Visualization method and system based on periodic recognition of tool wear

Abstract

The invention provides a visualization method and system based on periodic recognition of tool wear, and relates to the technical field of data processing, and the method comprises the steps: collecting a cutting signal, and forming an original time domain data stream; calculating a single-circle period based on the rotating speed, and determining the theoretical blade number of the cutter in the visual time window; in the visual time window, calculating a peak value and a dynamic threshold value, and screening feature cutting points; periodic folding is carried out on the feature cutting points, and phase mapping and sector storage are completed; performing sector division and sector statistics on a periodic folding result, and calculating a centroid for dynamic label anchoring; constructing a multi-dimensional visual view and generating a polar coordinate mapping contour; according to the polar coordinate mapping contour, based on a percentage over-limit rule of baseline comparison, carrying out wear identification and blade positioning; and rendering a recognition result on an interface of the upper computer and outputting a real-time alarm and interaction prompt. The tool wear monitoring accuracy can be improved.

Description

A visualization method and system for periodic identification of tool wear Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a visualization method and system based on periodic identification of tool wear. Background Technology

[0002] Cutting tools are critical or high-value components in automated machining equipment. If a tool breaks during production, it will affect product quality and cause unplanned equipment downtime, resulting in unnecessary losses. Therefore, monitoring tool breakage is essential to ensure product quality and avoid unplanned downtime. Simultaneously, to further improve production efficiency and mitigate the significant impact of machining anomalies (tool breakage), it is necessary to add real-time monitoring to the data monitoring system and provide effective algorithms and criteria. This digital and intelligent approach should allow users to intuitively assess the machining status.

[0003] Traditional tool breakage monitoring technology typically relies on a single characteristic signal to determine whether a tool has broken. However, this judgment mechanism is somewhat limited and prone to misjudgment, resulting in inaccurate tool breakage detection results.

[0004] Meanwhile, existing tool wear monitoring also considers deploying multiple sensors and extracting valid outliers based on data for judgment. However, since there are many factors that affect the feature values ​​during machining, the fluctuations in feature values ​​are difficult to interpret. Therefore, high-resolution data is required to ensure the reliability of the detection results. As a result, the effect is not good when performing fault detection with a small amount of data. In most cases, the effective duration of relevant data in the machining process is often insufficient, thus affecting the effect of machining fault detection. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a visualization method based on periodic identification of tool wear, which can solve the technical problem of insufficient accuracy in tool wear monitoring in the prior art.

[0006] A first aspect of this invention provides a visualization method based on periodic identification of tool wear, comprising:

[0007] S1: Acquire cutting signals and generate raw time-domain data stream;

[0008] S2: Calculate the single-cycle period based on the rotational speed and determine the theoretical number of cutting edges of the tool within the visible time window;

[0009] S3: Within the visible time window, calculate the peak value and dynamic threshold and filter the feature cutting points;

[0010] S4: Perform periodic folding on the feature cutting points and complete phase mapping and sector storage;

[0011] S5: Perform sector division and sector statistics on the periodic folding results, and calculate the centroid for dynamic tag anchoring;

[0012] S6: Construct a multi-dimensional visualization view and generate a polar coordinate mapping profile;

[0013] S7: Based on the polar coordinate mapping profile, wear identification and blade positioning are performed according to the percentage over-limit rule of baseline comparison;

[0014] S8: Render the recognition results on the host computer interface and output real-time alarms and interactive prompts.

[0015] A second aspect of the present invention provides a visualization system based on periodic identification of tool wear, comprising: a processor and a memory;

[0016] The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the visualization method based on periodic identification of tool wear as described in the first aspect.

[0017] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0018] (1) In this embodiment of the invention, the anomaly can be precisely located to a certain cutting edge by sector positioning, rather than just giving a coarse-grained conclusion of tool anomaly. The multi-dimensional comprehensive evidence chain of sector positioning and relative change of contour baseline can reduce the risk of misjudgment and improve the accuracy of tool wear monitoring.

