Multi-axis machine tool part quality tracing method and system

Abstract

The invention relates to the technical field of machine tool precision analysis, in particular to a multi-axis machine tool part quality tracing method and system. The method comprises the steps that a time sequence cutting point location data set and corresponding label data in the part machining process are acquired; constructing a bidirectional long-short term memory network model based on an attention mechanism; on the basis of a multi-body system theory, a homogeneous transformation matrix is adopted to represent a pose transformation relation between adjacent topological structures of the machine tool, and a multi-axis machine tool space error model is established; and inputting the motion axis feed amount of the defect cutting point identified by the defect classification model into the geometric error sensitivity analysis model, quantifying the contribution degree of the geometric error to the machining error, and obtaining a part quality traceability result based on a statistical result of the machining process. According to the method, accurate traceability analysis of the multi-axis machine tool from defect cutting points to key geometric errors is realized, weight information of influence of various geometric errors on machining precision is provided, and optimization design of machine tool precision is effectively guided.

Description

Technical Field

[0001] The present invention relates to the technical field of machine tool precision analysis, and more specifically, to a multi-axis machine tool part quality tracing method and system. Background Art

[0002] CNC machine tools are basic equipment in the manufacturing industry and play a vital role in improving the innovation capability and market competitiveness of the basic manufacturing equipment industry.

[0004] At present, mid- and low-end CNC machine tools still dominate the market, and there is still much room for improvement in processing accuracy and core component performance.

[0005] Multi-axis CNC machine tools, especially five-axis CNC machine tools, have been widely used in modern manufacturing due to their high level of automation, good flexibility and high processing accuracy. Compared with traditional three-axis machine tools, five-axis machine tools significantly improve processing performance and accuracy by adding two rotating axes, and are particularly suitable for fine processing of complex parts.

[0006] Although five-axis machine tools have many advantages, quality traceability still faces major challenges in the production of high-end five-axis machine tools (such as double rotary table machine tools). At present, quality traceability mainly relies on manual references and manual experience, lacking a systematic, scientific method and digital quality management system.

[0007] In view of the complexity of five-axis CNC machining, solving the problem of quality traceability is of decisive significance for promoting the further development of this technology. Establishing a complete digital quality traceability system will help improve the accuracy and efficiency of component processing, help break through the various limitations faced in the current production process, and provide solid technical support for the continuous innovation and upgrading of the manufacturing industry. Summary of the invention

[0008] The purpose of the present invention is to provide a multi-axis machine tool part quality traceability method and system to solve the problem that the prior art is difficult to achieve quality traceability of processed parts for multi-axis CNC machine tools, especially five-axis CNC machine tools.

[0009] In order to achieve the above object, the present invention provides a multi-axis machine tool part quality traceability method, comprising the following steps:

[0010] Step S1, obtaining a sequential cutting point data set and corresponding label data during part processing;

[0011] Step S2, constructing a bidirectional long short-term memory network model based on an attention mechanism, using the cutting point data and the adjacent point data combined with the corresponding label data for training, to obtain a defect classification model;

[0012] Step S3, based on the multi-body system theory, a homogeneous transformation matrix is ​​used to characterize the posture conversion relationship between adjacent topological structures of the machine tool, and a spatial error model of the multi-axis machine tool is established;

[0013] Step S4, establish a geometric error sensitivity analysis model based on the multi-axis machine tool spatial error model, input the moving axis feed amount of the defective cutting point identified by the defect classification model into the geometric error sensitivity analysis model, quantify the contribution of geometric error to machining error, and obtain the part quality traceability result based on the statistical results of the machining process.

[0014] In some embodiments, the step S1 further comprises:

[0015] Step S101, loading the CAD model of the multi-axis machine tool, the tool model and the workpiece CAD model through the VERICUT simulation software, constructing a virtual machining environment, and setting the assembly relationship and relative position between the models;

[0016] Step S102, generating geometric error values ​​of each motion axis according to a preset geometric error range, so as to adjust the relative positions between the motion axes of the multi-axis machine tool and simulate the geometric error in the actual machining process;

[0017] Step S103, importing the NC program of the part to be processed, setting simulation parameters and performing simulation, and recording and exporting the time series data of the processing parameters;

[0018] Step S104, preprocessing the collected time series data, and adding corresponding quality labels to the preprocessed data according to the processing quality evaluation standard.

[0019] In some embodiments, the step S2 further comprises:

[0020] Step S201, constructing an input layer, wherein the input layer is used to input the original data of the cutting state of the cutting point and the adjacent points;

[0021] Step S202, constructing a feature extraction layer, wherein the feature extraction layer is used to extract features from the original data and generate a feature matrix, and the feature extraction layer at least includes a bidirectional long short-term memory network;

[0022] Step S203, constructing a multi-head attention mechanism module, wherein the multi-head attention mechanism module is used to perform weight optimization on the feature matrix to generate an optimized feature matrix;

[0023] Step S204, constructing a classification layer, classifying the quality status of the cutting points based on the optimized feature matrix, and outputting the quality status of the cutting points after classification and identification.

[0024] In some embodiments, the bidirectional long short-term memory network outputs a latent vector representing the defect classification type through forward long short-term memory and backward long short-term memory operations.

[0025] In some embodiments, the multi-head attention mechanism module adds the output features to the input features through a residual connection, and finally outputs an optimized feature matrix.

[0026] In some embodiments, the classification layer includes a fully connected layer and a Softmax classifier:

[0027] The fully connected layer converts the optimized feature matrix output by the multi-head attention mechanism module into a one-dimensional feature sequence;

[0028] The Softmax classifier classifies the one-dimensional feature sequence processed by the fully connected layer and outputs the probability distribution of the defect classification type.

[0029] In some embodiments, in step S204, the Focal Loss loss function is used as the classification loss function, and the corresponding expression is:

[0030] L=-α t (1-p t )γlog(p t );

[0031] Among them, α t is the balance factor, p t is the model’s predicted probability for the true category, and γ is the focusing parameter.

[0032] In some embodiments, in step S204, a binary cross entropy loss function is used as a classification loss function, and the corresponding expression is:

[0033]

[0034] Where S is the number of samples, y j is the true label of the jth sample, Represents the predicted probability value of the jth sample.

[0035] In some embodiments, after step S204, the method further includes:

[0036] Evaluate the established defect classification model based on performance evaluation indicators;

[0037] The performance evaluation indicators include precision, recall and F1-score.

