Regulation and control method and regulation and control system for melt forming precision of polymer laser powder bed and method for improving forming precision

Abstract

The invention discloses a method and a system for regulating and controlling melting forming precision of a polymer laser powder bed and a method for improving the forming precision, and belongs to the technical field of additive manufacturing. Comprising the following steps: acquiring temperature distribution data in a polymer LPBF process; constructing a database based on the temperature distribution data and process parameters adopted for printing the current layer; based on machine learning, a polymer LPBF technological process accurate regulation and control model is established; training of the polymer LPBF technological process accurate regulation and control model is completed through the database, the mapping relation of technological parameters, the printing layer thickness and layer-by-layer temperature field distribution is established, a negative feedback mechanism of the technological parameters and the printing layer real-time temperature field is obtained, and accurate regulation and control of polymer laser powder bed fusion forming precision are completed. According to the method, the excellence of the reinforcement learning algorithm in the aspect of big data processing is utilized, the heat accumulation phenomenon in the forming process is improved, the defects of warping, caking and the like are eliminated, and the method has guiding significance for improving the precision and performance of polymer material LPBF process forming parts.

Description

Technical Field

[0001] The present invention belongs to the technical field of additive manufacturing, and relates to a method for regulating the precision of polymer laser powder bed fusion forming, a regulation system, and a method for improving the forming precision. Specifically, it relates to a method for improving the heat accumulation effect in the laser powder bed fusion process of polymer materials based on machine learning, reducing thermal warping deformation, and improving forming precision. Background Art

[0002] Laser Powder Bed Fusion (LPBF) is an advanced additive manufacturing method based on the principle of powder bed fusion. It uses a powder feeding mechanism to evenly lay a layer of powder materials such as metal and polymer on the forming platform. A high-power laser is used to scan and melt the powder according to the path planned by the software. The molten powder solidifies during the cooling process. The powder bed then moves down a layer thickness. The powder feeding mechanism re-spreads the powder and repeats the laser scanning, melting, cooling and solidification processes until the three-dimensional component is formed. Compared with other manufacturing technologies, the LPBF process has many advantages, including high forming efficiency, excellent part strength and wide applicability of forming materials. It can also manufacture parts with complex shapes without the need to add additional support structures. Therefore, it is widely used in aerospace, automotive, medical and other fields, providing new solutions for fast, flexible and efficient manufacturing, while also promoting the development of customization and innovative design.

[0003] The chemical structure of polymer materials is mostly linear or cross-linked. They have high specific strength, excellent processing performance, low thermal conductivity, good chemical stability, good electrical insulation, and good decorative performance. With the improvement of people's awareness of environmental protection and the growing demand for resources and the environment, polymer material components prepared by LPBF process have gradually become one of the most ideal components in various fields, and have broad application prospects in electronics, aerospace, automobiles, biomedicine and other fields. However, how to improve the forming accuracy of polymer LPBF is still a major problem. The main reason for the insufficient precision of formed components is the defects such as part warping and powder agglomeration during the polymer LPBF process. Among them, the real-time temperature field of the printing layer is the most direct factor affecting warping and agglomeration, and the key factors affecting the real-time temperature field of the printing layer are the heat accumulation effect and the real-time process parameters: in the early stage of printing, the laser heats and melts the powder on the surface of the formed part, and the heat of the formed part and the powder near it increases; as the printing progresses, the heat absorbed by the formed part in the early stage is gradually released; when a new layer of powder is laid, the powder accumulated in the forming cylinder will transfer heat to the newly laid powder, causing the temperature of the newly laid powder layer to be processed to rise rapidly, so that the processed solid part actually forms a new heat source for the processed powder - the bottom heat source, causing heat accumulation. The existing polymer LPBF forming process cannot realize the adaptive regulation of process parameters during printing, and the heat accumulation phenomenon in the process cannot be solved, resulting in uneven distribution of the real-time temperature field of the printing layer during the polymer LPBF process, and then defects such as part warping and powder agglomeration occur, which ultimately leads to insufficient precision and performance consistency of the formed part.

