3D printing component surface roughness prediction method and device based on deep learning

By using deep learning technology and sensors to collect multi-dimensional working parameters of 3D printers, and combining a surface roughness prediction model with random forest algorithm and attention mechanism, the problems of accuracy and efficiency in predicting the surface roughness of 3D printed components are solved, achieving more efficient prediction results.

CN115223267BActive Publication Date: 2024-11-26XUZHOU MEDICAL UNIVERSITY
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202210843111.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-18
Publication Date
2024-11-26
Estimated Expiration
2042-07-18

AI Technical Summary

Technical Problem

Existing technologies lack accuracy and efficiency in predicting the surface roughness of 3D printed components, making it difficult to improve them effectively.

Method used

By employing a deep learning-based approach, multidimensional operating parameters of a 3D printer are obtained, data is collected in real time using sensors, and a surface roughness prediction model combining a random forest algorithm and an attention mechanism is used for preprocessing and feature extraction to generate surface roughness predictions.

Benefits of technology

It significantly improves the accuracy and efficiency of predicting the surface roughness of 3D printed components, overcoming the shortcomings of traditional methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115223267B_ABST
    Figure CN115223267B_ABST
Patent Text Reader

Abstract

The present application proposes a method and device for predicting the surface roughness of a 3D printed component based on deep learning, wherein the surface roughness prediction method includes: obtaining multi-dimensional working parameters of a 3D printer during the printing of a 3D printed component; preprocessing the multi-dimensional working parameters to generate target feature information of the multi-dimensional working parameters; and predicting the surface roughness of the target feature information according to a surface roughness prediction model to generate the surface roughness of the 3D printed component. Thus, the prediction accuracy and efficiency of the surface roughness of the 3D printed component can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of 3D printing technology, and in particular to a method and device for predicting the surface roughness of a 3D printed component based on deep learning. Background Art

[0002] 3D printing (3DP) is a type of rapid prototyping technology. It is a technology that uses digital model files as the basis, uses powdered metal or plastic and other bondable materials to construct objects by printing layer by layer. It is a bottom-up printing method.

[0003] 3D printing is usually achieved by using digital technology material printers. Compared with traditional subtractive manufacturing, 3D printing can achieve efficient resource utilization. Among them, the quality indicators of 3D printed components can be mainly divided into the degree of shape realization: such as corner realization, surface realization, etc.; printing molding accuracy: such as component total size accuracy, printing layer size accuracy, etc.; surface texture quality: such as texture uniformity, surface roughness, etc.; mechanical properties: such as material tensile strength, interlayer adhesion, etc.

[0004] Regarding surface roughness, related technologies mainly adopt methods such as changing printing machine parameters, changing printing direction, and predicting the surface roughness of objects through STL (Stereo Lithography) file geometric data. Summary of the invention

[0005] The first aspect of the present application proposes a method for predicting the surface roughness of a 3D printed component based on deep learning, which can improve the prediction accuracy and efficiency of the surface roughness of a 3D printed component.

[0006] The second embodiment of the present application proposes a 3D printed component surface roughness prediction device based on deep learning.

[0007] A third aspect of the present application provides an electronic device.

[0008] The fourth aspect of the present application provides a computer-readable storage medium.

[0009] The first aspect of the present application proposes a method for predicting the surface roughness of a 3D printed component based on deep learning, comprising: obtaining multidimensional working parameters of a 3D printer during the printing of a 3D printed component; preprocessing the multidimensional working parameters to generate target feature information of the multidimensional working parameters; and predicting the surface roughness of the target feature information according to a surface roughness prediction model to generate the surface roughness of the 3D printed component.

[0010] According to the deep learning-based 3D printed component surface roughness prediction method of the embodiment of the present application, firstly, the multi-dimensional working parameters of the 3D printer in the process of printing the 3D printed component are obtained, and the multi-dimensional working parameters are pre-processed to generate target feature information of the multi-dimensional working parameters, and the surface roughness prediction of the target feature information is performed according to the surface roughness prediction model to generate the surface roughness of the 3D printed component. Thus, the prediction accuracy and prediction efficiency of the surface roughness of the 3D printed component can be improved.

[0011] In addition, the 3D printed component surface roughness prediction method based on deep learning according to the above embodiment of the present application may also have the following additional technical features:

[0012] In one embodiment of the present application, the multi-dimensional working parameters include the temperature of the extruded deposition material, the temperature of the building plate, the temperature of the nozzle, the vibration data of the building plate, the vibration data of the extruder, the working environment pressure and the working environment humidity.

[0013] In one embodiment of the present application, the obtaining of multi-dimensional operating parameters of the 3D printer during the printing of a 3D printed component includes: obtaining the multi-dimensional operating parameters by means of a plurality of sensors disposed on the 3D printer, wherein the plurality of sensors include an infrared temperature sensor, an acceleration sensor, a pressure sensor, a humidity sensor, and a thermocouple temperature sensor.

