Tool wear detection method, device, equipment and medium for numerical control machine tool

Abstract

The application provides a tool wear detection method, device, equipment and medium of a numerical control machine tool, and relates to the technical field of engineering machinery.The method comprises the following steps: acquiring sensor standard data features corresponding to a tool of a numerical control machine tool; inputting the sensor standard data features into a pre-trained wear detection neural network model to obtain a wear detection result of the tool output by the wear detection neural network model; and the wear detection neural network model is obtained based on sensor standard data feature samples and corresponding labels.The technical scheme of the application performs tool wear detection through the pre-trained wear detection neural network model, and the detection of tool wear does not need to be performed by an operator according to experience or a traditional mechanical model, so that the labor cost is reduced, the tool wear state can be detected in a complex machining environment, the detection accuracy of the tool wear state is improved, and therefore, the waste of resources in the machining process of the numerical control machine tool is reduced, and the machining precision is improved.

Description

Technical Field

[0001] This invention relates to the field of engineering machinery technology, and in particular to a method, apparatus, equipment and medium for detecting tool wear on CNC machine tools. Background Technology

[0002] In manufacturing, CNC machine tools, as one of the core pieces of equipment, play an indispensable role in improving production efficiency and ensuring machining accuracy and quality. However, in actual production processes, unavoidable factors such as tool wear on CNC machine tools often have a significant impact on machining efficiency and quality, becoming a major factor restricting machining accuracy and efficiency. According to statistics, approximately 15% of CNC machine tool maintenance work stems from excessive tool wear, reflecting the enormous impact of tool wear on machining efficiency and production costs.

[0003] In CNC machine tool machining, premature tool replacement not only increases production costs but also wastes tool life; conversely, delayed tool replacement can lead to decreased machining accuracy and even equipment failure or production accidents. Therefore, replacing tools based on their wear level is crucial. Tool wear is influenced by various factors, such as machining parameters, tool material, and the material being machined, resulting in different patterns and degrees of tool wear. Currently, CNC machine tool replacement still relies on operators' experience to judge the degree of tool wear, or on operators calculating tool wear using traditional mechanical formulas.

[0004] This manual method is highly uncertain and has high labor costs. Traditional mechanical models also struggle to accurately predict and measure tool wear in complex machining environments, often leading to resource waste and decreased machining accuracy. Summary of the Invention

[0005] This invention provides a method, apparatus, equipment, and medium for detecting tool wear on CNC machine tools. It addresses the shortcomings of existing technologies that rely on manual methods, which suffer from significant uncertainty and high labor costs. Furthermore, traditional mechanical models struggle to accurately predict and measure tool wear in complex machining environments, often leading to resource waste and decreased machining accuracy. This invention utilizes a pre-trained wear detection neural network model for tool wear detection, eliminating the need for operators to rely on experience or traditional mechanical models. This reduces labor costs, makes the invention applicable to complex machining environments, and improves the accuracy of tool wear detection, thereby reducing resource waste and increasing machining accuracy during CNC machine tool processing.

[0006] This invention provides a method for detecting tool wear on CNC machine tools, comprising the following steps.

[0007] Obtain the standard sensor data characteristics corresponding to the cutting tools of CNC machine tools;

[0008] The standard data features of the sensor are input into a pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on sensor standard data feature samples and corresponding labels.

[0009] According to the present invention, a method for detecting tool wear in a CNC machine tool includes acquiring sensor standard data features corresponding to the tool of the CNC machine tool, comprising:

[0010] Acquire multiple initial sensor data corresponding to the cutting tool; the initial sensor data includes vibration signals, acoustic signals, and current signals;

[0011] For each of the initial sensor data, the initial sensor data is preprocessed to obtain the sensor data features corresponding to the initial sensor data;

[0012] The sensor data features are normalized to obtain the standard data features of the sensors.

[0013] According to the present invention, a method for detecting tool wear in a CNC machine tool includes preprocessing comprising: noise reduction, data cleaning, and feature extraction.

