Digital twin and stress prediction method and device for large component additive manufacturing process

Abstract

The application discloses a kind of digital twin and stress prediction method and device of large component additive manufacturing process, and time sequence image is obtained by scanning in time sequence by slidable line laser head, corresponding three-dimensional real-time model is constructed and surface displacement difference in molten pool formation to cooling solidification process is obtained;It is input to three-dimensional real-time model as surface displacement load and is carried out numerical simulation, temperature field, stress field are obtained by numerical simulation, and the temperature measured by infrared thermal imager, the stress measured by stress sensor and the temperature, stress obtained by numerical simulation are compared, the data obtained are input into three-dimensional real-time model, and the digital twin of real-time molten pool and deposition layer is formed.The application can well reduce simulation time, improve the efficiency of printing process numerical simulation, and effectively monitor and predict the deformation and cracks generated in large component printing process.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of additive manufacturing. Specifically relates to a digital twin and stress prediction method and device for large component additive manufacturing process. BACKGROUND

[0002] In the additive manufacturing process, stress will cause the printed part to deform and crack, thereby affecting the quality of the printed part. In large component (super large component) printing, this not only affects the manufacturing efficiency, but also increases the manufacturing cost, which seriously affects the development of additive manufacturing technology in manufacturing large components. By constructing a real-time three-dimensional model digital twin, not only can the printing quality be controlled during printing, but also the stress of the large component can be predicted, so that the deformation and cracks generated during printing can be effectively controlled.

[0003] In the additive manufacturing process, numerical simulation is usually used to simulate the printing process, so that the printing process can be optimized, printing defects can be predicted, and trial and error costs can be reduced according to the simulation results. In the large component printing process, due to the large model, the calculation amount of numerical simulation is large, and the conventional numerical simulation method cannot be completely suitable for large components, resulting in waste of time and effort. SUMMARY

[0004] The purpose of the present application is to solve the problem that in the large component printing process, due to the large model, the calculation amount of numerical simulation is large, and the conventional numerical simulation method cannot be completely suitable for large components, resulting in waste of time and effort. The present application provides a digital twin and stress prediction method for large component additive manufacturing process. By adopting pooling partition simulation for large components, the simulation time can be greatly reduced, the efficiency of numerical simulation of the printing process can be improved, and the deformation and cracks generated during the large component printing process can be effectively monitored and predicted.

[0005] Technical scheme: In order to achieve the above purpose, the technical scheme adopted by the present application is:

[0006] A digital twin method for large component additive manufacturing process, comprising the following steps:

[0007] Step 1: In the large component additive manufacturing process, a time sequence is scanned by a slidable line laser head to obtain a time-dependent image of the deposition process, denoted as time sequence image M(t), which records the topography height h(x,y,z,t) of the deposit, x,y,z are spatial coordinates, and t is time.

[0008] Step 2: Construct a three-dimensional real-time model corresponding to the time sequence image M(t).

[0009] Step 3, the initial topography height h of the molten pool is obtained according to the time sequence image M(t) initial (x, y, z) and the topography height h after the molten pool is cooled and solidified final (x, y, z) and the topography height h after the molten pool is cooled and solidified initial (x, y, z) and the topography height h after the molten pool is cooled and solidified final (x, y, z) and the topography height h after the molten pool is cooled and solidified

[0010] Step 4, the surface displacement difference Δh(x, y, z) in the cooling and solidification process is taken as the surface displacement load u(x, y, z), and the surface displacement load u(x, y, z) is used as an input load to the three-dimensional real-time model for numerical simulation, and the stress generated in the cooling and shrinkage process of the molten pool is calculated by numerical simulation.

[0011] Step 5, the temperature field and stress field are obtained by numerical simulation, and the temperature measured by the infrared thermal imager and the stress measured by the stress sensor are compared with the numerical simulation temperature and stress, if the error of the two results is within the error range, the simulation result is input to the three-dimensional real-time model. If it is not within the range error, the simulation result and the result measured by the stress sensor are processed by taking the average value, and input to the three-dimensional real-time model to form the digital twin of the real-time molten pool and the deposited layer.