[0038] In some embodiments, step S3 further comprises:

[0039] Step S301, establishing the topological structure of the multi-axis machine tool, treating each component of the machine tool as a rigid body, and defining the relative motion relationship;

[0040] Step S302, performing geometric error analysis on the multi-axis CNC machine tool to obtain error elements of the translation axis and the rotation axis;

[0041] Step S303, based on the geometric error analysis results, a homogeneous transformation matrix is ​​used to describe the relative position and posture relationship between adjacent topological structures of the machine tool, a characteristic matrix of the CNC machine tool is established, and finally a spatial error model of the multi-axis machine tool is obtained.

[0042] In some embodiments, the step S303 further includes:

[0043] For the BC dual-rotary table model of the five-axis machine tool, the actual forward kinematics model of the whole kinematic chain of the machine tool is constructed by integrating the actual forward kinematics model of the workpiece chain and the tool chain;

[0044] Based on the ideal forward kinematics model of the workpiece chain and tool chain of the machine tool, the ideal forward kinematics model of the whole kinematic chain of the machine tool is established;

[0045] A geometric error-tool posture error model is established to analyze the influence of geometric error on the relative posture error between the tool, grinding wheel and workpiece gear.

[0046] In some embodiments, the step S4 further comprises:

[0047] Step S401, quantifying the error correlation of each cutting point according to the multi-axis machine tool spatial error model, and establishing a geometric error sensitivity analysis model;

[0048] Step S402, inputting the defective cutting point data identified in step S2 into a geometric error sensitivity analysis model for calculation, quantifying the influence of geometric errors on the machining accuracy of each cutting point, and obtaining the error contribution of all defective cutting points;

[0049] Step S403, based on the quantified results of the error contribution, data standardization is performed to identify the error source that has the greatest impact on the processing quality, clarify the key geometric errors, and obtain the part quality traceability results.

[0050] In order to achieve the above object, the present invention provides a multi-axis machine tool parts quality traceability system, comprising:

[0051] a memory for storing instructions executable by a processor;

[0052] The processor is used to execute the instructions to implement the above method.

[0053] In order to achieve the above object, the present invention provides a computer readable medium having computer instructions stored thereon, wherein:

[0054] When the computer instructions are executed by a processor, the above method is performed.

[0055] The present invention proposes a multi-axis machine tool part quality traceability method and system, innovatively combining the BiLSTM neural network with the machine tool error modeling technology, realizing accurate traceability analysis from defective cutting points to key geometric errors, providing machine tool design engineers with weight information on the impact of various geometric errors on machining accuracy, effectively guiding the optimal design of machine tool accuracy, and laying a solid theoretical foundation for improving the accuracy of CNC machine tools. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The above and other features, properties and advantages of the present invention will become more apparent through the following description in conjunction with the accompanying drawings and embodiments, in which the same reference numerals always represent the same features, wherein:

[0057] Figure 1 A step diagram of a multi-axis machine tool part quality traceability method according to an embodiment of the present invention is disclosed;

[0058] Figure 2 A detailed flow chart of a multi-axis machine tool part quality traceability method according to an embodiment of the present invention is disclosed;

[0059] Figure 3 A specific structural diagram of a BiLSTM layer according to an embodiment of the present invention is disclosed;

[0060] Figure 4 A structural diagram of a multi-head attention mechanism according to an embodiment of the present invention is disclosed;

[0061] Figure 5 A topological structure diagram of a BC double-turntable five-axis CNC machine tool according to an embodiment of the present invention is disclosed;

[0062] Figure 6 A schematic diagram of the kinematic chain of a BC dual-turret five-axis CNC machine tool according to an embodiment of the present invention is disclosed;

[0063] Figure 7 A schematic diagram of error correlation according to an embodiment of the present invention is disclosed;

[0064] Figure 8 A schematic diagram of the correlation coefficient of the main errors in the X direction according to an embodiment of the present invention is disclosed;

[0065] Fig. 9 A schematic diagram of the main error correlation coefficient in the Y direction according to an embodiment of the present invention is disclosed;

[0066] Fig.10 A schematic diagram of the correlation coefficient of the main errors in the Z direction according to an embodiment of the present invention is disclosed. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the invention and are not used to limit the invention.

[0068] In view of the complexity of multi-axis CNC machining, solving the quality traceability problem is crucial to promoting the advancement of this technology. Establishing a complete digital quality traceability system will help improve the accuracy and efficiency of component processing and break through the limitations of current production processes. The present invention provides a part quality traceability method and system based on BiLSTM and geometric error modeling, establishes a complete digital quality traceability system, can realize accurate traceability analysis from defective cutting points to key geometric errors, and effectively guide the precision optimization design of machine tools.

[0069] The basic framework of the present invention includes four main parts: using VERICUT simulation software to obtain a sequential cutting point data set, constructing a BiLSTM model based on an attention mechanism, establishing a multi-axis machine tool spatial error model based on multi-body system theory, and constructing a machine tool key geometric error tracing analysis model. The following embodiments mainly describe a five-axis machine tool, especially a BC double-turntable five-axis CNC machine tool. The BC double-turntable five-axis CNC machine tool is a high-precision, high-efficiency CNC machining equipment that is widely used in the machining of complex parts. It combines double turntables and five-axis linkage technology.

[0070] However, the present invention can actually be extended to all multi-axis CNC machine tools and can be widely used in various CNC machining scenarios.

[0071] Figure 1 The following is a step diagram of a method for tracing the quality of multi-axis machine tool parts according to an embodiment of the present invention. Figure 1 As shown, a multi-axis machine tool parts quality traceability method proposed by the present invention comprises the following steps:

[0072] Step S1, obtaining a sequential cutting point data set and corresponding label data during part processing;

[0073] Step S2, constructing a bidirectional long short-term memory network model (Bidirectional Long Short-Term Memory, BiLSTM) based on the attention mechanism, using the cutting point data and the adjacent point data combined with the corresponding label data for training, to obtain a defect classification model;

[0074] Step S3, based on the multi-body system theory, a homogeneous transformation matrix is ​​used to characterize the posture conversion relationship between adjacent topological structures of the machine tool, and a spatial error model of the multi-axis machine tool is established;

[0075] Step S4, establish a geometric error sensitivity analysis model based on the multi-axis machine tool spatial error model, input the moving axis feed amount of the defective cutting point identified by the defect classification model into the geometric error sensitivity analysis model, quantify the contribution of geometric error to machining error, and obtain the part quality traceability result based on the statistical results of the machining process.