[0014] In one embodiment of the present application, the preprocessing of the multidimensional working parameters to generate target feature information of the multidimensional working parameters includes: obtaining a printing start time of the 3D printed component, and obtaining a printing end time of the 3D printed component; determining a target multidimensional working parameter from the multidimensional working parameters according to the printing start time and the printing end time; converting the target multidimensional working parameter to generate a target multidimensional working parameter to be processed; and extracting features of the target multidimensional working parameter to be processed based on a random forest algorithm to generate the target feature information.

[0015] In one embodiment of the present application, determining the target multidimensional working parameters from the multidimensional working parameters based on the printing start time and the printing end time includes: obtaining the acquisition time of the multidimensional working parameters; determining the target multidimensional working parameters from the multidimensional working parameters based on the printing start time, the printing end time and the acquisition time.

[0016] In one embodiment of the present application, the surface roughness prediction model is generated in the following manner: acquiring sample data, wherein the sample data includes a calibrated surface roughness; inputting the sample data into the surface roughness prediction model to generate a predicted surface roughness; generating a loss value based on the predicted surface roughness and the calibrated surface roughness, and training the surface roughness prediction model based on the loss value.

[0017] In one embodiment of the present application, the surface roughness prediction model includes an encoder, an attention mechanism module and a decoder.

[0018] The second aspect embodiment of the present application proposes a 3D printed component surface roughness prediction device based on deep learning, including: an acquisition module, used to obtain multidimensional working parameters of the 3D printer during the printing of the 3D printed component; a preprocessing module, used to preprocess the multidimensional working parameters to generate target feature information of the multidimensional working parameters; a prediction module, used to predict the surface roughness of the target feature information according to a surface roughness prediction model to generate the surface roughness of the 3D printed component.

[0019] The 3D printed component surface roughness prediction device based on deep learning in the embodiment of the present application first obtains the multi-dimensional working parameters of the 3D printer in the process of printing the 3D printed component through the acquisition module, and pre-processes the multi-dimensional working parameters through the pre-processing module to generate target feature information of the multi-dimensional working parameters, and predicts the surface roughness of the target feature information according to the surface roughness prediction model through the prediction module to generate the surface roughness of the 3D printed component. In this way, the prediction accuracy and prediction efficiency of the surface roughness of the 3D printed component can be improved.

[0020] In addition, the 3D printed component surface roughness prediction device based on deep learning according to the above embodiment of the present application may also have the following additional technical features:

[0021] In one embodiment of the present application, the multi-dimensional working parameters include the temperature of the extruded deposition material, the temperature of the building plate, the temperature of the nozzle, the vibration data of the building plate, the vibration data of the extruder, the working environment pressure and the working environment humidity.

[0022] In one embodiment of the present application, the acquisition module is specifically used to: acquire the multi-dimensional working parameters by means of a plurality of sensors arranged on the 3D printer, wherein the plurality of sensors include an infrared temperature sensor, an acceleration sensor, a pressure sensor, a humidity sensor and a thermocouple temperature sensor.

[0023] In one embodiment of the present application, the preprocessing module includes: an acquisition unit, used to acquire the printing start time of the 3D printing component, and to acquire the printing end time of the 3D printing component; a determination unit, used to determine the target multidimensional working parameters from the multidimensional working parameters according to the printing start time and the printing end time; a conversion unit, used to convert the target multidimensional working parameters to generate target multidimensional working parameters to be processed; a feature extraction unit, used to extract features of the target multidimensional working parameters to be processed based on a random forest algorithm to generate the target feature information.

[0024] In one embodiment of the present application, the determination unit is specifically used to: obtain the acquisition time of the multidimensional working parameter; and determine the target multidimensional working parameter from the multidimensional working parameters according to the printing start time, the printing end time and the acquisition time.

[0025] In one embodiment of the present application, the surface roughness prediction model includes an encoder, an attention mechanism module and a decoder.

[0026] In one embodiment of the present application, the above-mentioned deep learning-based 3D printed component surface roughness prediction device also includes: a model training module, used to obtain sample data, wherein the sample data includes a calibrated surface roughness; input the sample data into the surface roughness prediction model to generate a predicted surface roughness; generate a loss value based on the predicted surface roughness and the calibrated surface roughness, and train the surface roughness prediction model based on the loss value.

[0027] The third aspect embodiment of the present application proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the surface roughness prediction method of a 3D printed component based on deep learning as described in the first aspect embodiment above is implemented.

[0028] The electronic device of the embodiment of the present application can improve the prediction accuracy and prediction efficiency of the surface roughness of a 3D printed component by executing a computer program stored in a memory through a processor.

[0029] The fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for predicting the surface roughness of a 3D printed component based on deep learning as described in the first aspect of the present application is implemented.

[0030] The computer-readable storage medium of the embodiment of the present application can improve the prediction accuracy and prediction efficiency of the surface roughness of a 3D printed component by storing a computer program and being executed by a processor.

