A method, device, medium and equipment for identifying chatter in a part machining process

Abstract

This application discloses a method, apparatus, medium, and equipment for chatter identification during parts processing, relating to the field of quality control technology, and aims to solve the problem of poor performance of existing chatter identification and analysis methods. This application first optimizes the original vibration sensor data by screening it to obtain data that is more conducive to feature expression in the processing feature direction. Secondly, it performs noise reduction processing on the screened data to improve the signal-to-noise ratio and make the chatter signal characteristics clearer. Then, it introduces data decomposition and transformation technology to visualize the chatter signal characteristics as images. Finally, it uses a chatter identification model to quickly and accurately identify the images, analyze the patterns of the chatter signals, and output chatter prediction results, effectively improving the effect of chatter identification and analysis, and more accurately guiding chatter prevention and control work, thus playing a positive role in improving the overall technical level of the aerospace manufacturing industry.

Description

Technical Field

[0001] This application relates to the field of quality control technology, specifically to a method, device, medium, and equipment for identifying chatter during parts processing. Background Technology

[0002] In the aerospace field, the machining quality of aircraft parts directly affects the safety and reliability of aircraft. Simultaneously, with the development of aerospace technology, the requirements for the precision and performance of aircraft parts are becoming increasingly stringent. This not only demands high-precision and high-efficiency machining equipment but also requires effective control and prevention of various abnormal situations during machining, particularly chatter abnormalities generated during part cutting. Chatter is a dynamic instability phenomenon that typically occurs in high-speed rotating or high-speed moving mechanical systems, such as the machining of critical components like aircraft engine blades, aluminum alloy parts, and airfoils. When the interaction force between the machining tool and the workpiece exceeds a certain threshold, it leads to intensified vibration of the machining system, affecting machining accuracy and, in severe cases, even damaging the workpiece and machine tool, causing economic losses and safety hazards.

[0003] Traditional chatter control methods primarily rely on experienced operators' intuitive judgment and monitoring based on sounds and vibrations during processing. This approach is highly subjective and uncertain, making it difficult to meet the requirements of 24 / 7 monitoring in automated production lines. In recent years, with the development of sensor technology, data processing technology, and artificial intelligence, chatter anomaly detection and control methods based on real-time monitoring and intelligent analysis have gradually become a research hotspot. These methods collect multi-dimensional data such as vibration, temperature, and pressure during processing using various sensors installed on the processing equipment. They then utilize advanced signal processing algorithms and machine learning models to analyze the data, achieving early warning and precise control of chatter phenomena. This improves processing quality and production efficiency while reducing production costs and safety risks.

[0004] However, existing flutter identification and analysis methods are affected by factors such as the quality, diversity, and volume of collected data, resulting in poor flutter identification performance. Therefore, designing a more efficient, accurate, and reliable method for identifying and analyzing flutter anomalies in parts manufacturing processes is of great significance for improving the overall technical level of the aerospace manufacturing industry. Summary of the Invention

[0005] The main objective of this application is to provide a method, device, medium, and equipment for chatter identification during parts processing, aiming to solve the problem of poor performance of existing chatter identification and analysis methods.

[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows:

[0007] In a first aspect, embodiments of this application provide a chatter identification method during part manufacturing, comprising the following steps:

[0008] Based on the expression of processing features, the original vibration sensor data of the target part processing process is filtered to obtain the first vibration sensor data;

[0009] The noise reduction process is applied to the first vibration sensor data to obtain the target vibration sensor data.

[0010] Data decomposition and transformation are performed on the target vibration sensor data to generate a flutter feature map;

[0011] Input the flutter feature map into the flutter recognition model, and output the flutter prediction result of the target part.

[0012] In one possible implementation of the first aspect, data decomposition and transformation are performed based on the target vibration sensor data to generate a flutter feature map, including:

[0013] Based on the target vibration sensor data, the data is segmented in the time domain to obtain multiple segmented data.

[0014] Data decomposition and transformation are performed on multiple segmented data to obtain the frequency components of the segmented data at different time points;

[0015] The frequency components of multiple segmented data at different time points are combined to generate a flutter feature map.

[0016] In one possible implementation of the first aspect, before filtering the raw vibration sensor data of the target part's machining process based on the expression of machining features to obtain the first vibration sensor data, the method further includes:

[0017] Based on the changes in machine tool machining parameters between the current time and the previous adjacent time, it is determined whether the machine tool is in a machining state. If the determination result is that the machine tool is in a machining state, the expression based on machining characteristics is executed to filter the original vibration sensor data of the target part machining process and obtain the first vibration sensor data.