[0004] Therefore, it is very necessary to improve the heat accumulation effect of polymer materials in the LPBF process and realize the adaptive control of real-time process parameters, which can not only effectively improve the precision and performance of the formed parts, but also has important significance for the stable printing and practical application of engineering components of high-melting-point polymer materials such as PEEK, PEAK, PU and other high-melting-point polymers and their composite materials. However, the current control methods for heat accumulation are mostly limited to the control of the temperature range of the forming cavity and the uniformity of the preheating temperature field, and the research focusing on the precise control of the real-time temperature field of the printing layer during the forming process is far from enough. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a method for controlling the precision of polymer laser powder bed fusion forming, a control system and a method for improving the forming precision, so as to solve the current technical problem that the real-time temperature field of the printing layer in the polymer LPBF process cannot be accurately controlled, which leads to defects such as warping and agglomeration in the forming process, thereby achieving high-precision and high-performance printing of the polymer LPBF process.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention discloses a method for controlling the precision of polymer laser powder bed melting forming, comprising the following steps:

[0008] S1: Obtain temperature distribution data during polymer LPBF process;

[0009] S2: Building a database based on the temperature distribution data and the process parameters used to print the current layer;

[0010] S3: Establish a precise control model for the polymer LPBF process based on machine learning;

[0011] S4: Use the database to complete the training of the precise control model of the polymer LPBF process, establish the mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, obtain the negative feedback mechanism between process parameters and real-time temperature field of the printing layer, and complete the precise control of the polymer laser powder bed fusion forming accuracy.

[0012] Preferably, in S1, the temperature distribution data includes the thickness of the printed layer and the temperature distribution state of each pixel point of the real-time printed layer of the formed part;

[0013] According to the difference between the temperature distribution state of each pixel point on a single laser scanning path during the printing process and the target temperature distribution state, the laser energy density and the lamp power are adjusted in real time. Among them, the laser energy density is jointly controlled by the laser power, scanning speed and scanning spacing.

[0014] Preferably, in S2, the process parameters include preheating temperature, lamp power, laser power, scanning speed and scanning interval.

[0015] Preferably, in S3, the machine learning adopts a reinforcement learning algorithm, and its database is composed of multiple sets of episodes, wherein each set of episodes is defined as the entire process of completing a single-channel laser scanning.

[0016] Further preferably, in S3, the established polymer LPBF process precise control model includes an environmental state function S, a model action range A, an environmental reward R and a value function Q; the environmental state function S, the model action range A, and the environmental reward R are independent of each other, and the value function Q is related to the first three functions;

[0017] The environmental state function is S=f[T(x,y,z)], where x and y are used to determine the position of each pixel point of the real-time printing layer of the formed part during the printing process, z is the current printing layer thickness, and T(x,y,z) represents the temperature distribution state of the real-time printing layer of the formed part;

[0018] The action range A of the model is defined as a set of process parameters that can be adjusted in real time, including lamp power, laser power, scanning speed, and scanning spacing;

[0019] The reward R of the environment is the difference between the temperature distribution state of each pixel in the next layer and the target temperature distribution state after completing a single laser scan;

[0020] The value function Q is Q π (s, a) = E π [r(s,a)+γQ π (s′, a′)], where: π represents the strategy executed by the individual, that is, when the temperature distribution state of the printed layer of the molded part is s, the probability of adjusting a certain controllable process parameter in the action range A to a, which is used to explore better parameters; s′ represents the temperature distribution state of the next layer of the molded part after performing operation a; a′ is the process parameter that needs to be adjusted in the next state s′; r(s, a) is the reward observed in state s after performing operation a; y is the attenuation factor, which is used to control the weight of the current reward and the subsequent state reward; E π It represents the expected value of the value function, which is used to measure the size of the theoretical reward and provide a basis for the model to select subsequent actions.

[0021] Further preferably, the reward maximization characteristic of the value function is utilized to sum the absolute errors between the temperature distribution on a single laser scanning path and the target temperature distribution as a decreasing term of the value function.

[0022] Preferably, in S4, a negative feedback mechanism is provided between the process parameters and the real-time temperature field of the printing layer.