[0019] (2) In this embodiment of the invention, periodic folding brings together multiple cycles of data in a short period of time into the same phase domain, enhances periodicity and exposes random anomalies, improves detectability in scenarios with insufficient data duration, and uses the percentage over-limit rule based on baseline comparison as the core anomaly criterion, reducing dependence on absolute amplitude and making it more adaptable. Attached Figure Description

[0020] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0021] Figure 1 is a flowchart illustrating a visualization method for periodically identifying tool wear provided by an embodiment of the present invention.

[0022] Figure 2 is a time-domain waveform diagram of an original time-domain data stream provided in an embodiment of the present invention.

[0023] Figure 3 is a scatter plot of periodic folding provided by an embodiment of the present invention.

[0024] Figure 4 is a polar coordinate mapping diagram provided by an embodiment of the present invention.

[0025] Figure 5 is a schematic diagram of the structure of a visualization system based on periodic identification of tool wear provided in an embodiment of the present invention. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0027] The visualization method for periodic identification of tool wear provided by the present invention will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0028] Referring to Figure 1 in the specification, a flowchart illustrating a visualization method for periodically identifying tool wear provided by an embodiment of the present invention is shown.

[0029] This invention provides a visualization method based on periodic identification of tool wear, which may include the following steps:

[0030] S1: Acquire cutting signals and form a raw time-domain data stream.

[0031] In one possible implementation, S1 specifically includes sub-steps S101 and S102:

[0032] S101: During the milling process, continuous cutting-related signals serve as the data source.

[0033] Specifically, the signal type can be any one or more of strain, acceleration, current, or power.

[0034] Furthermore, during signal acquisition, the spindle rotates at a constant speed (e.g., 1908 RPM), which gives the cutting signal a very strong periodicity.

[0035] S102: The acquired time-domain data is arranged in chronological order to form the original time-domain data stream.

[0036] Referring to Figure 2 in the specification, a time-domain waveform diagram of a raw time-domain data stream provided by an embodiment of the present invention is shown.

[0037] S2: Calculate the single-cycle period based on the rotational speed and determine the theoretical number of cutting edges of the tool within the visible time window.

[0038] In one possible implementation, S2 specifically includes sub-steps S201 to S203:

[0039] S201: Obtain the actual machining speed (RPM) of the spindle and calculate the single-cycle cycle based on physical parameters:

[0040]

[0041] Among them, T rev RPM indicates the cycle time corresponding to one revolution of the spindle, while RPM indicates the machining speed set during the actual machining process.

[0042] S202: Determine the X-axis range of the current view as the visible time window, and obtain... , used to limit the data range for subsequent threshold statistics and feature point extraction, where t start Indicates the start time of the current view's time window, t end Indicates the end time of the current view's time window.

[0043] S203: Calculate the theoretical number of cutting edges of the tool within the visible time window:

[0044]

[0045] Where ExpectedCuts represents the theoretical number of cutting edges within the time window, and Flutes represents the number of cutting edges of the tool.

[0046] In this invention, on the one hand, a strict physical correlation is established between the original time-domain data and the actual kinematic parameters of the tool, so that subsequent cycle folding and phase mapping have an accurate time reference. On the other hand, by calculating the theoretical number of cutting edges, it can be determined whether the amount of data in the current window matches the theoretical number of cutting operations, providing a reasonable basis for subsequent anomaly detection, sector statistics, and wear identification, thereby improving the accuracy and reliability of the overall judgment.

[0047] S3: Within the visible time window, calculate the peak value and dynamic threshold, and filter the feature cutting points.

[0048] In one possible implementation, S3 specifically includes sub-steps S301 to S303:

[0049] S301: Within the visible time window, find the maximum peak value within the visible range based on the collected data, and simultaneously record the global maximum value and the maximum value of each sector as the basis for threshold filtering.

[0050] S302: Set the threshold coefficient for the peak ratio threshold method and calculate the dynamic threshold:

[0051]

[0052] Among them, V thresh V represents the dynamic threshold. max This represents the global maximum value within the current view, and α represents the threshold coefficient.

[0053] Those skilled in the art can set the value of the threshold coefficient according to the actual situation, and the present invention does not impose any limitations.

[0054] S303: Based on a dynamic threshold, check whether the amplitude of each sampling point within the time window is greater than the dynamic threshold. If so, use the sampling point as the feature cutting point.

