboard cutter as the primary optimization objective, a first optimization task is established to optimize the cutting motion trajectory. The first optimization task is as follows:

[0019] P t =BP(M t );

[0020] In the formula, Pt Let M be the cutting trajectory point at time t. t Let t be the shape of the plasterboard cut at time t, BP be the mapping relationship between the shape of the plasterboard cut and the cutting trajectory position represented by the BP neural network, and t be the counting variable.

[0021] Taking the wear rate of the plasterboard cutter as the primary optimization objective, a second optimization task is established to optimize the cutting motion trajectory. The second optimization task is as follows:

[0022] P t =LSTM({P1,P2,…,P) t-1});

[0023] In the formula, P t Let P1, P2, P be the cutting trajectory points at time t. t-1 These are the cutting trajectory points at time 1, time 2, and time t-1, respectively. LSTM is an LSTM neural network, and t is a counting variable.

[0024] As a preferred embodiment of the present invention, simulation learning is performed on multiple optimization tasks to obtain the optimized trajectory of the plasterboard cutter's cutting motion, including:

[0025] Based on the first and second optimization tasks, establish learning objectives for multiple optimization tasks;

[0026] The output layer, learning objective, and bottom network layer of the first optimization task and the second optimization task are shared. Based on the shared output layer, learning objective, and bottom network layer, multi-task learning is performed on the first optimization task and the second optimization task to obtain the trajectory optimization model.

[0027] The model expression for the trajectory optimization model is:

[0028]

[0029] In the formula, P t_share Let M be the optimized value of the cutting trajectory point at time t, share be the network volume of the trajectory optimization model, and M be the value of the cutting trajectory point at time t. t Let P1, P2, P be the shape of the plasterboard cut at time t. t-1 These are the cutting trajectory points at time 1, time 2, and time t-1, respectively, where t is the counting variable.

[0030] As a preferred embodiment of the present invention, the construction of the learning objective includes:

[0031] The similarity between the output items of the first optimization task and the output items of the second optimization task is calculated to obtain the similarity between the output items of the first optimization task and the output items of the second optimization task.

[0032] The learning objective is obtained by maximizing the similarity between the outputs of the first optimization task and the outputs of the second optimization task. The learning objective is:

[0033] Goal = max[sim(P) BP ,P LSTM )];

[0034] In the formula, Goal is the learning objective, max is the maximization operator, sim is the similarity operator, and P BP ,P LSTM These are the output terms of the first optimization task and the second optimization task, respectively, sim(P) BP ,P LSTM ) is P BP and P LSTM Similarity between them.

[0035] As a preferred embodiment of the present invention, the mapping relationship between the gypsum board cutting shape and the cutting motion trajectory point is obtained by a BP neural network through big data learning between the gypsum board cutting shape and the cutting motion trajectory point.

[0036] As a preferred embodiment of the present invention, the cutting blade digital twin platform feeds back the optimized trajectory to the physical control terminal of the plasterboard cutting blade through the data channel.

[0037] As a preferred embodiment of the present invention, the present invention provides an optimization system applying the aforementioned cutter motion trajectory optimization method, comprising:

[0038] The data acquisition unit is used to acquire real-world spatial data of the cutting motion area of ​​the gypsum board cutter, as well as the physical spatial data of the gypsum board cutter.

[0039] The 3D modeling unit is used to construct a 3D model of the gypsum board cutter's cutting motion area based on the real-world spatial data of the gypsum board cutter's cutting motion area, and to construct a 3D model of the gypsum board cutter based on the solid spatial data of the gypsum board cutter.

[0040] The twin digital unit is used to construct a cutting tool digital twin platform by using a three-dimensional model of the gypsum board cutting motion area and a three-dimensional model of the gypsum board cutter through digital twin technology, which is used to optimize the cutting motion trajectory of the gypsum board cutter through virtual simulation.