[0031] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0033] Figure 1 It is a schematic diagram of a process of predicting the surface roughness of a 3D printed component based on deep learning according to an embodiment of the present application;

[0034] Figure 2 A schematic diagram of the installation positions of multiple sensors in a 3D printer according to a specific embodiment of the present application;

[0035] Figure 3 It is a flowchart of a method for predicting the surface roughness of a 3D printed component based on deep learning according to another embodiment of the present application;

[0036] Figure 4 It is a flowchart of a method for predicting the surface roughness of a 3D printed component based on deep learning according to another embodiment of the present application;

[0037] Figure 5 It is a flowchart of a method for predicting the surface roughness of a 3D printed component based on deep learning according to another embodiment of the present application;

[0038] Figure 6 is a schematic diagram of a surface roughness prediction model according to a specific embodiment of the present application;

[0039] Figure 7 Schematic diagram of a block diagram of a 3D printed component surface roughness prediction device based on deep learning according to one embodiment of the present application;

[0040] Figure 8 It is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0041] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0042] The following describes a method and device for predicting the surface roughness of a 3D printed component based on deep learning according to an embodiment of the present application with reference to the accompanying drawings.

[0043] The deep learning-based 3D printed component surface roughness prediction method provided in the embodiment of the present application can be executed by an electronic device, which may be a PC (Personal Computer), a tablet computer or a server, etc., without any limitation here.

[0044] In the embodiment of the present application, the electronic device may be provided with a processing component, a storage component and a driving component. Optionally, the driving component and the processing component may be integrated, the storage component may store an operating system, an application program or other program modules, and the processing component implements the 3D printing component surface roughness prediction method based on deep learning provided in the embodiment of the present application by executing the application program stored in the storage component.

[0045] Figure 1 The present invention is a flowchart of a method for predicting the surface roughness of a 3D printed component based on deep learning according to an embodiment of the present application.

[0046] The deep learning-based 3D printed component surface roughness prediction method of the embodiment of the present application can also be executed by the deep learning-based 3D printed component surface roughness prediction device provided by the embodiment of the present application. The device can be configured in an electronic device (for example, a PC) to obtain the multi-dimensional working parameters of the 3D printer during the printing of the 3D printed component, and pre-process the multi-dimensional working parameters to generate target feature information of the multi-dimensional working parameters, and perform surface roughness prediction on the target feature information according to the surface roughness prediction model to generate the surface roughness of the 3D printed component, thereby improving the prediction accuracy and prediction efficiency of the surface roughness of the 3D printed component.

[0047] As a possible scenario, the deep learning-based 3D printed component surface roughness prediction method of the embodiment of the present application can also be executed on the server side. The server can be a cloud server, and the deep learning-based 3D printed component surface roughness prediction method can be executed on the cloud.

[0048] like Figure 1 As shown, the 3D printing component surface roughness prediction method based on deep learning may include:

[0049] Step 101, obtaining multi-dimensional working parameters of the 3D printer during the process of printing a 3D printed component, wherein the multi-dimensional working parameters may include the temperature of the extruded deposition material, the temperature of the build plate, the temperature of the nozzle, the vibration data of the build plate, the vibration data of the extruder, the working environment air pressure and the working environment humidity, etc.

[0050] It should be noted that the deposition material described in this embodiment may be a material or one of the materials of a printing component (i.e., a 3D printing component). The extruder described in this embodiment may be an extruder built into a 3D printer. The working environment air pressure and working environment humidity described in this embodiment may be the air pressure and humidity of the environment where the 3D printer is located.

[0051] To clearly explain the previous embodiment, in one embodiment of the present application, obtaining the multi-dimensional operating parameters of the 3D printer during the process of printing a 3D printed component may include obtaining the multi-dimensional operating parameters through a plurality of sensors disposed on the 3D printer, wherein the plurality of sensors may include an infrared temperature sensor, an acceleration sensor, a pressure sensor, a humidity sensor, and a thermocouple temperature sensor.

[0052] It should be noted that the infrared temperature sensor, acceleration sensor, pressure sensor, humidity sensor and thermocouple temperature sensor described in this embodiment may be multiple.

[0053] In the examples of this application, see Figure 2 , an infrared temperature sensor can be arranged (set) on the nozzle (extrusion head) to measure the temperature of the extruded deposited material; a thermocouple temperature sensor can be arranged on the printer base to measure the temperature of the build plate; a thermocouple temperature sensor can also be arranged on the nozzle to measure the nozzle temperature; an acceleration sensor can be arranged on the build plate and the nozzle, respectively, to measure the vibration of the build plate and the extruder; a pressure sensor and a humidity sensor can be arranged on the outer casing of the 3D printer, respectively, to collect environmental parameters (i.e., working environment air pressure and working environment humidity).

[0054] As a possible scenario, the pressure sensor and humidity sensor mentioned above may also be arranged in the room where the 3D printer is located, so as to collect the air pressure and humidity of the environment where the 3D printer is located.

[0055] Specifically, the electronic device (for example, a computer connected to a 3D printer) can obtain multi-dimensional working parameters such as the temperature of the extruded deposition material, the building plate temperature, the nozzle temperature, the vibration data of the building plate, the vibration data of the extruder, the working environment air pressure and the working environment humidity in real time through the above-mentioned infrared temperature sensor, acceleration sensor, pressure sensor, humidity sensor and thermocouple temperature sensor during the process of printing a component by the 3D printer (that is, during the operation of the 3D printer).