[0055] In this invention, the dynamic threshold can adaptively change with the current processing state, avoiding the failure of the fixed threshold due to fluctuations in working conditions, material differences, or changes in sensor gain, thereby improving the stability and robustness of feature point extraction. Simultaneously, only key cutting events above a proportional threshold are extracted, effectively compressing the data scale and highlighting the main stress moments. This provides more representative input data for subsequent periodic folding, sector statistics, and polar coordinate contour construction, improving the accuracy and computational efficiency of wear identification.

[0056] S4: Perform periodic folding on the feature cutting points and complete phase mapping and sector loading.

[0057] In one possible implementation, S4 specifically includes sub-steps S401 to S403:

[0058] S401: Using one cycle as the period, the collected data points are folded into a visible time window to enhance the periodicity and expose random outliers.

[0059] S402: Record the absolute time t for each feature cutting point. abs And a modulo mapping is used to convert absolute time into periodic phase:

[0060]

[0061] Among them, t phaseThe periodic phase time, t abs T represents the absolute time of the feature points. rev This represents the cycle time of a single revolution, and fmod represents the modulo operation function.

[0062] S403: Write the phase-mapped feature points into the corresponding containers according to the sector identifier to obtain a set of periodically folded feature points organized by sector, which is used for subsequent sector statistics and centroid annotation.

[0063] In this invention, on the one hand, the periodic structural features are significantly enhanced, enabling stable cutting behavior to form a high-density cluster in the phase domain, while random anomalies will be revealed due to phase inconsistencies, thereby improving the identifiability of anomalies. On the other hand, by establishing a "phase-sector-physical cutting edge" correspondence through a sector-based data storage mechanism, a structured data foundation is provided for subsequent sector statistics, centroid calculation, and cutting edge-level positioning. This not only improves data expression efficiency but also creates conditions for accurately locating wear or breakage positions.

[0064] S5: Perform sector division and sector statistics on the periodic folding results, and calculate the centroid for dynamic tag anchoring.

[0065] In one possible implementation, S5 specifically includes sub-steps S501 to S504:

[0066] S501: For multi-blade end mills, a 360° logical circle is divided into multiple sectors, each sector corresponding to a specific physical cutting edge, so as to achieve cutting edge-level positioning by monitoring the peak value of a specific sector.

[0067] S502: Perform sector statistics on the periodically folded feature point set to obtain the number and proportion of feature points in each sector, and use this information to generate sector label content.

[0068] S503: Calculate the geometric center of the feature points within each sector as the label display position. ,in, This represents the geometric center coordinates of the feature point within the sector along the X-axis (phase / time axis). This indicates the geometric center coordinates of the feature point within the sector along the Y-axis (phase / time axis).

[0069] S504: Dynamically display sector label content at the label display position, so that the sector label content moves with the data group.

[0070] Referring to Figure 3 in the specification, a periodic folded scatter plot provided by an embodiment of the present invention is shown. The X-axis length is fixed to the time of one revolution (e.g., 0 ~ 0.0314s), and the Y-axis represents the amplitude. The data is presented as a "stack" of thousands of cutting points within a narrow periodic window over a period of time (e.g., the past 5 seconds). The sector marking method uses a vertical grid to divide the period into N equal parts, while the centroid marking method calculates the geometric center of each point within each sector. The label is then dynamically displayed at that location (e.g., S1: 15pts 20%). This method solves the problem of rigid label placement, ensuring the label always moves with the data set. Figure 3 also provides auxiliary information: if the cutting condition is good, the scatter points will cluster tightly in a specific phase; conversely, if the tool is worn or the spindle wobbles abnormally, the corresponding scatter points will exhibit phase shifts or amplitude divergence.

[0071] In this invention, on the one hand, abnormal features can be directly associated with specific physical cutting edges, achieving cutting edge-level positioning and improving the precision of fault diagnosis. On the other hand, by statistically analyzing the number and proportion of feature points in each sector, the stress activity or abnormal concentration of different cutting edges can be quantified, making the judgment basis more structured. Simultaneously, dynamic centroid labeling is employed, allowing the labels to move with the data cluster, avoiding fixed labels obscuring or deviating from the actual data area, improving the accuracy and readability of visualization, thereby enhancing the intuitive judgment capability and interactive experience on-site.