[0018] In one possible implementation of the first aspect, based on the expression of processing features, the raw vibration sensor data of the processing of the target part is filtered to obtain first vibration sensor data, including:

[0019] Based on the direction that is more favorable to the expression of processing features, the variance of the target part in the original vibration sensor data of the processing of the target part is compared.

[0020] The data with the largest variance in the target portion is selected and retained together with the data in the original vibration sensor data excluding the target portion to obtain the first vibration sensor data.

[0021] In one possible implementation of the first aspect, before filtering the raw vibration sensor data of the target part's machining process based on the expression of machining features to obtain the first vibration sensor data, the method further includes:

[0022] Acquire useful signals for processing the target part;

[0023] The useful signal is sampled to obtain the raw vibration sensor data.

[0024] In one possible implementation of the first aspect, the flutter prediction result represents the probability of flutter occurrence. If the probability represented by the flutter prediction result exceeds a set threshold, after inputting the flutter feature map into the flutter recognition model and outputting the flutter prediction result for the target part, the method further includes:

[0025] Based on the image orientation gradient factor and its distribution, the flutter prediction results are analyzed to obtain the flutter prediction target results.

[0026] In one possible implementation of the first aspect, the flutter prediction results are analyzed based on the flutter anomalies of the image orientation gradient factor and its distribution to obtain the flutter prediction target result, including:

[0027] Global high-frequency feature extraction is performed on the flutter prediction results based on the image orientation gradient factor to obtain a feature element image;

[0028] Non-edge maxima suppression is applied to the feature element image to obtain the first feature element image;

[0029] The moment feature is obtained by describing the feature information of the first feature element image using moments.

[0030] Based on the analysis of the moment characteristics of flutter anomalies in the distribution, the flutter prediction target results are obtained.

[0031] Secondly, embodiments of this application provide a chatter identification device for a part processing procedure, comprising:

[0032] The filtering module is used to filter the raw vibration sensor data of the target part's machining process based on the expression of machining features to obtain the first vibration sensor data.

[0033] The noise reduction module is used to perform noise reduction processing on the data from the first vibration sensor to obtain the data from the target vibration sensor.

[0034] The generation module is used to perform data decomposition and transformation based on the target vibration sensor data to generate a flutter feature map.

[0035] The identification module is used to input the flutter feature map into the flutter identification model and output the flutter prediction result of the target part.

[0036] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when loaded and executed by a processor, implements a chatter recognition method for part processing as provided in any of the first aspects above.

[0037] Fourthly, embodiments of this application provide an electronic device, including a processor and a memory, wherein,

[0038] Memory is used to store computer programs;

[0039] The processor is used to load and execute a computer program to cause the electronic device to perform a chatter identification method for a part manufacturing process as provided in any of the first aspects above.

[0040] Compared with the prior art, the beneficial effects of this application are:

[0041] This application proposes a method, apparatus, medium, and device for chatter identification during the machining process of a part. The method includes: filtering the original vibration sensor data of the target part's machining process based on the expression of machining features to obtain first vibration sensor data; performing noise reduction processing on the first vibration sensor data to obtain target vibration sensor data; performing data decomposition and transformation based on the target vibration sensor data to generate a chatter feature map; inputting the chatter feature map into a chatter identification model to output the chatter prediction result of the target part. This application first optimizes the original vibration sensor data by filtering it to obtain data that is more conducive to feature expression in the machining feature direction. Secondly, it performs noise reduction processing on the filtered data to improve the signal-to-noise ratio, making the chatter signal features clearer. Then, it introduces data decomposition and transformation technology to visualize the chatter signal features as images. Finally, it uses a chatter identification model to quickly and accurately identify the images to analyze the patterns of the chatter signals and output chatter prediction results. This effectively improves the effect of chatter identification and analysis, more accurately guides chatter prevention and control work, and plays a positive role in improving the overall technical level of the aerospace manufacturing industry. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of this application;

[0043] Figure 2 A flowchart illustrating the chatter identification method for the part processing procedure provided in this application embodiment;

[0044] Figure 3 A schematic diagram of a chatter feature map in the chatter identification method for the part processing process provided in the embodiments of this application;

[0045] Figure 4A schematic diagram of a first feature element image in the chatter recognition method for the part processing process provided in the embodiments of this application;

[0046] Figure 5 This is a schematic diagram illustrating the analysis of moment features in the chatter identification method for the part processing process provided in this application embodiment;

[0047] Figure 6 A flowchart illustrating one implementation of the chatter identification method for part processing provided in this application embodiment;

[0048] Figure 7 A schematic diagram of a chatter recognition device for part processing provided in an embodiment of this application;

[0049] The diagram is labeled as follows: 101-Processor, 102-Communication bus, 103-Network interface, 104-User interface, 105-Memory. Detailed Implementation

[0050] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0051] See attached document Figure 1 , attached Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of this application. The electronic device may include: a processor 101, such as a central processing unit (CPU), a communication bus 102, a user interface 104, a network interface 103, and a memory 105. The communication bus 102 is used to realize the connection and communication between these components. The user interface 104 may include a display screen and an input unit such as a keyboard. Optionally, the user interface 104 may also include a standard wired interface and a wireless interface. The network interface 103 may optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface). The memory 105 may be a storage device independent of the aforementioned processor 101. The memory 105 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as at least one disk storage device. The processor 101 may be a general-purpose processor, including a central processing unit, a network processor, etc., or it may be a digital signal processor, an application-specific integrated circuit, a field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component.