[0014] The preprocessing of the initial sensor data to obtain the sensor data features corresponding to the initial sensor data includes:

[0015] The initial sensor data is denoised using wavelet transform to obtain the denoised sensor data corresponding to the initial sensor data.

[0016] The outlier detection method is used to clean the denoised sensor data to obtain the cleaned sensor data corresponding to the denoised sensor data.

[0017] Feature extraction is performed on the cleaning sensor data to obtain the sensor data features corresponding to the cleaning sensor data.

[0018] According to the present invention, a method for detecting tool wear on a CNC machine tool includes a step of using wavelet transform to denoise the initial sensor data to obtain denoised sensor data corresponding to the initial sensor data, comprising:

[0019]

[0020] in, This represents the noise-reducing sensor data. The scaling factor represents the corresponding frequency feature. The scale factor represents the corresponding time feature. This represents the initial sensor data. To represent conjugate complex numbers, Indicates the time.

[0021] According to the present invention, a method for detecting tool wear on a CNC machine tool includes inputting the standard data features of the sensor into a pre-trained wear detection neural network model to obtain the wear detection result corresponding to the tool output by the wear detection neural network model, comprising:

[0022] The sensor standard data features are position-encoded to obtain the encoded data features corresponding to the sensor standard data features;

[0023] The encoded data features are input into the Transformer encoder in the wear detection neural network model to obtain the feature tensor corresponding to the encoded data features output by the Transformer encoder.

[0024] The feature tensor is input into the stacked bidirectional long short-term memory network in the wear detection neural network model to obtain the dimension-reduced feature vector corresponding to the feature tensor output by the stacked bidirectional long short-term memory network.

[0025] The reduced-dimensional feature vector is input into the fully connected layer of the wear detection neural network model to obtain the wear detection result corresponding to the reduced-dimensional feature vector output by the fully connected layer.

[0026] According to the present invention, a method for detecting tool wear on a CNC machine tool further includes:

[0027] During the training process of the wear detection neural network model, the Bayesian optimization method is used to adjust the number of layers in the Transformer encoder, the number of attention heads in the Transformer encoder, and the number of hidden unit layers in the stacked bidirectional long short-term memory network; the loss function corresponding to the wear detection neural network model is determined based on the mean square error loss and the focus loss.

[0028] According to the present invention, a method for detecting tool wear on a CNC machine tool further includes:

[0029] Based on the wear detection results, determine the change curve of the detection results and the remaining service life of the tool;

[0030] The wear detection results, the change curve of the detection results, and the remaining service life of the tool are displayed in real time through a graphical interactive interface.

[0031] Based on the wear detection result and its location within a target preset threshold range from multiple preset threshold ranges, an alarm response is generated corresponding to the target preset threshold range using a graphical user interface.

[0032] The present invention also provides a tool wear detection device for CNC machine tools, comprising the following modules:

[0033] The acquisition module is used to acquire the standard sensor data features corresponding to the cutting tools of CNC machine tools;

[0034] The detection module is used to input the standard data features of the sensor into a pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on the standard data feature samples of the sensor and the corresponding labels.

[0035] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the tool wear detection method for any of the CNC machine tools described above.

[0036] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the tool wear detection method for CNC machine tools as described above.

[0037] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements a tool wear detection method for a CNC machine tool as described above.

[0038] This invention provides a method, apparatus, equipment, and medium for detecting tool wear in CNC machine tools. It acquires standard sensor data features corresponding to the tool of the CNC machine tool; inputs these standard sensor data features into a pre-trained wear detection neural network model to obtain the wear detection result corresponding to the tool output by the wear detection neural network model; the wear detection neural network model is trained based on standard sensor data feature samples and corresponding labels. This invention's technical solution uses a pre-trained wear detection neural network model for tool wear detection, eliminating the need for operators to detect tool wear based on experience or traditional mechanical models, reducing labor costs, and making it applicable to complex machining environments. It improves the accuracy of tool wear detection, thereby reducing resource waste during CNC machine tool machining and improving machining precision. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating the tool wear detection method for CNC machine tools provided by the present invention.