[0076] Figure 2 A detailed flow chart of a method for tracing the quality of multi-axis machine tool parts according to an embodiment of the present invention is disclosed. The following takes a five-axis CNC machine tool as an example and combines Figure 1 and Figure 2 It should be understood that within the scope of the present invention, the above-mentioned technical features of the present invention and the technical features specifically described below (such as embodiments) can be combined and related to each other to form a preferred technical solution.

[0077] Step S1, obtaining a sequential cutting point data set and corresponding label data during part processing;

[0078] In this embodiment, VERICUT simulation software is used to obtain a sequential cutting point data set and its corresponding label data of a part simulation cutting process.

[0079] VERICUT software is a CNC machining simulation system in the prior art. It consists of NC (Numerical Control) program verification module, machine tool motion simulation module, optimization path module and other modules. It can simulate the CNC machining process of various machining equipment such as CNC lathes, milling machines, machining centers, wire cutting machines and multi-axis machine tools.

[0080] Furthermore, the step S1 further comprises the following steps:

[0081] Step S101, loading the CAD (Computer-Aided Design) model of the five-axis CNC machine tool, the tool model and the workpiece CAD model through VERICUT simulation software, building a complete virtual machining environment, and setting the assembly relationship and relative position between the models;

[0082] Step S102, using a random number generator to generate geometric error values ​​of each motion axis according to a preset geometric error range, so as to adjust the relative positions between the motion axes of the machine tool, thereby simulating the geometric errors that may occur in the actual machining process;

[0083] Step S103, importing the NC program of the part to be processed into VERICUT software, setting appropriate simulation parameters and performing simulation, and recording and exporting the time series data of processing parameters such as cutting points and feed rates during the simulation process;

[0084] Step S104, preprocessing the collected time series data, and adding corresponding quality labels to the processed data according to the processing quality evaluation standard;

[0085] The preprocessing process includes cleaning the data, removing outliers and redundant information, and standardizing the data.

[0086] Through this series of steps, the accuracy and reliability of the simulation data are ensured.

[0087] Step S2, constructing a bidirectional long short-term memory network model based on an attention mechanism, using the cutting point data and the adjacent point data combined with the corresponding label data for training, to obtain a defect classification model;

[0088] Furthermore, the step S2 further comprises the following steps:

[0089] Step S201: constructing an input layer, wherein the input layer is used to input the original data of the cutting state of the cutting point and the adjacent points;

[0090] Step S202, constructing a feature extraction layer, wherein the feature extraction layer is used to extract features from the original data and generate a feature matrix, and the feature extraction layer at least includes a BiLSTM neural network;

[0091] The output of the feature extraction layer is a feature matrix, which compresses the data of the cutting point into a meaningful latent vector representing the characteristics of the input data.

[0092] It should be noted that the BiLSTM neural network is based on the LSTM (Long Short-Term Memory Network). It draws on the idea of ​​human understanding of the connection between words and introduces the deformation structure formed by the concepts of positive and negative time directions. It can capture the connection between time series data. It processes the input sequence simultaneously through two LSTMs (one forward and one backward), enhances the learning of the front-to-backward dependency in the time series, and thus improves the accuracy of the model in tasks such as defect classification.

[0093] Figure 3 The specific structure diagram of the BiLSTM neural network according to an embodiment of the present invention is disclosed, such as Figure 3 As shown, the BiLSTM neural network is used to input the original vector x 0 、x 1 、x 2 , ..., x t, output the latent vector h representing the defect classification type 0 、h 1 、h 2 , ..., h t .

[0094] Each hidden layer unit of the BiLSTM neural network structure stores two state information, namely the hidden state of the forward LSTM and the backward LSTM at the time step. The corresponding expression is:

[0095]

[0096] in, represents the hidden state of the forward LSTM at time step t, represents the hidden state of the backward LSTM at time step t, x t represents the input vector at the tth time step, h t represents the hidden state of BiLSTM at time step t.

[0097] The BiLSTM neural network uses known time series and reverse position sequences, and deepens the feature extraction level of the original sequence through forward and backward bidirectional operations, which can effectively improve the accuracy of the model output results.

[0098] In the feature extraction network layer, its input is an image and its output is the feature of the image, that is, the feature map.

[0099] The feature extraction network module layer is the cornerstone of the entire target detection process. Its main purpose is to reduce the data dimension without losing important features of the data, and effectively organize the features that can be used by the subsequent region candidate network layer and the fully connected layer.

[0100] Step S203: construct a multi-head attention mechanism module, wherein the multi-head attention mechanism module is used to optimize the weights of the feature matrix and generate an optimized feature matrix.

[0101] The multi-head attention mechanism module is used to strengthen the latent vector output by the BiLSTM module to enhance the relevance and conformity of information, and finally obtain a more optimized feature vector.

[0102] The multi-head attention mechanism refers to the use of multiple attention heads in parallel, with each head learning a set of weights independently. Each attention head captures the relevance of input features from a different perspective, thus providing richer information. Ultimately, these features from each head are concatenated together to form the final optimized feature vector.

[0103] Figure 4 The structure diagram of the multi-head attention mechanism according to one embodiment of the present invention is disclosed, Figure 4The multi-head attention mechanism shown in the figure is applied in the BiLSTM network. The attention results of each head are obtained by linearly transforming the input vectors V (value), K (key), and Q (query), and then calculating the attention results of each head through the shrinking point active attention. The input information is retained through the residual connection, while strengthening the model's attention to key features. Finally, the outputs of multiple heads are combined to obtain an optimized feature vector, which helps to improve the performance and stability of the model.

[0104] MultiHead(Q,K,V)=Concat(head 1 , head 2 , ..., head h )W o

[0105]

[0106] Among them, head i represents the output of the i-th attention head, W o represents the linear transformation weight matrix of the output, is the weight matrix of the ith head, is a scaling factor used to prevent the dot product from being too large;

[0107] Concat refers to concatenating the outputs of multiple attention heads along a certain dimension, and MultiHead refers to a multi-head attention mechanism used to capture different features of the input;

[0108] Attention refers to the attention mechanism, which performs a linear transformation on the input vectors V (value), K (key), and Q (query), followed by a scaled dot product attention calculation;

[0109] Softmax refers to the Softmax activation function, which is used to calculate the attention weights and convert the dot product between K (key) and Q (query) into a probability distribution.

