On-machine unsupervised real-time monitoring method for hob wear status in shifting cutter mode

Abstract

The present invention discloses an unsupervised, real-time monitoring method for the wear status of a hob cutter in a gear shifting mode. S1 collects hob vibration information during the gear hobbing process; S2 extracts multi-domain features from the vibration information corresponding to each processed gear; S3 constructs the statistical features corresponding to the processed gears within a gear shifting cycle into a gear shifting cycle monitoring feature set; S4 calculates the Q statistic limit corresponding to the gear shifting cycle; S5 calculates the Q statistic corresponding to the monitoring statistical features of all gears in the subsequent gear shifting cycle; and S6 counts the number of gears processed in the subsequent gear shifting cycle whose Q statistic exceeds the Q statistic limit of the previous cycle. When the Q statistic exceeds a set percentage of the total number of gears processed in the gear shifting cycle, it is considered that the hob wear level has changed. The present invention does not require image acquisition or the establishment of a labeled hob full life cycle dataset in advance, reducing the difficulty of data collection and model training, and can achieve real-time monitoring of the wear status of the hob cutter on the machine.

Description

Technical Field

[0001] The present invention relates to hob wear detection, and in particular to an on-machine unsupervised real-time monitoring method for hob wear status based on multi-domain statistical features and Q statistics in a tool shifting processing mode, belonging to the technical field of gear hobbing processing status monitoring. Background Art

[0002] Hobs are one of the most direct tools involved in gear machining. Wear is inevitable during the long, continuous gear machining process, leading to a decline in gear quality. However, the definition of hob wear is unclear. Severe hob wear can exacerbate machine tool chatter, leading to irreparable consequences if not detected promptly. To prevent hob wear from causing gear quality degradation, premature hob replacement reduces its utilization, increasing machine downtime and hob repair costs. To reduce tool replacement frequency, gear hobbing is often based on a tool-shifting strategy. Hobs are manufactured to be long and narrow. After each gear is machined, the hob moves along its axial direction to process new teeth. After a certain number of gears are machined, the hob returns to its original position. This process is considered a tool-shifting cycle. While this tool-shifting strategy reduces hob replacement frequency to some extent, it cannot meet the requirements for real-time monitoring of hob wear on the machine.

[0003] Patent publication number CN115541610A, "A Tool Wear Measurement Device," aims to enable a single microscope to capture images of worn tools at multiple specific angles while the machine is in operation. Patent publication number CN115488696A, "A Tool Wear State Prediction Method Based on Variational Mode Decomposition and Neural Networks," collects machining force and vibration information, measures tool wear through optical images, constructs a milling tool wear state dataset, and trains a three-layer BP neural network to predict tool wear. Patent publication number CN114102260A, "A Variable-Condition Tool Wear State Monitoring Method Driven by Mechanism-Data Fusion," constructs a milling process dynamics equation based on collected vibration and current information, derives cutting forces from a mechanistic perspective, and establishes a tool wear state prediction model through deep learning. Patent publication number CN115509178A, "A Tool Wear Monitoring Method and CNC Machine Tool Equipment Driven by Digital Twins," constructs a deep learning model based on convolutional neural networks and long short-term memory (LSTM) based on historical machining data collected by sensors to obtain an optimized tool wear monitoring model.

[0004] The above achievements demonstrate tool wear monitoring using both image-based and signal-based methods. However, the harsh gear hobbing environment, characterized by chips and oil contamination, can easily damage the camera, limiting the application of image-based hob wear monitoring. Frequent machine downtime for photo-taking also impacts production efficiency. Furthermore, the aforementioned signal-based hob wear status monitoring method relies on a pre-built, labeled dataset of the tool's entire lifecycle. In industrial environments, most monitoring data is unlabeled, normal status data, making it prohibitively expensive to obtain labeled data on worn hobs. Summary of the Invention

[0005] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to propose a method for on-machine unsupervised real-time monitoring of the hob wear status in the cutter shifting mode. This method does not rely on graphic acquisition or hob wear label data. By mining the difference characteristics of the hob wear status contained in the monitoring sensor signal, it can realize on-machine real-time monitoring of the hob wear status, thereby improving the tool utilization rate and reducing the downtime and maintenance time of the gear hobbing machine.