[0041] Figure 2 This is a schematic diagram of the structure of the CNC machine tool and the execution body of the present invention.

[0042] Figure 3 This is a schematic diagram of the wear detection neural network model provided by the present invention.

[0043] Figure 4 This is a schematic diagram of the tool wear detection device for CNC machine tools provided by the present invention.

[0044] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0046] To address the aforementioned problems in the prior art, this invention provides a method for detecting tool wear in CNC machine tools. It should be noted that the executing entity of this invention's technical solution can be an industrial control computer or other electronic devices, and the embodiments of this invention are not specifically limited here. Figure 1 This is a flowchart illustrating the tool wear detection method for CNC machine tools provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps 110 to 120.

[0047] Step 110: Obtain the standard data features of the sensors corresponding to the cutting tools of the CNC machine tool.

[0048] Specifically, standard sensor data features corresponding to the cutting tool of the CNC machine tool can be obtained. It is easy to understand that these standard sensor data features can reflect the current state of the cutting tool. In this embodiment, the standard sensor data features obtained can be within a time window. In this case, the final wear detection result is also the wear detection result within that time window.

[0049] In one embodiment, acquiring the sensor standard data features corresponding to the cutting tool of the CNC machine tool includes:

[0050] Acquire multiple initial sensor data corresponding to the cutting tool; the initial sensor data includes vibration signals, acoustic signals, and current signals;

[0051] For each of the initial sensor data, the initial sensor data is preprocessed to obtain the sensor data features corresponding to the initial sensor data;

[0052] The sensor data features are normalized to obtain the standard data features of the sensors.

[0053] Specifically, Figure 2 This is a structural schematic diagram of the CNC machine tool and the execution body of the present invention, as shown below. Figure 2 As shown, the executing entity of this invention is an industrial control computer. The data acquisition card can acquire data from the vibration sensor, the acoustic emission sensor, and the current loop, respectively. That is, the data acquisition card obtains vibration signals from the vibration sensor, sound wave signals from the acoustic emission sensor, and current signals from the current loop. To ensure the synchronization and high timeliness of various signals, the sampling frequency of the data acquisition card can be set to no less than 10 kHz. Furthermore, the data acquisition card can perform synchronization alignment processing on multiple types of signals using a unified timestamp. After synchronization alignment processing, multiple initial sensor data can be obtained. The industrial control computer can then acquire all the initial sensor data from the data acquisition card.

[0054] Furthermore, since the initial sensor data may contain noise and outliers, the industrial control computer can preprocess each initial sensor data separately to obtain the corresponding sensor data features. Then, a normalization transformation can be performed on all sensor data features, mapping them to a unified feature space to obtain standard sensor data features. The normalization transformation can employ the z-score method.

[0055] In the above embodiments, by preprocessing and normalizing the initial sensor data, the stability and expressive power of the data can be improved, providing high-quality data support for subsequent wear prediction.

[0056] In one embodiment, the preprocessing includes: noise reduction, data cleaning, and feature extraction;

[0057] The preprocessing of the initial sensor data to obtain the sensor data features corresponding to the initial sensor data includes:

[0058] The initial sensor data is denoised using wavelet transform to obtain the denoised sensor data corresponding to the initial sensor data.

[0059] The outlier detection method is used to clean the denoised sensor data to obtain the cleaned sensor data corresponding to the denoised sensor data.

[0060] Feature extraction is performed on the cleaning sensor data to obtain the sensor data features corresponding to the cleaning sensor data.

[0061] Specifically, the preprocessing of the initial sensor data can sequentially include denoising, data cleaning, and feature extraction. The industrial control computer can use wavelet transform to denoise the initial sensor data, obtaining denoised sensor data. Then, after obtaining the denoised sensor data, outlier detection can be used to clean the data, obtaining cleaned sensor data. Finally, after obtaining the cleaned sensor data, feature extraction can be performed to obtain the corresponding sensor data features. This feature extraction can be performed in the time domain, frequency domain, and time-frequency domain, respectively.