[0012] A stress prediction method for monitoring the additive manufacturing process of a large component, which adopts a digital twin method for the additive manufacturing process of the large component, and calculates the stress generated in the cooling and shrinkage process of the molten pool by numerical simulation, comprising the following steps:

[0013] Step 41, the printing area A is preliminarily divided into several parts according to the size of the established three-dimensional real-time model, and the three-dimensional real-time model is processed by pooling partition, and each pooling partition is marked as P i .

[0014] Step 42, the first pooling partition P1 is selected according to the printing path, the stress field of the first pooling partition P1 is simulated according to the deposition process, and the obtained stress is marked as S 1,sim , the stress measured by the stress sensor is marked as S 1,meas , the stress S 1,sim obtained by numerical simulation is compared with the stress S 1,meas measured by the stress sensor, and the absolute difference is marked as ΔS1.

[0015] Step 43, the dynamic adjustment error limit ε i is determined, i = 1, 2, 3, …, n, and n is the total number of pooling partitions. In the first pooling partition, if ΔS1≤ε1, the stress S1,sim Input into the corresponding part of the three-dimensional real-time model. If ΔS1> ε1, the mean value of the two data is processed to obtain the mean value of the stress result, and the mean value of the stress result is The mean value of the stress result S 1,avg Input into the corresponding part of the three-dimensional real-time model. According to the printing path, the second pooling partition P2 is selected, and the numerical simulation stress result of the first pooling area is The stress S bod is loaded as the boundary condition of the second pooling area. Then, the solution of the second pooling partition is performed in the same way as the first pooling partition. According to the printing path, the pooling partition simulation is performed in sequence.

[0016] Preferably, if there are multiple completed numerical simulation pooling partitions in the selected pooling partition, all completed numerical simulation stress results should be considered, and all completed numerical simulation stress results are summed up and recorded as cumulative stress S sum as the boundary condition of the next area, S sum,i represents the boundary condition of the i-th pooling partition. After completing all partition simulations, the sum is obtained to obtain the actual total stress size of the entire three-dimensional real-time model

[0017] Preferably, the obtained stress S sum is used to train a deep learning model to predict the stress of the unprinted area.

[0018] Preferably, the dynamic adjustment error limit is determined as:

[0019] ε i = α + β · SD(S prev )

[0020] wherein ε i is the dynamic adjustment error limit, i represents the i-th pooling partition, α is the basic error tolerance, β is the adjustment coefficient, and SD(S prev ) is the standard deviation of the stress measurement of the previous pooling partition.

[0021] Another object of the present application is to provide a large component gantry printing device for realizing the stress prediction method of the large component additive manufacturing process monitoring, comprising a reversible laser head, an infrared thermal imager, a substrate table, a slidable line laser head, a gantry column, a substrate fixing shaft, a control console, a printing base table, a rear cross beam, and a front cross beam, wherein:

[0022] The rear crossbeam and the front crossbeam are installed above the gantry column, the gantry column is loaded with a movable shaft, the base fixed shaft is installed on the movable shaft, the base table is fixed on the base fixed shaft, the printing base is arranged on the base table, and the base fixed shaft is loaded with a movable shaft driving motor and a stress sensor.

[0023] The reversible laser head is loaded on the rear crossbeam through a sliding and flipping mechanism, and the infrared thermal imager is installed on the reversible laser head.

[0024] The slidable linear laser head is installed on the front crossbeam through a sliding mechanism.

[0025] The reversible laser head, the infrared thermal imager, the slidable linear laser head and the stress sensor are connected with the control console.

[0026] Preferably, the sliding mechanism comprises a linear guide rail one, a sliding block one and a linear motor one, the linear guide rail one is installed on the front crossbeam, the linear guide rail one and the sliding block one are slidably connected with each other, the linear motor one is installed on the front crossbeam, and the linear motor one is drivingly connected with the sliding block one.

[0027] Preferably, the sliding and flipping mechanism comprises a linear guide rail two, a sliding block two, a linear motor two, a vertical rotating bearing, a powder conveying pipe, a motor output shaft, a coupling, a speed reducer input shaft, a speed reducer, a speed reducer output shaft, a rotating bearing sleeve, a horizontal rotating bearing, a universal joint, a laser generator and a motor two, the linear guide rail two is arranged on the rear crossbeam, the linear guide rail two and the sliding block two are slidably connected with each other, the linear motor two is installed on the rear crossbeam, and the linear motor two is drivingly connected with the sliding block two. The motor two is installed on the sliding block two, and the motor two, the coupling, the speed reducer, the horizontal rotating bearing, the universal joint, the vertical rotating bearing and the laser generator are drivingly connected.