[0006] The technical solution of the present invention is achieved as follows:

[0007] The method for monitoring the wear state of the hob cutter in the cutter shifting mode in real time without supervision on the machine includes the following steps:

[0008] S1: Vibration monitoring data acquisition: Install a vibration sensor on the gear hobbing machine to obtain the vibration information of the hob during the gear hobbing process, thereby obtaining multiple monitoring samples;

[0009] S2: Multi-domain monitoring feature extraction: extract multi-domain statistical features for each monitoring sample obtained in S1;

[0010] S3: Construct a tool shifting cycle monitoring feature set. During a tool shifting cycle, all teeth on the hob perform the same cutting operation, and all positions on the hob are considered to be equally worn. The monitoring samples within the same tool shifting cycle extracted in S2 are combined into a tool shifting cycle monitoring feature set, thereby obtaining tool shifting cycle monitoring feature sets within different tool shifting cycles. The tool shifting cycle monitoring feature set is shown in the following formula:

[0011]

[0012] Among them, F i represents the feature set corresponding to the i-th hob shifting cycle, M represents the number of gears processed in a hob shifting cycle, N represents the number of multi-domain statistical features extracted from each monitoring sample, and t m,n Indicates the nth feature corresponding to the mth machined gear in the tool shifting cycle;

[0013] S4: Calculate the Q statistic limit of the i-th knife-crossing cycle;

[0014] S5: Calculate the Q statistics corresponding to all the gears processed in the (i+1)th tool shifting cycle;

[0015] S6: Determine whether the wear level has changed; in the (i+1)th tool shifting cycle, count the number of gears whose Q statistic exceeds the Q statistic limit of the i-th tool shifting cycle. When the number of gears does not exceed the set percentage of the number of gears processed in the (i+1)th tool shifting cycle, it is considered that the wear level of the hob in the (i+1)th tool shifting cycle has not changed relative to the i-th tool shifting cycle; otherwise, it is considered that the wear level of the hob in the (i+1)th tool shifting cycle has changed.

[0016] The multi-domain statistical features described in S2 include time domain features, frequency domain features, and wavelet domain features.

[0017] Preferably, when obtaining the vibration information of the hob during the gear hobbing process in S1, only the vibration signal of the hob during the stable processing stage of each gear tooth blank is intercepted to form a monitoring sample.

[0018] Furthermore, the step of calculating the Q statistic limit of the i-th tool-slip cycle in S4 is:

[0019] S41: Normalize each feature in the i-th tool shifting cycle monitoring feature set. The normalization formula is as follows:

[0020]

[0021] Among them, t' m,n Indicates t m,n After normalization, the features and σ n They represent the mean and standard deviation of the n-th dimension feature in the current knife-shifting cycle, and the standardized monitoring feature set F i ' means as follows

[0022]

[0023] S42: The standardized monitoring feature set F i 'Perform principal component analysis;

[0024] First, construct the corresponding covariance matrix, which is expressed as follows;

[0025] X i =(F i ') T ·F i '

[0026] where X i is the covariance matrix corresponding to the standardized monitoring feature set, (F i ') T It's Fi 'The transposed matrix;

[0027] Then calculate the eigenvalues and corresponding eigenvectors of the covariance matrix, and arrange the eigenvalues in descending order;

[0028] λ1≥λ2≥…≥λ N-1 ≥λ N

[0029] The number of the first l eigenvalues of the covariance matrix is determined again so that the cumulative sum of the eigenvalues of the first l covariance matrix is not less than the set proportion p of the sum of all eigenvalues;

[0030] S43: Based on the last (Nl) eigenvalues of the covariance matrix, the Q statistic limit of the i-th hob cutter shifting cycle is obtained, and its expression is as follows:

[0031]

[0032] Among them, Q α is the Q statistic limit obtained in the current knife-crossing cycle, C α is the confidence limit of the standard normal distribution, and the expressions of θ1, θ2 and h are as follows:

[0033]

[0034] Furthermore, the step of calculating the Q statistics corresponding to all the processed gears in the (i+1)th tool shifting cycle in S5 is:

[0035] S51: Obtaining the monitoring feature set of the (i+1)th tool shifting cycle according to step S3;

[0036] S52: performing normalization processing on the monitoring features corresponding to each gear being processed in the (i+1)th tool shifting cycle according to the method of step S41 to obtain a normalized monitoring feature set for the (i+1)th tool shifting cycle;

[0037] S53: Based on the standardized monitoring feature set of the (i+1)th tool shifting cycle, the calculation formula of the Q statistic corresponding to the mth machined gear in the (i+1)th tool shifting cycle is as follows

[0038]

[0039] Where I is an identity matrix, p is the reconstructed principal element matrix obtained by the eigenvectors corresponding to the first l eigenvalues of the covariance matrix in S42, and t' m It represents the normalized monitoring feature vector corresponding to the mth gear processed in the tool shifting cycle, Q m is the Q statistic corresponding to the mth gear processed in the tool shifting cycle.