[0072] S6: Construct a multidimensional visualization view and generate a polar coordinate mapping profile.

[0073] In one possible implementation, S6 specifically includes sub-steps S601 to S604:

[0074] S601: Generates a time-domain waveform of the raw data to show the macroscopic trend of the signal and preserve the time context, and overlays threshold lines and over-threshold scatter points on the waveform to show which moments of excessive cutting force occurred.

[0075] S602: Generate a periodic folded scatter plot, fixing the X-axis length to a time range of one circle, the Y-axis to the amplitude, and stacking the feature cutting points within a time period in a single circle window, while also overlaying sector grids and centroid labels.

[0076] S603: The periodic phase time obtained by phase mapping is mapped to an angle value, and the amplitude is mapped to a radius to generate a polar coordinate mapping profile, which is used to simulate the force evolution behavior of the tool in actual machining.

[0077] S604: Set up a comparison between the baseline profile and the real-time profile in the polar coordinate mapping graph. The baseline profile is used to represent the baseline of the health status, and the real-time profile is used to represent the current real-time status, providing a visual input for subsequent baseline percentage exceedance determination.

[0078] Referring to Figure 4 in the specification, a polar coordinate mapping diagram provided by an embodiment of the present invention is shown. This can effectively simulate the force evolution behavior of the tool during actual machining. Wherein, t... phase The amplitude is mapped to the angle θ and the radius r. Users can perform intuitive analysis based on this image. The green dashed line represents the baseline contour of the "health status" or "previous moment". The red / colored solid line represents the current real-time contour. The WORN alarm indicates that when the current amplitude exceeds the baseline by a certain percentage (e.g., 30%), it will automatically mark [WORN] and turn red as an alarm. Here, the 30% exceedance percentage is used as an example. Users can set the evaluation ratio parameter themselves, from 0-100%, without limitation.

[0079] In this invention, the time-domain waveform diagram preserves the macroscopic trend and temporal context of the signal, enabling users to identify instantaneous impacts or abnormal peaks. The periodic folded scatter plot compresses multi-cycle data into a single-cycle phase space, enhancing periodic consistency and visually presenting phase shift and amplitude divergence. The polar coordinate mapping contour diagram maps phase to angle and amplitude to radius, forming a petal-like force evolution structure, presenting the force state of each cutting edge in a spatial form. By superimposing the baseline contour and the real-time contour on the polar coordinate diagram for comparison, not only can the wear trend be intuitively judged, but a clear visual basis is also provided for subsequent baseline percentage exceeding limit determination, thereby improving the accuracy, interpretability, and real-time interactive capabilities in the engineering field.

[0080] S7: Based on the polar coordinate mapping profile and the percentage over-limit rule based on baseline comparison, wear identification and blade positioning are performed.

[0081] In one possible implementation, S7 specifically includes sub-steps S701 and S702:

[0082] S701: Using the baseline profile as a reference, perform a baseline comparison on the current real-time profile, and use the deviation from the baseline ratio threshold as the anomaly criterion. When the current amplitude is detected to exceed the baseline ratio threshold, an anomaly marker result is generated. The marker result is used to trigger subsequent alarm display.

[0083] Specifically, it can provide configurable evaluation ratio parameters, allowing users to set the ratio parameters within the range of 0 to 100% to adapt to the wear evaluation needs of different processing scenarios.

[0084] S702: Backtrack the anomaly marking results to the sector division and sector statistics results to determine the sector where the anomaly occurred.

[0085] Based on the sector where the anomaly occurred, the specific physical blade where the anomaly occurred is located, thus achieving a closed-loop correlation from "anomaly occurrence" to "anomaly blade location".

[0086] This invention shifts the focus from absolute numerical judgment to relative rate of change judgment, freeing the judgment results from the direct influence of factors such as material differences, load fluctuations, or sensor gain variations, thereby improving the stability and adaptability of the identification. Simultaneously, by tracing abnormal results back to sector division and statistical structures, overall anomalies at the contour level can be accurately mapped to specific sectors, and then corresponding to the actual physical cutting edge, achieving cutting edge-level positioning and forming a closed-loop correlation mechanism of "contour change—sector anomaly—physical cutting edge." Furthermore, the proportional parameter is adjustable within the range of 0~100%, allowing the system to flexibly set sensitivity according to different processing conditions, ensuring accuracy while also considering false alarm control and engineering practicality.