[0023] Further preferably, the data in the database is distributed in a ratio of nine to one, where ninety percent of the data is used as a training set and the remaining tenth of the data is used as a validation set. The PPO (Proximal Policy Optimization) algorithm is used to iteratively optimize the data, and the excellence of the algorithm in policy selection is used to complete the training of the model. After the model is trained, the validation set is used to test the model to ensure the accuracy of the model, and the model is processed by L2 regularization to optimize the weights of each feature data in the environmental state function S to prevent the model from being overly complex and distorted due to overfitting problems.

[0024] Preferably, the polymer includes one or more of polyaryletherketone, polyetherketone, polyetherketoneketone, polyetheretherketoneketone, polyimide, polycarbonate, polyethylene, polypropylene, polystyrene and nylon.

[0025] The present invention also discloses a control system based on the above-mentioned control method for polymer laser powder bed fusion forming accuracy, comprising:

[0026] Data acquisition module, used to obtain temperature distribution data during the polymer LPBF process;

[0027] A database construction module is used to construct a database based on the temperature distribution data and the process parameters used for printing the current layer;

[0028] The control model building module is used to establish a precise control model for the polymer LPBF process based on machine learning;

[0029] The model training module is used to use the database to complete the training of the precise control model of the polymer LPBF process, establish the mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, and obtain the negative feedback mechanism between process parameters and the real-time temperature field of the printing layer.

[0030] Preferably, it includes: building a precise control model of the polymer LPBF process on the LPBF equipment, improving the heat accumulation effect in the polymer LPBF process by real-time control of printing process parameters, reducing the warping deformation of the polymer material during the forming process, and improving the precision of the formed parts.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention discloses a method for regulating the precision of polymer laser powder bed fusion forming. On the one hand, the present invention constructs a database based on the temperature distribution data obtained in the polymer LPBF process and the process parameters adopted for printing the current layer, and then fully trains the precise control model of the polymer LPBF process established in the machine learning process through the database, so that the environmental state function obtained can more accurately map the influence of the heat accumulation effect on the formed part itself and the surrounding powder; on the other hand, the present invention also considers the improvement effect of the process parameters on the polymer LPBF forming precision, couples a variety of process parameters, establishes a mapping relationship between the process parameters, the printing layer thickness and the layer-by-layer temperature field distribution, obtains the negative feedback mechanism of the process parameters and the real-time temperature field of the printing layer, and then realizes the adaptive regulation of the LPBF forming process parameters, improves the heat accumulation effect in the forming process, and finally improves the LPBF process forming precision of the polymer material.

[0033] Furthermore, the environmental state function in the polymer LPBF process described in the present invention is obtained by extracting the temperature distribution state of each pixel point of the real-time printing layer of the formed part during the process by an infrared thermal imager. Compared with directly extracting the overall temperature inside the forming cavity, the environmental state function obtained by this method can more accurately map the influence of the thermal accumulation effect on the formed part itself and the surrounding powder.

[0034] Furthermore, in the database for training the model described in the present invention, each group of episodes is defined as the entire process of completing a single-channel laser scan. A large number of data samples can be efficiently collected in a single data collection experiment, avoiding the large amount of repetitive and cumbersome data collection process required for general machine learning.

[0035] Furthermore, the present invention takes into account the active exploration and optimization of strategies and defines an individual execution strategy π in the model, wherein the model has a probability π of selecting a new strategy within the process window while adopting the optimal strategy to explore better possibilities. The model can continuously improve the improvement effect on the polymer LPBF forming accuracy over time.

[0036] Furthermore, the present invention takes into account the overfitting problem that may exist in algorithm learning and introduces a regularization term in the value function. The term is defined as the difference between the lowest and highest temperatures of each pixel observed during an episode. This avoids the abnormal strategy in which model selection will cause a sudden increase in the temperature of the printed layer of the molded part during the process, thereby improving the safety performance of the model.