[0052] Those skilled in the art will understand that the appendix Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0053] As attached Figure 1 As shown, the memory 105, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and a chatter recognition device for the part processing.

[0054] In the appendix Figure 1 In the electronic device shown, the network interface 103 is mainly used for data communication with the network server; the user interface 104 is mainly used for data interaction with the user; the processor 101 and the memory 105 in this application can be set in the electronic device. The electronic device calls the chatter identification device of the part processing process stored in the memory 105 through the processor 101 and executes the chatter identification method of the part processing process provided in the embodiment of this application.

[0055] See attached document Figure 2 Based on the hardware device described in the foregoing embodiments, embodiments of this application provide a method for chatter identification during part manufacturing, comprising the following steps:

[0056] S10: Based on the expression of processing features, the original vibration sensor data of the processing of the target part is screened to obtain the first vibration sensor data.

[0057] In the specific implementation process, the target part is the part currently being processed on the machine tool, and the original vibration sensor data is the data fed back by multiple sensors installed on the machine tool. After receiving the feedback data, since the data quality of each sensor is significantly different during the processing of aircraft parts, and the data quality of each sensor also varies with the changes in the characteristics of the processed parts, it is necessary to make targeted selections, optimize and evaluate the data, and screen out more suitable and more conducive data information for accurate analysis.

[0058] Specifically, based on the expression of processing features, the raw vibration sensor data of the target part's processing is filtered to obtain the first vibration sensor data, including:

[0059] Based on the direction that is more favorable to the expression of processing features, the variance of the target part in the original vibration sensor data of the processing of the target part is compared.

[0060] The data with the largest variance in the target portion is selected and retained together with the data in the original vibration sensor data excluding the target portion to obtain the first vibration sensor data.

[0061] In the specific implementation process, assuming that the received original vibration sensor data consists of 5 sets of sensor data (C1, C2, C3, C4, C5), these 5 sets of data need to be filtered. When the processed parts are installed on the machine tool, they may not be in the processing state due to factors such as the processing flow. Therefore, when the machine tool is not processing, there is no need to analyze the data. When the machine tool is processing, the target part is compared, such as the Div(C1,C2,C3) variance of the corresponding Data(C1), Data(C2), and Data(C3) data in Data(C1,C2,C3). The larger the variance, the more conducive the data is to the expression of the current processing feature in the corresponding direction. Then, the data of the corresponding channel is used for further analysis. After filtering, it becomes Data(Co,C4,C5), where Co represents the optimized data, that is, the one with the largest variance among C1, C2, and C3. The corresponding data is represented by Co(n).

[0062] In one embodiment, before obtaining the first vibration sensor data by filtering the raw vibration sensor data of the target part's machining process based on the expression of machining features, the method further includes:

[0063] Based on the changes in machine tool machining parameters between the current time and the previous adjacent time, it is determined whether the machine tool is in a machining state. If the determination result is that the machine tool is in a machining state, the expression based on machining characteristics is executed to filter the original vibration sensor data of the target part machining process and obtain the first vibration sensor data.

[0064] In the specific implementation process, since the screening needs to be carried out while the machine tool is in the processing state, other collected data can be used to determine whether processing is in progress. The machine tool processing parameter data does not need to be screened, and the complete data obtained is directly saved and denoted as Data(M). A corresponding data segment is acquired every unit time T. Based on the data Data(M), the current machine tool speed S and feed F information are read. The S and F data corresponding to the current time and the previous adjacent time are compared. If the corresponding F has changed and S is greater than 0, it means that the machine tool is currently processing; otherwise, the machine tool is not processing.

[0065] In one embodiment, before obtaining the first vibration sensor data by filtering the raw vibration sensor data of the target part's machining process based on the expression of machining features, the method further includes:

[0066] Acquire useful signals for processing the target part;

[0067] The useful signal is sampled to obtain the raw vibration sensor data.