[0110] In this embodiment, the scaled dot product attention represents the core calculation module in the multi-head attention mechanism, which determines how to focus on each input feature by the correlation between Q (query) and K (key) (usually calculated by dot product). The attention layer determines which parts of the feature information are most important to the current task by calculating weights.

[0111] In this embodiment, in the output of the multi-head attention mechanism, a residual network connection is used to ensure the integrity of information during feature transmission and prevent information loss problems during deep network training, thereby more effectively retaining key features and allowing the original information to directly participate in subsequent calculations, thereby improving the stability of the network.

[0112] Specifically, the residual connection adds the output features of the multi-head attention mechanism module to the input features to form the following expression:

[0113] Output norm =LayerNorm(MultiHead(X,X,X)+X);

[0114] Among them, LayerNorm refers to layer normalization, which is used to normalize the input of all features in each sample, MultiHead refers to the multi-head attention mechanism, which is used to capture different features of the input, and Output norm It refers to the output after multi-head attention processing and layer normalization, and X is the input feature.

[0115] This structure ensures that after multi-head attention calculation, the original features are retained and fused with the newly generated output features, allowing the network to better learn complex feature representations.

[0116] In addition, residual connections can help alleviate the gradient vanishing problem in deep networks, thereby accelerating the convergence process of the model and improving the overall performance.

[0117] After multi-head attention calculation and residual connection, an optimized feature vector is finally obtained, which better expresses the key features in the input data for subsequent tasks.

[0118] Step S204: construct a classification layer, classify the quality status of the cutting points based on the optimized feature matrix, and output the quality status of the cutting points.

[0119] The purpose of step S204 is to construct a classification layer for classifying the quality status of the cutting points based on the optimized feature matrix, and finally outputting the quality status of the cutting points.

[0120] The classification layer includes a fully connected layer and a Softmax classifier:

[0121] The fully connected layer is responsible for converting the optimized feature matrix output by the multi-head attention mechanism module into a one-dimensional feature sequence, thereby preparing data for subsequent classification tasks.

[0122] The Softmax classifier classifies the one-dimensional feature sequence after being processed by the fully connected layer, and outputs the probability distribution of the defect classification type to help identify whether there is a defect at each cutting point.

[0123] In order to optimize the performance of the classification layer, a loss function is used in step S204 to measure the difference between the model prediction result and the actual category.

[0124] In some embodiments, in step S204, a Focal Loss loss function is used as a classification loss function, and the expression corresponding to the Focal Loss loss function L is:

[0125] L=-α t (1-p t ) γ log(p t )

[0126] Among them, α t is a balancing factor used to adjust the relative importance of different categories, p t is the model’s predicted probability for the true category, and γ is a focusing parameter used to adjust the loss difference between easy-to-classify samples and difficult-to-classify samples.

[0127] The traditional cross entropy loss function does not work well when dealing with category imbalanced data. The Focal Loss loss function is particularly suitable for solving the category imbalance problem. By adjusting the loss weight, the model can better focus on samples that are difficult to classify, thereby improving the generalization ability of the model.

[0128] In other embodiments, in step S204, a binary cross entropy (BCE) loss function is used as the classification loss function, and the expression corresponding to the BCE loss function L is:

[0129]

[0130] Where S is the number of samples, y j is the true label of the jth sample, which takes a value of 0 or 1. Represents the predicted probability value of the jth sample.

[0131] After training, the model is evaluated on the test set to assess the model’s defect detection accuracy. The test set contains labeled data, including normal and overcut cutting points.

[0132] Precision, recall and F1-score are used as evaluation indicators.

[0133] Precision measures the proportion of cutting points predicted by the model to be defects that actually have defects. The corresponding expression is:

[0134]

[0135] The recall rate Recall measures the proportion of all cutting points with actual defects that are correctly predicted by the model. The corresponding expression is:

[0136]

[0137] Fl-score is the harmonic mean of precision and recall, which comprehensively measures the performance of the model. The corresponding expression is:

[0138]

[0139] Among them, TP is the number of samples that correctly identify defective targets, FN is the number of samples that do not identify defective targets, and FP is the number of samples that incorrectly identify defective targets.

[0140] Taking precision as an example, the test set data and each generation model parameters obtained during the training process are input into the BiLSTM model based on the attention mechanism to identify defects for each test sample. By comparing the number of defects identified by the model with the actual number of defects, the defect detection accuracy of the model is calculated, and the precision is used as a performance evaluation standard to measure the accuracy of the model in defect detection.

[0141] An example is given below. In step S2, a plurality of data sets of different geometric error defects obtained based on VERICUT simulation software are used, among which there are more than 130,000 normal cutting point samples and more than 30,000 overcut cutting point samples.

[0142] 70% of the samples are selected as training set data and 30% of the sample data are selected as test set data.

[0143] After obtaining the training set and test set, the BiLSTM model based on the attention mechanism is input for training to obtain a high-precision defect classification model.

[0144] In this embodiment, the detection results of the test set are shown in Table 1.

[0145] Table 1 Test set target detection accuracy table

[0146]

[0147] In this embodiment, the time series data of the cutting point position, feed rate and other processing parameters of the complete cutting point obtained by overcutting the impeller part are processed, and then input into the trained BiLSTM model based on the attention mechanism for testing. The results are shown in Table 2.

[0148] Table 2 Example target detection accuracy table

[0149]

[0150] Step S3, based on the multi-body system theory, a homogeneous transformation matrix is ​​used to characterize the posture conversion relationship between adjacent topological structures of the machine tool, and a spatial error model of the multi-axis machine tool is established;

[0151] Based on the multi-body system theory, the homogeneous transformation matrix is ​​used to accurately characterize the posture conversion relationship between adjacent topological structures of the machine tool, and the spatial error model of the five-axis machine tool is established.

[0152] More specifically, the step S3 further comprises the following steps:

[0153] Step S301, establishing the topological structure of the multi-axis machine tool, treating each component of the machine tool as a rigid body, and defining the relative motion relationship;

[0154] By establishing the topological structure between machine tool components, the multi-body system theory is used to describe the relative motion of each component, and finally provides a basis for the spatial error modeling of the machine tool. This process uses the homogeneous transformation matrix to accurately characterize the posture transformation relationship between machine tool components and provide support for subsequent error propagation analysis and machine tool accuracy optimization.

