Big data-based 3D printing tool clamp data acquisition and analysis method and system

Abstract

The invention provides an acquisition and analysis method and system for 3D printing tool clamp data based on big data, and relates to the technical field of 3D printing and intelligent manufacturing, and the method comprises the steps: obtaining a clamp historical design scheme, machining parameters and sensor data, carrying out the feature extraction and data cleaning, and forming structured data; then, the design intention is analyzed based on a BERT-CNN model, and a clamp applicability classification database is constructed in combination with the classification model; and then, generating a lightweight clamp structure by adopting a generative adversarial network and topological optimization, and optimizing the topological structure of the lightweight clamp structure through an optimization model. Thirdly, simulation analysis is carried out, stress, deformation and thermal stress are evaluated, errors are optimized based on simulation parameters, and optimized clamp design data are obtained; and finally, the clamp is manufactured, stress data are collected in real time, and machining process parameters are optimized through dynamic analysis. According to the method, clamp design and manufacturing are optimized through data driving, and design efficiency and clamp performance are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of 3D printing and intelligent manufacturing, and in particular to a method and system for collecting and analyzing 3D printing tooling and fixture data based on big data. Background Art

[0002] With the continuous pursuit of precision and efficiency in the field of industrial manufacturing, the application of 3D printing technology in the manufacturing of fixtures has gradually become a trend. Traditional fixture design and manufacturing methods rely on manual experience and fixed design rules. This method often cannot provide flexible and efficient solutions when faced with complex processing technology and changing production needs. Existing fixture design methods usually cannot make full use of big data and advanced algorithm technology for optimization, which easily leads to data redundancy in the design process, low design efficiency and insufficient fixture applicability.

[0003] Therefore, there is an urgent need for a method and system for collecting and analyzing 3D printing tooling and fixture data based on big data to solve the problems of incomplete data processing, lack of dynamic optimization and applicability classification in the existing technology. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for collecting and analyzing 3D printing fixture data based on big data to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows: In the first aspect, the present application provides a method for collecting and analyzing 3D printing fixture data based on big data, including: Acquiring first information, the first information including historical fixture design solutions, machining process parameters, and sensor data; Performing feature extraction and data cleaning on the first information to obtain second information, wherein the second information includes the first information after data cleaning; The design intent of the second information is analyzed based on the BERT-CNN dual channel, and the historical design schemes of the fixture are classified in combination with the preset classification model to obtain the fixture applicability classification database; According to the fixture applicability classification database, a lightweight fixture structure is generated through generative adversarial networks and topology optimization, and the lightweight fixture structure is optimized through a preset optimization model to obtain an optimized fixture topology structure; Perform simulation according to the optimized fixture topology, and perform error optimization based on the simulation parameters to obtain the fixture design data after error optimization; The fixture is manufactured according to the fixture design data after error optimization, and the fixture force data is collected for dynamic optimization to obtain the optimized fixture processing parameters.

[0005] In the second aspect, the present application also provides a 3D printing fixture data collection and analysis system based on big data, including: An acquisition unit, configured to acquire first information, wherein the first information includes a historical design scheme of the fixture, processing parameters, and sensor data; a cleaning unit, configured to perform feature extraction and data cleaning on the first information to obtain second information, wherein the second information includes the first information after data cleaning; A classification unit is used to analyze the design intent of the second information based on the BERT-CNN dual channel, and classify the historical design schemes of the fixture in combination with a preset classification model to obtain a fixture applicability classification database; The optimization unit is used to generate a lightweight fixture structure according to the fixture applicability classification database through generative adversarial networks and topology optimization, and optimize the lightweight fixture structure through a preset optimization model to obtain an optimized fixture topology structure; A simulation unit is used to perform simulation according to the optimized fixture topology and perform error optimization based on the simulation parameters to obtain the fixture design data after error optimization; The analysis unit is used to manufacture the fixture based on the fixture design data after error optimization, and collect the fixture force data for dynamic optimization to obtain the optimized fixture processing parameters.

[0006] The beneficial effects of the present invention are: The present invention comprehensively improves the efficiency and adaptability of fixture design and manufacturing through innovative algorithm combinations and data processing methods. The method obtains the first information including the fixture's historical design scheme, processing parameters and sensor data, and optimizes the information using a variety of data processing techniques (such as feature extraction, data cleaning, dimensionality reduction, clustering, etc.), then analyzes the design intent based on the BERT-CNN dual-channel, and classifies the fixture applicability in combination with a preset classification model. By combining the generative adversarial network with topological optimization, a lightweight fixture structure is generated, and topological optimization and simulation are performed on it to further optimize the error, and finally the dynamic optimization and efficient manufacturing of the fixture are achieved. In addition, based on the dynamically collected force data, time series analysis and multi-objective optimization methods are used to achieve real-time optimization of fixture processing parameters. The present invention provides a complete closed-loop system from design to manufacturing to optimization, which significantly improves the accuracy and production efficiency of fixture design, overcomes the limitations of traditional methods, and has high innovation and practicality.

[0007] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A schematic diagram of a flow chart of a method for collecting and analyzing 3D printing fixture data based on big data according to an embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of the 3D printing fixture data collection and analysis system based on big data described in an embodiment of the present invention.