[0068] In the specific implementation process, all useful signals of the current part being processed are acquired to provide a basis for anomaly identification and analysis. Potential useful signals include the machine tool information, NC information, and multi-sensor data of the current processing. Machine tool information, such as machine tool processing parameters, may include the machine tool's speed, feed, coordinates, power, etc. NC information mainly consists of the CNC program information for part processing, including tool name, number of tool teeth, program line number, etc. Multi-sensor data includes vibration data, torque data, and bending moment data from X-axis, Y-axis, and Z-axis vibration sensors. As in the example in the aforementioned embodiment, a total of 5 sets of sensor data may be included.

[0069] Data acquisition and sampling are fundamental, ensuring that no important information of the signal is lost during the sampling process. To ensure that the transmitted data accurately reflects the actual processing state and that continuous analog data can be accurately converted into discrete data to guarantee data quality, the data acquisition principle is as follows: Let X be the signal acquired by the sensor, whose spectrum is zero when |f|>fm, where fm is the maximum frequency of the signal. According to the sampling theorem, if the sampling frequency fs satisfies [fs≥2fm], the signal can be reconstructed from the sampled signal X = x(nTs) using an ideal low-pass filter, where n=1, 2, 3, …Ls, Ls is the number of sampling points, and Ts = 1 / fs is the sampling period. After acquiring and sampling the multi-sensor data from the useful signal, the original vibration sensor data is obtained.

[0070] In one embodiment, after the useful signal is acquired, the data transmission and communication method can be set to send out the corresponding raw data so that it can be used as input data for subsequent analysis and processing algorithms. The data transmission and communication method is set to use a non-blocking send and receive mode for data transmission. That is, in non-waiting mode, when a socket attempts to send a message, if the message queue is full or the network connection is unavailable, the send operation will return immediately instead of blocking and waiting until the message is successfully sent. This allows the sender to continue to perform other tasks without waiting for the processing result of the current message.

[0071] Specifically, first, a data receiving and sending object OBJ is created. Creating data receiving and sending objects is the foundation for managing and maintaining resources. Each receiving port only needs one receiving and sending object OBJ.

[0072] Create and configure a socket, select the appropriate socket type, and perform the necessary configurations. For non-wait mode, push sockets are typically used, which are suitable for non-blocking transmissions.

[0073] Next, bind and connect the socket BindC, select to bind (as the sender) or connect (as the receiver) the socket, and set the same sender and receiver IP addresses, and the receiver's IP address must be consistent with the sender's IP address;

[0074] Set the Hflag flag to limit the maximum length of the message queue and prevent memory exhaustion;

[0075] Then send the message using the Send method, and specify a signal flag to ensure that the send operation is non-blocking;

[0076] When the message queue is full, the Send method will throw an exception. To handle this exception, you can clear the message queue and resend the message.

[0077] Finally, clean up resources. After completing all operations, close the socket (`close`) and release the resource objects (`release`). The machine tool, NC, and multi-sensor data are configured with the same IP address but different corresponding ports, such as 5551, 5552, and 5553 respectively. The ports can be flexibly modified, but the sending and receiving ports must be consistent to ensure data reception.

[0078] S20: Perform noise reduction processing on the first vibration sensor data to obtain the target vibration sensor data.

[0079] In practical implementation, filtering and noise reduction can improve the signal-to-noise ratio of the data, making the characteristics of the flutter signal clearer. By performing numerical filtering on the original data, high-frequency noise and random interference can be removed, making the flutter signal easier to identify and analyze in the frequency and time domains. The first vibration sensor data after noise reduction is the target vibration sensor data. Specifically, this application provides a method for noise reduction of data based on third-order filtering. The filter can be represented by a linear constant-coefficient difference equation, let Co(n) = x(n):

[0080]

[0081] in, and Here are the filter coefficients, N and M are the filter orders, i and j are subscripts, n = 1, 2, ... N. Design a third-order filter with the following transfer function:

[0082]

[0083] in, Represents angular frequency. = *2π, The cutoff frequency, The result of the filtering is obtained by transforming the complex variable using the Laplace transform. .

[0084] S30: Based on the target vibration sensor data, perform data decomposition and transformation to generate a flutter feature map.

[0085] In practical implementation, the filtered data forms the basis for generating the flutter feature map, which is generated based on data decomposition and transformation. The core principle is to segment the signal in the time domain and perform data decomposition and transformation on each segment to obtain the frequency components of the data signal at different time points. These components are then combined to generate the flutter feature image. That is, based on the target vibration sensor data, data decomposition and transformation are performed to generate the flutter feature map, including:

[0086] Based on the target vibration sensor data, the data is segmented in the time domain to obtain multiple segmented data.

[0087] Data decomposition and transformation are performed on multiple segmented data to obtain the frequency components of the segmented data at different time points;

[0088] The frequency components of multiple segmented data at different time points are combined to generate a flutter feature map.