[0062] In the above embodiments, noise reduction, data cleaning, and feature extraction are used to improve the quality of the initial sensor data, thereby laying the foundation for accurate modeling and prediction of tool wear.

[0063] In one embodiment, the step of denoising the initial sensor data using a wavelet transform method to obtain denoised sensor data corresponding to the initial sensor data includes:

[0064]

[0065] in, This represents the noise-reducing sensor data. The scaling factor represents the corresponding frequency feature. The scale factor represents the corresponding time feature. This represents the initial sensor data. To represent conjugate complex numbers, Indicates the time.

[0066] Specifically, the wavelet transform is a mathematical tool widely used in the field of time-frequency analysis, which can provide local characteristics of a signal at different times and frequencies.

[0067] In the above embodiments, wavelet transform is used to denoise the initial sensor data. Compared with traditional denoising methods such as low-pass filtering and median filtering, wavelet transform denoising can better preserve signal details while more effectively removing noise and improving the signal-to-noise ratio.

[0068] Step 120: Input the standard data features of the sensor into the pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on the standard data feature samples of the sensor and the corresponding labels.

[0069] Specifically, standard sensor data feature samples can be pre-constructed. It's easy to understand that the method for obtaining these standard sensor data feature samples is the same as the method for obtaining standard sensor data features. The standard sensor data feature samples and their corresponding labels can be pre-labeled. After obtaining the standard sensor data feature samples and their corresponding labels, the initial neural network model can be trained to obtain the wear detection neural network model.

[0070] Furthermore, during the tool wear detection process, the industrial control computer can input the standard data features of the sensor into the pre-trained wear detection neural network model to obtain the wear detection result of the tool corresponding to the current time window, which is output by the wear detection neural network model.

[0071] In one embodiment, the step of inputting the standard data features of the sensor into a pre-trained wear detection neural network model to obtain the wear detection result corresponding to the tool output by the wear detection neural network model includes:

[0072] The sensor standard data features are position-encoded to obtain the encoded data features corresponding to the sensor standard data features;

[0073] The encoded data features are input into the Transformer encoder in the wear detection neural network model to obtain the feature tensor corresponding to the encoded data features output by the Transformer encoder.

[0074] The feature tensor is input into the stacked bidirectional long short-term memory network in the wear detection neural network model to obtain the dimension-reduced feature vector corresponding to the feature tensor output by the stacked bidirectional long short-term memory network.

[0075] The reduced-dimensional feature vector is input into the fully connected layer of the wear detection neural network model to obtain the wear detection result corresponding to the reduced-dimensional feature vector output by the fully connected layer.

[0076] Specifically, Figure 3 This is a schematic diagram of the wear detection neural network model provided by the present invention, as shown below. Figure 3 As shown, data preprocessing refers to the preprocessing of the original data (the data used to train the model) during the training phase of the wear detection neural network model. After data augmentation, feature extraction, and data standardization, the original data yields sensor standard data feature samples and corresponding labels. These sensor standard data feature samples and corresponding labels are then divided into training, validation, and test sets. The wear detection neural network model is then trained based on these sets, resulting in the trained wear detection neural network model. The sequence length of a single feature vector in the training set is L, and this feature vector includes... Subsequences, Indicates subsequence 1, Indicates subsequence 2, Representing a subsequence A single feature vector has a feature dimension of D, and includes a total of One dimension, Indicates the first dimension. Indicates the second dimension. Indicates the first The wear detection neural network model comprises a Transformer encoder, a stacked bidirectional long short-term memory (BiLSTM) network, and an output layer. The Transformer encoder can include a multi-head self-attention structure based on a self-attention mechanism, which effectively captures the global correlations between key features in the input data. In this embodiment, while maintaining the standard Transformer encoder framework, a channel-wise attention module can be further introduced to model the interaction relationships between different modal features in the standard sensor data, thereby enhancing the multimodal information fusion capability.