[0010] In the figure: 701, acquisition unit; 702, cleaning unit; 703, classification unit; 704, optimization unit; 705, simulation unit; 706, analysis unit. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0012] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0013] Embodiment 1: This embodiment provides a method for collecting and analyzing 3D printing tooling and fixture data based on big data.

[0014] See also Figure 1 , the figure shows that the method includes step S1, step S2, step S3, step S4, step S5 and step S6.

[0015] Step S1, obtaining first information, wherein the first information includes a historical design scheme of a fixture, processing parameters and sensor data; It can be understood that the fixture design history in this step includes previous design examples, manufacturing processes, and applicable engineering parameters. These data can provide valuable references for the current design. The processing parameters cover the various parameters involved in the fixture manufacturing process, such as cutting speed, tool selection, coolant usage, etc. These are key factors affecting the quality and accuracy of the fixture. Sensor data comes from various sensors installed on the fixture during use, collecting real-time data such as force, deformation, vibration, temperature, etc. of the fixture. These data can reflect the dynamic behavior of the fixture at work and provide real and objective information for subsequent performance analysis and optimization.

[0016] Step S2, extracting features and cleaning data of the first information to obtain second information, where the second information includes the first information after data cleaning; It can be understood that this step not only effectively eliminates noise while cleaning the data, but also enhances the effective information in the data and extracts more accurate and representative features. The processed data is more usable and can help subsequent algorithms provide higher accuracy and robustness in optimizing fixture design and manufacturing. In this step, step S2 includes step S21, step S22, step S23 and step S24.

[0017] Step S21: According to the first information, high-dimensional feature dimensionality reduction and abnormal data detection processing are performed, wherein the first information is embedded with high-dimensional features by a variational autoencoder, and abnormal data in the first information is identified in combination with an isolation forest, and the distribution pattern of the abnormal data is analyzed by using a DBSCAN clustering algorithm, and isolated points and low-density area data are removed to obtain a denoised feature data set; It is understandable that this step first extracts original features from the first information, but the dimensions of these features are usually very high and contain redundant information. Direct use of these data will result in excessive computational complexity and may interfere with subsequent analysis and optimization. In order to reduce this complexity, a variational autoencoder is used for high-dimensional feature embedding. A variational autoencoder is a deep learning model that maps data to a latent space through a neural network and reduces the dimension through a reconstruction process. The variational autoencoder compresses the input high-dimensional data and embeds it into the latent space while maintaining the important features of the data for subsequent processing.

[0018] This step detects abnormal data for the embedded feature data. Among them, high-dimensional data contains some outliers. If these outliers are not removed, they will affect the accuracy of subsequent analysis. For this reason, the isolation forest algorithm is used for processing, which is specially used to process abnormal data. The working principle of the isolation forest is to recursively divide the isolated points by randomly dividing the data space. The abnormal data is easy to identify because it is relatively isolated. Therefore, it can efficiently process a large number of features and quickly identify isolated points without being affected by the data dimension.

[0019] However, after anomaly detection, there will still be some noise data or low-density areas, which may be mistakenly considered as abnormal data and removed, but in fact they are just a normal distribution pattern in the data. Therefore, this step also uses the DBSCAN clustering algorithm to analyze the data, identify low-density areas and isolated points in the data, and remove them. The DBSCAN clustering algorithm can determine which data belongs to the "group" and which data is distributed on the "edge" or isolated based on the density characteristics of the data.

[0020] After high-dimensional feature reduction, abnormal data detection and cluster analysis, the denoised feature data set can more accurately reflect the key factors in fixture design and processing, and the data quality is significantly improved.

[0021] Step S22, performing multi-scale feature extraction and alignment processing according to the denoised feature data set, wherein the features of different time scales in the fixture design and processing process are extracted by wavelet transform to obtain a feature-extracted data set; It can be understood that the input data set in this step is a feature data set after denoising, which contains various feature data at different time points in the fixture design and processing process. These data include information such as the mechanical parameters, temperature changes, and processing accuracy of the fixture, which are usually expressed as time series signals. Then, the Daubechies wavelet is selected to perform multi-layer decomposition of the original signal to achieve multi-scale analysis, decomposing the signal into a series of low-frequency (approximation) and high-frequency (details) parts, among which the Daubechies wavelet is used to convolve the signal and calculate the projection of the signal on different time scales; then the downsampling operation (this step is to reduce the data length by half each time decomposition) makes the low-frequency part obtained in each layer represent the larger scale changes of the signal, while the high-frequency part retains the smaller scale changes.

[0022] During the decomposition process, each layer of the signal represents the characteristics of different time scales. For example, the low-frequency part obtained by the first layer of decomposition corresponds to the large-scale changes in the signal (such as the gradual wear of the fixture over a long period of time), while the high-frequency part corresponds to the sudden changes in the signal (such as the instantaneous failure of the fixture). By continuing to decompose, more levels of signal details are obtained. The decomposition effect of each layer is shown through the different characteristics of the signal on the time scale.

[0023] Step S23, performing local consistency enhancement and structure preservation processing according to the feature-extracted data set, wherein the local structure of the fixture geometric features is preserved by Laplace feature mapping, and the data distribution is optimized in a low-dimensional space in combination with t-SNE to obtain feature data after structural optimization; It can be understood that this step uses the data set after wavelet transformation and feature extraction as input data. These input data contain key feature information in the fixture design and processing process (such as multi-scale features of time series data such as temperature, pressure, displacement, etc.), and then constructs an adjacency matrix based on the input data, in which each data point is connected to its most similar neighbor point in the feature space. The elements of the adjacency matrix represent the similarity between data points. In this step, a weighted graph is used for representation, and the weights are calculated by the Gaussian kernel function.