[0089] In the specific implementation process, the filtered signal The data is divided into multiple segments, denoted by k, with each segment corresponding to a signal s(k), representing segmented data of length H. For local signal processing, to reduce discontinuities at segment edges and mitigate the impact of feature leakage leading to incomplete feature representation, each segment is multiplied by a function w(k), expressed as 0.5 - 0.5 * cos(2 * 2π * k / (H - 1)). This performs signal processing in the time domain, yielding the processed result x'(k). Each segment undergoes a Fast Fourier Transform (FFT) to convert x'(k) to frequency X'(f). Finally, all segments are combined using a linear combination of complex exponential functions to form a series of two-dimensional feature maps, Fimg, as shown in the appendix. Figure 3 As shown, the number of images is k, which is the same as the number of segments after the data is divided. The horizontal axis of the image represents time, the vertical axis represents frequency, and the color of each point in the image represents the data feature. Based on the differences in global and local color features of the image, the variation law of the flutter signal over time can be analyzed.

[0090] S40: Input the flutter feature map into the flutter recognition model and output the flutter prediction result of the target part.

[0091] In the specific implementation process, the identification and analysis of flutter anomalies can utilize a trained flutter anomaly identification model based on a memory mechanism to predict flutter risk from the flutter feature map, extracting features identical to those in the training data to obtain the predicted flutter risk. The model output is the probability threshold corresponding to the presence or absence of flutter. If the corresponding value is greater than the set threshold Thred, it indicates a higher probability of flutter; if the corresponding value is less than or equal to the threshold, it indicates a lower probability of flutter. For results with a higher probability of flutter, further identification is required to ensure the accuracy of shutdown, i.e., flutter anomaly reanalysis based on image orientation gradient factors and distribution; otherwise, no flutter has occurred. The data within the next unit time T is received cyclically, and judgments are made, with the corresponding results stored in a log file. That is, the flutter prediction result represents the probability of flutter occurrence. When the probability represented by the flutter prediction result exceeds the set threshold, the flutter feature map is input into the flutter identification model, and after outputting the flutter prediction result of the target part, the method also includes:

[0092] Based on the image orientation gradient factor and its distribution, the flutter prediction results are analyzed to obtain the flutter prediction target results.

[0093] In one embodiment, flutter prediction results are analyzed based on image orientation gradient factors and distribution anomalies to obtain flutter prediction target results, including:

[0094] Global high-frequency feature extraction is performed on the flutter prediction results based on the image orientation gradient factor to obtain a feature element image;

[0095] Non-edge maxima suppression is applied to the feature element image to obtain the first feature element image;

[0096] The moment feature is obtained by describing the feature information of the first feature element image using moments.

[0097] Based on the analysis of the moment characteristics of flutter anomalies in the distribution, the flutter prediction target results are obtained.

[0098] In the specific implementation process, for results that may exhibit chatter, a re-analysis of chatter anomalies based on image orientation gradient factors and their distribution is performed. Further identification is then conducted based on the results output by the analysis model, resulting in refined identification and analysis. This improves the accuracy of chatter identification, significantly reducing the risk of production line downtime due to misidentification, and ensuring the quality, reliability, and manufacturing efficiency of the processing. Specifically:

[0099] Image orientation gradient factor feature extraction method: For the Result feature map Fimg that may have flutter, perform global high-frequency feature extraction, as shown in the following formula:

[0100]

[0101] in, The calculation is as follows:

[0102]

[0103] The corresponding feature direction is:

[0104]

[0105]

[0106] in, and Let x and y represent the partial derivatives along the X and Y axes, respectively, and let x and y represent the corresponding coordinates in the image Fimg. Represents the feature gradient; This represents the partial derivative. Based on the above formula, the feature extraction results are then obtained.

[0107] Non-Edge Maximal Suppression (NMS) Strategy. After the above steps, an image containing feature elements is obtained, i.e., the feature element image. Considering that the edge width of these features may exceed 1 pixel, in order to simplify subsequent processing, ensure region closure, and improve localization accuracy, the image needs to be thinned by non-edge maximal suppression. This thinning process follows the following principles: retaining the external edge features of the image while suppressing the internal edge features, ensuring that the edge lines in the final image are both clear and continuous. Regions that do not meet the size requirements are further filtered based on the area enclosed by each independent feature. The corresponding image Iout is shown in the attached figure. Figure 4 As shown.