[0041] The data processing unit is used to establish multiple optimization tasks for the cutting motion trajectory of the plasterboard cutter on the digital twin platform of the cutter, and then perform simulation learning on the multiple optimization tasks to obtain the optimized trajectory of the cutting motion of the plasterboard cutter.

[0042] The data processing unit and the physical control terminal of the gypsum board cutter communicate with each other through a data channel;

[0043] The digital twin platform of the cutter feeds back the optimized trajectory to the physical control terminal of the plasterboard cutter, so as to control the physical plasterboard cutter to perform optimal cutting motion along the optimized trajectory in the real space of the plasterboard cutter cutting motion area.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] This invention utilizes digital twin technology to construct a digital twin platform for optimizing the cutting trajectory of a plasterboard cutter through virtual simulation. This allows for extensive testing of trajectory optimization in a virtual space that closely resembles a real-world environment, effectively improving the safety of the cutter's cutting motion in real-world conditions. Simultaneously, it reduces the hardware and computer load on the cutter's control end while increasing the efficiency of the cutter's task execution. By employing a multi-task learning mechanism for optimizing the plasterboard cutter's cutting trajectory, the burden of training and inference for multiple independent tasks is reduced, computational efficiency is improved, and ultimately, the generalization ability and efficiency of trajectory optimization are enhanced. Attached Figure Description

[0046] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0047] Figure 1 This is a flowchart of the cutter motion trajectory optimization method provided in an embodiment of the present invention;

[0048] Figure 2 An optimized system block diagram provided for an embodiment of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] like Figure 1 As shown, the present invention provides a method for optimizing the motion trajectory of a cutter, comprising the following steps:

[0051] Acquire real-world spatial data of the cutting motion area of ​​the plasterboard cutter, as well as the physical spatial data of the plasterboard cutter;

[0052] A three-dimensional model of the gypsum board cutter's cutting motion area was constructed based on the real-world spatial data of the gypsum board cutter's cutting motion area, and a three-dimensional model of the gypsum board cutter was constructed based on the solid spatial data of the gypsum board cutter.

[0053] By using digital twin technology, a digital twin platform for the gypsum board cutter is constructed using a 3D model of the gypsum board cutting motion area and a 3D model of the gypsum board cutter. This platform is used to optimize the cutting motion trajectory of the gypsum board cutter through virtual simulation.

[0054] On the digital twin platform for the gypsum board cutter, multiple optimization tasks are established for the cutting motion trajectory of the gypsum board cutter. Then, simulation learning is performed on the multiple optimization tasks to obtain the optimized trajectory of the gypsum board cutter's cutting motion.

[0055] The digital twin platform of the gypsum board cutter feeds back the optimized trajectory to the physical control end of the gypsum board cutter, so as to control the physical gypsum board cutter to perform optimal cutting motion along the optimized trajectory in the real space of the gypsum board cutter cutting motion area.

[0056] To improve the safety of the cutting motion of the plasterboard cutter in a real environment, and to effectively reduce the hardware and computer load on the cutter control terminal while improving the efficiency of the cutter's task execution, this invention constructs a digital twin platform for the cutter using digital twin technology. This platform optimizes the cutting trajectory of the plasterboard cutter through virtual simulation. The trajectory optimization is completed in a digital virtual environment without occupying the hardware resources of the running cutter control terminal, thus avoiding interference with the cutter's operation. The trajectory simulation in the virtual environment avoids the need for real-world cutting trajectory simulation, thereby preventing interference with the cutter's operation. Furthermore, using the optimized trajectory obtained in the virtual environment for cutter cutting planning ensures the operational safety of the cutter in the real environment.

[0057] To improve the generalization ability and efficiency of trajectory optimization, this invention utilizes a multi-task learning mechanism to jointly learn different optimization tasks, thereby achieving optimized learning of the cutting trajectory of the plasterboard cutter. This reduces the training and inference burden of multiple independent tasks and improves computational efficiency.