[0024] The adjacency matrix is ​​used to calculate the Laplacian matrix of the graph. The Laplacian matrix is ​​used to describe the local structural information of each node in the graph, which can help reveal the local geometric structure in the data set. By performing eigendecomposition on the matrix, eigenvectors are obtained, which capture the geometric relationship of the local neighborhood in the data. And through eigendecomposition, low-dimensional embedded feature data is obtained. The Laplacian eigenmap can more effectively reflect the local structural changes in the data by retaining local similarities rather than global variance.

[0025] The data set processed by Laplace eigenmap will be passed into the t-SNE algorithm as input for processing. By calculating the similarity between high-dimensional data points, the probability distribution is used to describe the degree of similarity. Among them, for each pair of data points, t-SNE first calculates their similarity in high-dimensional space, and then calculates their similarity in low-dimensional space, and makes the similarity in low-dimensional space consistent with that in high-dimensional space as much as possible. The optimization process of t-SNE is achieved by iteratively minimizing the Kullback-Leibler divergence (KL divergence), that is, the distribution of data points in low-dimensional space reflects the relative relationship of data points in high-dimensional space as much as possible. Through this optimization, the relative distance between data points can be retained, and the group structure and cluster structure in the original data will be effectively displayed in the low-dimensional space. Finally, the feature data with optimized structure is obtained.

[0026] Step S24: performing noise suppression and data smoothing processing according to the characteristic data after structural optimization to obtain second information.

[0027] It is understandable that there will be some abnormal fluctuations or high-frequency noise in the feature data after the structural optimization of this step. This step uses the physical model of the fixture processing process (such as the dynamic change law of the fixture displacement). The Kalman filter can predict the state of the fixture at a certain moment and update the predicted value by fusing the measurement data of the sensor, thereby effectively suppressing the noise in the sensor data. For example, if the temperature sensor data has instantaneous fluctuations due to changes in the external environment, the Kalman filter can smooth these instantaneous fluctuations and eliminate the high-frequency noise in the data. After the Kalman filter suppresses most of the noise, this step smoothes the data through local weighted regression to obtain a smoother and more accurate feature data set.

[0028] Step S3, analyzing the design intent of the second information based on the BERT-CNN dual channel, and classifying the fixture historical design schemes in combination with a preset classification model to obtain a fixture applicability classification database; It can be understood that this step uses the dual-channel analysis method of BERT and CNN. The BERT-CNN dual-channel analysis means that BERT is first responsible for extracting rich semantic representations from the input text, and then CNN performs further feature extraction and spatial structure optimization on these representations, thereby achieving a more efficient model for feature learning and classification tasks. This step can comprehensively consider the text semantics and structural features in the fixture design scheme, and comprehensively and accurately judge the applicability of each fixture design. This method not only improves the accuracy of classification, but also can efficiently handle the complexity of the fixture design scheme, ensuring that the most suitable design scheme can be quickly found during the fixture selection and optimization process, improving production efficiency and reducing resource waste. In this step, step S3 includes step S31, step S32 and step S33.

[0029] Step S31: Perform text and semantic understanding processing on the second information after data cleaning, wherein the historical fixture design scheme in the second information is encoded by the BERT model, the context information and potential semantic features of the fixture design are extracted, and the design features after semantic encoding are obtained; It can be understood that in this step, the input data is the second information after feature extraction, denoising and data smoothing. This information includes the text description of the fixture design, the technical parameters of the processing process, the processing records of the historical design, etc. In this process, the input data is mainly text information (such as the description of the fixture design), which contains important semantic information such as design intent, use environment, and technical requirements.

[0030] In order for the BERT model to efficiently process input text, the text data is first preprocessed, including removing stop words, segmenting, and standardizing. These processing steps can reduce the computational burden of the model while improving the accuracy of semantic understanding.

[0031] The BERT model receives this text data and captures the contextual information in the text through its bidirectional encoding capabilities. For example, BERT can identify the relationship between "precision milling" and "high-precision clamping" and then understand the core requirements of the fixture design plan. Through the processing of the BERT model, each word and sentence in the fixture design plan is converted into a high-dimensional semantic vector, from which BERT can extract potential semantic features such as fixture functions, applicable scenarios, and design constraints.

[0032] After encoding processing by the BERT model, the obtained design features are the deep semantic representation of the fixture design scheme. These features contain information such as the functional requirements, potential constraints, and applicable scenarios of the fixture design.