[0077] Stacked BiLSTM networks can construct bidirectional information transmission structures, simultaneously capturing the dynamic dependencies between forward and backward directions in time series, thus improving the model's responsiveness to wear-out stage characteristics and anomalous mutation trends. To enhance modeling depth and feature representation capabilities, this stacked BiLSTM network can employ a two- or multi-layer stacked structure, introducing residual connections between layers to mitigate gradient decay issues that may occur during deep network training. To avoid information overload or redundant interference from global features extracted by the Transformer encoder on the time series modeling of the stacked BiLSTM network, this embodiment can also introduce a feature compression and reconstruction module (Feature Bottleneck) between the output of the Transformer encoder and the input of the stacked BiLSTM network. This module reconstructs and filters feature dimensions through a set of trainable fully connected layers, thereby enhancing the discriminativeness and specificity of feature representations.

[0078] The output layer can include the input layer and a fully connected layer. It can also introduce a time gating module to filter and control the feature representations at different time points, thereby further compressing invalid feature redundancy and improving the model's responsiveness to key nodes.

[0079] Furthermore, the standard data features can be position-encoded to obtain the coded data features corresponding to the standard data features of the sensor. In the position encoding design, temporal positional embedding can be introduced to embed time information into the sequence in the form of a vector, thereby improving the model's ability to recognize the time dynamic features during the processing.

[0080] Further, the wear detection results of the tool can be obtained through the following steps (1) to (4):

[0081] (1) The pre-trained wear detection neural network model can be lightweighted and deployed on an industrial control computer. The wear detection neural network model can also be retrained based on real-time data to improve its accuracy.

[0082] (2) The encoded data features can be input into the Transformer encoder in the wear detection neural network model, and then pass through the multi-head attention mechanism, first layer normalization, feedforward neural network, and second layer normalization in the Transformer encoder to obtain the feature tensor corresponding to the encoded data features output by the Transformer encoder.

[0083] (3) The feature tensor can be input into the stacked BiLSTM network in the wear detection neural network model to obtain the dimensionality-reduced feature vector corresponding to the feature tensor output by the stacked BiLSTM network. The stacked BiLSTM network can preserve the temporal structure and dynamic features of the original data.

[0084] (4) Input the dimension-reduced feature vector into the fully connected layer in the output layer of the wear detection neural network model to obtain the wear detection result corresponding to the dimension-reduced feature vector output by the fully connected layer.

[0085] In the above embodiments, the sequential processing of the Transformer encoder, stacked BiLSTM network, and fully connected layer effectively enhances the deep neural network's ability to model complex temporal structures, enabling high-precision prediction of tool wear state and its development trend. Furthermore, the introduction of residual connections and layer normalization mechanisms during feature propagation maintains gradient stability and improves the convergence speed of model training.

[0086] This invention provides a tool wear detection method for CNC machine tools. The method acquires standard sensor data features corresponding to the tool of the CNC machine tool; inputs these standard sensor data features into a pre-trained wear detection neural network model to obtain the tool wear detection result output by the wear detection neural network model; the wear detection neural network model is trained based on standard sensor data feature samples and corresponding labels. This invention's technical solution uses a pre-trained wear detection neural network model for tool wear detection, eliminating the need for operators to detect tool wear based on experience or traditional mechanical models. This reduces labor costs, is applicable to complex machining environments, and improves the accuracy of tool wear detection, thereby reducing resource waste during CNC machine tool machining and improving machining precision.

[0087] In one embodiment, the method further includes:

[0088] During the training process of the wear detection neural network model, the Bayesian optimization method is used to adjust the number of layers in the Transformer encoder, the number of attention heads in the Transformer encoder, and the number of hidden unit layers in the stacked bidirectional long short-term memory network; the loss function corresponding to the wear detection neural network model is determined based on the mean square error loss and the focus loss.