[0037] Furthermore, the heat received by the real-time printing layer of the formed part during the scanning process is regulated by adaptively controlling the laser energy density. The laser energy density is controlled by the laser power, scanning speed and scanning interval. The overall temperature inside the forming cavity is regulated by adaptively controlling the lamp power. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of the control method of the present invention;

[0039] Figure 2 A flow chart of a method for improving the precision of polymer laser powder bed fusion forming according to the present invention;

[0040] Figure 3 This is a schematic diagram of the thermal accumulation effect in an embodiment of the present invention;

[0041] Figure 4 It is a scatter diagram of the change of the internal temperature of the forming cavity during the printing process in the embodiment of the present invention;

[0042] Figure 5 This is a morphology diagram of the initial formed part printed in the embodiment of the present invention;

[0043] Figure 6 Extraction principle diagram for each episode of the database in the embodiment of the present invention;

[0044] Figure 7 This is a morphology diagram of a formed part after printing 10 layers to form a bottom heat source in an embodiment of the present invention;

[0045] Figure 8This is a morphology diagram of the final formed part obtained in an embodiment of the present invention;

[0046] Fig. 9 This is a morphology diagram of the final formed part obtained in the comparative example of the present invention. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0048] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0049] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0050] See also Figure 1 The present invention discloses a method for controlling the precision of polymer laser powder bed fusion forming, which comprises the following steps:

[0051] S1: Obtain temperature distribution data during polymer LPBF process;

[0052] S2: Building a database based on the temperature distribution data and the process parameters used to print the current layer;

[0053] S3: Establish a precise control model for the polymer LPBF process based on machine learning;

[0054] S4: Use the database to complete the training of the precise control model of the polymer LPBF process, establish the mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, obtain the negative feedback mechanism between process parameters and real-time temperature field of the printing layer, and complete the precise control of the polymer laser powder bed fusion forming accuracy.

[0055] The present invention is based on the control system of the above-mentioned polymer laser powder bed fusion forming precision control method, comprising:

[0056] Data acquisition module, used to obtain temperature distribution data during the polymer LPBF process;

[0057] A database construction module is used to construct a database based on the temperature distribution data and the process parameters used for printing the current layer;

[0058] The control model building module is used to establish a precise control model for the polymer LPBF process based on machine learning;

[0059] The model training module is used to use the database to complete the training of the precise control model of the polymer LPBF process, establish the mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, and obtain the negative feedback mechanism between process parameters and the real-time temperature field of the printing layer.

[0060] Furthermore, the present invention adopts the above-mentioned control system to improve the method for the precision of polymer laser powder bed fusion forming, including: building a precise control model of the polymer LPBF process on the LPBF equipment, improving the heat accumulation effect in the polymer LPBF process by real-time control of printing process parameters, reducing the warping deformation of the polymer material during the forming process, and improving the precision of the formed parts.

[0061] Specifically, see Figure 2 , the control strategy adopted by the present invention is as follows:

[0062] According to the difference between the temperature distribution state of each pixel point on a single laser scanning path during the printing process and the target temperature distribution state, the laser energy density and the lamp power are adjusted in real time. Among them, the laser energy density is jointly controlled by the laser power, scanning speed and scanning spacing.

[0063] Preferably, the algorithm used is a reinforcement learning algorithm; the language used is Python; and the software used is Pycharm.

[0064] Preferably, the polymer material can be one or a mixture of polyaryletherketone (PEAK), polyetherketone (PEK), polyetherketoneketone (PEKK), polyetheretherketoneketone (PEEKK), polyimide (PI), polycarbonate (PC), polyethylene (PE), polypropylene (PP), polystyrene (PS), and nylon (PA).

[0065] Preferably, the database is composed of multiple episodes, wherein each episode is defined as the entire process of completing a single-pass laser scanning.

[0066] Preferably, the environmental state function is S=f[T(x, y, z)], wherein x and y are used to determine the position of each pixel point of the printed layer of the formed part during the printing process, z is the current printing layer thickness, and T(x, y, z) represents the temperature distribution state of the printed layer of the formed part.

[0067] Preferably, the action range A of the model is defined as a set of process parameters that can be controlled in real time, including lamp power, laser power, scanning speed, and scanning spacing.

[0068] Preferably, the reward R of the environment is defined as the difference between the temperature distribution state of each pixel point in the next layer and the target temperature distribution state after completing a single laser scan.