[0108] Abstract description of feature information. After non-edge maximum suppression processing, feature element images can be obtained. To ensure the identification of closed regions, the moment features of these feature images need to be calculated. Moment features can help identify and describe closed regions in an image. The expression of moment features is as follows: if there is a variable x and a constant c, then E[(xc)k] is called the k-th moment of x with respect to point c. For an Iout image processed by non-edge maximum suppression, the pixel coordinates can be regarded as a two-dimensional random variable (x, y), so moments can be used to describe the features of the image. For an image Iout of size rows*col, the (i+j)-th moment in the image space can be expressed as:

[0109]

[0110] The sum of i+j determines the order of the moments; the centroid can be calculated when the sum of i+j is 1. , .

[0111] The final anomaly detection analysis considers the corresponding moment features to form continuous feature regions that appear at intervals, with each independent region exhibiting a characteristic of intermittent appearance along the horizontal direction, as shown in the attached figure. Figure 5 As shown, if the number of consecutive occurrences of the corresponding feature is greater than the set threshold Tnum, it indicates the presence of a flutter anomaly. The corresponding result is directly connected to the counter, which controls the final alarm shutdown. If it is less than or equal to the set threshold Tnum, it indicates that there is no flutter and it is in a normal state. The next unit time T of cyclic monitoring is started. Both the abnormal and normal results must be recorded in the log file.

[0112] The tremor detection model can be constructed based on a tremor anomaly detection algorithm using a memory mechanism. It analyzes the generated sequence feature map Fimg for each segment, making a preliminary judgment on potential data features that may cause tremor, providing input for further refined detection. The design, model construction, and training of the tremor anomaly detection algorithm based on the memory mechanism include: an input layer, using a sequence of feature images of number k as input; a memory mechanism layer, constructed based on multiple memory mechanism units to handle the correlation between time-series images; and an output layer, outputting the predicted tremor risk. The forget gate of the memory mechanism layer determines which information needs to be discarded, and its output is an information data between 0 and 1, representing how much information should be retained for each unit state; and a fully connected layer, mapping the output of the memory mechanism layer to the predicted tremor risk value. The specific formula is as follows:

[0113]

[0114] The input layer determines which information features need to be stored in memory.

[0115]

[0116] The next step is to generate candidate cell status data, which represents new information that can be added to the existing cell status.

[0117]

[0118] The state of the memory cell is updated using the following formula:

[0119]

[0120] The memory mechanism layer outputs a decision on which part of the current unit's information to output, and uses the result of non-linear processing using an activation function as the output.

[0121]

[0122] in, This represents the input information at the current moment; This indicates the hidden state at the previous moment; This represents the candidate state at the current moment; This represents the activation value of the forget gate; This indicates the activation value of the input gate; Indicates the current state of the cell; Indicates the cell state at the previous moment; This indicates the activation value of the output gate; The sigmoid function represents the activation function; tanh represents the activation function. Represents the weight matrix; This represents the bias vector. The loss function, chosen as the cross-entropy loss function, is used to evaluate the difference between the model's predictions and the true labels. The optimization algorithm, employing algorithms such as Adam, is used to update the model's parameters.

[0123] Model training optimizes the model's parameters using training data to accurately predict flutter risk. Batch training divides the training data into multiple batches, each containing a certain number of samples, and iteratively trains multiple batches to optimize the model. Early stopping can be used to monitor performance on the validation set during training, stopping training when performance no longer improves to prevent overfitting.

[0124] Model evaluation and optimization assess the model's predictive performance and optimize it based on the evaluation results. Cross-validation is employed, using K-fold cross-validation to evaluate the model's stability and generalization ability. Performance metrics include accuracy, recall, and F1 score. Hyperparameter tuning involves adjusting hyperparameters (such as the number of memory mechanism layers and hidden units) using methods like grid search and random search to optimize model performance.

[0125] See attached document Figure 6 In the attached Figure 6 The present application will be further described below with reference to the embodiments shown:

[0126] First, useful signals are acquired, namely, the acquisition and purchase strategies for machine tool, NC, and multi-sensor data are formulated, and the data transmission and communication methods are set. Through the design of a multi-sensor data optimization and evaluation mechanism, vibration data is filtered, namely Data (C1, C2, C3, C4, C5). Then, raw data is processed and noise is reduced based on the signal evaluation strategy. Flutter feature map is generated based on data decomposition and transformation. A recognition model is introduced, namely, a flutter anomaly recognition algorithm based on a memory mechanism is designed and recognized to determine whether the Result of the flutter feature map is abnormal. If it is not abnormal, the real-time flutter prediction result is obtained, and the process returns to the multi-sensor data optimization step to analyze the data in the next unit time T. The current prediction result is saved as log data.

[0127] If an anomaly is detected, a flutter anomaly reanalysis based on the image orientation gradient factor and its distribution is performed. If an anomaly is confirmed, a shutdown alarm control strategy is designed and saved as log data. If no anomaly is confirmed, the log data is saved. After the reanalysis is completed, the process returns to the multi-sensor data optimization step to analyze the data for the next unit time T, thus achieving real-time cyclic analysis.