[0089] Specifically, in the training process of the wear detection neural network model, the Transformer encoder and the stacked BiLSTM network can be pre-trained separately. This allows the Transformer encoder to learn local feature relationships, and the stacked BiLSTM network to learn temporal evolution patterns. Then, the Transformer encoder and the stacked BiLSTM network can be jointly fine-tuned. After this joint fine-tuning, Bayesian optimization methods are used to adjust the number of layers in the Transformer encoder, the number of attention heads in the Transformer encoder, and the number of hidden unit layers in the stacked bidirectional long short-term memory network.

[0090] Furthermore, during the training process of the wear detection neural network model, the loss function corresponding to the wear detection neural network model is determined based on the mean squared error loss and the focus loss. For example, to further enhance the synergistic effect of the Transformer encoder and the stacked BiLSTM network during training, a joint loss function design can be adopted. This involves setting an auxiliary loss branch at the end of the Transformer encoder to achieve preliminary discrimination of local wear trends. The main loss function is driven by the wear prediction results output by the stacked BiLSTM network, using a joint optimization strategy to simultaneously guide the learning direction of both the Transformer encoder and the stacked BiLSTM network. Through this fusion mechanism, the wear detection neural network model can simultaneously capture long-range dependencies and local temporal evolution patterns during the machining process, thereby significantly improving the accuracy, robustness, and generalization ability of tool wear prediction.

[0091] In addition, the system introduces a dynamic learning rate decay mechanism and an early stopping strategy to improve training convergence speed and prevent overfitting.

[0092] In the above embodiments, Bayesian optimization is used to adjust the parameters during the training of the wear detection neural network model, which can improve the model training efficiency and final performance. Determining the loss function based on mean squared error loss and focus loss enhances the model's sensitivity to wear mutation points and extreme operating conditions. This fusion modeling and optimization mechanism significantly improves the predictive model's adaptability and stability under multiple operating conditions and data sources.

[0093] In one embodiment, the method further includes:

[0094] Based on the wear detection results, determine the change curve of the detection results and the remaining service life of the tool;

[0095] The wear detection results, the change curve of the detection results, and the remaining service life of the tool are displayed in real time through a graphical interactive interface.

[0096] Based on the wear detection result and its location within a target preset threshold range from multiple preset threshold ranges, an alarm response is generated corresponding to the target preset threshold range using a graphical user interface.

[0097] Specifically, the wear detection result change curve and remaining tool life can be determined based on the wear detection results. For example, the wear detection result change curve can be plotted based on wear detection results from multiple time periods, and the remaining tool life can be determined based on the expected total tool life and the current wear detection results. Figure 2 As shown, the industrial computer can also control a graphical user interface, which can be a LabVIEW platform. LabVIEW can display wear detection results (i.e., ... Figure 2 (Predicted wear value). In addition, the graphical interface can display the change curve of the detection results and the remaining tool life.

[0098] Furthermore, multiple preset threshold ranges can be pre-set (e.g., corresponding to light, moderate, and heavy wear levels, respectively). Different preset threshold ranges correspond to different alarm responses. Based on the wear detection result and the target preset threshold range among the multiple preset threshold ranges, the alarm response corresponding to the target preset threshold range is performed through a graphical interactive interface.

[0099] In the above embodiments, the wear detection results, the change curve of the detection results and the remaining service life of the tool are displayed in real time based on the graphical interactive interface, and an alarm response is generated based on the correspondence between the wear detection results and the preset threshold range. This facilitates the operator to make maintenance decisions or tool replacement operations, and avoids the waste of resources and processing risks caused by premature or delayed replacement.

[0100] The tool wear detection device for CNC machine tools provided by the present invention will be described below. The tool wear detection device for CNC machine tools described below can be referred to in correspondence with the tool wear detection method for CNC machine tools described above.

[0101] Figure 4 This is a schematic diagram of the tool wear detection device for CNC machine tools provided by the present invention, as shown below. Figure 4 As shown, the tool wear detection device 400 for this CNC machine tool includes the following modules:

[0102] The acquisition module 410 is used to acquire the sensor standard data characteristics corresponding to the cutting tool of the CNC machine tool;

[0103] The detection module 420 is used to input the standard data features of the sensor into a pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on the standard data feature samples of the sensor and the corresponding labels.