[0056] It should be noted that the 3D printer described in the above embodiment can be controlled by an electronic device (e.g., a computer) connected thereto, wherein the above-mentioned multiple sensors can be connected to the electronic device using a copper cable with a DuPont connector, and the multiple sensors send the multi-dimensional working parameters acquired (collected) in real time to the electronic device through the copper cable. In addition, after receiving the multi-dimensional working parameters, the electronic device can save them for subsequent calls.

[0057] As a possible scenario, the above-mentioned multiple sensors can also be connected to the host of the 3D printer using a copper cable with a DuPont connector, wherein the multiple sensors will acquire (collect) multi-dimensional working parameters in real time and send them to the host of the 3D printer, which is then forwarded to the electronic device by the host of the 3D printer.

[0058] It should be noted that the 3D printer described in the above embodiment can print components based on the Fused Deposition Modeling (FDM) process, wherein FDM manufactures thermoplastic components by heating, extruding and thermoplastic materials, and printing them layer by layer.

[0059] As a possible situation, the 3D printer described in the above embodiment can also print related components based on other printing processes.

[0060] Step 102: preprocess the multi-dimensional working parameters to generate target feature information of the multi-dimensional working parameters.

[0061] In the embodiment of the present application, the multi-dimensional working parameters can be preprocessed by a preset preprocessing strategy to generate target feature information of the multi-dimensional working parameters. The preset preprocessing strategy can be calibrated according to actual conditions.

[0062] Specifically, after acquiring the multi-dimensional working parameters of the 3D printer during the component printing process, the electronic device may pre-process the multi-dimensional working parameters according to a preset pre-processing strategy to generate target feature information of the multi-dimensional working parameters.

[0063] Step 103: Predict the surface roughness of the target feature information according to the surface roughness prediction model to generate the surface roughness of the 3D printed component.

[0064] It should be noted that the surface roughness prediction model described in this embodiment can be trained in advance and pre-stored in the storage space of the electronic device for easy retrieval and application. The storage space is not limited to a physical storage space, such as a hard disk, but can also be a storage space of a network hard disk connected to the electronic device (cloud storage space).

[0065] The training and generation of the surface roughness prediction model can be performed by a related training server, which can be a cloud server or a computer host. A communication connection is established between the training server and an electronic device that can execute the 3D printing component surface roughness prediction method based on deep learning provided in the application embodiment. The communication connection can be at least one of a wireless network connection and a wired network connection. The training server can send the trained surface roughness prediction model to the electronic device so that the electronic device can call it when needed, thereby greatly reducing the computing pressure of the electronic device.

[0066] It should be noted that the surface roughness prediction model described in this embodiment can be obtained by training the LSTM (Long Short Term Memory) model.

[0067] Specifically, after obtaining the above-mentioned target feature information, the electronic device can call out the surface roughness prediction model from its own storage space, and input the target feature information into the surface roughness prediction model, thereby performing surface roughness prediction on the target feature information through the surface roughness prediction model to obtain the surface roughness of the 3D printed component output by the surface roughness prediction model.

[0068] Therefore, compared with the methods of predicting the surface roughness of objects by changing printing machine parameters, changing printing direction, and using STL file geometric data in related technologies, the method for predicting the surface roughness of 3D printed components based on deep learning in the embodiments of the present application fully considers the environmental parameters in actual operation and the thermal cycle of the printing process, thereby greatly improving the prediction accuracy and efficiency of the surface roughness of 3D printed components.

[0069] In the embodiment of the present application, the multi-dimensional working parameters of the 3D printer in the process of printing the component are first obtained, and the multi-dimensional working parameters are pre-processed to generate target feature information of the multi-dimensional working parameters, and the surface roughness prediction of the target feature information is performed according to the surface roughness prediction model to generate the surface roughness of the 3D printed component. In this way, the prediction accuracy and prediction efficiency of the surface roughness of the 3D printed component can be improved.

[0070] In order to clearly explain the above embodiment, in one embodiment of the present application, Figure 3 As shown, preprocessing the multidimensional working parameters to generate target feature information of the multidimensional working parameters may include:

[0071] Step 301, obtaining the printing start time of the 3D printing component and obtaining the printing end time of the 3D printing component.

[0072] In an embodiment of the present application, the electronic device can obtain the printing start time and printing end time of the 3D printer when printing a component through a relevant API (Application Programming Interface), that is, obtain the start time and end time of printing the component when the 3D printer is working.

[0073] Specifically, after obtaining the multi-dimensional working parameters of the 3D printer during the component printing process, the electronic device may first obtain the printing start time and the printing end time of the 3D printer during the component printing process through a related API.

[0074] Step 302: Determine target multi-dimensional working parameters from the multi-dimensional working parameters according to the printing start time and the printing end time.

[0075] In order to clearly explain the above embodiment, in one embodiment of the present application, Figure 4 As shown, according to the printing start time and the printing end time, determining the target multi-dimensional working parameters from the multi-dimensional working parameters may include:

[0076] Step 401, obtaining the acquisition time of the multi-dimensional working parameters.

[0077] In the embodiment of the present application, the above-mentioned multi-dimensional working parameters may include the acquisition time of each parameter, and each parameter and its corresponding acquisition time may be stored in a one-to-one correspondence.

