A low-power light-curing 3D printing system and method thereof

Abstract

The present invention discloses a low-power light-curing 3D printing system, including an input data conversion unit, which acquires 3D printing data of a model to be printed, converts the 3D printing data into layered data, then converts the layered data into a plurality of RGB image data, and inputs the converted RGB image data into a constructed neural network model; a neural network processing unit, where the neural network model outputs the corresponding feature map classification situation according to the input RGB image data, and maps the feature map classification and the corresponding energy-saving level; an energy-saving control unit, which changes the working state of the light source corresponding to the sliced data according to the mapped energy-saving level. In order to more precisely control the light source brightness and light source power, the present invention defines a convolution kernel with invariant weights to ensure that the light source power corresponding to the output energy regulation scheme can achieve the lowest light source intensity for light curing. Such a design realizes the refined management of the light source power while ensuring the lowest curing power.

Description

Technical Field

[0001] The present invention relates to the technical field of 3D printing energy management, and particularly relates to a low-power stereolithography 3D printing system and method thereof. Background Art

[0002] A 3D printer, also known as a three-dimensional printer, is called additive manufacturing technology. It is a machine that uses rapid prototyping technology. Based on a digital model file, it uses a forming material to construct a three-dimensional entity by layer-by-layer printing. Before printing, it is necessary to use computer modeling software to model to form a 3D model to be printed, and then "partition" the built 3D model into cross-sections layer by layer, that is, slicing, so as to guide the 3D printer to print layer by layer. 3D printers have been widely used in product manufacturing. The working principle of 3D printers is basically the same as that of traditional printers, consisting of a control component, a mechanical component, a print head, consumables (i.e., forming materials), and a medium, etc., and the printing principle is also basically similar.

[0003] At present, the light sources of LCD (Liquid Crystal Display) stereolithography 3D (3Dimensions) printers are all integrally driven. When printing each layer structure, the machine needs to turn on all LED (Light Emitting Diode) lights at a relatively high power to cure the resin. Moreover, the light source is the part with the largest power in the LCD stereolithography 3D printer. Poor control of the light source power consumption often causes a large amount of waste of electric energy.

[0004] In the prior art, Chinese Patent CN201910910013.5 discloses an energy-saving method and system for an LCD light-curing 3D printer. S11. Obtain the sliced cross-section data of the target model, and the sliced cross-section data is used to determine the LCD light-transmitting area. Generally, it can be obtained by reading the slicing result of the slicing software of the 3D printer. The 3D printer is often equipped with software such as slicing software to slice the model to obtain the slicing result. When the sliced cross-section data is determined, the LCD light-transmitting area can be uniquely determined, and the shape and position of the LCD light-transmitting area are the same as those of the sliced cross-section. S12. Generate a light source control instruction corresponding to the sliced cross-section data and send it. The light source control instruction is used to control the working states of a plurality of light sources so that the brightness of the LCD light-transmitting area is not lower than a preset value. The working states include a first working state and a second working state different from the first working state. The generated light source control instruction includes the work of the working state of each light source, such as working in the first working state or the second working state. Of course, further, it can also include setting more detailed working states, such as subdividing the first working state into multiple gears, and each gear works at a different power. The second working state can be regarded as the basic working state, and the first working state is the working state with higher power. This can be achieved by controlling the voltage, current, etc. of each light source to control the working state of the light source. In the embodiment of the present application, the light source control instruction is used to control the working states of a plurality of light sources. For example, the 5th to 10th light sources, the 25th to 30th light sources, and the 35th to 40th light sources are in the first working state (such as the preset high-power working state), and other light sources are in the second working state (such as the preset basic working state). After generating the light source control instruction, the instruction can be sent to the light source to control the working state of the light source. For example, control the 5th to 10th light sources, the 25th to 30th light sources, and the 35th to 40th light sources to be in the first working state, and other light sources to be in the second working state. Make the brightness of the LCD light-transmitting area not lower than the preset value. The preset value is not lower than the brightness that cures the resin within the time of printing this layer. In the prior art, although it is disclosed how to control the working state of the light source by controlling the voltage, current, etc. of each light source and formulating the light source state according to the layering data, however, the implementation of the above method cannot accurately formulate the situation of the light source and cannot further reduce the energy consumption. Summary of the Invention

