Search

CN-121973449-A - 3D printing intelligent orientation method based on virtual experiment

CN121973449ACN 121973449 ACN121973449 ACN 121973449ACN-121973449-A

Abstract

The invention discloses a 3D printing intelligent orientation method based on a virtual experiment, which belongs to the technical field of intelligent manufacturing and comprises the steps of constructing a 3D printing oriented virtual simulation environment model, constructing a part supporting rule module, generating a supporting structure required by 3D printing of a part, calculating the supporting structure volume and the contact area between the supporting structure and the part in the construction direction of the corresponding part, designing an orientation intelligent body, designing an intelligent orientation optimization reinforcement learning algorithm in the construction direction under multiple targets, enabling the orientation intelligent body to interact with the 3D printing oriented virtual simulation environment model in real time, outputting the optimal part construction direction under multiple targets, printing an actual part model based on the part construction direction, and comparing the printing quality of the actual part model and an original part grid model. The invention can reduce the volume of the supporting structure and the contact area between the supporting structure and the parts required by printing the parts, and realize the efficient and accurate intelligent orientation optimization of the parts.

Inventors

  • YI LI
  • LIU RENMING
  • TAO FEI
  • SUN XUEMIN

Assignees

  • 天目山实验室
  • 北京航空航天大学

Dates

Publication Date
20260505
Application Date
20260407

Claims (10)

  1. 1. The intelligent 3D printing orientation method based on the virtual experiment is characterized by comprising the following steps of: s1, constructing a 3D printing directional virtual simulation environment model, wherein the 3D printing directional virtual simulation environment model comprises an original part grid model, a printing equipment working space model and a constraint relation between the original part grid model and the printing equipment working space model; S2, constructing a part supporting rule module, generating a supporting structure required by 3D printing of the part, and calculating the volume of the supporting structure and the contact area between the supporting structure and the part in the construction direction of the corresponding part; S3, designing a directional agent comprising an evaluation network and a target network, calculating long-term rewards after the directional agent executes the actions selected according to the action selection strategy, updating network parameters of the evaluation network and the target network, selecting actions which can obtain the maximum long-term rewards in all the actions, and executing intelligent directional optimization of parts; S4, designing an intelligent directional optimization reinforcement learning algorithm of the construction direction under multiple targets, enabling the directional intelligent body to interact with the 3D printing directional virtual simulation environment model in real time, and outputting the optimal construction direction of the parts under the multiple targets; S5, based on the component construction direction output by the S4, printing an actual component model, comparing the printing quality of the actual component model and the printing quality of the original component grid model, and verifying and feeding back the intelligent directional optimization reinforcement learning effect.
  2. 2. The virtual experiment-based 3D printing intelligent orientation method according to claim 1, wherein S1 comprises: S11, constructing a triangular mesh model of the part based on the geometric features of the 3D printed part, wherein triangular mesh information of the triangular mesh model of the part comprises vertex coordinates and normal vectors of a triangular patch; S12, limiting the coordinate change range of the parts in the working space of the 3D printing equipment according to the constraint relation that the parts are required to be always contained in the working space of the 3D printing equipment, and constructing a 3D printing directional virtual simulation environment model; S13, the 3D printing directional virtual simulation environment model takes triangle patch vertex coordinates and normal vectors of the triangle mesh model of the component in different states as input, calculates and outputs the triangle patch vertex coordinates and normal vectors after rotation translation, and achieves the position information of the triangle mesh model of the component in the 3D printing directional virtual simulation environment model, the mapping of translation rotation motion and the real-time updating of the state.
  3. 3. The intelligent 3D printing orientation method according to claim 2, wherein in S12, under the condition of the known rotation matrix and displacement, the constraint relation is that the part outer envelope body obtained by the rotation matrix and displacement action is contained in the intersection of the working space and the safe working space of the printing equipment.
  4. 4. The intelligent orientation method for 3D printing based on virtual experiments according to claim 3, wherein in S12, the normal vector of the surface patch of the component after the rotation matrix and the displacement action is the dot product of the displacement plus the normal vector of the surface patch of the component before the rotation matrix and the action.
  5. 5. The virtual experiment-based 3D printing intelligent orientation method according to claim 4, wherein in S12, the rotation matrix comprises a third-order identity matrix, a rotation axis identity vector, an angle of rotation of an arbitrary identity vector, and a rotation axis identity vector in 、 、 Is a component of (a).
  6. 6. The virtual experiment-based 3D printing intelligent orientation method according to claim 5, wherein S2 comprises constructing a part supporting rule module based on a triangular mesh model of the part, a forming axis direction, a preset critical overhang angle threshold value and platform position coordinates, and comprises determining a supporting area and generating a supporting structure required by 3D printing of the part; The component supporting rule module comprises the steps of reading vertex coordinates and normal vectors of a triangular patch of a model, determining a supporting area according to a critical overhang angle threshold value, calculating the supporting structure volume and the contact area between the supporting structure and the component in the corresponding component constructing direction, and finally generating a supporting structure required by 3D printing of the component.
  7. 7. The intelligent 3D printing orientation method based on the virtual experiment according to claim 6 is characterized in that in S2, a rule that the opposite number of the unit vector of the supporting area in the plane sheet and the unit vector dot product of the construction direction is larger than the cosine value of the boundary angle of the overhang is determined, the volume of the supporting structure and the contact area of the supporting structure and the parts are calculated based on triangular plane sheet information, and the supporting structure required by the 3D printing of the parts is generated.
  8. 8. The intelligent 3D printing orientation method according to claim 7 wherein S3 comprises defining an action space, a state space and a rewarding function, wherein the action space is a set of action sequences in which parts rotate or remain motionless around orthogonal coordinate axes, the state space is a six-dimensional array containing part construction direction information, support structure information and time steps, and the rewarding function is related to reduction of volume and contact area of the support structure.
  9. 9. The virtual experiment-based 3D printing intelligent orientation method according to claim 8, wherein S3 comprises the steps of on the basis of defining action space, state space and rewarding function, the orientation agent calculates long-term rewards after the orientation agent performs actions selected according to action selection strategies in the 3D printing orientation virtual simulation environment model by utilizing the evaluation network and the target network based on the orientation of the components in the 3D printing orientation virtual simulation environment model constructed by S1 and the calculated supporting structure volume and the contact area between the supporting structure and the components by S2, and updating network parameters of the evaluation network and the target network.
  10. 10. The virtual experiment-based 3D printing intelligent orientation method according to claim 9, wherein S4 comprises the steps of designing a multi-objective intelligent orientation optimization reinforcement learning algorithm of the construction direction, performing real-time interactive training on an orientation agent and a 3D printing orientation virtual simulation environment model until an action selection strategy converges, outputting an optimal construction direction according to a pareto optimization method, and achieving training of the intelligent orientation optimization reinforcement learning algorithm of the construction direction.