[0128] The shutdown alarm control strategy is designed as follows: If the data obtained in a unit time T, after initial flutter anomaly identification based on a memory mechanism and refined flutter anomaly reanalysis based on image orientation gradient factors and distribution, both result in anomalies, then the final shutdown alarm counter value is incremented by 1, and the alarm counter is marked as W. If both corresponding results are not completely identified as anomalies, the corresponding value is decremented by 1. Initially, the alarm counter value is 0, and the value in the alarm counter is set to Wt. If the value of W is greater than or equal to Wt, where Wt is the alarm threshold, then the final shutdown alarm command is issued. It is worth noting that the result of the analysis of the data obtained in each unit time T will change the W value of the alarm counter, incrementing or decrementing it by one. Finally, the corresponding alarm information is input into the log file to provide traceable process data for subsequent analysis.

[0129] All log data is saved to obtain a complete log record. The designed log recording mode is to generate an analysis result once within a unit time T. Each result is saved as a line of log file data. At the same time, each line of data records detailed information such as time information, feature value information, and identification result information. One monitored object is recorded in one log file. If the size of the log file exceeds the set size, a new log file is saved.

[0130] In this embodiment, the original vibration sensor data is first optimized by screening to obtain data that is more conducive to feature expression in the processing feature direction. Secondly, the screened data is denoised to improve the signal-to-noise ratio and make the flutter signal features clearer. Then, data decomposition and transformation technology is introduced to visualize the flutter signal features as images. Finally, the flutter recognition model is used to quickly and accurately identify the images to analyze the pattern of the flutter signal and output flutter prediction results, effectively improving the effect of flutter recognition and analysis, and more accurately guiding flutter prevention and control work, which plays a positive role in improving the overall technical level of the aviation manufacturing industry.

[0131] See attached document Figure 7 Based on the same inventive concept as in the foregoing embodiments, this application also provides a chatter identification device for a part processing procedure, comprising:

[0132] The filtering module is used to filter the raw vibration sensor data of the target part's machining process based on the expression of machining features to obtain the first vibration sensor data.

[0133] The noise reduction module is used to perform noise reduction processing on the data from the first vibration sensor to obtain the data from the target vibration sensor.

[0134] The generation module is used to perform data decomposition and transformation based on the target vibration sensor data to generate a flutter feature map.

[0135] The identification module is used to input the flutter feature map into the flutter identification model and output the flutter prediction result of the target part.

[0136] Those skilled in the art should understand that the division of the various modules in the embodiments is merely a logical functional division. In actual applications, they can be fully or partially integrated into one or more actual carriers. These modules can be implemented entirely in software through processing unit calls, entirely in hardware, or a combination of software and hardware. It should be noted that each module in the part processing chatter identification device in this embodiment corresponds one-to-one with each step in the part processing chatter identification method in the aforementioned embodiments. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned part processing chatter identification method, which will not be repeated here.

[0137] Based on the same inventive concept as in the foregoing embodiments, embodiments of this application also provide a computer-readable storage medium storing a computer program. When the computer program is loaded and executed by a processor, it implements the chatter recognition method for the part processing process provided in the embodiments of this application.

[0138] Based on the same inventive concept as in the foregoing embodiments, embodiments of this application also provide an electronic device, including a processor and a memory, wherein,

[0139] Memory is used to store computer programs;

[0140] The processor is used to load and execute a computer program to enable the electronic device to perform a chatter recognition method for part processing as provided in the embodiments of this application.

[0141] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a device including one or any combination of the above-mentioned memories. The computer may be a variety of computing devices, including smart terminals and servers.

[0142] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0143] As an example, executable instructions may, but do not necessarily, correspond to files in the file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborative files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0144] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0145] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0146] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a multimedia terminal device (which may be a mobile phone, computer, television receiver, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0148] In summary, the embodiments of this application provide a method, apparatus, medium, and device for chatter identification during the machining process of a part. The method includes: filtering the original vibration sensor data of the machining process of the target part based on the expression of machining features to obtain first vibration sensor data; performing noise reduction processing on the first vibration sensor data to obtain target vibration sensor data; performing data decomposition and transformation based on the target vibration sensor data to generate a chatter feature map; inputting the chatter feature map into a chatter identification model to output the chatter prediction result of the target part. This application first optimizes the data by filtering the original vibration sensor data to obtain data that is more conducive to feature expression in the direction of machining features. Secondly, it performs noise reduction processing on the filtered data to improve the signal-to-noise ratio of the data, making the chatter signal features clearer. Then, it introduces data decomposition and transformation technology to visualize the chatter signal features as images. Finally, it uses a chatter identification model to quickly and accurately identify the images to analyze the pattern of chatter signals and output chatter prediction results, effectively improving the effect of chatter identification and analysis, and more accurately guiding chatter prevention and control work, playing a positive role in improving the overall technical level of the aerospace manufacturing industry.