[0104] In one embodiment, the acquisition module 410 is specifically used for:

[0105] Acquire multiple initial sensor data corresponding to the cutting tool; the initial sensor data includes vibration signals, acoustic signals, and current signals;

[0106] For each of the initial sensor data, the initial sensor data is preprocessed to obtain the sensor data features corresponding to the initial sensor data;

[0107] The sensor data features are normalized to obtain the standard data features of the sensors.

[0108] In one embodiment, the preprocessing includes: noise reduction, data cleaning, and feature extraction; the acquisition module 410 is further configured to:

[0109] The initial sensor data is denoised using wavelet transform to obtain the denoised sensor data corresponding to the initial sensor data.

[0110] The outlier detection method is used to clean the denoised sensor data to obtain the cleaned sensor data corresponding to the denoised sensor data.

[0111] Feature extraction is performed on the cleaning sensor data to obtain the sensor data features corresponding to the cleaning sensor data.

[0112] In one embodiment, the acquisition module 410 is further configured to:

[0113]

[0114] in, This represents the noise-reducing sensor data. The scaling factor represents the corresponding frequency feature. The scale factor represents the corresponding time feature. This represents the initial sensor data. To represent conjugate complex numbers, Indicates the time.

[0115] In one embodiment, the detection module 420 is specifically used for:

[0116] The sensor standard data features are position-encoded to obtain the encoded data features corresponding to the sensor standard data features;

[0117] The encoded data features are input into the Transformer encoder in the wear detection neural network model to obtain the feature tensor corresponding to the encoded data features output by the Transformer encoder.

[0118] The feature tensor is input into the stacked bidirectional long short-term memory network in the wear detection neural network model to obtain the dimension-reduced feature vector corresponding to the feature tensor output by the stacked bidirectional long short-term memory network.

[0119] The reduced-dimensional feature vector is input into the fully connected layer of the wear detection neural network model to obtain the wear detection result corresponding to the reduced-dimensional feature vector output by the fully connected layer.

[0120] In one embodiment, during the training of the wear detection neural network model, a Bayesian optimization method is used to adjust the number of layers in the Transformer encoder, the number of attention heads in the Transformer encoder, and the number of hidden unit layers in the stacked bidirectional long short-term memory network; the loss function corresponding to the wear detection neural network model is determined based on the mean squared error loss and the focus loss.

[0121] In one embodiment, the tool wear detection device for the CNC machine tool further includes an interaction module, which is specifically used for:

[0122] Based on the wear detection results, determine the change curve of the detection results and the remaining service life of the tool;

[0123] The wear detection results, the change curve of the detection results, and the remaining service life of the tool are displayed in real time through a graphical interactive interface.

[0124] Based on the wear detection result and its location within a target preset threshold range from multiple preset threshold ranges, an alarm response is generated corresponding to the target preset threshold range using a graphical user interface.

[0125] The tool wear detection device for CNC machine tools provided by this invention acquires standard sensor data features corresponding to the tool of the CNC machine tool; inputs the standard sensor data features into a pre-trained wear detection neural network model to obtain the wear detection result corresponding to the tool output by the wear detection neural network model; the wear detection neural network model is trained based on sensor standard data feature samples and corresponding labels. This invention's technical solution performs tool wear detection through a pre-trained wear detection neural network model, eliminating the need for operators to detect tool wear based on experience or traditional mechanical models, reducing labor costs, and making it applicable to complex machining environments. It improves the accuracy of tool wear detection, thereby reducing resource waste during CNC machine tool machining and improving machining precision.

[0126] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a tool wear detection method for a CNC machine tool, the method including:

[0127] Obtain the standard sensor data characteristics corresponding to the cutting tools of CNC machine tools;

[0128] The standard data features of the sensor are input into a pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on sensor standard data feature samples and corresponding labels.