[0033] Step S32: performing image feature extraction processing according to the semantically encoded design features, wherein a CNN convolutional neural network is used to perform multi-level convolution and pooling operations on the semantically encoded design features, wherein the multi-level convolution includes extracting spatial correlation features through a sliding window mechanism, and fusing these features in combination with a fully connected layer to obtain visual features of the fixture history design scheme; It can be understood that this step takes the semantically encoded design features as input, i.e., the high-dimensional vector representation from the BERT model, which encodes the text information of the fixture design scheme and contains key features such as the fixture’s functional description, processing environment, and design parameters; In this step, the semantic features are arranged into a two-dimensional matrix according to preset rules. The preset rules are to map semantic features of different categories to fixed rows / columns; through the sliding window mechanism, a convolution operation is performed on the input semantic feature matrix to extract local spatial correlation features. For example: the first layer of convolution extracts low-level features (local associations between semantic units in the present invention). The second layer of convolution extracts intermediate features (syntactic structures and common design patterns in the present invention). The third layer of convolution extracts high-level features (complex semantic relationships in the present invention, such as "matching of clamping force and stiffness"). The convolution kernel of each layer will automatically learn the pattern of the data. For example: a small-sized convolution kernel (such as 3×3) can extract local semantic relationships, and a large-sized convolution kernel (such as 5×5) can identify larger semantic structures, such as the design intent expressed by the entire sentence. After each layer of convolution, a pooling layer is used to reduce the data dimension, improve computational efficiency, and retain key features. After multiple layers of convolution and pooling, the extracted features are flattened and input into the fully connected layer. The function of the fully connected layer is to fuse features at different levels to generate the final visual feature representation. For example, through the fully connected layer, the model can associate features such as "clamping force", "clamp material", and "processing accuracy" to form an overall fixture design feature representation. After being processed by the CNN convolutional neural network, the data of the fixture design solution can be represented in a graphical way, making the matching of similar solutions more intuitive.

[0034] Step S33, concatenate the semantically encoded design features with the visual features of the fixture historical design scheme to form a multimodal feature vector, and classify the fixture historical design scheme in combination with a support vector machine to obtain the applicability classification result of the fixture historical design scheme.

[0035] It can be understood that the design features after semantic encoding in this step include semantic information of the fixture design scheme, such as fixture type, design purpose, use environment and other features, and visual features of the fixture historical design scheme, including spatial correlation features of the fixture scheme extracted by convolutional neural network, such as clamping method, processing error control, etc.; This step uses a weighted splicing method to assign different weights to semantic features and visual features, and then splices the weighted features to obtain spliced ​​features; Then, the support vector machine training data set contains a large number of labeled fixture design solutions, annotated with their applicability categories (such as "suitable for aluminum alloy processing" and "suitable for precision processing", etc.). The optimal hyperplane is trained to maximize the classification interval of different types of fixture solutions. Then a trained support vector machine model is obtained, which can classify the newly input fixture design solutions and output their applicability categories. The use of support vector machines for classification enables the system to accurately determine the applicability of new fixture solutions based on historical fixture data. The automation capability of fixture design has been greatly improved, providing strong support for efficient fixture solution recommendation and optimization.

[0036] Step S4, generating a lightweight fixture structure through a generative adversarial network and topology optimization according to a fixture applicability classification database, and optimizing the lightweight fixture structure through a preset optimization model to obtain an optimized fixture topology structure; It can be understood that this step utilizes deep learning to generate adversarial networks and topology optimization, combined with a preset optimization model, to generate a lightweight fixture structure that meets specific manufacturing requirements, improves the mechanical properties of the fixture, and reduces material consumption. In this step, step S4 includes step S41, step S42, and step S43.

[0037] Step S41, performing lightweight fixture structure generation processing according to the fixture applicability classification database, wherein a candidate fixture structure is generated by a generative adversarial network, the generator network generates different fixture structure schemes according to the design parameters, material properties and process requirements in the applicability classification database, and the discriminator network performs authenticity judgment on these generated fixture structures to obtain a fixture structure that meets the preset lightweight design requirements; It can be understood that the fixture applicability classification database in this step includes design parameters (fixture size, fixing method, load capacity), material properties (density, strength, toughness, corrosion resistance) and process requirements (whether it is suitable for 3D printing, CNC processing, etc.). After taking it as input, the design parameters, material properties and process requirements are initially fused through the fully connected layer, and the three-dimensional form of the fixture is constructed layer by layer using a convolutional neural network. The structural rationality is improved through jump connections to avoid information loss, and then a forged fixture design scheme is obtained; The authenticity of the forged fixture design is evaluated by the discriminator, and then whether the fixture structure meets the lightweight design requirements is predicted. The adversarial loss is used for optimization as shown below: in, represents the adversarial loss function, represents the actual fixture design, represents a data sample of real fixture design solutions, represents the discriminant output of the real data, represents a random noise vector, represents the probability distribution of the noise variable, Indicates a forged fixture design. represents the judgment result of the forged fixture design. It is the expectation operator, which represents the mathematical expectation calculation of data distribution.

[0038] Step S42, performing topology optimization processing according to the generated lightweight fixture structure, wherein the shape and material distribution of the fixture structure are dynamically adjusted according to preset loading conditions and constraints to obtain a fixture structure after topology optimization; It can be understood that in this step, topology optimization is a method based on mathematical optimization, and its goal is to find the optimal material distribution scheme under the premise of given design space, load, boundary conditions and manufacturing constraints, so that the fixture can achieve the maximum lightweight while ensuring stiffness. This step uses the density method to optimize the material distribution, initialize the design area of ​​the fixture, discretize it into a finite element grid, define the material density variable and initialize it to 1 (that is, the initial model is full of material), calculate the stress and strain distribution of the fixture under external force according to the loading conditions, and finally iteratively remove the inefficient material area to generate the optimized topology structure. The optimized topology structure includes the fixture geometry and material distribution scheme obtained after topology optimization.