[0069] Preferably, the value function is Q π (s, a) = E π [r(s,a)+γQ π (s′, a′)], where π represents the strategy executed by the individual, that is, when the temperature distribution state of the printed layer of the molded part is s, the probability of adjusting a controllable process parameter within the action range A to a, which is used to explore better parameters; s′ refers to the temperature distribution state of the next layer of the molded part after performing operation a; a′ refers to the process parameter that needs to be adjusted in the next state s′; r(s, a) is the reward observed in state s after performing operation a; y is the attenuation factor, which is used to control the weight of the current reward and the reward of the subsequent state; Eππ represents the expected value of the value function, which is used to measure the size of the theoretical reward and give the model a basis for selecting subsequent actions.

[0070] Preferably, the reward maximization property of the value function is utilized, and the absolute errors between the temperature distribution on a single laser scanning path and the target temperature distribution are summed as the decreasing term of the value function.

[0071] Preferably, a regularization term is designed in the value function to reduce the variance in the model error and solve the overfitting problem.

[0072] The following embodiment of the present invention takes 10% h-BN / PEEK composite material powder as an example to carry out relevant forming precision control experiments. The melting point of the composite material powder is 340°C. The thermal accumulation effect is significant during the LPBF process, and the precision of the formed component is significantly affected. The specific preparation method includes the following steps:

[0073] Use a dryer to dry the PEEK powder at 120°C for 12 hours to ensure that the PEEK powder is dry. Select a 120-mesh screen, pour the PEEK powder evenly into the screen, place it on a vibrating sieving machine and work for 300 seconds before taking it out to obtain PEEK fine powder; follow the above process again to obtain h-BN fine powder; weigh 2700 grams of PEEK powder and 300 grams of h-BN powder respectively, mix them preliminarily and pour them into a powder mixer, mix them for 3 hours and take them out to obtain 10% h-BN / PEEK composite material powder. The prepared powder should be sealed and stored in a dry, light-proof environment to prevent moisture absorption or contamination.

[0074] Example 1

[0075] 1) A standard LPBF forming experiment was carried out using 10% h-BN / PEEK composite material powder, and an infrared thermal imager was used to obtain temperature distribution data during the process. The process parameters include preheating temperature, laser power, scanning speed, scanning spacing, and lamp power. Among them, the preheating temperature was selected as 300℃, 310℃, and 320℃, the laser power was selected as 3W, 6W, 9W, 12W, 15W, 21W, 24W, 27W, and 30W, the scanning speed was selected as 2000mm / s, 2500mm / s, 3000mm / s, and 3500mm / s, and the scanning spacing was selected as 0.1mm, 0.2mm, and 0.3mm. The lamp power was adjusted according to the aggravation of the heat accumulation phenomenon during the experiment (see Figure 4 , which is a scatter plot of the temperature inside the molding cavity as the printing process changes). When the molded part does not warp and the surrounding powder does not agglomerate (see Figure 5 , which is the topography of the formed part when the bottom heat source is not formed at the initial stage of printing), extract the temperature distribution state corresponding to the single laser scanning path of the printed layer of the formed part as a set of data and the current corresponding process parameters to form a set of episodes (see Figure 6 , extract the schematic diagram for each episode in the database); as the printing progresses, when the formed part and the surrounding powder have defects such as warping and agglomeration (see Figure 7 , which is the morphology of the formed part after printing 10 layers to form the bottom heat source), extract the data again to form a set of episodes, repeat the above standard experiments, and complete the establishment of the database.

[0076] 2) Based on machine learning, a precise control model for the polymer LPBF process is established.

[0077] The environmental state function is S=f[T(x, y, z)], where x and y are used to determine the position of each pixel point in the printed layer of the formed part during the printing process, z is the current printing layer thickness, and T(x, y, z) represents the temperature distribution state of the printed layer of the formed part.

[0078] The action range A of the model is defined as a set of process parameters that can be controlled in real time, including lamp power, laser power, scanning speed, and scanning spacing.

[0079] The reward R of the environment is defined as the difference between the temperature distribution state of each pixel point in the next layer and the target temperature distribution state after completing a single laser scan.