[0149] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for chatter identification during part manufacturing, characterized in that, Includes the following steps: Based on the expression of processing features, the original vibration sensor data of the target part processing process is filtered to obtain the first vibration sensor data; The first vibration sensor data is subjected to noise reduction processing to obtain the target vibration sensor data; Data decomposition and transformation are performed on the target vibration sensor data to generate a flutter feature map; The flutter feature map is input into the flutter recognition model, and the flutter prediction result of the target part is output. The flutter prediction result represents the probability of flutter occurrence. If the probability represented by the flutter prediction result exceeds a set threshold, after inputting the flutter feature map into the flutter recognition model and outputting the flutter prediction result of the target part, the method further includes: The flutter prediction results are analyzed based on the image directional gradient factor and the distribution of flutter anomalies to obtain the flutter prediction target result, including: extracting global high-frequency features from the flutter prediction results based on the image directional gradient factor to obtain a feature element image; performing non-edge maxima suppression on the feature element image to obtain a first feature element image; describing the feature information of the first feature element image using moments to obtain moment features; and analyzing the moment features based on the distribution of flutter anomalies to obtain the flutter prediction target result.

2. The chatter identification method for part processing according to claim 1, characterized in that, The step of performing data decomposition and transformation based on the target vibration sensor data to generate a flutter feature map includes: Based on the target vibration sensor data, the data is segmented in the time domain to obtain multiple segmented data. Data decomposition and transformation are performed on multiple segments of data to obtain the frequency components of the segments at different time points; The frequency components of multiple segmented data at different time points are combined to generate a flutter feature map.

3. The chatter identification method for part processing according to claim 1, characterized in that, Before obtaining the first vibration sensor data by filtering the raw vibration sensor data of the target part's machining process based on the expression of machining features, the method further includes: Based on the changes in machine tool processing parameters between the current time and the previous adjacent time, it is determined whether the machine tool is in a processing state. If the determination result is that the machine tool is in a processing state, the expression based on processing characteristics is executed to filter the original vibration sensor data of the target part processing process and obtain the first vibration sensor data.

4. The chatter identification method for part processing according to claim 1, characterized in that, The expression based on processing features filters the original vibration sensor data of the target part's processing to obtain the first vibration sensor data, including: Based on the direction that is more favorable to the expression of processing features, the variance of the target part in the original vibration sensor data of the processing of the target part is compared. The data with the largest variance in the target portion is selected and retained together with the data in the original vibration sensor data excluding the target portion to obtain the first vibration sensor data.

5. The chatter identification method for part processing according to claim 1, characterized in that, Before obtaining the first vibration sensor data by filtering the raw vibration sensor data of the target part's machining process based on the expression of machining features, the method further includes: Acquire useful signals for processing the target part; The useful signal is sampled to obtain the original vibration sensor data.

6. A chatter detection device for parts processing, characterized in that, include: The filtering module is used to filter the raw vibration sensor data of the target part's machining process based on the expression of machining features to obtain the first vibration sensor data. The noise reduction module is used to perform noise reduction processing on the first vibration sensor data to obtain the target vibration sensor data; The generation module is used to perform data decomposition and transformation based on the target vibration sensor data to generate a flutter feature map; The identification module is used to input the flutter feature map into the flutter identification model and output the flutter prediction result of the target part; the flutter prediction result represents the probability of flutter occurring, and when the probability represented by the flutter prediction result exceeds a set threshold, after inputting the flutter feature map into the flutter identification model and outputting the flutter prediction result of the target part, the module further includes: The flutter prediction results are analyzed based on the image directional gradient factor and the distribution of flutter anomalies to obtain the flutter prediction target result, including: extracting global high-frequency features from the flutter prediction results based on the image directional gradient factor to obtain a feature element image; performing non-edge maxima suppression on the feature element image to obtain a first feature element image; describing the feature information of the first feature element image using moments to obtain moment features; and analyzing the moment features based on the distribution of flutter anomalies to obtain the flutter prediction target result.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is loaded and executed by the processor, it implements the chatter identification method for the part machining process as described in any one of claims 1-5.

8. An electronic device, characterized in that, Including processor and memory, among which, The memory is used to store computer programs; The processor is used to load and execute the computer program to enable the electronic device to perform the chatter identification method for the part machining process as described in any one of claims 1-5.

Images