[0155] Figure 5 The topological structure diagram of the BC double-turntable five-axis CNC machine tool according to an embodiment of the present invention is disclosed, as shown in FIG. Figure 5 The BC double-rotary table five-axis CNC machine tool shown is composed of multiple components and motion axes, including a bed 0, a worktable, a tool 8, a workpiece 3, an X-axis 4, a Y-axis 5, a Z-axis 6, a B-axis 1, a C-axis 2, and a spindle 7.

[0156] The various components of the machine tool are regarded as rigid bodies, and based on their relative motion relationships, the topological connection relationships between the rigid bodies are defined to establish a complete topological structure diagram.

[0157] Assume that bed 0 is B 0 body, and then move away from B 0 direction, and then assign numbers to each rigid body in turn according to the relative positions and motion relationships of the various components of the machine tool, following the natural growth sequence, from one branch to another.

[0158] Optional Body B j For any n typical entities in the system, B j The serial number of the n-th order low-order body is defined as:

[0159] L n (j) = i;

[0160] In the formula, L is a low-order body operator and is called body B j For body B i The n-order high-order body satisfies the following relationship:

[0161] L n (j) = L(Ln-1 (j));

[0162] L 0 (j) = j;

[0163] L n (0) = 0

[0164] The low-order body array of the entire multi-body system can be obtained, as shown in Table 3.

[0165] j represents the serial number of the object, j=1, 2, ..., n, and n represents the number of typical bodies contained in the machine tool.

[0166] Table 3 Low-order array of BC double rotary machine tool

[0167] j 1 2 3 4 5 6 7 8 <![CDATA[L 0 (j)]]> 1 2 3 4 5 6 7 8 <![CDATA[L 1 (j)]]> 0 1 2 0 4 5 6 7 <![CDATA[L 2 (j)]]> 0 0 1 0 0 4 5 6 <![CDATA[L 3 (j)]]> 0 0 0 0 0 0 4 5 <![CDATA[L 4 (j)]]> 0 0 0 0 0 0 0 4 <![CDATA[L 5 (j)]]> 0 0 0 0 0 0 0 0

[0168] The calculation of the low-order body array takes into account the relative position and movement of each rigid body, so that the error transmission relationship between different components can be effectively represented, which can help understand how the geometric errors of each component affect the accuracy of the entire machine tool.

[0169] Step S302: Perform geometric error analysis on the multi-axis CNC machine tool to obtain error elements of the translation axis and the rotation axis.

[0170] According to the principles of spatial kinematics, each axis of a machine tool has six degrees of freedom under unconstrained conditions, and six errors will inevitably occur during the movement process, namely three translation errors and three rotation errors. These errors are all independent of the position point.

[0171] For example, on the X-axis, the three translation errors are:

[0172] Positioning error δ x (X), Y direction straightness error δ y (X), Z direction straightness error δ x (X);

[0173] On the X-axis, the three rotation errors are:

[0174] Roll error ε x (X), runout error ε y (X), pitch error ε z (X).

[0175] In addition, there are three verticality errors between the X, Y, and Z rails, namely S xy , S yz , S zx .

[0176] There are two radial movement errors δ between the C axis and the X and Y axes Cx ,δCy And two deflection and verticality errors S aoc , S boc There are two radial runout errors δ between the B axis and the X and Z axes. Bx ,δ Bz And two deflection and verticality errors S aob , S cob .

[0177] In the embodiment, the error elements of the translation axis of the BC double-turret type five-axis machine tool are shown in Table 4, and the error elements of the rotation axis of the BC double-turret type five-axis machine tool are shown in Table 5.

[0178] Table 4 Error elements of the translation axis of the BC double-rotary five-axis machine tool

[0179]

[0180] Table 5 Error elements of the translation axis of the BC double-rotary table five-axis machine tool

[0181]

[0182] Step S302 analyzes the geometric errors of the multi-axis machine tool, including translation error, rotation error, guide error, and error propagation between axes. The main contents include:

[0183] The translation error and rotation error of each moving axis; the perpendicularity error between the guide rails; the radial play error, deflection error, perpendicularity error between the axes, etc.

[0184] The analysis of these errors provides a theoretical basis for subsequent machine tool accuracy optimization and error compensation.

[0185] Step S303: Based on the results of the geometric error analysis, a homogeneous transformation matrix is ​​used to describe the relative position and posture relationship between adjacent topological structures of the machine tool, a characteristic matrix of the CNC machine tool is established, and finally a spatial error model of the multi-axis machine tool is obtained.

[0186] Based on the geometric error analysis results in step S302, a homogeneous transformation matrix is ​​used to accurately describe the posture conversion relationship between adjacent topological structures of the machine tool.

[0187] The homogeneous transformation matrix is ​​used to effectively and uniformly represent the translation and rotation relationship between the various parts of the machine tool, so that the relative position and posture between the various parts of the machine tool can be calculated and described through matrix operations.

[0188] First, establish the ideal posture matrix for each motion axis, which represents the ideal posture (position and attitude) of the axis.

[0189] Then, by introducing the geometric error in step S302 into the ideal pose matrix, the actual pose matrix of each motion axis is obtained, that is, the pose after considering the error.

[0190] For a five-axis CNC machine tool, the spatial error model can be obtained by layer-by-layer matrix multiplication. Each layer of the matrix represents the error accumulation effect of a motion axis. Therefore, by continuously multiplying the actual position matrix of each motion axis, the spatial error model of the entire machine tool can be gradually obtained.

[0191] To explain the specific establishment process of the motion axis geometric error module, the construction of the N-axis geometric error module is taken as an example.

[0192] Assuming that the Q-axis and the N-axis are in the same forward kinematic chain and the Q-axis is installed before the N-axis, the N-axis geometric error module T QN , that is, the actual posture transformation relationship of the N axis relative to the Q axis, can be established by the following expression:

[0193] T QN =T pQN T peQN T mQN T meQN ;

[0194] In the formula,

[0195] T peQN =I,

[0196]

[0197] Among them, T QN Represents the geometric error module of the N-axis, T pQN 、F peQN , T mQN and T meQN Represents the installation pose matrix, installation pose error matrix, motion pose matrix and motion pose error matrix from the Q coordinate system to the N coordinate system;

[0198] {q x ,q y ,q z} and N represent the installation offset and CNC motion command of the N-axis respectively;

[0199] {δ x (N),δ y (N),δ z (N), ε x (N), ε y (N), ε z (N)} represents the six errors of the N axis.