[0078] It should be noted that the acquisition time described in this embodiment may be the acquisition time of each parameter in the above-mentioned multi-dimensional working parameters, for example, the acquisition time of the temperature of the extruded deposition material, the temperature of the build plate, the temperature of the nozzle, the vibration data of the build plate, the vibration data of the extruder, the working environment air pressure and the working environment humidity, etc. In addition, when each of the multiple sensors described in the above embodiments acquires (detects) the corresponding data (multi-dimensional working parameters), it can record the time of acquisition (detection), and feed back the time and the acquired (detected) data (multi-dimensional working parameters) to the host of the electronic device or the 3D printer.

[0079] Specifically, after obtaining the printing start time and the printing end time of the 3D printer, the electronic device can parse the above-mentioned multi-dimensional working parameters to obtain the acquisition time of the multi-dimensional working parameters.

[0080] Step 402: Determine target multi-dimensional working parameters from the multi-dimensional working parameters according to the printing start time, the printing end time and the acquisition time.

[0081] Specifically, after the electronic device analyzes the above multi-dimensional working parameters and obtains the acquisition time of the multi-dimensional working parameters, it can compare and retrieve the acquisition time of each parameter in the multi-dimensional working parameters according to the printing start time and the printing end time, so as to delete the multi-dimensional working parameters before the 3D printing component starts printing and after the printing ends, so as to obtain the above target multi-dimensional working parameters. In this way, unnecessary data (multi-dimensional working parameters) can be removed to improve the accuracy and efficiency of subsequent predictions.

[0082] Step 303: convert the target multi-dimensional working parameters to generate target multi-dimensional working parameters to be processed.

[0083] It should be noted that the target multi-dimensional working parameter described in this embodiment may be a time domain signal, and the target multi-dimensional working parameter to be processed described in this embodiment may be a frequency domain signal.

[0084] Specifically, after determining the target multi-dimensional operating parameters, the electronic device can convert the target multi-dimensional operating parameters according to a preset conversion algorithm to generate the target multi-dimensional operating parameters to be processed. The preset conversion algorithm can be calibrated according to the actual situation. For example, the preset conversion algorithm can be a Fourier transform algorithm. Thus, the time domain signal can be converted into a frequency domain signal to provide data support for subsequent feature extraction.

[0085] Step 304 , extracting features of the multi-dimensional working parameters of the target to be processed based on the random forest algorithm to generate target feature information.

[0086] Specifically, after obtaining the above-mentioned multi-dimensional working parameters of the target to be processed, the electronic device may also perform feature extraction on the multi-dimensional working parameters of the target to be processed based on the random forest algorithm to generate target feature information.

[0087] As a possible situation, in order to improve the accuracy and speed of feature extraction, after obtaining the above-mentioned multidimensional working parameters of the target to be processed, the electronic device can first filter the multidimensional working parameters of the target to be processed to remove measurement noise, and perform noise reduction processing, so as to obtain the optimized multidimensional working parameters of the target to be processed, and then perform feature extraction on the optimized multidimensional working parameters of the target to be processed based on the random forest algorithm to generate target feature information. In this way, the target feature information obtained can be made more accurate, that is, the accuracy of feature extraction is improved.

[0088] In order to clearly illustrate the above embodiment, in one embodiment of the present application, as Figure 5 As shown in Figure 2, the surface roughness prediction model mentioned above is generated in the following way:

[0089] Step 501: Acquire sample data, wherein the sample data includes calibrated surface roughness.

[0090] It should be noted that the sample data described in this embodiment may be sample multi-dimensional working parameters, and the sample multi-dimensional working parameters may be multiple groups.

[0091] In the embodiment of the present application, there are multiple ways to obtain sample data. Among them, the multi-dimensional working parameters of the 3D printer during the printing component printing process can be collected through the above-mentioned sensors to obtain sample data. Sample data can also be created manually, for example, by software simulation.

[0092] Step 502: input sample data into a surface roughness prediction model to generate predicted surface roughness.

[0093] Step 503: Generate a loss value according to the predicted surface roughness and the calibrated surface roughness, and train the surface roughness prediction model according to the loss value. The loss value may be a root mean square error.

[0094] Specifically, after obtaining the sample data, the sample data can be input into the surface roughness prediction model to generate a predicted surface roughness, and a loss value can be generated based on the predicted surface roughness and the calibrated surface roughness, and the surface roughness prediction model can be trained based on the loss value, thereby optimizing the surface roughness prediction model and improving the accuracy of the prediction.

[0095] As a possible situation, after completing the training of the surface roughness prediction model, test sample data can also be obtained, and the surface roughness prediction model can be tested by the test sample data to determine whether the training of the surface roughness prediction model meets the training completion condition, wherein the training completion condition can be calibrated according to the actual situation. In this way, the training effect of the surface roughness prediction model can be guaranteed, thereby further improving the accuracy of the prediction.

[0096] For example, 486 sample data can be randomly divided into 350 and 136 for training and testing of the surface roughness prediction model. Five random tests were conducted, with the mean square error as the loss function (loss value). The training algorithm uses the gradient descent method and the learning rate is set to 0.005. The network structure of the attention-based encoder-decoder is set to: 30 neurons per LSTM unit and 30 neurons in the dense layer of the attention module, and the neural network model can be constructed based on Tensorflow (deep learning framework). In this way, the training effect of the surface roughness prediction model is guaranteed.