[0058] Specifically, when optimizing the trajectory, this invention takes improving cutting accuracy as an optimization goal. The object of this optimization goal is the object being cut by the cutter—the gypsum board. It is expected that the shape of the gypsum board after cutting will conform to the product standard specifications. The cutting accuracy of the gypsum board is also an indispensable optimization indicator for trajectory optimization. At the same time, this invention also takes reducing cutter wear as another optimization goal. The object of this optimization goal is the cutter itself. It is expected that the cutting trajectory can produce the least amount of cutter wear.

[0059] This invention optimizes the object and subject aspects of the cutter's motion trajectory into two optimization tasks, and performs synchronous optimization through a multi-task learning mechanism. The similarity between the outputs of the two optimization tasks is used as the learning target of the multi-task learning, so that the final output optimized trajectory can balance the object accuracy and the subject wear, thereby optimizing the cutting accuracy and protecting the cutter. Therefore, multi-task optimization achieves synchronous optimization of the object and subject, and improves the trajectory optimization effect.

[0060] To improve the safety of the cutting motion of the cutter in a real environment, and to effectively reduce the hardware and computer load on the cutter control terminal while improving the efficiency of the cutter in executing tasks, this invention constructs a digital twin platform for the cutter using digital twin technology, as detailed below:

[0061] The 3D model of the cutting motion area of ​​the plasterboard cutter and the 3D model of the plasterboard cutter were both obtained by CAD software based on real-scene space data and physical space data.

[0062] The digital twin platform for the plasterboard cutter uses a projection method to establish a positional mapping between the physical plasterboard cutter and the real-world space of the plasterboard cutter's cutting motion area, as well as to establish a data channel for data transmission between the physical plasterboard cutter and the physical plasterboard cutter.

[0063] Location mapping, including:

[0064] Select several ground control points from the 3D model of the gypsum board cutting movement area and record their 3D coordinates. Then, find the positions of the observation points corresponding to the selected ground control points in the real space and record their latitude and longitude coordinates. Finally, project the latitude and longitude coordinates containing height information of each observation point into 3D coordinates in the Cartesian coordinate system using the projection forward formula.

[0065] Based on the 3D coordinates of several control points and the 3D coordinates of the corresponding observation points after projection transformation, the coordinate system deviation of the center point of the 3D model of the gypsum board cutting motion area from the latitude and longitude coordinates to the origin of the Cartesian coordinate system is estimated using the least squares method. The 3D coordinates of each point in the real space after projection transformation using the forward projection formula are added to the coordinate system deviation to obtain the 3D coordinates of the corresponding point in the cutting knife digital twin platform in the Cartesian coordinate system. The coordinate system deviation is subtracted from the 3D coordinates of each point in the cutting knife digital twin platform in the Cartesian coordinate system, and the 3D coordinates after projection transformation using the inverse projection formula are used as the 3D coordinates of the corresponding point in the real space.

[0066] To improve the generalization ability and efficiency of trajectory optimization, this invention utilizes a multi-task learning mechanism to jointly learn different optimization tasks, thereby achieving optimized learning of the cutting trajectory of a plasterboard cutter. This reduces the training and inference burden of multiple independent tasks and improves computational efficiency, as detailed below:

[0067] On the digital twin platform for the cutting tool, multiple optimization tasks are established for the cutting motion trajectory of the plasterboard cutter, including:

[0068] With the object accuracy of the plasterboard cutter as the primary optimization objective, a first optimization task is established to optimize the cutting motion trajectory. The first optimization task is as follows:

[0069] P t =BP(M t );

[0070] In the formula, P t Let M be the cutting trajectory point at time t. t Let t be the shape of the plasterboard cut at time t, BP be the mapping relationship between the shape of the plasterboard cut and the cutting trajectory position represented by the BP neural network, and t be the counting variable.