[0080] The value function is Q π (s, a) = E π [r(s,a)+γQ π (s′, a′)], where π represents the strategy executed by the individual, that is, when the temperature distribution state of the printed layer of the molded part is s, the probability of adjusting a certain controllable process parameter in the action range A to a, which is used to explore better parameters; s′ refers to the temperature distribution state of the next layer of the molded part after performing operation a; a′ refers to the process parameter that needs to be adjusted in the next state s′; r(s, a) is the reward observed in state s after performing operation a; y is the attenuation factor, which is used to control the weight of the current reward and the reward of the subsequent state; Eππ represents the expected value of the value function, which is used to measure the size of the theoretical reward and give the model a basis for selecting subsequent actions. Using the characteristic of maximizing the reward of the value function, the absolute error between the temperature distribution on the single laser scanning path and the target temperature distribution is summed as the decreasing term of the value function.

[0081] 3) Use the database to complete the training of the precise control model of the polymer LPBF process, establish the mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, and construct a negative feedback mechanism of process parameters-real-time temperature field.

[0082] 4) The precise control model of the polymer LPBF process is built on the LPBF equipment. The model appropriately controls the process parameters used for the next layer of printing based on the real-time temperature distribution of the printing layer read by the infrared thermal imager after the completion of one layer of printing, and controls the temperature on the single laser scanning path within the optimal temperature window to improve the heat accumulation effect in the polymer LPBF process, and ultimately improve the precision of the formed parts ( Figure 8 ).

[0083] Comparative Example 1

[0084] The printing temperature range of the 10% h-BN / PEEK composite material powder used should be set between the initial melting temperature and the initial crystallization temperature, that is, between 302 and 320°C. Therefore, a layer of powder with a thickness of 1 mm should be pre-laid inside the forming cylinder, and the temperature inside the forming cavity should be set to 300°C for preheating for one hour. Subsequently, the laser power was set to 18W, the scanning speed to 3000mm / s, and the scanning spacing to 0.1mm. During the printing process, the temperature inside the forming cavity was kept constant at 320°C. After printing is completed, the temperature inside the forming cavity is slowly cooled to 300°C at a rate of 1°C / min, and then kept warm for half an hour before the formed part is taken out to improve the precision of the formed part ( Fig. 9 ).

[0085] By comparing the molded parts obtained by printing the same material in Example 1 and Comparative Example 1, it can be found that the precision of the molded parts obtained by the method of the present invention in Example 1 is significantly better than the precision of the molded parts obtained by direct printing in Comparative Example 1. This is because the influence of the thermal accumulation effect and the adaptive regulation of real-time process parameters are not considered in Comparative Example 1. As printing progresses, the heat of the molded part and the surrounding powder continues to increase, and the bottom heat source is gradually formed, resulting in defects such as warping of the molded part and powder agglomeration, which ultimately leads to a decrease in the precision of the molded part.

[0086] In summary, the present invention obtains the temperature distribution data in the polymer LPBF process; constructs a database based on the temperature distribution data and the process parameters used for printing the current layer; establishes a precise control model for the polymer LPBF process based on machine learning; uses the database to complete the training of the precise control model for the polymer LPBF process, establishes a mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, and constructs a negative feedback mechanism of process parameters-real-time temperature field. Furthermore, the precise control model for the polymer LPBF process constructed by the above control method is built on the LPBF equipment to achieve adaptive control of the polymer LPBF forming process parameters. The present invention can monitor and precisely control the temperature field distribution state in the polymer LPBF process in real time, utilize the excellence of the reinforcement learning algorithm in big data processing, improve the heat accumulation phenomenon in the forming process, eliminate defects such as warping and agglomeration, and has guiding significance for improving the accuracy and performance of polymer LPBF process formed parts.

[0087] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for controlling the precision of polymer laser powder bed fusion forming, characterized in that: The following steps are involved: S1: Obtain temperature distribution data during polymer LPBF process; S2: Building a database based on the temperature distribution data and the process parameters used to print the current layer; S3: Establish a precise control model for the polymer LPBF process based on machine learning; S4: Use the database to complete the training of the precise control model of the polymer LPBF process, establish the mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, obtain the negative feedback mechanism between process parameters and real-time temperature field of the printing layer, and complete the precise control of the polymer laser powder bed fusion forming accuracy.