[0040] Specifically, the mean value and standard deviation of the normalization process in step S52 are directly the mean value used in the calculation of the Q statistic limit value in the previous tool shifting cycle. and standard deviation σ n .

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention mines the tool wear status from the vibration monitoring signal, which can avoid the need for stopping the machine to take pictures based on the image method and the damage to the camera caused by oil stains, iron filings, etc., thereby reducing the downtime of the gear hobbing machine for maintenance.

[0043] 2. The present invention incorporates statistical features from multiple domains, such as time domain, frequency domain, wavelet domain, etc., which can more comprehensively describe the current wear status information of the hob and more accurately determine whether the wear level has changed.

[0044] 3. Most current tool wear monitoring methods rely on constructed labeled wear datasets. However, calibration of tool wear levels is difficult to achieve in industrial environments. This unsupervised process eliminates the need for a pre-built labeled dataset of the hob's entire lifecycle, reducing the complexity of data collection and model training. This reduces the high cost of building labeled datasets and makes it easier to apply to on-machine, real-time tool wear monitoring for machine tools of varying sizes and types.

[0045] 4. The present invention is an iterative statistical calculation process that does not require the acquisition of tool life cycle data in advance. It only identifies the jump in wear status based on the difference between the previous tool-shifting cycle and the next tool-shifting cycle, and has higher applicability for tool wear monitoring in the gear hobbing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Schematic diagram of the monitoring process of the present invention.

[0047] Figure 2 Schematic diagram of the eigenvalues of the covariance matrix in an embodiment of the present invention.

[0048] Figure 3 This is a Q statistic control chart of the hob cutter changing with the cutter shifting cycle in an embodiment of the present invention.

[0049] Figure 4 1 is a schematic diagram comparing the Q statistic of the hob in the 30th cutter shifting cycle and the limit value of the 29th cutter shifting cycle according to an embodiment of the present invention.

[0050] Figure 5 Schematic diagram comparing the Q statistic of the hob in the 88th cutter shifting cycle and the limit value of the 87th cutter shifting cycle according to an embodiment of the present invention.

[0051] Figure 6 14 is a schematic diagram comparing the Q statistic of the hob in the 145th cutter-shifting cycle and the limit value of the 144th cutter-shifting cycle according to an embodiment of the present invention.

[0052] Figure 7 1 is a schematic diagram comparing the Q statistic of the hob in the 204th cutter-shifting cycle and the limit value of the 203rd cutter-shifting cycle according to an embodiment of the present invention.

[0053] Figure 8 3 is a curve showing the change in surface quality detection value of the processed gear corresponding to 231 tool shifting cycles according to the embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention are described in more detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to help understand the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0055] The present invention provides a method for real-time monitoring of the wear state of the hob cutter in the tool shifting mode based on multi-domain features and Q statistics, which includes the following steps. Figure 1 The flowchart shown.

[0056] S1: Vibration Monitoring Data Acquisition. In this embodiment, a vibration sensor is installed along the Z-axis of the gear hobbing machine tool's tool carriage to capture the hob's vibration during the gear hobbing process. The vibration signal acquisition frequency is set to 10 kHz, and the monitoring sample length corresponding to each smooth cutting process of the gear is set to 10,000. Other parameters related to gear hobbing are shown in Table 1.

[0057] Table 1 Gear hobbing related parameters

[0058]

[0059] S2: Extract multi-domain monitoring features to describe the current state of the hob. In this embodiment, 15 time-domain features and 13 frequency-domain features are first extracted for each monitoring sample. The extracted time-domain features and frequency-domain features are shown in Tables 2 and 3, respectively.