[0071] Taking the wear rate of the plasterboard cutter as the primary optimization objective, a second optimization task is established to optimize the cutting motion trajectory. The second optimization task is as follows:

[0072] P t =LSTM({P1,P2,…,P) t-1});

[0073] In the formula, P t Let P1, P2, P be the points on the cutting motion trajectory at time t. t-1 These are the cutting trajectory points at time 1, time 2, and time t-1, respectively. LSTM is an LSTM neural network, and t is a counting variable.

[0074] Simulation learning was performed on multiple optimization tasks to obtain the optimized trajectory of the plasterboard cutter's cutting motion, including:

[0075] Based on the first and second optimization tasks, establish learning objectives for multiple optimization tasks;

[0076] The output layer, learning objective, and bottom network layer of the first optimization task and the second optimization task are shared. Based on the shared output layer, learning objective, and bottom network layer, multi-task learning is performed on the first optimization task and the second optimization task to obtain the trajectory optimization model.

[0077] The model expression for the trajectory optimization model is:

[0078]

[0079] In the formula, P t_share Let M be the optimized value of the cutting trajectory point at time t, share be the network volume of the trajectory optimization model, and M be the value of the cutting trajectory point at time t. t Let P1, P2, P be the shape of the plasterboard cut at time t. t-1 These are the cutting trajectory points at time 1, time 2, and time t-1, respectively, where t is the counting variable.

[0080] The construction of learning objectives includes:

[0081] The similarity between the output items of the first optimization task and the output items of the second optimization task is calculated to obtain the similarity between the output items of the first optimization task and the output items of the second optimization task.

[0082] The learning objective is obtained by maximizing the similarity between the outputs of the first optimization task and the outputs of the second optimization task. The learning objective is:

[0083] Goal = max[sim(P) BP ,P LSTM )];

[0084] In the formula, Goal is the learning objective, max is the maximization operator, sim is the similarity operator, and P BP ,P LSTM These are the output terms of the first optimization task and the second optimization task, respectively, sim(P) BP ,P LSTM ) is P BP and P LSTM Similarity between them.

[0085] The learning objective of this invention is to maintain the highest similarity between the output trajectory points of two tasks. This allows the trajectory optimization model jointly learned based on this learning objective to take into account the effects of both task lines, namely, the accuracy of the cutter and the loss of the cutter. This ensures that the shape of the gypsum board after cutting conforms to the product standard specifications and produces the least amount of cutter wear, thereby optimizing the cutting accuracy and protecting the cutter.

[0086] This invention utilizes a multi-task learning mechanism to fuse two optimization task lines. Multi-task learning trains the two optimization tasks simultaneously, improving the performance of each task through parameter sharing and interactive learning. Furthermore, multi-task learning combines the two optimization tasks into a single model, reducing the training and inference burden of multiple independent models, improving computational efficiency, and ultimately enhancing the generalization ability and efficiency of rehabilitation assessment.

[0087] Specifically, when optimizing the trajectory, this invention takes improving cutting accuracy as an optimization goal. The object of this optimization goal is the object being cut by the cutter—the gypsum board. It is expected that the shape of the gypsum board after cutting will conform to the product standard specifications. The cutting accuracy of the gypsum board is also an indispensable optimization indicator for trajectory optimization. At the same time, this invention also takes reducing cutter wear as another optimization goal. The object of this optimization goal is the cutter itself. It is expected that the cutting trajectory can produce the least amount of cutter wear.

[0088] This invention optimizes the object and subject aspects of the cutter's motion trajectory into two optimization tasks, and performs synchronous optimization through a multi-task learning mechanism. The similarity between the outputs of the two optimization tasks is used as the learning target of the multi-task learning, so that the final output optimized trajectory can balance the object accuracy and the subject wear, thereby optimizing the cutting accuracy and protecting the cutter. Therefore, multi-task optimization achieves synchronous optimization of the object and subject, and improves the trajectory optimization effect.