[0097] In one embodiment of the present application, Figure 6As shown, the above-mentioned surface roughness prediction model may include an encoder, an attention module and a decoder.

[0098] It should be noted that the surface roughness prediction model described in this embodiment can also introduce an Attention mechanism [i.e., adding an Attention mechanism] to further improve the prediction accuracy of the model. Among them, the Attention mechanism considers different weight parameters for each element of the input. In this application, the relevant network weights of each part are dynamically determined by considering the specific conditions during printing. Therefore, by using a deep learning framework, a surface roughness prediction model trained by a long short-term memory network (LSTM) combined with an attention mechanism (Attention) can improve the prediction accuracy and efficiency of the surface roughness of 3D printed components.

[0099] Specifically, see Figure 6 , the input data xi of the LSTM cell (unit) (i=1, 2, ..., n, n is the number of printing layers) can correspond to the sensor data collected by a single printing layer of FDM (i.e., the temperature of the extruded deposition material, the temperature of the build plate, the temperature of the nozzle, the vibration data of the build plate, the vibration data of the extruder, the working environment air pressure and the working environment humidity, etc. collected by the above multiple sensors). Among them, since FDM is a sequential task, printing layer by layer makes the printing effect of each layer affected by both the printed layer and the next layer to be printed in the future, so the encoder can use BiLSTM (bidirectional long short-term memory network), and the input signal can be the average value and standard deviation (target feature information) of the signal segment corresponding to each printing layer. It should be noted that the input data xi of the LSTM cell (unit) described in this embodiment can be an array containing a joint analysis of multi-dimensional parameters such as the temperature of the extruded deposition material, the temperature of the build plate, the temperature of the nozzle, the vibration data of the build plate, the vibration data of the extruder, the working environment air pressure and the working environment humidity, etc., the array is fused and then input into the Attention module for enhancement, and the result is made more accurate through the joint analysis of multi-dimensional data.

[0100] Furthermore, the inter-layer influence reflected by the inter-layer thermal history is simulated through the inter-unit communication of the LSTM forward and reverse paths. The forward LSTM processes the input data and obtains the forward LSTM output as After the backward LSTM processing, the backward LSTM output is The hidden vectors of the forward and backward outputs are concatenated to obtain the output of the bidirectional LSTM (i.e., encoder) as h t , where h tIt can be calculated by the following formula (1):

[0101]

[0102] Among them, h t Can be the output of bidirectional LSTM, W h Can be the weight of dense layer, b h Can be an offset value, can be the output of the forward LSTM, It can be the output of backward LSTM.

[0103] Furthermore, the attention mechanism allows weights to be determined by considering the relevant context network. In the process of printing components by a 3D printer, the context network corresponds to the specific printing conditions of each part (such as layer-by-layer thermal state, machine parameters and environmental parameters). In this application, the machine parameters can be set layer height, printing speed, and nozzle temperature. The environmental parameters can be the current ambient air pressure value collected by the pressure sensor and the current ambient humidity collected by the humidity sensor. For the current output position t, the Attention module can generate the corresponding output weight W according to the context network. t (t = 1, 2, ..., n). The corresponding non-normalized output weights can be generated by a dense layer (fully connected neural network layer) first, and then, in order to ensure that the relative influence of all printed layers adds up to 1, a softmax layer can be introduced to normalize the generated output weights. Mathematically, this process can be expressed as the following formula (2):

[0104]

[0105] Among them, W attn,dense Can be the weight of the dense layer in the Attention module, x machine and x environment They can be the machine parameters of the 3D printer and the environmental parameters of the 3D printer respectively. t Can be the output of bidirectional LSTM, w' t and w' j Both can be unstandardized output weights, w t Can be the normalized output, j and t They can all be the serial numbers of the encoders.

[0106] Furthermore, the decoder receives the weighted sum of all thermal states from the last LSTM layer in the encoder and then predicts the surface roughness of the component through a dense layer with a single neuron.

[0107] Therefore, in the Attention module, machine parameters (e.g., machine setting parameters of the 3D printer) and environmental parameters (e.g., air pressure, humidity, and temperature of the environment in which the 3D printer is located) can be used as supplementary inputs of the context network to further improve the accuracy of the surface roughness prediction model.

[0108] In summary, according to the deep learning-based 3D printed component surface roughness prediction method of the embodiment of the present application, firstly, the multi-dimensional working parameters of the 3D printer in the component printing process are obtained, and the multi-dimensional working parameters are pre-processed to generate target feature information of the multi-dimensional working parameters, and the surface roughness prediction of the target feature information is performed according to the surface roughness prediction model to generate the surface roughness of the 3D printed component. Thus, the prediction accuracy and prediction efficiency of the surface roughness of the 3D printed component can be improved.

[0109] Figure 7 Schematic diagram of a block diagram of a device for predicting surface roughness of a 3D printed component based on deep learning according to an embodiment of the present application.