[0087] S8: Render the recognition results on the host computer interface and output real-time alarms and interactive prompts.

[0088] In one possible implementation, S8 specifically includes sub-steps S801 to S804:

[0089] S801: The time-domain waveform, periodic folded scatter plot and polar coordinate mapping diagram generated in step S6 are displayed synchronously in the host computer software, so that users can intuitively observe the changes in cutting state under multiple views.

[0090] S802: Based on the abnormal marking result of step S7, perform alarm display in the polar coordinate mapping diagram. When the amplitude exceeds the baseline by a certain proportion, automatically mark [WORN] and present the alarm.

[0091] S803: Based on the abnormal sector location results output in step S7, the abnormal distribution of the corresponding sector is highlighted in the periodic folded scatter plot, and the abnormal blade edge and its statistical characteristics are indicated to the user in combination with the sector label (such as "S1: 15pts 20%)).

[0092] S804: Outputs judgment feedback while ensuring real-time performance, keeping the feedback delay of the algorithm system within 50 ms, thereby realizing real-time judgment and visual interaction.

[0093] In one possible implementation, after S6, the following is also included:

[0094] S9: Based on the polar coordinate mapping profile, the tool wear type is detected by a convolutional neural network that combines depthwise separable convolution with residual connections.

[0095] In one possible implementation, S9 specifically includes sub-steps S901 to S904:

[0096] S901: Map the polar coordinates to the contour map, and perform random regularization and normalization through the GELU layer and the batch normalization layer, respectively.

[0097] S902: Input the normalized image data into the multi-layer stacked feature extraction module to extract deep features.

[0098] The feature extraction module employs a mechanism combining depthwise separable convolution and residual connections, which is a significant innovation of this invention.

[0099] Specifically, the feature extraction module includes: a depthwise separable convolutional unit, a first GELU unit, a first BN unit, a pointwise convolutional unit, a second GELU unit, and a second BN unit. The depthwise separable convolutional unit, the first GELU unit, and the first BN unit are connected in series. The output of the first BN unit is concatenated with the input residual and then input to the pointwise convolutional unit. The pointwise convolutional unit, the second GELU unit, and the second BN unit are connected in series.

[0100] Depthwise Separable Convolution is a convolutional structure that breaks down traditional convolution operations into two steps: "channel-wise convolution" and "pointwise convolution." Pointwise convolution typically refers to a 1×1 convolution operation, whose main function is to reorganize features and perform linear mapping along the channel dimension. GELU (Gaussian Error Linear Unit) activation function is a non-linear activation function based on the idea of ​​Gaussian distribution. Its core idea is to determine the proportion of input values ​​that pass through a standard normal distribution based on the probability of the input value. Compared to the hard threshold truncation of the traditional ReLU function, GELU has smoother non-linear characteristics. BN (Batch Normalization) is a technique for standardizing the output of intermediate layers in a neural network. By normalizing the mean and variance of the current small batch of data, its distribution is stabilized within a preset range.

[0101] By combining depthwise separable convolution with residual connections, and configuring GELU activation units and batch normalization units before and after, a hierarchical structure is formed: spatial feature extraction – nonlinear enhancement – ​​normalization stabilization – residual fusion – channel recombination – further nonlinear enhancement and normalization. First, depthwise separable convolution significantly reduces the number of parameters and computational complexity, making the network more suitable for real-time tool wear detection scenarios. Second, residual connections superimpose input features with intermediate features, helping to alleviate gradient vanishing or exploding problems, improving training stability and preserving original structural information. Third, the GELU activation function enhances nonlinear expressive power, enabling the model to more finely characterize subtle morphological differences in polar coordinate contours. Simultaneously, batch normalization improves the network's convergence speed and enhances its generalization ability to different machining conditions. Finally, pointwise convolution achieves inter-channel feature fusion, effectively coupling spatial and channel features. The overall structure improves feature extraction depth, model stability, and classification accuracy while maintaining low computational overhead, balancing real-time performance and recognition performance.