[0089] The mapping relationship between the gypsum board cutting shape and the cutting motion trajectory point is obtained by a BP neural network through big data learning between the gypsum board cutting shape and the cutting motion trajectory point.

[0090] The digital twin platform for the cutting blade feeds back the optimized trajectory to the physical control terminal of the plasterboard cutting blade through a data channel.

[0091] like Figure 2 As shown, the present invention provides an optimization system for an application of a cutter motion trajectory optimization method, comprising:

[0092] The data acquisition unit is used to acquire real-world spatial data of the cutting motion area of ​​the gypsum board cutter, as well as the physical spatial data of the gypsum board cutter.

[0093] The 3D modeling unit is used to construct a 3D model of the gypsum board cutter's cutting motion area based on the real-world spatial data of the gypsum board cutter's cutting motion area, and to construct a 3D model of the gypsum board cutter based on the solid spatial data of the gypsum board cutter.

[0094] The twin digital unit is used to construct a cutting tool digital twin platform by using a three-dimensional model of the gypsum board cutting motion area and a three-dimensional model of the gypsum board cutter through digital twin technology, which is used to optimize the cutting motion trajectory of the gypsum board cutter through virtual simulation.

[0095] The data processing unit is used to establish multiple optimization tasks for the cutting motion trajectory of the plasterboard cutter on the digital twin platform of the cutter, and then perform simulation learning on the multiple optimization tasks to obtain the optimized trajectory of the cutting motion of the plasterboard cutter.

[0096] The data processing unit and the physical control terminal of the gypsum board cutter communicate with each other through a data channel;

[0097] The digital twin platform for the drywall cutter feeds back the optimized trajectory to the physical control unit of the drywall cutter, so as to control the physical drywall cutter to perform optimal cutting motion along the optimized trajectory in the real space of the drywall cutter's cutting motion area.

[0098] The data processing unit establishes learning objectives for multiple optimization tasks based on the first and second optimization tasks;

[0099] The data processing unit shares the output layer, learning objective, and bottom network layer of the first and second optimization tasks, and performs multi-task learning on the first and second optimization tasks based on the shared output layer, learning objective, and bottom network layer to obtain the trajectory optimization model.

[0100] The model expression for the trajectory optimization model is:

[0101]

[0102] In the formula, P t_share Let M be the optimized value of the cutting trajectory point at time t, share be the network volume of the trajectory optimization model, and M be the value of the cutting trajectory point at time t. t Let P1, P2, P be the shape of the plasterboard cut at time t. t-1 These are the cutting trajectory points at time 1, time 2, and time t-1, respectively, where t is the counting variable.

[0103] The data processing unit constructs learning objectives, including:

[0104] The similarity between the output items of the first optimization task and the output items of the second optimization task is calculated to obtain the similarity between the output items of the first optimization task and the output items of the second optimization task.

[0105] The learning objective is obtained by maximizing the similarity between the outputs of the first optimization task and the outputs of the second optimization task. The learning objective is:

[0106] Goal = max[sim(P) BP ,P LSTM )];

[0107] In the formula, Goal is the learning objective, max is the maximization operator, sim is the similarity operator, and P BP ,P LSTM These are the output terms of the first optimization task and the second optimization task, respectively, sim(P) BP ,P LSTM ) is P BP and P LSTM Similarity between them.

[0108] This invention utilizes digital twin technology to construct a digital twin platform for optimizing the cutting trajectory of a plasterboard cutter through virtual simulation. This allows for extensive testing of trajectory optimization in a virtual space that closely resembles a real-world environment, effectively improving the safety of the cutter's cutting motion in real-world conditions. Simultaneously, it reduces the hardware and computer load on the cutter's control end while increasing the efficiency of the cutter's task execution. By employing a multi-task learning mechanism for optimizing the plasterboard cutter's cutting trajectory, the burden of training and inference for multiple independent tasks is reduced, computational efficiency is improved, and ultimately, the generalization ability and efficiency of trajectory optimization are enhanced.