2. The method for controlling the precision of polymer laser powder bed fusion forming according to claim 1, characterized in that: In S1, the temperature distribution data includes the thickness of the printing layer and the temperature distribution state of each pixel point of the real-time printing layer of the formed part; According to the difference between the temperature distribution state of each pixel point on a single laser scanning path during the printing process and the target temperature distribution state, the laser energy density and the lamp power are adjusted in real time. Among them, the laser energy density is jointly controlled by the laser power, scanning speed and scanning spacing.

3. The method for controlling the precision of polymer laser powder bed fusion forming according to claim 1, characterized in that: In S2, the process parameters include preheating temperature, lamp power, laser power, scanning speed and scanning spacing.

4. The method for controlling the precision of polymer laser powder bed fusion forming according to claim 1, characterized in that: In S3, machine learning uses a reinforcement learning algorithm, and its database consists of multiple sets of episodes, where each set of episodes is defined as the entire process of completing a single-channel laser scan.

5. The method for controlling the precision of polymer laser powder bed fusion forming according to claim 4 is characterized in that: In S3, the established precise control model of the polymer LPBF process includes the environmental state function S, the action range A of the model, the environmental reward R and the value function Q; The environmental state function is S=f[T(x,y,z)], where x and y are used to determine the position of each pixel point of the real-time printing layer of the formed part during the printing process, z is the current printing layer thickness, and T(x,y,z) represents the temperature distribution state of the real-time printing layer of the formed part; The action range A of the model is defined as a set of process parameters that can be adjusted in real time, including lamp power, laser power, scanning speed, and scanning spacing; The reward R of the environment is the difference between the temperature distribution state of each pixel in the next layer and the target temperature distribution state after completing a single laser scan; The value function Q is Q π (s,a)=E π [r(s,a)+γQ π (s′, a′)], where: π represents the strategy executed by the individual, that is, when the temperature distribution state of the printed layer of the molded part is s, the probability of adjusting a certain controllable process parameter in the action range A to a, which is used to explore better parameters; s′ represents the temperature distribution state of the next layer of the molded part after performing operation a; a′ is the process parameter that needs to be adjusted in the next state s′; r(s, a) is the reward observed in state s after performing operation a; γ is the attenuation factor, which is used to control the weight of the current reward and the subsequent state reward; E π It represents the expected value of the value function, which is used to measure the size of the theoretical reward and provide a basis for the model to select subsequent actions.

6. The method for controlling the precision of polymer laser powder bed fusion forming according to claim 5, characterized in that: By utilizing the reward maximization characteristic of the value function, the absolute errors between the temperature distribution on a single laser scanning path and the target temperature distribution are summed as the decreasing term of the value function.

7. The method for controlling the precision of polymer laser powder bed fusion forming according to claim 1, characterized in that: In S4, the negative feedback mechanism between process parameters and the real-time temperature field of the printing layer.

8. The method for controlling the precision of polymer laser powder bed fusion forming according to any one of claims 1 to 7, characterized in that: The polymer includes one or more of polyaryletherketone, polyetherketone, polyetherketoneketone, polyetheretherketoneketone, polyimide, polycarbonate, polyethylene, polypropylene, polystyrene and nylon.

9. A control system based on the control method for polymer laser powder bed fusion forming accuracy according to any one of claims 1 to 8, characterized in that: include: Data acquisition module, used to obtain temperature distribution data during the polymer LPBF process; A database construction module is used to construct a database based on the temperature distribution data and the process parameters used for printing the current layer; The control model building module is used to establish a precise control model for the polymer LPBF process based on machine learning; The model training module is used to use the database to complete the training of the precise control model of the polymer LPBF process, establish the mapping relationship between process parameters, printing layer thickness and layer-by-layer temperature field distribution, and obtain the negative feedback mechanism between process parameters and the real-time temperature field of the printing layer.

10. A method for improving the precision of polymer laser powder bed fusion forming using the control system of claim 9, characterized in that: include: A precise control model of the polymer LPBF process is built on the LPBF equipment. By real-time control of printing process parameters, the heat accumulation effect in the polymer LPBF process is improved, the warping deformation of the polymer material during the forming process is reduced, and the precision of the formed parts is improved.

Images