[0200] In this embodiment, the geometric error modules of each motion axis of the BC dual-turret five-axis CNC machine tool are listed in Table 6, where the subscript R represents the bed, the subscript T represents the tool, and the Q axis and N axis can be substituted into the actual axis, for example, T BC It refers to the C-axis geometric error module, that is, the actual position transformation relationship of the C-axis relative to the B-axis; T RB It refers to the B-axis geometric error module, that is, the actual posture transformation relationship of the bed R relative to the B-axis.

[0201] Table 6 Geometric error modules of each motion axis of the machine tool

[0202]

[0203]

[0204] In the table, the motion posture error matrix (T meQN ) are:

[0205]

[0206] It can be seen that each axis (such as B axis, C axis, X axis, etc.) has a corresponding error matrix to describe the motion error of the axis. Through these error matrices, the influence of the error of each motion axis on the overall motion of the machine tool can be calculated.

[0207] In this embodiment, for a BC dual-turret five-axis CNC machine tool, the full kinematic chain model of the entire machine tool can be constructed by integrating the actual forward kinematic models of two sub-kinematic chains (workpiece chain and tool chain). The workpiece chain model and the tool chain model describe the motion of the workpiece and the tool, respectively.

[0208] According to the actual motion chain of the BC double-turntable five-axis, the actual forward kinematics model of the machine tool is quickly established, and the final spatial error model can be obtained by layer-by-layer matrix multiplication.

[0209] Figure 6 The schematic diagram of the kinematic chain of a BC dual-turret five-axis CNC machine tool according to an embodiment of the present invention is disclosed. Figure 6 As shown in the figure, according to the workpiece chain R (bed) → B (B axis) → C (C axis) → W (workpiece), the actual forward motion model of the workpiece chain can be obtained:

[0210]

[0211] According to the tool chain R (bed) → X (X axis) → Y (Y axis) → Z (Z axis) → T (tool), the actual forward kinematic model of the tool chain can be obtained:

[0212]

[0213] By integrating the actual forward kinematics model of the two sub-kinematic chains (workpiece chain and tool chain) of the profile gear grinding machine, the actual forward kinematics model of the entire kinematic chain of the machine tool can be further established:

[0214]

[0215] Similarly, based on the workpiece chain, tool chain and full motion chain of the machine tool, the ideal motion module of the motion axis is reconstructed in sequence. Based on the ideal forward kinematics model of the workpiece chain and tool chain of the machine tool, the ideal forward kinematics model of the full motion chain of the machine tool is established:

[0216]

[0217] In the formula, They represent the ideal forward kinematic models of the tool chain, workpiece chain and the entire kinematic chain respectively.

[0218] A model is constructed to analyze the influence of geometric error on the relative posture error between the tool, grinding wheel and workpiece gear. This model is also called the geometric error-tool posture error model (GE-TPEM) of the machine tool, which can also be called the spatial error model or volume error model. The corresponding expression is:

[0219]

[0220] In the formula, δ represents the position error vector of the tool center, and ε represents the attitude error vector of the tool;

[0221] δ x , δ y , δ z They are the position error elements of the tool center in the directions of the three coordinate axes;

[0222] ε x , ε y , ε z They are the angular error elements of the tool in the WCS (workpiece coordinate system) around the three coordinate axes;

[0223] [a, b, c, 1] T Indicates tool position data, [i, j, k, 0] T Indicates tool vector data.

[0224] Step S4, establish a geometric error sensitivity analysis model based on the multi-axis machine tool spatial error model, input the moving axis feed amount of the defective cutting point identified by the defect classification model into the geometric error sensitivity analysis model, quantify the contribution of geometric error to machining error, and obtain the part quality traceability result based on the statistical results of the machining process.

[0225] A geometric error sensitivity analysis model is established, and the feed amount of the moving axis of the defective cutting point identified in step S2 is substituted to quantify the contribution of each geometric error to the machining error. The part quality traceability result is obtained through statistical machining process.

[0226] The step S4 specifically comprises the following steps:

[0227] Step S401: quantify the error correlation of each cutting point according to the multi-axis machine tool spatial error model, and establish a geometric error sensitivity analysis model;

[0228] Based on step S3, the spatial error model E of the five-axis machine tool can be obtained, and the corresponding general expression is:

[0229] E=F(G,P w ,U,U w , U t )

[0230] G=(Δe 1 , ..., Δe i , ..., Δe n ) T

[0231] P w =(P wx , P wy , P wz , 1) T

[0232] U = (x, y, z, 1) T

[0233] U w =(x w ,y w , z w , 1) T

[0234] U t =(x t ,y t , z t , 1) T

[0235] Where G is the error vector composed of the geometric errors of each component;

[0236] P w is the coordinate vector of the machining point in the workpiece coordinate system;

[0237] U is the position vector of the machine tool motion axis;

[0238] U t is the coordinate vector of the tool position;

[0239] U w is the coordinate vector of the workpiece position;

[0240] Δe i is the geometric error of the component, i=1, 2, ..., n.

[0241] Among them, G, P w ,U,U w , U t Generally a fixed value.

[0242] Based on the spatial error model of the five-axis machine tool obtained in step S3, an error correlation matrix can be constructed to describe how the geometric error of each component affects the machining accuracy of the machine tool, which specifically includes the following steps:

[0243] First, install and fix the workpiece and tool, and set the initial value of the CNC machine tool;

[0244] Then, the workpiece machining trajectory of the machine tool is simulated under this condition, and the results are analyzed to determine the influence of the geometric errors of various components of the CNC machine tool on the machining accuracy of the machine tool;

[0245] Then calculate the correlation coefficient, and the corresponding expression is:

[0246]

[0247] Therefore, the geometric error sensitivity analysis model of multi-axis CNC machine tools is:

[0248]

[0249] In the above formula, S is the matrix expression of the error correlation degree common to five-axis CNC machine tools:

[0250] Normalize the error correlation, and the corresponding expression is as follows:

[0251]

[0252] In the formula, S mi =Δe i The corresponding error correlation coefficient, and the sum of all correlation coefficients is 1.

[0253] By analyzing the correlation coefficient, we can know that the error correlation coefficient S mi and the error element Δe i There is a clear positive relationship between the correlation of S mi The increase in Δe i The greater the correlation with the spatial error, the greater the impact of the error generated by the component on the overall machining accuracy of the machine tool.