[0109] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for optimizing the motion trajectory of a cutting tool, characterized in that, Includes the following steps: Acquire real-world spatial data of the cutting motion area of ​​the plasterboard cutter, as well as the physical spatial data of the plasterboard cutter; A three-dimensional model of the gypsum board cutter's cutting motion area was constructed based on the real-world spatial data of the gypsum board cutter's cutting motion area, and a three-dimensional model of the gypsum board cutter was constructed based on the solid spatial data of the gypsum board cutter. By using digital twin technology, a digital twin platform for plasterboard cutter is constructed using a 3D model of the cutting motion area of ​​the plasterboard cutter and a 3D model of the plasterboard cutter. This platform is used to optimize the cutting motion trajectory of the plasterboard cutter through virtual simulation. On the digital twin platform for the gypsum board cutter, multiple optimization tasks are established for the cutting motion trajectory of the gypsum board cutter. Then, simulation learning is performed on the multiple optimization tasks to obtain the optimized trajectory of the gypsum board cutter's cutting motion. The digital twin platform of the cutter feeds back the optimized trajectory to the physical control terminal of the plasterboard cutter, so as to control the physical plasterboard cutter to perform optimal cutting motion along the optimized trajectory in the real space of the plasterboard cutter cutting motion area; On the digital twin platform for the cutting tool, multiple optimization tasks are established for the cutting motion trajectory of the plasterboard cutter, including: Taking the object accuracy of the plasterboard cutter as the primary optimization objective, a first optimization task is established to optimize the cutting motion trajectory. The first optimization task is as follows: P t =BP(M t ); In the formula, P t Let M be the cutting trajectory point at time t. t Let t be the shape of the plasterboard cut at time t, BP be the mapping relationship between the shape of the plasterboard cut and the cutting trajectory position represented by the BP neural network, and t be the counting variable. Taking the wear rate of the plasterboard cutter as the second optimization objective, a second optimization task is established to optimize the cutting motion trajectory. The second optimization task is as follows: P t =LSTM({P1,P2,…,P t-1 }); In the formula, P t Let P1, P2, P be the points on the cutting motion trajectory at time t. t-1 These are the cutting trajectory points at time 1, time 2, and time t-1, respectively. LSTM is an LSTM neural network, and t is a counting variable.

2. The method for optimizing the motion trajectory of a cutter according to claim 1, characterized in that: The 3D model of the cutting motion area of ​​the plasterboard cutter and the 3D model of the plasterboard cutter are both obtained by 3D modeling based on the real-world spatial data and the physical spatial data using CAD software.

3. The method for optimizing the motion trajectory of a cutter according to claim 1, characterized in that: The digital twin platform for the gypsum board cutter uses a projection method to establish a positional mapping between the physical gypsum board cutter and the real-world space of the gypsum board cutter's cutting motion area, as well as to establish a data channel for data transmission with the physical gypsum board cutter.

4. The method for optimizing the motion trajectory of a cutter according to claim 3, characterized in that: The location mapping includes: Several ground control points are selected from the 3D model of the gypsum board cutting motion area and their 3D coordinates are recorded. Then, the positions of the selected ground control points and their corresponding observation points are found in the real space and their latitude and longitude coordinates are recorded. The latitude and longitude coordinates of each observation point, which contain height information, are then projected into 3D coordinates in the Cartesian coordinate system using the projection forward formula. Based on the 3D coordinates of several control points and the 3D coordinates of the corresponding observation points after projection transformation, the coordinate system deviation of the center point of the 3D model of the gypsum board cutter cutting motion area from the latitude and longitude coordinates to the origin of the Cartesian coordinate system is estimated using the least squares method. The 3D coordinates of each point in the real space after projection transformation using the forward projection formula are added to the coordinate system deviation to obtain the 3D coordinates of the corresponding point in the cutter digital twin platform in the Cartesian coordinate system. The coordinate system deviation is subtracted from the 3D coordinates of each point in the cutter digital twin platform in the Cartesian coordinate system, and the 3D coordinates after projection transformation using the inverse projection formula are used as the 3D coordinates of the corresponding point in the real space.