[0005] The present invention aims to at least solve one of the technical problems existing in the prior art. For this reason, the present invention discloses a low-power light-curing 3D printing system, and the system includes:

[0006] An input data conversion unit obtains 3D printing data of a model to be printed, converts the 3D printing data into layered data, then converts the layered data into multiple RGB image data, and inputs the converted RGB image data into a constructed neural network model;

[0007] A neural network processing unit, the neural network model outputs the corresponding feature map classification according to the input RGB image data, and maps the feature map classification and the corresponding energy-saving level;

[0008] An energy-saving control unit changes the working state of the light source corresponding to the sliced data according to the mapped energy-saving level.

[0009] Furthermore, a neural network model is constructed. The neural network model includes an input layer. The input layer inputs multiple RGB images converted from sliced data. Among them, the RGB image is 480×480 pixels. Then, the data obtained by the input layer is input into a convolutional layer. The convolutional layer includes multiple convolutional kernels. Define one or more convolutional kernels that do not change the parameter values to provide the light source guarantee for the lowest curing brightness. The convolutional layer performs convolutional operations on the RGB images input by the input layer and obtains multiple feature maps. Then, the obtained feature maps are input into a fully connected layer for classification to obtain the corresponding classification results. Finally, the energy-saving level corresponding to the classification result is output through the output layer.

[0010] Furthermore, during the training process, the corresponding weights of multiple convolutional kernels are changed by means of backpropagation learning and gradient descent, and converge when the loss function reaches a preset condition. During this process, the weights of the one or more convolutional kernels that do not change the parameter values do not change. The convolutional kernels with unchanged weights are used to ensure that the light source power corresponding to the output energy adjustment scheme can achieve the lowest light source intensity for light curing.

[0011] Furthermore, the conversion of the layered data into multiple RGB image data further includes: according to the 3D printing data of the model to be printed, the 3D printing data is layered into multiple layers with the same thickness. The thicker the discrete layer thickness, the more RGB image data is converted. Among them, the RGB image data is a projection image representing the layered data.

[0012] Furthermore, calculate the similarity of the RGB images converted from continuous layer data. When the similarity is less than the first preset value, the energy-saving level remains unchanged.

[0013] The present invention also discloses a low-power light-curing 3D printing method, and the method includes:

[0014] S1. Obtain the 3D printing data of the model to be printed, convert the 3D printing data into layered data, then convert the layered data into multiple RGB image data, and input the converted RGB image data into the constructed neural network model;

[0015] S2. The neural network model outputs the corresponding feature map classification situation according to the input RGB image data, and maps the feature map classification and the corresponding energy-saving level;

[0016] S3. Change the working state of the light source corresponding to the sliced data according to the mapped energy-saving level.

[0017] Furthermore, construct a neural network model. The neural network model includes an input layer, and the input layer inputs multiple RGB images converted from sliced data. Among them, the RGB image is 480×480 pixels. Then, input the data obtained by the input layer into the convolutional layer. There are multiple convolutional kernels in the convolutional layer. Define one or more convolutional kernels that do not change the parameter values to provide the light source guarantee for the lowest curing brightness. Perform convolutional operations on the RGB images input by the input layer through the convolutional layer to obtain multiple feature maps, and then input the obtained feature maps into the fully connected layer for classification to obtain the corresponding classification results. Finally, output the energy-saving level corresponding to the classification results through the output layer.

[0018] Furthermore, during the training process, change the corresponding weights of multiple convolutional kernels by means of backpropagation learning and gradient descent, and converge when the loss function reaches the preset condition. During this process, the weights corresponding to the defined one or more convolutional kernels that do not change the parameter values remain unchanged.

[0019] Furthermore, the conversion of the layered data into multiple RGB image data further includes: according to the 3D printing data of the model to be printed, layer the 3D printing data into multiple layers with the same thickness. The thicker the discrete layer thickness, the more RGB image data is converted. Among them, the RGB image data is a projection image representing the layered data.