[0110] The deep learning-based 3D printed component surface roughness prediction device of the embodiment of the present application can be configured in an electronic device to obtain multi-dimensional working parameters of the 3D printer during the printing of the 3D printed component, and pre-process the multi-dimensional working parameters to generate target feature information of the multi-dimensional working parameters, and perform surface roughness prediction on the target feature information according to the surface roughness prediction model to generate the surface roughness of the 3D printed component, thereby improving the prediction accuracy and prediction efficiency of the surface roughness of the 3D printed component.

[0111] like Figure 7 As shown, the deep learning-based 3D printing component surface roughness prediction device 700 may include: an acquisition module 710 , a preprocessing module 720 and a prediction module 730 .

[0112] The acquisition module 710 is used to acquire the multi-dimensional working parameters of the 3D printer during the process of printing the 3D printed component.

[0113] The preprocessing module 720 is used to preprocess the multi-dimensional working parameters to generate target feature information of the multi-dimensional working parameters.

[0114] The prediction module 730 is used to perform surface roughness prediction on the target feature information according to the surface roughness prediction model to generate the surface roughness of the 3D printed component.

[0115] In one embodiment of the present application, the multi-dimensional working parameters may include the temperature of the extruded deposition material, the temperature of the building plate, the temperature of the nozzle, the vibration data of the building plate, the vibration data of the extruder, the working environment pressure and the working environment humidity.

[0116] In one embodiment of the present application, the acquisition module 710 is specifically used to: acquire multi-dimensional working parameters through multiple sensors set on the 3D printer, wherein the multiple sensors include infrared temperature sensors, acceleration sensors, pressure sensors, humidity sensors and thermocouple temperature sensors.

[0117] In one embodiment of the present application, Figure 7 As shown, the preprocessing module 720 may include: an acquisition unit 721 , a determination unit 722 , a conversion unit 733 and a feature extraction unit 744 .

[0118] The acquisition unit 721 is used to acquire the printing start time of the 3D printing component and acquire the printing end time of the 3D printing component.

[0119] The determination unit 722 is used to determine the target multi-dimensional operating parameters from the multi-dimensional operating parameters according to the printing start time and the printing end time.

[0120] The conversion unit 723 is used to convert the target multi-dimensional working parameters to generate target multi-dimensional working parameters to be processed.

[0121] The feature extraction unit 724 is used to extract features of the multi-dimensional working parameters of the target to be processed based on the random forest algorithm to generate target feature information.

[0122] In one embodiment of the present application, the determination unit 722 is specifically used to: obtain the acquisition time of the multidimensional working parameters; and determine the target multidimensional working parameters from the multidimensional working parameters according to the printing start time, the printing end time and the acquisition time.

[0123] In one embodiment of the present application, Figure 7 As shown, the deep learning-based 3D printed component surface roughness prediction device 700 may also include a model training module 740, wherein the model training module 740 is used to obtain sample data, wherein the sample data includes a calibrated surface roughness; input the sample data into a surface roughness prediction model to generate a predicted surface roughness; generate a loss value based on the predicted surface roughness and the calibrated surface roughness, and train the surface roughness prediction model based on the loss value.

[0124] In one embodiment of the present application, the surface roughness prediction model includes an encoder, an attention mechanism module and a decoder.

[0125] It should be noted that the aforementioned explanation of the embodiment of the method for predicting the surface roughness of a 3D printed component based on deep learning is also applicable to the device for predicting the surface roughness of a 3D printed component based on deep learning in this embodiment, and will not be repeated here.

[0126] In summary, the 3D printed component surface roughness prediction device based on deep learning in the embodiment of the present application first obtains the multi-dimensional working parameters of the 3D printer in the process of printing the 3D printed component through the acquisition module, and pre-processes the multi-dimensional working parameters through the pre-processing module to generate the target feature information of the multi-dimensional working parameters, and predicts the surface roughness of the target feature information according to the surface roughness prediction model through the prediction module to generate the surface roughness of the 3D printed component. In this way, the prediction accuracy and prediction efficiency of the surface roughness of the 3D printed component can be improved.

[0127] In order to implement the above embodiment, Figure 8 As shown, the present application also proposes an electronic device 800, including a memory 810, a processor 820, and a computer program stored in the memory 810 and executable on the processor 820, and the processor 820 executes the program to implement the 3D printed component surface roughness prediction method based on deep learning proposed in the aforementioned embodiment of the present application.

[0128] The electronic device of the embodiment of the present application can improve the prediction accuracy and prediction efficiency of the surface roughness of a 3D printed component by executing a computer program stored in a memory through a processor.

[0129] In order to implement the above embodiments, the present application also proposes a non-temporary computer-readable storage medium on which a computer program is stored, and the program is executed by a processor to implement the deep learning-based 3D printed component surface roughness prediction method proposed in the aforementioned embodiments of the present application.

[0130] The computer-readable storage medium of the embodiment of the present application can improve the prediction accuracy and prediction efficiency of the surface roughness of a 3D printed component by storing a computer program and being executed by a processor.