[0028] Preferably, the motor two transmits power to the coupling through the motor output shaft, the coupling is matched with the speed reducer input shaft, the speed reducer output shaft is matched with the horizontal rotating bearing through the rotating bearing sleeve, the horizontal rotating bearing is matched with the universal joint, the universal joint is connected with the vertical rotating bearing, and the vertical rotating bearing is connected with the laser generator.

[0029] An electronic device comprises at least one processor, at least one memory and a communication interface. The processor, the memory and the communication interface communicate with each other. The memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the stress prediction method for monitoring the large component additive manufacturing process.

[0030] Compared with the prior art, the present application has the following beneficial effects:

[0031] 1. The digital twin of the present invention is based on the real-time morphology obtained by real-time scanning with a sliding line laser head, which corresponds to the actual deposition process.

[0032] 2. The digital twin proposed in this invention calculates stress by scanning and calculating the height difference between the height of the molten pool during formation and the height of the deposited layer after cooling using a sliding line laser head. This difference is used as a surface displacement load to calculate the stress field results, which is different from traditional thermo-mechanical coupling and significantly improves computational efficiency.

[0033] 3. The numerical simulation strategy proposed in this invention divides the completed printing area into several small regions, performs pooling partition simulation, and applies deep learning to the simulation results. This not only reduces the numerical simulation time for large components but also enables stress prediction for unprinted areas.

[0034] 4. This invention takes into account the error between the simulation results and the measurement results of the experimental stress sensor equipment, and processes the results that exceed the error. The average of the two results is then taken, so that the stress value assigned to the digital twin is more accurate and better simulates the real printing process. Attached Figure Description

[0035] Figure 1 This is a front view of a large component gantry printing device.

[0036] Figure 2 This is a top view of a large component gantry printing device.

[0037] Figure 3 It is a rotatable laser head.

[0038] Figure 4 It is a sliding line laser head.

[0039] Figure 5 This is a side view of the linear guide slider of the front crossbeam.

[0040] Figure 6 It is a digital twin flowchart.

[0041] Figure 7 This is a flowchart of the numerical simulation. Detailed Implementation

[0042] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0043] A large component gantry printing device, such as Figure 1 ,2 As shown, including a reversible laser head 1, an infrared thermal imager 2, a base station 3, a slidable line laser head 4, a gantry column 6, a base fixing shaft 7, a control console 9, a printing base station 12, a rear crossbeam 13, a front crossbeam 14, wherein:

[0044] The rear crossbeam 13 and the front crossbeam 14 are installed above the gantry column 6, the gantry column 6 is loaded with a movable shaft 5, the base fixing shaft 7 is installed on the movable shaft 5, the base station 3 is fixed on the base fixing shaft 7, the printing base station 12 is arranged on the base station 3, and the base fixing shaft 7 is loaded with a movable shaft driving motor 8 and a stress sensor 15, facilitating the up and down movement of the base station 3.

[0045] The reversible laser head 1 is loaded on the rear crossbeam 13 through a sliding and flipping mechanism, the infrared thermal imager 2 is installed on the reversible laser head 1, and the infrared thermal imager 2 measures the temperature of the deposited layer.

[0046] The slidable line laser head 4 is installed on the front crossbeam 14 through a sliding mechanism. The slidable line laser head 4 is used to measure the height of the molten pool and the height of the deposited layer after cooling, and the height difference is obtained through the computer, and then the height displacement difference generated in the process of the molten pool cooling and solidification is obtained, and the displacement difference is transmitted to the computer control console, and the displacement difference is loaded as the surface displacement load on the three-dimensional real-time model for numerical simulation to obtain the stress generated in the process of the molten pool cooling and shrinkage

[0047] The reversible laser head 1, the infrared thermal imager 2, the slidable line laser head 4, and the stress sensor 15 are connected with the control console 9.