[0060] Table 2 Extracted time domain features

[0061]

[0062] x i Represents the i-th value in the monitoring sample signal corresponding to a processed gear, where I is the sample length

[0063] Table 3 Extracted frequency domain features

[0064]

[0065] k=1,2,..., K represents the kth spectral line on the spectrum, s k represents the amplitude corresponding to the kth spectral line, f k Indicates the frequency corresponding to the kth spectral line

[0066] In addition, a 5-layer discrete wavelet decomposition is performed on each sample, using the biorthogonal wavelet bior3.3. Each sample obtains 6 sets of decomposed signals at different scales, and the energy values of the decomposed signals are calculated respectively, thus obtaining energy features at 6 different scales, which are expressed as follows:

[0067] [t 29 t 30 t 31 t 32 t 33 t 34 ]

[0068] Therefore, in this embodiment, when the hob processes each gear, its wear state is described by 34 characteristic values.

[0069] S3: Construct a tool-slip cycle monitoring feature set. For monitoring samples within the same tool-slip cycle, extract multi-domain monitoring features according to S2 to construct a tool-slip cycle monitoring feature set. In this embodiment, the constructed monitoring feature set for the first tool-slip cycle is shown in Table 4.

[0070] Table 4 Monitoring feature sets corresponding to each gear sample processed in the first tool shifting cycle

[0071]

[0072]

[0073] S4: Calculate the Q statistic limit of the i-th tool-slipping cycle. In this embodiment, the first tool-slipping cycle monitoring feature set is taken as an example and normalized. The normalization method is as follows:

[0074]

[0075] in, and σ n The standardized features of each gear processed in the first shifting cycle are shown in Table 5.

[0076] Table 5 Standardized monitoring feature sets corresponding to each gear processed in the first tool shifting cycle

[0077]

[0078]

[0079] The standardized monitoring feature set is further subjected to principal component analysis. First, the corresponding covariance matrix is constructed. In this embodiment, the dimension of the obtained covariance matrix is 34×34. The covariance matrix is subjected to singular value decomposition to obtain the eigenvalues and corresponding eigenvectors of the covariance, such as Figure 2 When the cumulative sum of the first l eigenvalues is not less than the set ratio p of the sum of all eigenvalues, it indicates that the first l features occupy the information of the current state of the hob, and the wear change effect is weak relative to the current state of the hob, which is included in the remaining component. In this embodiment, p = 0.85. Figure 2 It can be seen that the cumulative sum of the first four covariance eigenvalues is greater than 85% of the sum of all covariance matrix eigenvalues. Therefore, l=4. Further, the eigenvectors corresponding to the first four covariance eigenvalues are reconstructed to obtain the reconstructed principal element matrix p.

[0080] Furthermore, substituting l = 4, we can find the statistical limit of Q in the i-th hob cutter shifting cycle to obtain the control variable, which is expressed as follows:

[0081]

[0082] Among them, Q α is the control value limit obtained in the current knife shifting cycle, C α is the confidence limit of the standard normal distribution. In this embodiment, C α The confidence level is 0.99. The expressions of θ1, θ2 and h are as follows:

[0083]

[0084] Therefore, the Q statistic limit Q of the first channeling cycle can be obtained α It is 16.9371.

[0085] S5: Calculate the Q statistic corresponding to the monitoring data in the (i+1)th tool shifting cycle. Since step S4 calculates the Q statistic limit of the first tool shifting cycle, here we take (i+1)=2 as an example. For each monitoring statistical feature vector corresponding to the second tool shifting cycle, use the statistical feature average value calculated in step S4 and standard deviation σ n The monitoring feature set of the second tool shifting cycle is standardized in the same manner as described above.

[0086] Then calculate the Q statistic of each processing gear corresponding to the feature in the (i+1)th cutting cycle. The calculation formula of the Q statistic corresponding to the mth processing gear is as follows

[0087] Q m =||(I-pp T )t' m || 2

[0088] Where I is an identity matrix, p is the reconstructed principal element matrix obtained by the eigenvectors corresponding to the first l eigenvalues of the covariance matrix in S4, and t' m It represents the normalized monitoring feature vector corresponding to the mth gear processed in the tool shifting cycle, Q m is the Q statistic corresponding to the mth gear processed in the tool shifting cycle, which represents t' m The projection on the remaining principal element subspace. The Q statistics corresponding to each gear processed in the second tool shifting cycle are shown in Table 6.

[0089] Table 6 Q statistics of each gear processed in the second tool shifting cycle

[0090]

[0091] S6: Determine whether the wear level has changed. During the (i+1)th shifting cycle, the number of gears processed whose Q statistics exceed the Q statistic limit for the i-th shifting cycle is counted. If this number of gears does not exceed a set percentage (90% in the embodiment) of the number of gears processed during the (i+1)th shifting cycle, the wear level of the hob during the (i+1)th shifting cycle is considered unchanged relative to the i-th shifting cycle. Otherwise, the wear level of the hob during the (i+1)th shifting cycle is considered to have changed. Taking the second shifting cycle as an example, Table 6 shows that during the second shifting cycle, only seven gears corresponding to the Q statistics exceed the Q statistic limit for the first shifting cycle, falling short of the set percentage of 90%. Therefore, the second shifting cycle is considered to be at the same wear level as the first shifting cycle.