[0039] The mathematical expression of topology optimization in this step is as follows: in, represents the material density distribution vector, represents the objective function, is the deformation energy of the structure, represents the total number of units after the structure is discretized, Indicates The material density of each unit, p represents the penalty factor, Indicates The flexibility contribution of each element is represents the material volume of the current topology, Indicates The volume of a unit, Indicates the maximum allowed material volume, represents the minimum density value, Step S43: optimizing the parameters of the topology-optimized fixture structure based on a particle swarm optimization algorithm to obtain an optimized fixture topology structure.

[0040] In this step, the optimization objectives and parameters are defined by the particle swarm optimization algorithm. The optimization objectives include minimizing the quality of the fixture (reducing material usage and reducing costs), maximizing the stiffness of the fixture (improving structural strength and reducing deformation), optimizing the stress distribution (ensuring stable mechanical properties) and meeting the manufacturing process constraints (applicable to 3D printing and CNC processing). The optimization parameters include the position and shape of the key support structure, the hole layout and size (reducing weight), the material distribution density (determining the balance between stiffness and weight) and the thickness variation rules of the fixture structure. Then the particle swarm is initialized, and the deformation and stress distribution of the fixture under external force are evaluated using finite element analysis, and the fitness function is calculated. The fitness function is as follows: in, Represents the fixture design Fitness value, Indicates the quality of the fixture, Indicates the minimum value of the fixture quality, represents the maximum value of the fixture mass, represents the maximum value of the fixture stiffness, represents the fixture stiffness, represents the minimum value of the fixture stiffness, represents the maximum value of the variance of the fixture stress distribution, represents the variance of the fixture stress distribution, Represents the minimum value of the variance of the fixture stress distribution. Among them, , and are all weight coefficients.

[0041] Then, the particle velocity and position are iteratively updated through the fitness function to obtain a fixture topology structure with minimum mass, maximum stiffness and most uniform stress distribution.

[0042] Step S5, performing simulation according to the optimized fixture topology structure, and performing error optimization based on the simulation parameters to obtain fixture design data after error optimization; It can be understood that this step evaluates the mechanical properties of the fixture through finite element simulation, and combines error modeling and optimization algorithms to correct manufacturing and assembly errors. This step ensures the structural stability, accuracy and applicability of the fixture, and finally obtains the fixture design data after error optimization, providing a high-precision reference model for subsequent manufacturing and practical applications. In this step, step S5 includes step S51, step S52 and step S53.

[0043] Step S51, performing simulation processing according to the optimized fixture topology structure, wherein the force, deformation and thermal stress of the fixture under different working conditions are modeled and analyzed by finite element analysis to obtain simulation results, wherein the simulation results include the stress distribution, deformation and fault point location of the fixture during operation; It can be understood that this step adopts an adaptive meshing strategy to improve the calculation accuracy of high stress concentration areas. Select a suitable mesh unit type, which is a hexahedral unit in this step to ensure simulation accuracy and calculation efficiency. Then, set the fixed constraints of the fixture (the contact point with the workpiece in this step). Apply external loads, including clamping force, workpiece weight, and impact force, set thermal boundary conditions, such as thermal conductivity and convection heat transfer conditions, and then use static analysis calculation formulas to calculate the force conditions of the fixture, generate a stress distribution cloud map, and perform modal analysis to evaluate the vibration characteristics of the fixture, generate a deformation cloud map, and ensure that the fixture will not affect the processing accuracy due to excessive deformation. Perform thermal-mechanical coupling simulation to evaluate the impact of high temperature environment on fixture accuracy, identify fault points, find possible failure risks, and increase the service life of the fixture.

[0044] The static analysis calculation formula is as follows: in, represents the global stiffness matrix, which is the mechanical properties of the material and the geometric characteristics of the structure. represents the displacement vector, which is the displacement distribution of the structure under the force. Represents the external force load vector, which is the external load acting on the structure.

[0045] Step S52, performing error analysis processing according to the simulation results, wherein the errors that may be generated by the fixture under different working conditions are evaluated by constructing a Gaussian process regression model to obtain potential error sources in the fixture design; It can be understood that this step predicts the mean and variance of the error through GPR modeling, and clarifies the error variation range of the fixture under different working conditions. The error source of the fixture is analyzed through Gaussian process regression, the error range under different working conditions is predicted, and finally the potential error source in the fixture design is obtained. The error analysis results of this step will be used as error optimization to provide accurate optimization direction for fixture design and improve overall accuracy and stability.

[0046] Step S53: Optimize and adjust the design parameters of the error model based on the potential error sources in the fixture design to obtain the fixture design data after error optimization.

[0047] It is understandable that this step pre-adds compensation structures to the fixture design, including: adding reinforcing ribs to reduce thermal deformation. Optimizing the clamping force distribution to reduce elastic deformation. Adjusting the processing tolerance to match the assembly accuracy. The fixture design data after error optimization makes the fixture more in line with the actual application requirements and improves the processing accuracy. By optimizing the material distribution and error compensation design, the stability of the fixture during long-term use is improved. The fixture after error optimization can adapt to different temperatures, loads and assembly conditions, improve applicability, and the adjusted design parameters obtained through error analysis during simulation and actual manufacturing. These data cover the geometry, material properties, stress distribution, deformation and other mechanical performance indicators of the fixture. After error optimization processing, the design of the fixture is more in line with the actual working conditions, reducing potential error sources, and ensuring that the fixture has higher accuracy and reliability during manufacturing and use. In addition, the design data after error optimization also includes processing parameters based on real-time feedback and dynamic optimization, so as to facilitate more accurate manufacturing and adjustment of the fixture.