[0048] In another embodiment, as shown in Figure 4 , 5 The sliding mechanism includes a linear guide rail one 261, a sliding block one 271, and a linear motor one 161, the linear guide rail one 261 is installed on the front crossbeam 14, the linear guide rail one 261 and the sliding block one 271 are slidably connected with each other, the linear motor one 161 is installed on the front crossbeam 14, and the linear motor one 161 is drivingly connected with the sliding block one 271, realizing the left and right movement of the movable line laser head 4 on the front crossbeam 14.

[0049] In another embodiment, as shown in Figure 3As shown, the sliding turnover mechanism includes linear guide rail two 262, slider two 272, linear motor two, vertical rotating bearing 10, powder conveying pipe 11, motor output shaft 17, shaft coupling 18, speed reducer input shaft 19, speed reducer 20, speed reducer output shaft 21, rotating bearing bushing 22, horizontal rotating bearing 23, universal joint 24, laser generator 25, motor two 28, the linear guide rail two 262 is arranged on the rear cross beam 13, the linear guide rail two 262 and the slider two 272 are slidably connected with each other, the linear motor two 162 is installed on the rear cross beam 13, and the linear motor two 162 is drivingly connected with the slider two 272, realizing the left and right movement of the reversible laser head 1 on the rear cross beam 13. In another embodiment, the motor two 28 is installed on the slider two 272, and the motor two 28, the shaft coupling 18, the speed reducer 20, the horizontal rotating bearing 23, the universal joint 24, the vertical rotating bearing 10, and the laser generator 25 are drivingly connected. The motor two 28 transmits power to the shaft coupling 18 through the motor output shaft 17, the shaft coupling 18 cooperates with the speed reducer input shaft 19, the speed reducer output shaft 21 cooperates with the horizontal rotating bearing 23 through the rotating bearing bushing 22, the horizontal rotating bearing 23 cooperates with the universal joint 24, the universal joint 24 is connected with the vertical rotating bearing 10, and the vertical rotating bearing 10 is connected with the laser generator 25. The powder conveying pipe 11 conveys powder to the laser head, realizing the rotary conversion of the reversible laser head 1 at any angle.

[0050] The slidable linear laser head 4 is used to measure the height of the molten pool and the height of the deposited layer after cooling, and the height difference is obtained through the computer, and then the height displacement difference generated in the process of the molten pool cooling and solidification is obtained. The displacement difference is transmitted to the computer console 9, and the displacement difference is loaded as the surface displacement load on the three-dimensional real-time model for numerical simulation to obtain the stress generated in the molten pool cooling and shrinkage process.

[0051] A digital twin method of a large component additive manufacturing process, according to the slidable linear laser head 4, generates a three-dimensional real-time model by measuring the size and shape of the molten pool and the size and shape of the deposited layer in real time, and performs numerical simulation on the three-dimensional real-time model to obtain the temperature field and the stress field. The temperature measured by the infrared thermal imager 2 and the stress measured by the stress sensor 15 are subtracted from the simulated temperature and stress. If the error of the two results is within the relative error range, the simulation result is input into the three-dimensional real-time model; if it is not within the range error, the simulation result and the result measured by the stress sensor 13 are processed by taking the average value, and input into the three-dimensional real-time model, thereby forming the digital twin of the real-time deposited layer, as shown in Figure 6 Specifically, the method comprises the following steps:

[0052] Step 1, in the large component additive manufacturing process, the time sequence images M(t) of the deposition process are obtained by scanning with a slidable line laser head, which records the topography height h(x, y, z, t) of the deposit, x, y, z are spatial coordinates, and t is time.

[0053] Step 2, a three-dimensional real-time model corresponding to the time sequence images M(t) is constructed.

[0054] Step 3, in order to simulate the stress distribution in the deposition process, the topography height h(x, y, z) of the initial molten pool formation is defined, and the topography height h(x, y, z) after the molten pool cools and solidifies is obtained. initial Step 4, the surface displacement difference Ah(x, y, z) in the process from molten pool formation to cooling and solidification is obtained according to the topography height h(x, y, z) of the initial molten pool formation and the topography height h(x, y, z) after the molten pool cools and solidifies. final initial final initial final

[0055] Step 4, the surface displacement difference Ah(x, y, z) in the process from molten pool formation to cooling and solidification is obtained according to the topography height h(x, y, z) of the initial molten pool formation and the topography height h(x, y, z) after the molten pool cools and solidifies.