[0092] Furthermore, after the cutter wear level of the (i+1)th cutter shifting cycle is identified in step S6, i=i+1 is set, and the process returns to S4 to determine whether the cutter wear level corresponding to the next cutter shifting cycle changes.

[0093] In this embodiment, the hob used in the process processed 4158 gears, with a total of 231 shifting cycles. From the second shifting cycle to the 231st shifting cycle, the wear state of the hob changes as shown in the following figure: Figure 3As shown, the Q statistics for 17 gears in the 30th shifting cycle exceeded the Q statistic limit for the 29th shifting cycle; the Q statistics for 18 gears in the 88th shifting cycle exceeded the Q statistic limit for the 87th shifting cycle; the Q statistics for 17 gears in the 145th shifting cycle exceeded the Q statistic limit for the 144th shifting cycle; and the Q statistics for 18 gears in the 204th shifting cycle exceeded the Q statistic limit for the 203rd shifting cycle. This indicates that the hob underwent transitions in the 30th, 88th, 145th, and 204th shifting cycles. Therefore, the wear state of this hob machining process can be described as shown in Table 7.

[0094] Table 7 Identification of hob wear status

[0095]

[0096] Figure 4-Figure 7 1 is a schematic diagram comparing the Q statistics of the hob in the four tool-shifting cycles in which the hob changes according to an embodiment of the present invention with the Q statistics limit values of the corresponding previous tool-shifting cycle.

[0097] Furthermore, if Figure 8 As shown in the figure, since the hob wear can be directly reflected by the quality of the machined gear, in this embodiment, the measurement values of the tooth surface quality of the machined gear under different tool shifting cycles are collected. In order to avoid the randomness between the measurement values, the surface quality is fitted by the empirical mode decomposition method. It can be seen that the change point of the wear state obtained by the method of the present invention corresponds to the turning point of the fitting curve, which proves the rationality of the wear state division.

[0098] The present invention mines the difference characteristics of the hob wear status contained in the monitoring sensor signal, and identifies whether the hob wear status has changed by judging the difference between the two previous and subsequent cutter shifting cycles, thereby avoiding the need for stopping the machine to take pictures and the risk of camera damage based on the image method; there is no need to establish a labeled hob full life cycle data set in advance, which reduces the difficulty of data collection and model training, and can realize real-time monitoring of the hob wear status on the machine, thereby improving the tool utilization rate and reducing the downtime of the gear hobbing machine for maintenance.

[0099] Finally, it should be noted that the above examples of the present invention are merely illustrative of the present invention and are not intended to limit the embodiments of the present invention. Although the applicant has described the present invention in detail with reference to preferred embodiments, those skilled in the art will appreciate that other variations and modifications can be made based on the above description. It is not possible to enumerate all embodiments here. Any obvious variations or modifications derived from the technical solution of the present invention remain within the scope of protection of the present invention.

Claims

1. A method for monitoring the wear status of a hob cutter in a non-supervisory manner in real time in a machine under a cutter shifting mode, characterized by: The steps include: S1: Vibration monitoring data acquisition: Install a vibration sensor on the gear hobbing machine to obtain the vibration information of the hob during the gear hobbing process, thereby obtaining multiple monitoring samples; S2: Multi-domain monitoring feature extraction: extract multi-domain statistical features for each monitoring sample obtained in S1; S3: Construct a tool shifting cycle monitoring feature set. During a tool shifting cycle, all teeth on the hob perform the same cutting operation, and all positions on the hob are considered to be equally worn. The monitoring samples within the same tool shifting cycle extracted in S2 are combined into a tool shifting cycle monitoring feature set, thereby obtaining tool shifting cycle monitoring feature sets within different tool shifting cycles. The tool shifting cycle monitoring feature set is shown in the following formula: Among them, F i represents the feature set corresponding to the i-th hob shifting cycle, M represents the number of gears processed in a hob shifting cycle, N represents the number of multi-domain statistical features extracted from each monitoring sample, and t m,n Indicates the nth feature corresponding to the mth machined gear in the tool shifting cycle; S4: Calculate the Q statistic limit of the i-th knife-crossing cycle; S5: Calculate the Q statistics corresponding to all the gears processed in the (i+1)th tool shifting cycle; S6: Determine whether the wear level has changed; in the (i+1)th tool shifting cycle, count the number of gears whose Q statistic exceeds the Q statistic limit of the i-th tool shifting cycle. When the number of gears does not exceed the set percentage of the number of gears processed in the (i+1)th tool shifting cycle, it is considered that the wear level of the hob in the (i+1)th tool shifting cycle has not changed relative to the i-th tool shifting cycle; otherwise, it is considered that the wear level of the hob in the (i+1)th tool shifting cycle has changed.