Description

3D printing intelligent orientation method based on virtual experiment Technical Field The invention belongs to the technical field of intelligent manufacturing, and particularly relates to a 3D printing intelligent orientation method based on a virtual experiment. Background With the continuous deep application of metal and polymer additive manufacturing (3D printing) in the fields of aerospace, medical treatment, dies and the like, the requirements on the quality of parts are also higher and higher. Determining the build direction of the component is also the first step of the additive manufacturing process, and the build direction has a decisive influence on the support requirements, surface quality, residual stress and printing time of the component. In complex curved surfaces and thin-wall structures, the coupling between the direction, support, thermal field and path planning makes it difficult to obtain a stable and efficient solution of the component construction direction by experience or single index iteration only. The construction direction selection affects the support topology, surface roughness, construction time and cost at the same time, and belongs to a key bottleneck in the preparation stage of additive manufacturing. The traditional printing technology generally does not optimize the construction direction when processing small parts, and often prints according to the initial orientation of a model file or manually selects the construction direction by experience, so that the mode generally does not obtain the printing direction which is the most material-saving, but saves time, and therefore, the traditional printing technology is more commonly used in customized printing. When mass production is required, the printing cost is increased without optimizing the building direction, but the selection of the building direction by experience is delayed and the effect is not optimal. Therefore, the construction direction is oriented by an automatic means before printing, and then the printing mode gradually replaces the traditional construction direction orientation mode. The existing automatic technology is used for repeatedly testing based on simulation experiments, selecting an optimal printing direction or outputting a directional result based on a heuristic search algorithm, and although a certain result is obtained, the overall optimal result is difficult to ensure, the required time cost is high, and the construction period may be delayed. The heuristic search-based method is generally to combine the characteristics of parts under a specific environment, perform a directional experiment on a single target, thereby evaluating the advantages and disadvantages of the printing direction and optimizing the printing direction, and only consider the single target and the specific environment to have a certain gap with the actual production requirement although realizing automation. Therefore, it is needed to propose a 3D printing intelligent orientation method based on a virtual experiment, which combines the printing space and the forming axis constraint of the device under the simulation environment, and uses the advanced algorithm of reinforcement learning to rapidly evaluate and iteratively optimize the support related indexes of the candidate directions, so as to reduce the support consumption, control the cost and shorten the printing preparation period. Disclosure of Invention In order to solve the technical problems, the invention adopts the following technical scheme: A3D printing intelligent orientation method based on a virtual experiment comprises the following steps: s1, constructing a 3D printing directional virtual simulation environment model, wherein the 3D printing directional virtual simulation environment model comprises an original part grid model, a printing equipment working space model and a constraint relation between the original part grid model and the printing equipment working space model; S2, constructing a part supporting rule module, generating a supporting structure required by 3D printing of the part, and calculating the volume of the supporting structure and the contact area between the supporting structure and the part in the construction direction of the corresponding part; S3, designing a directional agent comprising an evaluation network and a target network, calculating long-term rewards after the directional agent executes the actions selected according to the action selection strategy, updating network parameters of the evaluation network and the target network, selecting actions which can obtain the maximum long-term rewards in all the actions, and executing intelligent directional optimization of parts; S4, designing an intelligent directional optimization reinforcement learning algorithm of the construction direction under multiple targets, enabling the directional intelligent body to interact with the 3D printing directional virtual simulation envi