[0020] Furthermore, calculate the similarity of the RGB images converted from consecutive layer data. When the similarity is less than the first preset value, the energy-saving level remains unchanged.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: In order to more precisely control the brightness and power of the light source, the present invention classifies the corresponding hierarchical data by inputting the model data into the trained neural network, and assigns precise control of the light source power to different categories of hierarchical data. At the same time, the neural network designed by the present invention is different from that in the prior art. In order to ensure that the minimum power of the light source control can still guarantee the light source intensity of photocuring, the present invention designs a layer content that does not change the weight, that is, defines that the weights corresponding to one or more convolution kernels that do not change the parameter values do not change. The convolution kernels with unchanged weights are used to ensure that the light source power corresponding to the output energy regulation scheme can achieve the light source intensity of photocuring at the lowest level. Such a design realizes the refined management of the light source power while ensuring the lowest curing power. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention can be further understood from the following description in conjunction with the drawings. The components in the drawings are not necessarily drawn to scale, but the emphasis is placed on showing the principles of the embodiments. In the drawings, the same reference numerals designate corresponding parts in different views.

[0023] Figure 1 is a flowchart of a low-power photocuring 3D printing method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The technical solutions of the present invention will be described in more detail below in conjunction with the drawings and embodiments.

[0025] Now, mobile terminals implementing various embodiments of the present invention will be described with reference to the drawings. In the following description, suffixes such as "module", "component", or "unit" used to denote elements are only for the convenience of describing the present invention, and have no specific meaning in themselves. Therefore, "module" and "component" can be used interchangeably.

[0026] The mobile terminal can be implemented in various forms. For example, the terminal described in the present invention may include mobile terminals such as mobile phones, smart phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), navigation devices, etc., and fixed terminals such as digital TVs, desktop computers, etc. Hereinafter, it is assumed that the terminal is a mobile terminal. However, those skilled in the art will understand that, except for elements specifically for mobile purposes, the configuration according to the embodiments of the present invention can also be applied to fixed-type terminals.

[0027] As Figure 1 shown, a low-power photocuring 3D printing method, the method includes:

[0028] S1. Obtain the 3D printing data of the model to be printed, convert the 3D printing data into layered data, then convert the layered data into multiple RGB image data, and input the converted RGB image data into the constructed neural network model;

[0029] S2. The neural network model outputs the corresponding feature map classification according to the input RGB image data, and maps the feature map classification and the corresponding energy-saving level;

[0030] S3. Change the working state of the light source corresponding to the sliced data according to the mapped energy-saving level.

[0031] Furthermore, construct a neural network model. The neural network model includes an input layer, and the input layer inputs multiple RGB images converted from sliced data. Among them, the RGB image is 480×480 pixels. Then, input the data obtained by the input layer into the convolutional layer. There are multiple convolutional kernels in the convolutional layer. Define one or more convolutional kernels that do not change the parameter values to provide the light source guarantee for the lowest curing brightness. Perform convolutional operations on the RGB images input by the input layer through the convolutional layer to obtain multiple feature maps, and then input the obtained feature maps into the fully connected layer for classification to obtain the corresponding classification results. Finally, output the energy-saving level corresponding to the classification results through the output layer.

[0032] Furthermore, during the training process, change the corresponding weights of multiple convolutional kernels by means of backpropagation learning and gradient descent, and converge when the loss function reaches the preset condition. During this process, the weights corresponding to the one or more convolutional kernels that do not change the parameter values do not change.

[0033] In this embodiment, it can be that the mobile terminal is communicatively connected to the 3D printing device, and the neural network is trained and constructed through the computing unit of the mobile terminal.

[0034] Furthermore, the conversion of the layered data into multiple RGB image data further includes: according to the 3D printing data of the model to be printed, layer the 3D printing data into multiple layers with the same thickness. The thicker the discrete layer thickness, the more RGB image data is converted. Among them, the RGB image data is a projection image representing the layered data.

[0035] Furthermore, calculate the similarity of the RGB images converted from consecutive layer data. When the similarity is less than the first preset value, the energy-saving level remains unchanged.