2. The method for monitoring the wear of the hob cutter in the shifting mode in real time without supervision on the machine according to claim 1, characterized in that: The multi-domain statistical features described in S2 include time domain features, frequency domain features, and wavelet domain features.

3. The method for monitoring the wear of the hob cutter in the shifting mode in real time without supervision on the machine according to claim 1, characterized in that: When obtaining the vibration information of the hob during the gear hobbing process in S1, only the vibration signal of the hob during the stable processing stage of each gear tooth blank is intercepted to form a monitoring sample.

4. The method for monitoring the wear of a hob cutter in a shifting cutter mode in real time without supervision on the machine according to claim 1, characterized in that: The steps for calculating the Q statistic limit of the i-th tool-slip cycle in S4 are: S41: Normalize each feature in the i-th tool shifting cycle monitoring feature set. The normalization formula is as follows: Among them, t' m,n Indicates t m,n After normalization, the features and σ n They represent the mean and standard deviation of the n-th dimension feature in the current knife-shifting cycle, and the standardized monitoring feature set F i ' represents the following S42: The standardized monitoring feature set F i 'Perform principal component analysis; First, construct the corresponding covariance matrix, which is expressed as follows; X i =(F i ') T ·F i ' where X i is the covariance matrix corresponding to the standardized monitoring feature set, (F i ') T It's F i 'The transposed matrix; Then calculate the eigenvalues and corresponding eigenvectors of the covariance matrix, and arrange the eigenvalues in descending order; λ1≥λ2≥…≥λ N-1 ≥λ N The number of the first l eigenvalues of the covariance matrix is determined again so that the cumulative sum of the eigenvalues of the first l covariance matrix is not less than the set proportion p of the sum of all eigenvalues; S43: Based on the last (Nl) eigenvalues of the covariance matrix, the Q statistic limit of the i-th hob cutter shifting cycle is obtained, and its expression is as follows: Among them, Q α is the Q statistic limit obtained in the current knife-crossing cycle, C α is the confidence limit of the standard normal distribution, and the expressions of θ1, θ2 and h are as follows:

5. The method for monitoring the wear of the hob cutter in the shifting mode in real time without supervision on the machine according to claim 4, characterized in that: The steps for calculating the Q statistics corresponding to all the gears processed in the (i+1)th tool shifting cycle in S5 are: S51: Obtaining the monitoring feature set of the (i+1)th tool shifting cycle according to step S3; S52: performing normalization processing on the monitoring features corresponding to each gear being processed in the (i+1)th tool shifting cycle according to the method of step S41 to obtain a normalized monitoring feature set for the (i+1)th tool shifting cycle; S53: Based on the standardized monitoring feature set of the (i+1)th tool shifting cycle, the calculation formula of the Q statistic corresponding to the mth machined gear in the (i+1)th tool shifting cycle is as follows Q m =||(I-pp T )t' m || 2 Where I is an identity matrix, p is the reconstructed principal element matrix obtained by the eigenvectors corresponding to the first l eigenvalues of the covariance matrix in S42, and t' m It represents the normalized monitoring feature vector corresponding to the mth gear processed in the tool shifting cycle, Q m is the Q statistic corresponding to the mth gear processed in the tool shifting cycle.

6. The method for monitoring the wear of the hob cutter in the shifting mode in real time without supervision on the machine according to claim 5, characterized in that: The average value and standard deviation of the standardization process in step S52 are directly the average value used in the calculation of the Q statistic limit in the previous tool shifting cycle. and standard deviation σ n .

7. The method for monitoring the wear of a hob cutter in a shifting cutter mode in real time without supervision on the machine according to claim 1, characterized in that: The percentage set in step S6 is 90%.

Images