[0102] It is particularly important to emphasize that by combining depthwise separable convolutions with residual connections, the number of network parameters and the risk of gradient explosion or gradient vanishing can be significantly reduced.

[0103] S903: Dimensionality reduction and expansion of deep features are achieved through global average pooling layers and flattening layers.

[0104] S904: Activates the tool wear type based on the extended features using the Softmax activation function.

[0105] In this embodiment of the invention, an extension from "rule-based anomaly detection" to "feature-based type classification" is achieved. Regularization and normalization using GELU layers and batch normalization layers enhance the model's generalization ability to contour changes under different working conditions. Employing a depthwise separable convolutional structure significantly reduces the number of parameters and computational load while maintaining feature extraction capabilities, making it suitable for real-time online detection scenarios. Introducing a residual connection mechanism alleviates the vanishing or exploding gradient problem, improving network training stability. Furthermore, global average pooling and Softmax output specific wear types, enabling the system not only to determine "whether it is abnormal" but also to distinguish different wear modes (such as wear, chipping, and breakage). Therefore, this step improves the system's intelligence and type recognition accuracy without compromising the original real-time performance and visualization framework, enhancing the overall monitoring system's functional scalability and engineering practical value.

[0106] It is important to emphasize that this invention proposes a visualization method for real-time judgment and identification of tool wear through periodic folding. The proposed periodic folding method can display data distribution patterns lasting for several seconds or even minutes within a very small screen space. Identifying tool wear does not affect real-time data acquisition; there is no mutual interference, and the acquisition frequency can still be maintained at 10 kHz. The judgment has high real-time performance, with a feedback time within 50 ms. The data source can be any signal type or any sensor, including strain, acceleration, current, or power. The method for identifying characteristic anomalies is the percentage exceeding the baseline; the percentage setting is not limited and can be adjusted by the user according to actual needs. This analysis theory no longer relies on absolute values ​​but focuses on relative change rates (Diff %), making it more adaptable.

[0107] Referring to Figure 5 in the specification, a schematic diagram of a visualization system based on periodic identification of tool wear provided by an embodiment of the present invention is shown.

[0108] This invention provides a visualization system 20 based on periodic identification of tool wear, comprising: a processor 201 and a memory 202.

[0109] The memory 202 stores programs or instructions that can run on the processor 201. When the program or instructions are executed by the processor 201, they implement the steps of the above-described visualization method based on periodic identification of tool wear and achieve the same technical effect. To avoid repetition, the present invention will not elaborate further.

[0110] This invention provides a readable storage medium that stores a program or instructions on the medium. When the program or instructions are executed by a processor, they implement the steps of the visualization method based on periodic identification of tool wear described above, and achieve the same technical effect. To avoid repetition, this invention will not elaborate further.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended 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. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A visualization method based on periodic identification of tool wear, characterized in that, include: S1: Acquire cutting signals and generate raw time-domain data stream; S2: Calculate the single-cycle period based on the rotational speed and determine the theoretical number of cutting edges of the tool within the visible time window; S3: Within the visible time window, calculate the peak value and dynamic threshold and filter characteristic cutting points; S4: Perform periodic folding on the feature cutting points and complete phase mapping and sector entry; S5: Divide the periodic folding results into sectors and perform sector statistics, and calculate the centroid for dynamic label anchoring; S6: Construct a multi-dimensional visualization view and generate polar coordinate mapping contours. S7: Based on the polar coordinate mapping profile, wear identification and blade positioning are performed according to the percentage over-limit rule based on baseline comparison; S8: The identification results are rendered on the host computer interface and real-time alarms and interactive prompts are output.

2. The visualization method for periodic identification of tool wear according to claim 1, characterized in that, S2 specifically includes: S201: obtaining the actual machining speed of the spindle and calculating the single-cycle period based on physical parameters; S202: determining the X-axis range of the current view as a visible time window, the visible time window being used to limit the data range for subsequent threshold statistics and feature point extraction; S203: calculating the theoretical number of cutting edges of the tool within the visible time window.