[0131] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0132] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0133] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for predicting surface roughness of 3D printed components based on deep learning, characterized in that: include: Obtaining multi-dimensional working parameters of the 3D printer during the process of printing the 3D printed component; Preprocessing the multidimensional operating parameters to generate target feature information of the multidimensional operating parameters, wherein preprocessing the multidimensional operating parameters to generate target feature information of the multidimensional operating parameters includes: Obtaining a printing start time of the 3D printing component, and obtaining a printing end time of the 3D printing component; Determining target multi-dimensional operating parameters from the multi-dimensional operating parameters according to the printing start time and the printing end time; Converting the target multi-dimensional working parameters to generate target multi-dimensional working parameters to be processed; Extracting features of the multi-dimensional working parameters of the target to be processed based on a random forest algorithm to generate the target feature information; The surface roughness of the target feature information is predicted according to a surface roughness prediction model to generate the surface roughness of the 3D printed component.

2. The method according to claim 1, characterized in that in, The multi-dimensional working parameters include the temperature of the extruded deposition material, the temperature of the building plate, the temperature of the nozzle, the vibration data of the building plate, the vibration data of the extruder, the working environment pressure and the working environment humidity.

3. The method according to claim 1, characterized in that The obtaining of multi-dimensional working parameters of the 3D printer during the process of printing the 3D printed component includes: The multi-dimensional working parameters are obtained by means of a plurality of sensors arranged on the 3D printer, wherein the plurality of sensors include an infrared temperature sensor, an acceleration sensor, a pressure sensor, a humidity sensor and a thermocouple temperature sensor.

4. The method according to claim 1, characterized in that: Determining target multi-dimensional operating parameters from the multi-dimensional operating parameters according to the printing start time and the printing end time includes: Obtaining the acquisition time of the multi-dimensional working parameters; The target multi-dimensional operating parameter is determined from the multi-dimensional operating parameters according to the printing start time, the printing end time and the acquisition time.

5. The method according to claim 1, characterized in that in, The surface roughness prediction model is generated in the following way: Acquire sample data, wherein the sample data includes a calibrated surface roughness; Inputting the sample data into the surface roughness prediction model to generate predicted surface roughness; A loss value is generated according to the predicted surface roughness and the calibrated surface roughness, and the surface roughness prediction model is trained according to the loss value.

6. The method according to claim 1, characterized in that in, The surface roughness prediction model includes an encoder, an attention mechanism module and a decoder.

7. A 3D printed component surface roughness prediction device based on deep learning, characterized in that: include: An acquisition module is used to acquire multi-dimensional working parameters of the 3D printer during the process of printing a 3D printed component; A preprocessing module is used to preprocess the multi-dimensional working parameters to generate target feature information of the multi-dimensional working parameters, wherein the preprocessing module includes: An acquiring unit, used for acquiring a printing start time of the 3D printing component and acquiring a printing end time of the 3D printing component; A determining unit, configured to determine a target multi-dimensional operating parameter from the multi-dimensional operating parameters according to the printing start time and the printing end time; A conversion unit, used for converting the target multi-dimensional working parameter to generate a target multi-dimensional working parameter to be processed; A feature extraction unit, used for extracting features of the multi-dimensional working parameters of the target to be processed based on a random forest algorithm to generate the target feature information; The prediction module is used to perform surface roughness prediction on the target feature information according to a surface roughness prediction model to generate the surface roughness of the 3D printed component.

8. The device according to claim 7, characterized in that in, The multi-dimensional working parameters include the temperature of the extruded deposition material, the temperature of the building plate, the temperature of the nozzle, the vibration data of the building plate, the vibration data of the extruder, the working environment pressure and the working environment humidity.

9. The device according to claim 7, characterized in that The acquisition module is specifically used for: The multi-dimensional working parameters are obtained by means of a plurality of sensors arranged on the 3D printer, wherein the plurality of sensors include an infrared temperature sensor, an acceleration sensor, a pressure sensor, a humidity sensor and a thermocouple temperature sensor.

10. The device according to claim 7, characterized in that The determining unit is specifically configured to: Obtaining the acquisition time of the multi-dimensional working parameters; The target multi-dimensional operating parameter is determined from the multi-dimensional operating parameters according to the printing start time, the printing end time and the acquisition time.

11. The device according to claim 7, characterized in that Also includes: A model training module, used for acquiring sample data, wherein the sample data includes a calibrated surface roughness; The sample data is input into the surface roughness prediction model to generate predicted surface roughness; a loss value is generated according to the predicted surface roughness and the calibrated surface roughness, and the surface roughness prediction model is trained according to the loss value.

12. The device according to claim 7, characterized in that in, The surface roughness prediction model includes an encoder, an attention mechanism module and a decoder.

13. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for predicting the surface roughness of a 3D printed component based on deep learning as described in any one of claims 1 to 6 is implemented.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a method for predicting the surface roughness of a 3D printed component based on deep learning as described in any one of claims 1 to 6 is implemented.

Citation Information

Patent Citations

  • Accurate three-dimensional printing

    CN108883575A

  • Additive manufacturing laser 3D printing temperature gradient detection method

    CN110186570A

  • Three-dimensional key point prediction and deep learning model training method, device and apparatus

    CN111160375A

  • 3D rock structural surface roughness intelligent extraction system and method

    CN111709924A

  • 3D printing organization prediction method and system, medium and terminal equipment

    CN111832610A