[0056] Numerical simulation process: according to the size of the laser power, the temperature field of the deposition layer is obtained by numerical simulation. According to the surface displacement difference obtained by scanning and calculation of the slidable line laser head, the surface displacement difference is taken as the displacement load, and the displacement load is taken as the boundary condition to load on the model for numerical simulation to obtain the stress field result.

[0057] ​​​​​In the numerical simulation of large components, a numerical simulation strategy: according to the real-time deposition layer topography obtained by the slidable line laser head scanning, a three-dimensional real-time model is established, and the printing area is preliminarily divided into several parts according to the model size. The first pooling partition is selected according to the printing path, and the stress field of the first pooling partition is simulated according to the above numerical simulation process. The results obtained are compared with the stress results measured by the stress sensor. If the results are within the error range, the simulation results are imported into the real-time model; if not, the pooling partition is optimized, and the simulation results and the stress sensor measurement results are averaged. The results of the average processing are imported into the real-time model. Then, the second pooling partition is selected according to the printing path, and the numerical simulation results of the first pooling area are used as the boundary conditions of the second pooling area for numerical simulation. Similarly, whether the obtained stress results and the stress sensor measurement results are within the error range is considered. If they are within the error range, the simulation results are imported into the real-time model; if not, the simulation results and the stress sensor measurement results are averaged. The results of the average processing are imported into the real-time model. According to the printing path, the pooling partition simulation is carried out in turn. If there are multiple completed simulation pooling partitions in the selected pooling area, all the completed simulation results should be considered and simulated together as boundary conditions. The numerical simulation results can be learned deeply to predict the stress size of large components.

[0058] Step 5, obtain the temperature field and stress field by numerical simulation, and compare the temperature measured by the infrared thermal imager and the stress measured by the stress sensor with the numerical simulation temperature and stress. If the error of the two results is within the error range, take the simulation results as input to the three-dimensional real-time model. If not, take the simulation results and the stress sensor measurement results to do average processing, input the three-dimensional real-time model, and form the digital twin of the real-time molten pool and the deposition layer.

[0059] A stress prediction method for monitoring the additive manufacturing process of large components, which adopts the digital twin method of the additive manufacturing process of large components, and calculates the stress size generated in the molten pool cooling and shrinkage process by numerical simulation, as shown in Figure 7 , comprising the following steps:

[0060] Step 41, according to the size of the established three-dimensional real-time model, the printing area A is preliminarily divided into several parts, and the three-dimensional real-time model is processed by pooling partition. Each pooling partition is marked as P i .

[0061] Step 42, the first pooling partition P1 is selected according to the printing path, and the stress field of the first pooling partition P1 is simulated according to the numerical simulation of the deposition process. The obtained stress is marked as S 1,sim, the stress measured by the stress sensor 15 is denoted as S 1,meas , the stress obtained by numerical simulation is denoted as S 1,sim , the stress measured by the stress sensor 15 is denoted as S 1,meas , the absolute difference is denoted as AS1, AS1 = |S 1,sim -S 1,meas |.

[0062] Step 43, determine the dynamic adjustment error limit ε i , i = 1, 2, 3, …, n, n is the total number of pooling partitions.

[0063] The dynamic adjustment error limit is determined as follows:

[0064] ε i = α + β · SD(S prev )

[0065] In the formula, ε i is the dynamic adjustment error limit, i represents the ith pooling partition, α is the basic error tolerance, β is the adjustment coefficient, and SD(S prev ) is the standard deviation of the stress measured in the previous pooling partition.

[0066] In the first pooling partition, if AS1 ≤ ε1, the stress S 1,sim obtained by numerical simulation is input to the corresponding part of the three-dimensional real-time model. If AS1 > ε1, the mean value of the two data is obtained by mean value processing, and the mean value processed stress result is The mean value processed stress result S 1,avg is input to the corresponding part of the three-dimensional real-time model. According to the print path, the second pooling partition P2 is selected, and the numerical simulation stress result of the first pooling area is taken as the boundary condition load stress S bod of the second pooling area, and then the solution of the second pooling partition is performed in the same way as the first pooling partition. According to the print path, the pooling partition simulation is performed in turn.