[0048] Step S6: manufacturing the fixture according to the fixture design data after error optimization, and collecting the fixture force data for dynamic optimization to obtain optimized fixture processing parameters.

[0049] It is understandable that the fixture is manufactured according to the fixture design data after error optimization, and the force data is collected for dynamic optimization. By real-time monitoring and optimizing the process parameters, the optimized fixture processing parameters are obtained. By dynamically adjusting the fixture processing parameters, the processing accuracy and stability of the fixture can be improved, the adaptability and durability of the fixture can be enhanced, and the efficiency of the manufacturing process and the product quality can be finally improved. In this step, step S6 includes step S61, step S62 and step S63.

[0050] Step S61: Perform fixture manufacturing processing according to the fixture design data after error optimization, and collect force data of the fixture during the processing in real time based on a preset sensor, wherein the force data includes pressure, deformation and vibration data of the fixture during the working process; It can be understood that this step is to carry out the actual manufacture of the fixture according to the fixture design data after error optimization, to ensure that the fixture meets the design requirements and has high precision. Based on the preset sensors, the force data of the fixture during the processing is collected in real time, including the pressure, deformation and vibration data of the fixture. Through the installed sensors, the force data of the fixture during the processing is collected in real time. These data include: Pressure data: record the clamping force applied by the fixture during the working process to ensure that the clamping force of the fixture on the workpiece during the processing is uniform and stable. Deformation data: monitor the deformation of the fixture through strain gauges or other sensors to ensure that the fixture does not deform too much during the processing to avoid affecting the processing accuracy. Vibration data: monitor the vibration of the fixture during processing through vibration sensors, especially in high-speed processing, to ensure that the fixture will not cause processing instability due to vibration.

[0051] Step S62: Perform dynamic data analysis and processing based on the real-time collected clamp force data, wherein the force data is modeled by dynamic time series analysis, the force change trend of the clamp at different operation stages is analyzed, and potential abnormalities and uneven force conditions are identified by a decision tree to obtain dynamic analysis results; It can be understood that this step smoothes, denoises and normalizes the force data to ensure the quality of the data and remove interference factors. The long short-term memory network is used to analyze the time series characteristics of the force data and identify the force change trend of the fixture in different operation stages (such as clamping, processing, loosening, etc.). Based on the time series model, the mechanical changes of the fixture during the working process are analyzed to reveal the changes in the force of the fixture over time, including pressure, deformation and vibration. On the basis of time series analysis, the change trend of the force data of the fixture is monitored to detect abnormal fluctuations in force in a timely manner. For example: a sudden increase or decrease in force data may indicate that the clamp has a problem of too much or too little clamping force. Too fast or unstable force changes may indicate deformation or vibration problems of the fixture. Time series anomalies may reflect unstable operation of the fixture under different working conditions, or abnormal wear or damage.

[0052] Then, various types of force data (such as pressure, deformation, vibration, etc. of the fixture) obtained by time series analysis are used as input features of the decision tree model. Based on historical data and actual operation experience, the decision tree model is trained to determine the decision rules for various force changes. For example: if the pressure of the fixture suddenly exceeds the predetermined range, the decision tree is judged as "abnormal". If the deformation of the fixture is greater than the preset threshold, it is judged as "uneven force". The results of the dynamic analysis are obtained. Through dynamic time series analysis and decision tree models, the force change trend of the fixture at different stages can be more accurately analyzed, abnormalities and uneven force problems can be discovered in time, and the accuracy of fixture design and operation can be improved. And through real-time dynamic data analysis, the force anomaly during the fixture operation can be fed back instantly, potential problems can be identified in advance, and production interruptions or product quality problems can be avoided.

[0053] Step S63: Optimize the fixture processing parameters according to the dynamic analysis results to obtain optimized fixture processing parameters.

[0054] It can be understood that this step selects parameters closely related to the fixture processing process, and the fixture processing parameters include the fixture clamping force, the design of the support point, the geometry of the fixture, the material properties, the working pressure, etc. The selected parameters are optimized by using the simulated annealing algorithm as the optimization algorithm. The simulated annealing algorithm can efficiently find the optimal solution in the multi-dimensional parameter space. By optimizing the fixture processing parameters, the fixture is ensured to operate stably during the working process, and the processing errors caused by uneven force, deformation and other problems are reduced during the processing process, thereby improving the processing accuracy.

[0055] Embodiment 2: like Figure 2 As shown, this embodiment provides a 3D printing fixture data collection and analysis system based on big data, see Figure 2 The system includes an acquisition unit 701 , a cleaning unit 702 , a classification unit 703 , an optimization unit 704 , a simulation unit 705 and an analysis unit 706 .