[0254] Step S402: inputting the defective cutting point data identified in step S2 into the geometric error sensitivity analysis model for calculation, quantifying the influence of the geometric error on the machining accuracy of each cutting point, and obtaining the error contribution of all defective cutting points;

[0255] The defective cutting point data identified in step S2 includes the position vector U of each motion axis of the machine tool, which is input into the geometric error sensitivity analysis model;

[0256] Through model calculation, the influence of geometric error on the machining accuracy of each cutting point is quantified;

[0257] Repeat step S402 until the error contributions of all defective cutting points are obtained.

[0258] Figure 7 A schematic diagram of error correlation according to an embodiment of the present invention is disclosed, such as Figure 7 As shown, taking the 805th cutting point in the model running result obtained in the above example as an example, the error correlation degree of the cutting point can be obtained by calculation.

[0259] Figure 8 A schematic diagram of the main error correlation coefficient in the X direction according to an embodiment of the present invention is disclosed. After normalizing the correlation coefficient, the main error correlation coefficient in the X direction is obtained respectively;

[0260] Fig. 9 A schematic diagram of the main error correlation coefficients in the Y direction according to an embodiment of the present invention is disclosed. After normalizing the correlation coefficients, the main error correlation coefficients in the Y direction are obtained respectively;

[0261] Fig.10 A schematic diagram of the main error correlation coefficients in the Z direction according to an embodiment of the present invention is disclosed. After normalizing the correlation coefficients, the main error correlation coefficients in the Z direction are obtained respectively.

[0262] These coefficients are normalized and can accurately reflect the impact of geometric errors in each direction.

[0263] Step S403: Based on the quantified results of the error contribution, data standardization is performed to identify the error source that has the greatest impact on the processing quality, clarify the key geometric errors, and obtain the part quality traceability results.

[0264] First, based on the quantitative results of error contribution, data standardization is performed to eliminate dimension and unit differences and ensure the comparability of each error contribution.

[0265] Then, the weighted average method is used to identify the error sources that have the greatest impact on the processing quality, clarify the key geometric errors, and obtain the part quality traceability results, which provides an important basis for subsequent process improvements.

[0266] In this embodiment, the key geometric errors obtained in various directions according to step S403 are shown in Table 7.

[0267] Table 7 Key geometric errors in quality traceability analysis

[0268]

[0269]

[0270] After identifying the critical geometric errors, the process can be further improved, the design and manufacturing of the machine tool can be optimized, and the machining accuracy can be improved.

[0271] A multi-axis machine tool part quality traceability method proposed in the present invention innovatively combines the BiLSTM neural network with machine tool error modeling, realizes accurate traceability analysis from defective cutting points to key geometric errors, provides machine tool design engineers with weight information of the impact of various geometric errors on machining accuracy, effectively guides the machine tool accuracy optimization design, and lays a theoretical foundation for improving the accuracy of CNC machine tools.

[0272] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0273] Based on the above-mentioned multi-axis machine tool part quality traceability method, the present invention also proposes a multi-axis machine tool part quality traceability system, including an internal communication bus, a processor, a read-only memory (ROM), a random access memory (RAM), a communication port, and a hard disk. The internal communication bus can realize data communication between components of the multi-axis machine tool part quality traceability system. The processor can make judgments and issue prompts. In some embodiments, the processor can be composed of one or more processors.

[0274] The communication port can realize data transmission and communication between the multi-axis machine tool part quality traceability system and external input / output devices. In some embodiments, the multi-axis machine tool part quality traceability system can send and receive information and data from the network through the communication port. In some embodiments, the multi-axis machine tool part quality traceability system can transmit and communicate data with external input / output devices in a wired form through the input / output terminal.

[0275] The multi-axis machine tool parts quality traceability system may also include different forms of program storage units and data storage units, such as hard disks, read-only memories (ROMs) and random access memories (RAMs), which can store various data files used for computer processing and / or communication, as well as possible program instructions executed by the processor. The processor executes these instructions to implement the main part of the method. The results of the processor processing are transmitted to an external output device through a communication port and displayed on a user interface of the output device.

[0276] For example, the implementation process file of the multi-axis machine tool part quality traceability system mentioned above can be a computer program, which is stored in a hard disk and can be recorded in a processor for execution to implement the method of the present invention.

[0277] When the implementation process file of the multi-axis machine tool part quality traceability method is a computer program, it can also be stored in a computer-readable storage medium as a product. For example, a computer-readable storage medium may include, but is not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips), optical disks (e.g., compact disks (CDs), digital versatile disks (DVDs)), smart cards, and flash memory devices (e.g., electrically erasable programmable read-only memories (EPROMs), cards, sticks, key drives). In addition, the various storage media described herein can represent one or more devices and / or other machine-readable media for storing information. The term "machine-readable medium" may include, but is not limited to, wireless channels and various other media (and / or storage media) that can store, contain and / or carry code and / or instructions and / or data.

[0278] Compared with the prior art, the multi-axis machine tool parts quality traceability method and system proposed in the present invention have the following beneficial effects:

[0279] 1) In order to solve the problem of scarcity of part processing data set and insufficient defect data caused by geometric errors, the present invention innovatively uses VERICUT simulation software to obtain the time-series cutting point data and its labels of the cutting process. By reasonably setting simulation parameters and recording the time-series data of key processing parameters, sufficient data support is provided for subsequent geometric error analysis;

[0280] 2) The present invention constructs a BiLSTM model based on the attention mechanism, uses the cutting point and its neighboring data obtained by simulation for training, and optimizes it in combination with labels. It obtains the classification information of the cutting point by testing with actual processing data. By introducing a multi-head attention mechanism and a residual network, it can optimize the feature weights, improve the classification accuracy, enhance the sensitivity to important features, and improve the processing accuracy.

[0281] 3) Compared with the traditional AC double-turret five-axis machine tool geometric error tracing, the present invention establishes a spatial error model specifically for the BC double-turret five-axis CNC machine tool, revealing the relationship between geometric error and tool posture error.

[0282] By constructing a CNC machine tool correlation model and using the direct matrix differentiation method, the key geometric errors that affect the machining quality can be accurately obtained. The weighted average method can be used to identify the main error sources of the machining quality, providing data support for subsequent process improvements and optimizing machining accuracy.

[0283] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0284] Those skilled in the art will appreciate that information, signals, and data may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips cited throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

[0285] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0286] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0287] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0288] The above embodiments are provided for persons familiar with the art to implement or use the present invention. Personnel familiar with the art can make various modifications or changes to the above embodiments without departing from the inventive concept of the present invention. Therefore, the protection scope of the present invention is not limited to the above embodiments, but should be the maximum scope of the innovative features mentioned in the claims.