5. The method for optimizing the motion trajectory of a cutter according to claim 4, characterized in that: Simulation learning was performed on multiple optimization tasks to obtain the optimized trajectory of the plasterboard cutter's cutting motion, including: Based on the first and second optimization tasks, establish learning objectives for multiple optimization tasks; The output layer, learning objective, and bottom network layer of the first optimization task and the second optimization task are shared. Based on the shared output layer, learning objective, and bottom network layer, multi-task learning is performed on the first optimization task and the second optimization task to obtain the trajectory optimization model. The model expression for the trajectory optimization model is: ； In the formula, P t_share Let M be the optimized value of the cutting trajectory point at time t, share be the network volume of the trajectory optimization model, and M be the value of the cutting trajectory point at time t. t Let P1, P2, P be the shape of the plasterboard cut at time t. t-1 These are the cutting trajectory points at time 1, time 2, and time t-1, respectively, where t is the counting variable.

6. The method for optimizing the motion trajectory of a cutter according to claim 5, characterized in that: The construction of the learning objectives includes: The similarity between the output items of the first optimization task and the output items of the second optimization task is calculated to obtain the similarity between the output items of the first optimization task and the output items of the second optimization task. The learning objective is obtained by maximizing the similarity between the outputs of the first optimization task and the outputs of the second optimization task. The learning objective is: Goal=max[sim(P BP ,P LSTM )]; In the formula, Goal is the learning objective, max is the maximization operator, sim is the similarity operator, and P BP ,P LSTM These are the output terms of the first optimization task and the second optimization task, respectively, sim(P) BP ,P LSTM ) is P BP and P LSTM Similarity between them.

7. The method for optimizing the motion trajectory of a cutter according to claim 1, characterized in that: The mapping relationship between the gypsum board cutting shape and the cutting motion trajectory point is obtained by a BP neural network through big data learning between the gypsum board cutting shape and the cutting motion trajectory point.

8. The method for optimizing the motion trajectory of a cutter according to claim 3, characterized in that: The digital twin platform for the cutting blade feeds back the optimized trajectory to the physical control terminal of the plasterboard cutting blade through the data channel.

9. An optimization system applying the cutter motion trajectory optimization method according to any one of claims 1-8, characterized in that, include: The data acquisition unit is used to acquire real-world spatial data of the cutting motion area of ​​the gypsum board cutter, as well as the physical spatial data of the gypsum board cutter. The 3D modeling unit is used to construct a 3D model of the gypsum board cutter's cutting motion area based on the real-world spatial data of the gypsum board cutter's cutting motion area, and to construct a 3D model of the gypsum board cutter based on the solid spatial data of the gypsum board cutter. The digital twin unit is used to construct a digital twin platform for the plasterboard cutter by using a 3D model of the cutting motion area of ​​the plasterboard cutter and a 3D model of the plasterboard cutter through digital twin technology. This platform is used to optimize the cutting motion trajectory of the plasterboard cutter through virtual simulation. The data processing unit is used to establish multiple optimization tasks for the cutting motion trajectory of the plasterboard cutter on the digital twin platform of the cutter, and then perform simulation learning on the multiple optimization tasks to obtain the optimized trajectory of the cutting motion of the plasterboard cutter. The data processing unit and the physical control terminal of the gypsum board cutter communicate with each other through a data channel; The digital twin platform of the cutter feeds back the optimized trajectory to the physical control terminal of the plasterboard cutter, so as to control the physical plasterboard cutter to perform optimal cutting motion along the optimized trajectory in the real space of the plasterboard cutter cutting motion area.

Images