[0056] An acquisition unit 701 is used to acquire first information, wherein the first information includes a historical design scheme of a fixture, processing parameters and sensor data; A cleaning unit 702, configured to perform feature extraction and data cleaning on the first information to obtain second information, wherein the second information includes the first information after data cleaning; The classification unit 703 is used to analyze the design intent of the second information based on the BERT-CNN dual channel, and classify the historical design schemes of the fixture in combination with a preset classification model to obtain a fixture applicability classification database; The optimization unit 704 is used to generate a lightweight fixture structure through a generative adversarial network and topology optimization according to the fixture applicability classification database, and optimize the lightweight fixture structure through a preset optimization model to obtain an optimized fixture topology structure; The simulation unit 705 is used to perform simulation according to the optimized fixture topology structure and perform error optimization based on the simulation parameters to obtain the fixture design data after error optimization; The analysis unit 706 is used to manufacture the fixture using the fixture design data after error optimization, and collect the fixture force data for dynamic optimization to obtain the optimized fixture processing parameters.

[0057] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention. The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for collecting and analyzing 3D printing fixture data based on big data, characterized in that: include: Acquiring first information, the first information including historical fixture design solutions, machining process parameters, and sensor data; Performing feature extraction and data cleaning on the first information to obtain second information, wherein the second information includes the first information after data cleaning; The design intent of the second information is analyzed based on the BERT-CNN dual channel, and the historical design schemes of the fixture are classified in combination with the preset classification model to obtain the fixture applicability classification database; According to the fixture applicability classification database, a lightweight fixture structure is generated through generative adversarial networks and topology optimization, and the lightweight fixture structure is optimized through a preset optimization model to obtain an optimized fixture topology structure; Perform simulation according to the optimized fixture topology, and perform error optimization based on the simulation parameters to obtain the fixture design data after error optimization; The fixture is manufactured according to the fixture design data after error optimization, and the fixture force data is collected for dynamic optimization to obtain the optimized fixture processing parameters.

2. The method for collecting and analyzing 3D printing fixture data based on big data according to claim 1 is characterized in that , extracting features and cleaning data of the first information to obtain second information, wherein the second information includes the first information after data cleaning, including: According to the first information, high-dimensional feature dimensionality reduction and abnormal data detection processing are performed, wherein the first information is embedded with high-dimensional features by a variational autoencoder, and abnormal data in the first information is identified in combination with an isolation forest, and the distribution pattern of the abnormal data is analyzed by using a DBSCAN clustering algorithm, and isolated points and low-density area data are removed to obtain a denoised feature data set; Based on the denoised feature data set, multi-scale feature extraction and alignment processing is performed, wherein the features of different time scales in the fixture design and processing process are extracted by wavelet transform to obtain a feature-extracted data set; According to the feature-extracted data set, local consistency enhancement and structure preservation processing are performed, wherein the local structure of the fixture geometric features is maintained through Laplace eigenmapping, and the data distribution is optimized in low-dimensional space in combination with t-SNE to obtain feature data with optimized structure; According to the characteristic data after structural optimization, noise suppression and data smoothing processing are performed to obtain the second information.

3. The method for collecting and analyzing 3D printing fixture data based on big data according to claim 2 is characterized in that ,Based on the BERT-CNN dual-channel analysis of the design intent of the second information, and combined with the preset classification model to classify the historical design schemes of the fixture, the fixture applicability classification database is obtained, including: According to the second information after data cleaning, text and semantic understanding processing is performed, wherein the historical fixture design scheme in the second information is encoded through the BERT model, the context information and potential semantic features of the fixture design are extracted, and the design features after semantic encoding are obtained; According to the semantically encoded design features, image feature extraction processing is performed, wherein the semantically encoded design features are subjected to multi-level convolution and pooling operations through a CNN convolutional neural network, wherein the multi-level convolution includes extracting spatial correlation features through a sliding window mechanism, and fusing these features in combination with a fully connected layer to obtain visual features of the fixture's historical design scheme; The semantically encoded design features are spliced ​​with the visual features of the historical fixture design schemes to form a multimodal feature vector. The historical fixture design schemes are classified using a support vector machine to obtain the applicability classification results of the historical fixture design schemes.

4. The method for collecting and analyzing 3D printing fixture data based on big data according to claim 1 is characterized in that ,According to the fixture applicability classification database, a lightweight fixture structure is generated through generative adversarial networks and topology optimization, and the lightweight fixture structure is optimized through a preset optimization model to obtain the optimized fixture topology structure, including: According to the fixture applicability classification database, lightweight fixture structure generation processing is performed, wherein candidate fixture structures are generated through a generative adversarial network, and the generator network generates different fixture structure schemes according to the design parameters, material properties and process requirements in the applicability classification database, while the discriminator network makes authenticity judgments on these generated fixture structures to obtain fixture structures that meet the preset lightweight design requirements; Based on the generated lightweight fixture structure, a topology optimization process is performed, wherein the shape and material distribution of the fixture structure are dynamically adjusted according to the preset loading conditions and constraints to obtain a fixture structure after topology optimization; The parameters of the fixture structure after topology optimization are optimized based on the particle swarm optimization algorithm to obtain the optimized fixture topology structure.