[0067] It should be noted that if there are multiple completed numerical simulation pooling partitions in the selected pooling partition, all completed numerical simulation stress results should be considered, and the sum of all completed numerical simulation stress results is denoted as cumulative stress S sum as the boundary condition of the next area, S sum,i represents the boundary condition of the ith pooling partition. After completing all partition simulations, the sum is obtained to obtain the actual total stress size of the entire three-dimensional real-time model Using the obtained stress S sum , the deep learning model is trained to predict the stress of the unprinted area.

[0068] An electronic device is provided in another embodiment, comprising: at least one processor, at least one memory, and a communication interface. The processor, memory, and communication interface communicate with each other. The memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the stress prediction method for large component additive manufacturing process monitoring.

[0069] The above merely describes the preferred embodiments of the present application, and it should be noted that those skilled in the art can make several improvements and refinements without departing from the principles of the present application, and these improvements and refinements should also be considered within the protection scope of the present application.

Claims

1. A method for digital twinning of a large component additive manufacturing process, characterized in that, It comprises the following steps: Step 1: In the large component additive manufacturing process, the time sequence scanning is carried out by the slidable line laser head (4) to obtain the time-dependent image of the deposition process, denoted as time sequence image M(t), which records the topographic height h(x, y, z, t) of the deposit, x, y, z are spatial coordinates, and t is time; Step 2: According to the time sequence image M(t), a three-dimensional real-time model corresponding to it is constructed; Step 3, obtaining the height h of the molten pool at the initial stage of formation from the time series images M(t) initial (x,y,z) and the height h of the molten pool after cooling and solidification final (x,y,z) from the height h of the molten pool at the initial stage of formation initial (x,y,z) and the height h of the molten pool after cooling and solidification final (x,y,z) to obtain the surface displacement difference Ah(x,y,z) during the process from the formation of the molten pool to the cooling and solidification Step 4: The surface displacement difference Δh(x, y, z) in the cooling and solidification process is taken as the surface displacement load u(x, y, z), and the surface displacement load u(x, y, z) is used as input to the three-dimensional real-time model for numerical simulation, and the stress generated in the cooling and shrinkage process of the molten pool is calculated by numerical simulation; Step 5: The temperature field and stress field are obtained by numerical simulation, and the temperature measured by the infrared thermal imager and the stress measured by the stress sensor are compared with the numerical simulation temperature and stress, if the error between the two results is within the error range, take the simulation result as input to the three-dimensional real-time model; if it is not within the range of error, take the simulation result and the result measured by the stress sensor to do mean value processing, input to the three-dimensional real-time model, form the digital twin of the real-time molten pool and the deposited layer.

2. A stress prediction method for large component additive manufacturing process monitoring, characterized in that, The digital twin method of the large component additive manufacturing process of claim 1, the method for calculating the stress generated in the cooling and shrinkage process of the molten pool by numerical simulation, comprising the following steps: Step 41, according to the established three-dimensional real-time model size, the printing area A is preliminarily divided into several parts, and the three-dimensional real-time model is processed by pooling partition, and each pooling partition is marked as P i ; Step 42, according to the printing path, the first pooling partition P1 is selected, and the stress field of the first pooling partition P1 is simulated according to the deposition process to obtain the stress S 1,sim , the stress measured by the stress sensor (15) is recorded as S 1,meas , the stress S 1,sim obtained by numerical simulation is compared with the stress S 1,meas measured by the stress sensor (15) to obtain the absolute difference ΔS1; Step 43, determine the dynamic adjustment error limit ε i Let i = 1, 2, 3, ..., n, where n is the total number of pooling partitions; within the first pooling partition, if ΔS1 ≤ ε1, then the stress S obtained from the numerical simulation is... 1,sim Input the data into the corresponding part of the 3D real-time model; if ΔS1>ε1, perform mean processing on the two data to obtain the mean-processed stress result, which is: The stress result S after mean value processing 1,avg Input the data into the corresponding part of the 3D real-time model; select the second pooling partition P2 according to the printing path, and input the numerical simulation stress results of the first pooling region. The boundary condition load stress S of the second pooling region bod Then, the second pooling partition is solved in the same way as the first pooling partition; the pooling partition simulation is performed sequentially according to the printing path.