3. The visualization method for periodic identification of tool wear according to claim 1, characterized in that, S3 specifically includes: S301: Within the visible time window, find the maximum peak value within the visible range based on the collected data, and simultaneously record the global maximum value and the maximum value of each sector as the basis for threshold filtering; S302: Set the threshold coefficient of the peak ratio threshold method and calculate the dynamic threshold; S303: Based on the dynamic threshold, check whether the amplitude of each sampling point within the time window is greater than the dynamic threshold; if so, use the sampling point as the feature cutting point.

4. The visualization method for identifying tool wear based on periodicity as described in claim 1, characterized in that, S4 specifically includes: S401: Folding the collected data points into a visible time window with one cycle as the period to enhance the periodic features and expose random outliers; S402: Recording the absolute time of each feature cutting point and converting the absolute time into a periodic phase using modulo mapping; S403: Writing the phase-mapped feature points into corresponding containers according to sector identifiers to obtain a set of periodically folded feature points organized by sector, which is used for subsequent sector statistics and centroid labeling.

5. The visualization method for periodic identification of tool wear according to claim 1, characterized in that, S5 specifically includes: S501: For a multi-blade end mill, a 360° circle is logically divided into multiple sectors, each sector corresponding to a specific physical cutting edge, so as to achieve cutting edge-level positioning by monitoring the peak value of a specific sector; S502: Sector statistics are performed on the periodic folding feature point set to obtain the number and proportion of feature points in each sector, and used to generate sector label content; S503: The geometric center of the feature points in each sector is calculated as the label display position; S504: The sector label content is dynamically displayed at the label display position, so that the sector label content moves with the data group.

6. The visualization method for identifying tool wear based on periodicity as described in claim 1, characterized in that, S6 specifically includes: S601: Generating a time-domain waveform of the raw data to show the macroscopic trend of the signal and retain the time context, and superimposing a threshold line and over-threshold scatter points on the waveform to show which moments have excessive cutting forces; S602: Generating a periodic folded scatter plot, fixing the X-axis length to a time range of one circle, the Y-axis to the amplitude, and stacking the characteristic cutting points within a time period in a single-circle window, while superimposing sector grids and centroid labels; S603: Mapping the periodic phase time obtained by phase mapping to an angle value, and mapping the amplitude to a radius, generating a polar coordinate mapping contour map to obtain a real-time contour, used to simulate the force evolution behavior of the tool in actual machining; S604: Setting a comparison between the baseline contour and the real-time contour in the polar coordinate mapping map, wherein the baseline contour is used to represent the baseline of the health state, and the real-time contour is used to represent the current real-time state, providing visual input for subsequent baseline percentage over-limit determination.

7. The visualization method for periodic identification of tool wear according to claim 1, characterized in that, Following S6, the method further includes: S9: Based on the polar coordinate mapping profile, the tool wear type is detected by a convolutional neural network that combines depth-separable convolution with residual connections.

8. The visualization method for identifying tool wear based on periodicity as described in claim 7, characterized in that, S9 specifically includes: S901: Mapping the polar coordinates to the contour map, and performing random regularization and normalization through a GELU layer and a batch normalization layer, respectively; S902: Inputting the normalized image data into a multi-layer stacked feature extraction module to extract deep features. The feature extraction module adopts a mechanism based on a combination of depthwise separable convolution and residual connections; S903: Reducing and expanding the deep features through a global average pooling layer and a flattening layer; S904: Outputting the tool wear type based on the expanded features using the Softmax activation function.

9. The visualization method for identifying tool wear based on periodicity as described in claim 8, characterized in that, The feature extraction module specifically includes: a depthwise separable convolutional unit, a first GELU unit, a first BN unit, a pointwise convolutional unit, a second GELU unit, and a second BN unit; the depthwise separable convolutional unit, the first GELU unit, and the first BN unit are connected in series; the output of the first BN unit is connected to the input residual and then input to the pointwise convolutional unit; the pointwise convolutional unit, the second GELU unit, and the second BN unit are connected in series.

10. A visualization system based on periodic identification of tool wear, characterized in that, include: Processor and memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the visualization method based on periodic identification of tool wear as described in any one of claims 1 to 9.

Images