[0129] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0130] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the tool wear detection method for CNC machine tools provided by the above methods, the method comprising:

[0131] Obtain the standard sensor data characteristics corresponding to the cutting tools of CNC machine tools;

[0132] The standard data features of the sensor are input into a pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on sensor standard data feature samples and corresponding labels.

[0133] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a tool wear detection method for a CNC machine tool provided by the methods described above, the method comprising:

[0134] Obtain the standard sensor data characteristics corresponding to the cutting tools of CNC machine tools;

[0135] The standard data features of the sensor are input into a pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on sensor standard data feature samples and corresponding labels.

[0136] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0137] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, 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 can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting tool wear on a CNC machine tool, characterized in that, include: Obtain the standard sensor data characteristics corresponding to the cutting tools of CNC machine tools; The standard data features of the sensor are input into a pre-trained wear detection neural network model to obtain the wear detection result of the tool output by the wear detection neural network model; the wear detection neural network model is trained based on sensor standard data feature samples and corresponding labels; The acquisition of sensor standard data features corresponding to the cutting tool of the CNC machine tool includes: Acquire multiple initial sensor data corresponding to the cutting tool; the initial sensor data includes vibration signals, acoustic signals, and current signals; For each of the initial sensor data, the initial sensor data is preprocessed to obtain the sensor data features corresponding to the initial sensor data; Normalize all the sensor data features to obtain the standard sensor data features; The preprocessing includes: noise reduction, data cleaning, and feature extraction; The preprocessing of the initial sensor data to obtain the sensor data features corresponding to the initial sensor data includes: The initial sensor data is denoised using wavelet transform to obtain the denoised sensor data corresponding to the initial sensor data. The outlier detection method is used to clean the denoised sensor data to obtain the cleaned sensor data corresponding to the denoised sensor data. Feature extraction is performed on the cleaning sensor data to obtain the sensor data features corresponding to the cleaning sensor data; The step of denoising the initial sensor data using wavelet transform to obtain denoised sensor data corresponding to the initial sensor data includes: ； in, This represents the noise-reducing sensor data. The scaling factor represents the corresponding frequency feature. The scale factor represents the corresponding time feature. This represents the initial sensor data. To represent conjugate complex numbers, Indicates time; The step of inputting the standard data features of the sensor into a pre-trained wear detection neural network model to obtain the wear detection result corresponding to the tool output by the wear detection neural network model includes: The sensor standard data features are position-encoded to obtain the encoded data features corresponding to the sensor standard data features; The encoded data features are input into the Transformer encoder in the wear detection neural network model to obtain the feature tensor corresponding to the encoded data features output by the Transformer encoder. The feature tensor is input into the stacked bidirectional long short-term memory network in the wear detection neural network model to obtain the dimension-reduced feature vector corresponding to the feature tensor output by the stacked bidirectional long short-term memory network. The reduced-dimensional feature vector is input into the fully connected layer of the wear detection neural network model to obtain the wear detection result corresponding to the reduced-dimensional feature vector output by the fully connected layer.

2. The method for detecting tool wear on CNC machine tools according to claim 1, characterized in that, The method further includes: During the training process of the wear detection neural network model, the Bayesian optimization method is used to adjust the number of layers in the Transformer encoder, the number of attention heads in the Transformer encoder, and the number of hidden unit layers in the stacked bidirectional long short-term memory network; the loss function corresponding to the wear detection neural network model is determined based on the mean square error loss and the focus loss.

3. The method for detecting tool wear on CNC machine tools according to claim 1, characterized in that, The method further includes: Based on the wear detection results, determine the change curve of the detection results and the remaining service life of the tool; The wear detection results, the change curve of the detection results, and the remaining service life of the tool are displayed in real time through a graphical interactive interface. Based on the wear detection result and its location within a target preset threshold range from multiple preset threshold ranges, an alarm response is generated corresponding to the target preset threshold range using a graphical user interface.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the tool wear detection method for CNC machine tools as described in any one of claims 1 to 3.

5. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the tool wear detection method for CNC machine tools as described in any one of claims 1 to 3.

Images