3. The stress prediction method for monitoring of the large component additive manufacturing process according to claim 2, characterized in that, If there are multiple completed numerical simulation of the selected pooled partition, all completed numerical simulation stress results should be considered, and all completed numerical simulation stress results are summed up as cumulative stress S sum As the boundary condition of the next region, S sum,i denotes the i-th partition boundary condition; after all the partition simulations are completed, the actual total size of the entire three-dimensional real-time model can be obtained by summing up 4. The stress prediction method for monitoring of the large component additive manufacturing process according to claim 3, characterized in that, Using the stress S that has been obtained sum Training a deep learning model to predict the stress of unprinted areas.

5. The stress prediction method for monitoring of the large component additive manufacturing process according to claim 4, characterized in that, The dynamic adjustment error limit is determined as: ε i = a + b - SD(S prev ) where ε i is the dynamic adjustment error limit, i represents the i-th pooled partition, a is the base error tolerance, β is the adjustment coefficient, SD(S prev ) is the standard deviation of the previous pooled partition stress measurements.

6. A gantry printing apparatus for large components, characterized by: The stress prediction method for monitoring the large component additive manufacturing process of claim 2, comprising a reversible laser head (1), an infrared thermal imager (2), a base table (3), a slidable line laser head (4), a gantry column (6), a base fixed shaft (7), a control console (9), a printing base table (12), a rear cross beam (13), and a front cross beam (14), wherein: The rear cross beam (13) and the front cross beam (14) are installed above the gantry column (6), the gantry column (6) is loaded with an up-down movable shaft (5), the base fixed shaft (7) is installed on the up-down movable shaft (5), the base table (3) is fixed on the base fixed shaft (7), the printing base table (12) is arranged on the base table (3), and the base fixed shaft (7) is loaded with an up-down movable shaft driving motor (8) and a stress sensor (15); The reversible laser head (1) is loaded on the rear cross beam (13) through a sliding and flipping mechanism, and the infrared thermal imager (2) is installed on the reversible laser head (1); The slidable line laser head (4) is installed on the front cross beam (14) through a sliding mechanism; The reversible laser head (1), the infrared thermal imager (2), the slidable line laser head (4), and the stress sensor (15) are connected with the control console (9).

7. The gantry printer for printing large components of claim 6, wherein: The sliding mechanism comprises a linear guide rail one (261), a sliding block one (271) and a linear motor one (161), the linear guide rail one (261) is installed on the front cross beam (14), the linear guide rail one (261) is slidably connected with the sliding block one (271), the linear motor one (161) is installed on the front cross beam (14), and the linear motor one (161) is drivingly connected with the sliding block one (271).

8. The gantry printer for printing large components of claim 7, wherein: The sliding and overturning mechanism comprises a linear guide rail two (262), a sliding block two (272), a linear motor two, a vertical rotary bearing (10), a powder conveying pipe (11), a motor output shaft (17), a shaft coupling (18), a speed reducer input shaft (19), a speed reducer (20), a speed reducer output shaft (21), a rotary bearing sleeve (22), a horizontal rotary bearing (23), a universal joint (24), a laser generator (25) and a motor two (28), the linear guide rail two (262) is arranged on the rear cross beam (13), the linear guide rail two (262) is slidably connected with the sliding block two (272), the linear motor two (162) is installed on the rear cross beam (13), and the linear motor two (162) is drivingly connected with the sliding block two (272); the motor two (28) is installed on the sliding block two (272), and the motor two (28), the shaft coupling (18), the speed reducer (20), the horizontal rotary bearing (23), the universal joint (24), the vertical rotary bearing (10) and the laser generator (25) are drivingly connected.

9. The gantry printer for printing large components of claim 8, wherein: The motor two (28) transmits power to the shaft coupling (18) through the motor output shaft (17), the shaft coupling (18) cooperates with the speed reducer input shaft (19), the speed reducer output shaft (21) cooperates with the horizontal rotary bearing (23) through the rotary bearing sleeve (22), the horizontal rotary bearing (23) cooperates with the universal joint (24), the universal joint (24) is connected with the vertical rotary bearing (10), and the vertical rotary bearing (10) is connected with the laser generator (25).

10. An electronic device, comprising: Comprise: At least one processor, at least one memory and a communication interface; the processor, the memory and the communication interface communicate with each other; The memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the stress prediction method of the large component additive manufacturing process monitoring according to any one of claims 2-5.

Images