[0036] Another embodiment discloses a low-power stereolithography 3D printing system from the perspective of hardware description. The system includes:

[0037] An input data conversion unit acquires 3D printing data of a model to be printed, converts the 3D printing data into layered data, then converts the layered data into multiple RGB image data, and inputs the converted RGB image data into a constructed neural network model;

[0038] A neural network processing unit, the neural network model outputs the corresponding feature map classification according to the input RGB image data, and maps the feature map classification and the corresponding energy-saving level;

[0039] An energy-saving control unit changes the working state of the light source corresponding to the sliced data according to the mapped energy-saving level.

[0040] Furthermore, a neural network model is constructed. The neural network model includes an input layer. The input layer inputs multiple RGB images converted from sliced data. Among them, the RGB images are 480×480 pixels. Then, the data obtained by the input layer is input into a convolutional layer. The convolutional layer includes multiple convolutional kernels. Define one or more convolutional kernels that do not change the parameter values to provide the light source guarantee for the lowest curing brightness. The convolutional layer performs a convolutional operation on the RGB images input by the input layer to obtain multiple feature maps, and then inputs the obtained feature maps into a fully connected layer for classification to obtain the corresponding classification result. Finally, the energy-saving level corresponding to the classification result is output through the output layer.

[0041] Furthermore, during the training process, the corresponding weights of multiple convolutional kernels are changed by means of backpropagation learning and gradient descent, and converge when the loss function reaches a preset condition. During this process, the weights of the defined one or more convolutional kernels that do not change the parameter values do not change. The convolutional kernels with unchanged weights are used to ensure that the light source power corresponding to the output energy adjustment scheme can achieve the light source intensity of light curing at the lowest.

[0042] In this embodiment, in order to more precisely control the light source brightness and light source power, the model data is input into the trained neural network to classify the corresponding layered data, and fine control of the light source power is allocated to different categories of layered data. At the same time, the neural network designed in the present invention is different from the prior art. In order to ensure that the lowest power of the light source control can still ensure the light source intensity of light curing, the present invention designs a layer content that does not change the weight, that is, the weights of the defined one or more convolutional kernels that do not change the parameter values do not change. The convolutional kernels with unchanged weights are used to ensure that the light source power corresponding to the output energy adjustment scheme can achieve the light source intensity of light curing at the lowest. Such a design realizes the refined management of the light source power while ensuring the lowest curing power.

[0043] In this embodiment, in order to more finely control the light source brightness and light source power, the model data is input into a trained neural network for classification of corresponding hierarchical data, and fine control of the light source power is assigned to different categories of hierarchical data. At the same time, the designed neural network is different from the prior art. In order to ensure that the minimum power of the light source control can still ensure the light source intensity of photocuring, the layer content that does not change the weight is designed, that is, one or more convolution kernels that do not change the parameter values are defined, and the corresponding weights do not change. The convolution kernels with unchanged weights are used to ensure that the light source power corresponding to the output energy adjustment scheme can achieve the light source intensity of photocuring at the lowest level. Such a design realizes the refined management of the light source power while ensuring the lowest curing power.

[0044] Further, the conversion of the hierarchical data into a plurality of RGB image data further includes: according to the 3D printing data of the to-be-printed model, the 3D printing data is layered into multiple layers with the same thickness. The thicker the discrete layer thickness, the more RGB image data is converted, where the RGB image data is a projection image representing the hierarchical data.

[0045] Further, the similarity of the RGB images converted from consecutive layer data is calculated. When the similarity is less than the first preset value, the energy saving level is not changed.

[0046] It should also be noted that the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, commodity or device including a series of elements not only includes those elements but also other elements not expressly listed, or elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "including a..." does not exclude the presence of additional identical elements in the process, method, commodity or device including the element.