Claims

1. A method for tracing the quality of multi-axis machine tool parts, characterized in that: The following steps are involved: Step S1, obtaining a sequential cutting point data set and corresponding label data during part processing; Step S2, constructing a bidirectional long short-term memory network model based on an attention mechanism, using the cutting point data and the adjacent point data combined with the corresponding label data for training to obtain a defect classification model; Step S3, based on the multi-body system theory, a homogeneous transformation matrix is ​​used to characterize the posture conversion relationship between adjacent topological structures of the machine tool, and a spatial error model of the multi-axis machine tool is established; Step S4, establish a geometric error sensitivity analysis model based on the multi-axis machine tool spatial error model, input the moving axis feed amount of the defective cutting point identified by the defect classification model into the geometric error sensitivity analysis model, quantify the contribution of geometric error to machining error, and obtain the part quality traceability result based on the statistical results of the machining process.

2. The multi-axis machine tool parts quality traceability method according to claim 1 is characterized in that: The step S1 further comprises: Step S101, loading the CAD model of the multi-axis machine tool, the tool model and the workpiece CAD model through the VERICUT simulation software, constructing a virtual machining environment, and setting the assembly relationship and relative position between the models; Step S102, generating geometric error values ​​of each motion axis according to a preset geometric error range, so as to adjust the relative positions between the motion axes of the multi-axis machine tool and simulate the geometric error in the actual machining process; Step S103, importing the NC program of the part to be processed, setting simulation parameters and performing simulation, and recording and exporting the time series data of the processing parameters; Step S104, preprocessing the collected time series data, and adding corresponding quality labels to the preprocessed data according to the processing quality evaluation standard.

3. The multi-axis machine tool parts quality traceability method according to claim 1 is characterized in that: The step S2 further comprises: Step S201, constructing an input layer, wherein the input layer is used to input the original data of the cutting state of the cutting point and the adjacent points; Step S202, constructing a feature extraction layer, wherein the feature extraction layer is used to extract features from the original data and generate a feature matrix, and the feature extraction layer at least includes a bidirectional long short-term memory network; Step S203, constructing a multi-head attention mechanism module, wherein the multi-head attention mechanism module is used to perform weight optimization on the feature matrix to generate an optimized feature matrix; Step S204, constructing a classification layer, classifying the quality status of the cutting points based on the optimized feature matrix, and outputting the quality status of the cutting points after classification and identification.

4. The multi-axis machine tool parts quality traceability method according to claim 3 is characterized in that: The bidirectional long short-term memory network outputs a latent vector representing the defect classification type through forward long short-term memory and backward long short-term memory operations.

5. The multi-axis machine tool parts quality traceability method according to claim 3 is characterized in that: The multi-head attention mechanism module adds the output features to the input features through residual connections, and finally outputs the optimized feature matrix.

6. The multi-axis machine tool parts quality traceability method according to claim 3 is characterized in that: The classification layer includes a fully connected layer and a Softmax classifier: The fully connected layer converts the optimized feature matrix output by the multi-head attention mechanism module into a one-dimensional feature sequence; The Softmax classifier classifies the one-dimensional feature sequence processed by the fully connected layer and outputs the probability distribution of the defect classification type.

7. The multi-axis machine tool parts quality traceability method according to claim 3 is characterized in that: In step S204, the Focal Loss function is used as the classification loss function, and the corresponding expression is: L=-a t (1-p t ) γ log(p t ); Among them, α t is the balance factor, p t is the model’s predicted probability for the true category, and γ is the focusing parameter.

8. The multi-axis machine tool parts quality traceability method according to claim 3 is characterized in that: In step S204, a binary cross entropy loss function is used as the classification loss function, and the corresponding expression is: Where S is the number of samples, y i is the true label of the jth sample, Represents the predicted probability value of the jth sample.

9. The multi-axis machine tool parts quality traceability method according to claim 3, characterized in that: After step S204, the method further includes: Evaluate the established defect classification model based on performance evaluation indicators; The performance evaluation indicators include at least precision, recall and F1-score.

10. The multi-axis machine tool parts quality traceability method according to claim 1, characterized in that: The step S3 further comprises: Step S301, establishing the topological structure of the multi-axis machine tool, treating each component of the machine tool as a rigid body, and defining the relative motion relationship; Step S302, performing geometric error analysis on the multi-axis CNC machine tool to obtain error elements of the translation axis and the rotation axis; Step S303, based on the geometric error analysis results, a homogeneous transformation matrix is ​​used to describe the relative position and posture relationship between adjacent topological structures of the machine tool, a characteristic matrix of the CNC machine tool is established, and finally a spatial error model of the multi-axis machine tool is obtained.

11. The multi-axis machine tool parts quality traceability method according to claim 10, characterized in that: The step S303 further comprises: For the BC dual-rotary table model of the five-axis machine tool, the actual forward kinematics model of the whole kinematic chain of the machine tool is constructed by integrating the actual forward kinematics model of the workpiece chain and the tool chain; Based on the ideal forward kinematics model of the workpiece chain and tool chain of the machine tool, the ideal forward kinematics model of the whole kinematic chain of the machine tool is established; A geometric error-tool posture error model is established to analyze the influence of geometric error on the relative posture error between the tool, grinding wheel and workpiece gear.

12. The multi-axis machine tool parts quality traceability method according to claim 1, characterized in that: The step S4 further comprises: Step S401, quantifying the error correlation of each cutting point according to the multi-axis machine tool spatial error model, and establishing a geometric error sensitivity analysis model; Step S402, inputting the defective cutting point data identified in step S2 into a geometric error sensitivity analysis model for calculation, quantifying the influence of geometric errors on the machining accuracy of each cutting point, and obtaining the error contribution of all defective cutting points; Step S403, based on the quantified results of the error contribution, data standardization is performed to identify the error source that has the greatest impact on the processing quality, clarify the key geometric errors, and obtain the part quality traceability results.

13. A multi-axis machine tool parts quality traceability system, comprising: a memory for storing instructions executable by a processor; A processor, configured to execute the instructions to implement the method according to any one of claims 1 to 12.

14. A computer readable medium having computer instructions stored thereon, wherein: When the computer instructions are executed by a processor, the method according to any one of claims 1 to 12 is performed.

Images