5. The method for collecting and analyzing 3D printing fixture data based on big data according to claim 1, characterized in that ,According to the optimized fixture topology structure, simulation is performed, and error optimization is performed based on the simulation parameters to obtain the fixture design data after error optimization, including: According to the optimized fixture topology, simulation processing is performed, wherein the force, deformation and thermal stress of the fixture under different working conditions are modeled and analyzed by finite element analysis to obtain simulation results, which include the stress distribution, deformation and fault point location of the fixture during operation; According to the simulation results, error analysis is performed. A Gaussian process regression model is constructed to evaluate the possible errors of the fixture under different working conditions and obtain the potential error sources in the fixture design. Based on the potential error sources in the fixture design, the design parameters of the error model are optimized and adjusted to obtain the fixture design data after error optimization.

6. A 3D printing fixture data collection and analysis system based on big data, characterized in that: include: An acquisition unit, configured to acquire first information, wherein the first information includes a historical design scheme of the fixture, processing parameters, and sensor data; a cleaning unit, configured to perform feature extraction and data cleaning on the first information to obtain second information, wherein the second information includes the first information after data cleaning; A classification unit is used to analyze the design intent of the second information based on the BERT-CNN dual channel, and classify the historical design schemes of the fixture in combination with a preset classification model to obtain a fixture applicability classification database; The optimization unit is used to generate a lightweight fixture structure according to the fixture applicability classification database through generative adversarial networks and topology optimization, and optimize the lightweight fixture structure through a preset optimization model to obtain an optimized fixture topology structure; A simulation unit is used to perform simulation according to the optimized fixture topology and perform error optimization based on the simulation parameters to obtain the fixture design data after error optimization; The analysis unit is used to manufacture the fixture based on the fixture design data after error optimization, and collect the fixture force data for dynamic optimization to obtain the optimized fixture processing parameters.

7. The system for collecting and analyzing 3D printing fixture data based on big data according to claim 6 is characterized in that: The cleaning unit comprises: A first cleaning subunit is used to perform high-dimensional feature dimensionality reduction and abnormal data detection processing according to the first information, wherein the first information is embedded with high-dimensional features by a variational autoencoder, and abnormal data in the first information is identified in combination with an isolation forest, and a distribution pattern of the abnormal data is analyzed by using a DBSCAN clustering algorithm, and isolated points and low-density area data are removed to obtain a denoised feature data set; The second cleaning subunit is used to perform multi-scale feature extraction and alignment processing based on the denoised feature data set, wherein the features of different time scales in the fixture design and processing process are extracted by wavelet transform to obtain a feature-extracted data set; The third cleaning subunit is used to perform local consistency enhancement and structure preservation processing according to the feature-extracted data set, wherein the local structure of the fixture geometric features is maintained by Laplace feature mapping, and the data distribution is optimized in the low-dimensional space in combination with t-SNE to obtain feature data after structural optimization; The fourth cleaning subunit is used to perform noise suppression and data smoothing processing according to the characteristic data after structural optimization to obtain second information.

8. The system for collecting and analyzing 3D printing fixture data based on big data according to claim 6, characterized in that: The classification unit includes: The first classification subunit is used to perform text and semantic understanding processing based on the second information after data cleaning, wherein the historical fixture design scheme in the second information is encoded by the BERT model, the context information and potential semantic features of the fixture design are extracted, and the design features after semantic encoding are obtained; The second classification subunit is used to perform image feature extraction processing based on the semantically encoded design features, wherein the semantically encoded design features are subjected to multi-level convolution and pooling operations through a CNN convolutional neural network, wherein the multi-level convolution includes extracting spatial correlation features through a sliding window mechanism, and fusing these features in combination with a fully connected layer to obtain visual features of the fixture history design scheme; The third classification subunit is used to splice the semantically encoded design features with the visual features of the fixture historical design scheme to form a multimodal feature vector, and classify the fixture historical design scheme in combination with the support vector machine to obtain the applicability classification result of the fixture historical design scheme.

9. The system for collecting and analyzing 3D printing fixture data based on big data according to claim 6, characterized in that: The optimization unit comprises: The first optimization subunit is used to generate a lightweight fixture structure according to the fixture applicability classification database, wherein a candidate fixture structure is generated by a generative adversarial network, the generator network generates different fixture structure solutions according to the design parameters, material properties and process requirements in the applicability classification database, and the discriminator network performs authenticity judgment on these generated fixture structures to obtain a fixture structure that meets the preset lightweight design requirements; The second optimization subunit is used to perform topology optimization processing according to the generated lightweight fixture structure, wherein the shape and material distribution of the fixture structure are dynamically adjusted according to preset loading conditions and constraints to obtain a fixture structure after topology optimization; The third optimization subunit is used to optimize the parameters of the fixture structure after topology optimization based on the particle swarm optimization algorithm to obtain the optimized fixture topology structure.

10. The 3D printing fixture data collection and analysis system based on big data according to claim 6, characterized in that: The simulation unit comprises: The first simulation subunit is used to perform simulation processing according to the optimized fixture topology structure, wherein the force, deformation and thermal stress of the fixture under different working conditions are modeled and analyzed by finite element analysis to obtain simulation results, and the simulation results include the stress distribution, deformation and fault point location of the fixture during operation; The second simulation sub-unit is used to perform error analysis based on the simulation results, wherein the possible errors caused by the fixture under different working conditions are evaluated by constructing a Gaussian process regression model to obtain potential error sources in the fixture design; the third simulation sub-unit is used to optimize and adjust the design parameters of the error model based on the potential error sources in the fixture design to obtain the fixture design data after error optimization.

Images