[0047] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, system or computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] Although the present invention has been described above with reference to various embodiments, it should be understood that many changes and modifications can be made without departing from the scope of the present invention. Therefore, it is intended that the above detailed description be considered illustrative rather than restrictive, and it should be understood that the following claims (including all equivalents) are intended to define the spirit and scope of the present invention. These embodiments should be understood to be only for illustrating the present invention and not for limiting the scope of protection of the present invention. After reading the content described in the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A low-power stereolithography 3D printing system, characterized in that, the system includes: An input data conversion unit, which acquires 3D printing data of a model to be printed, converts the 3D printing data into layer data, then converts the layer data into multiple RGB image data, and inputs the converted RGB image data into a constructed neural network model; A neural network processing unit, the neural network model outputs the corresponding feature map classification according to the input RGB image data, and maps the feature map classification and the corresponding energy-saving level; An energy-saving control unit, which changes the working state of the light source corresponding to the slice data according to the mapped energy-saving level; The neural network model includes an input layer, and the input layer inputs multiple RGB images converted from slice data. Among them, the RGB image is 480×480 pixels. Then, the data obtained by the input layer is input into a convolutional layer. The convolutional layer includes multiple convolutional kernels. Define one or more convolutional kernels that do not change the parameter values to provide the light source guarantee for the lowest curing brightness. The convolutional layer performs convolutional operations on the RGB images input by the input layer to obtain multiple feature maps, and then inputs the obtained feature maps into a fully connected layer for classification to obtain the corresponding classification result. Finally, the energy-saving level corresponding to the classification result is output through the output layer; During the training process, the corresponding weights of multiple convolutional kernels are changed by means of backpropagation learning and gradient descent, and converge when the loss function reaches a preset condition. During this process, the weights of the one or more convolutional kernels that do not change the parameter values do not change. The convolutional kernels with unchanged weights are used to ensure that the light source power corresponding to the output energy adjustment scheme can achieve the lowest light source intensity for stereolithography.

2. A low-power stereolithography 3D printing system according to claim 1, characterized in that, the conversion of the layer data into multiple RGB image data further includes: according to the 3D printing data of the model to be printed, the 3D printing data is layered into multiple layers with the same thickness. The thicker the discrete layer thickness, the more RGB image data is converted. Among them, the RGB image data is a projection image representing the layer data.

3. A low-power stereolithography 3D printing system according to claim 2, characterized in that, Calculate the similarity of the RGB images converted from consecutive layer data. When the similarity is less than the first preset value, the energy-saving level remains unchanged.

4. A low-power stereolithography 3D printing method, characterized in that, the method includes: S1, acquiring 3D printing data of a model to be printed, converting the 3D printing data into layer data, then converting the layer data into multiple RGB image data, and inputting the converted RGB image data into a constructed neural network model; S2, the neural network model outputs the corresponding feature map classification according to the input RGB image data, and maps the feature map classification and the corresponding energy-saving level; S3, changing the working state of the light source corresponding to the slice data according to the mapped energy-saving level; The neural network model includes an input layer that inputs multiple RGB images converted from slice data. Among them, the RGB images are 480×480 pixels. Then, the data obtained by the input layer is input into a convolutional layer. The convolutional layer includes multiple convolutional kernels. One or more convolutional kernels with unchanged parameter values are defined to provide the light source guarantee for the lowest curing brightness. The convolutional layer performs a convolutional operation on the RGB images input by the input layer to obtain multiple feature maps. Then, the obtained feature maps are input into a fully connected layer for classification to obtain corresponding classification results. Finally, the energy-saving level corresponding to the classification result is output through the output layer; During the training process, the corresponding weights of multiple convolutional kernels are changed through backpropagation learning and gradient descent, and converge when the loss function reaches a preset condition. During this process, the weights corresponding to the one or more convolutional kernels with unchanged parameter values remain unchanged.

5. A low-power light-curing 3D printing method according to claim 4, characterized in that the conversion of the layered data into multiple RGB image data further includes: according to the 3D printing data of the model to be printed, the 3D printing data is layered into multiple layers with the same thickness. The thicker the discrete layer thickness, the more RGB image data is converted. Among them, the RGB image data is a projection image representing the layered data.

6. A low-power light-curing 3D printing method according to claim 5, characterized in that the similarity of the RGB images converted from consecutive layer data is calculated, and when the similarity is less than a first preset value, the energy-saving level remains unchanged.

Images