An online detection system based on a gear hobbing machine

Abstract

The present application relates to the field of metal cutting machine tool manufacturing, and specifically discloses an online detection system based on a gear hobbing machine, comprising: a cutter tooth phase locking multi-modal in-situ sensing module, a cutter tooth-tooth surface implicit defect decoupling and tracing module, and a cutter tooth phase directional process parameter regulation module; a cutter tooth phase reference signal is obtained through key phase marking, micro-vibration spectrum, chip root jagged shear band morphology, and acoustic emission double-band sequence are synchronously collected and latched, a three-channel decoupling model is used to extract cutter tooth microscopic passivation radius, white layer generation tendency, and dynamic stiffness phase lag, and defect confidence distribution and cutter tooth position coordinates are output; according to the foregoing, a feed speed attenuation, a rotational speed reduction, a cutting fluid pressure increment, or a synchronous phase compensation instruction is generated, and is issued in advance to a numerical control system in a cutter tooth phase directional manner. The present application realizes cutter tooth level traceable, directional, and quantifiable gear hobbing machining quality closed-loop control.

Description

Technical Field

[0001] This invention relates to the field of metal cutting machine tool manufacturing technology, and more specifically to an online inspection system based on a gear hobbing machine. Background Technology

[0002] As the core equipment in gear machining, the gear hobbing machine's machining accuracy directly determines the transmission performance and service life of the gear pair. With the widespread application of high-speed dry hobbing technology, multiple error factors such as tool micro-wear, workpiece thermal deformation, and machine tool dynamic stiffness attenuation are coupled and transmitted in the process chain, placing higher demands on tooth profile accuracy, tooth direction accuracy, and tooth surface quality consistency. Existing online inspection technologies for gear hobbing machines mainly rely on offline contact measurement or overall condition monitoring. The former requires stopping the machine and disassembling the workpiece, resulting in low inspection efficiency and an inability to capture transient cutting process conditions; the latter judges overall tool wear or machine tool abnormalities through macroscopic signals such as vibration, temperature, or current, but cannot establish a precise mapping relationship between multi-source errors and specific defects in the tooth and tooth surface creation area, and it is even more difficult to achieve progressive closed-loop control from condition perception to error tracing and parameter adjustment during the machining process.

[0003] In particular, existing technologies lack effective online monitoring methods for deep-seated process mechanisms during gear hobbing, such as microscopic blunting of the cutting edge, morphological distortion of the serrated shear band at the chip root, friction-induced phase transformation white layer on the machined surface, and phase lag of the machine tool process chain frequency-varying stiffness at the tooth passage frequency. On the one hand, chip morphology analysis is mostly used for chip breaking control or tool macroscopic wear assessment, without establishing a mapping between the microscopic morphology of the chip root and the blunting state of the specific cutting edge of the tooth. On the other hand, acoustic emission monitoring in gear hobbing is only used to assess the overall wear degree of the tool, without separating and identifying the dual-frequency band characteristics of the tooth back face friction and workpiece surface phase transformation, making it impossible to identify the tendency of white layer formation on the tooth surface online. Furthermore, existing vibration monitoring does not specifically extract phase information at the tooth passage frequency, making it difficult to decouple the cumulative pitch error and tooth surface waviness caused by the decrease in machine tool dynamic stiffness. More critically, existing process parameter adjustments are all global adjustments, unable to provide targeted compensation for specific cutting teeth or tooth surface creation areas based on defect tracing conclusions, resulting in insufficient online suppression capability for hidden quality defects.

[0004] Therefore, there is an urgent need for an online inspection system for gear hobbing machines that can achieve tooth phase locking, in-situ sensing of multi-modal latent states, tooth-by-tooth decoupling and tracing of multi-source errors, and coordinated control of tooth phase orientation process parameters. This system would overcome the limitations of existing technologies that can only perform overall monitoring or offline sampling inspections, and achieve tooth-by-tooth closed-loop control from the micro-morphology of the chip root to the blunting of the tooth cutting edge, from the acoustic emission phase change frequency band to the generation of white layer on the tooth surface, and from the phase lag of the tooth passing frequency to the attenuation of machine tool stiffness. Summary of the Invention

[0005] The technical problem to be solved by this invention is that existing online detection of gear hobbing machines is difficult to achieve tooth-by-tooth traceability, and cannot identify hidden defects such as micro-passivation of the cutting edge, formation of white layer on the tooth surface, and phase lag of machine tool frequency-varying stiffness.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: An online inspection system based on a gear hobbing machine, specifically including the following steps:

[0007] S1. The tooth phase-locking multimodal in-situ sensing module is used to acquire the tooth phase reference signal through the key phase mark during the hobbing process, and simultaneously collect the micro-chatter spectrum, the morphology of the sawtooth shear band at the chip root, the fluctuation of the shear band spacing and thickness pulsation, the high-frequency friction energy and the acoustic emission sequence of the workpiece phase transition characteristics, perform tooth phase-locking alignment, and generate a multimodal latent state feature vector.

[0008] S2. Tooth-tooth surface latent defect decoupling and tracing module, used to input the multimodal latent state feature vector into the pre-trained tooth-tooth surface latent defect decoupling model, extract the tooth micro-passivation radius and micro-chipping depth, tooth surface thermal accumulation and white layer generation tendency, dynamic stiffness phase lag and meshing impact non-uniformity, and output the tooth surface latent defect type confidence distribution and the tooth position coordinates where the defect occurs;

[0009] S3. The tooth phase orientation process parameter control module is used to call the control strategy library according to the confidence distribution of the hidden defect type on the tooth surface and the coordinates of the tooth where the defect occurs, generate the corresponding process parameter correction instruction according to the defect type and confidence level, and send it to the gear hobbing machine CNC system in the tooth phase orientation mode, so that the gear hobbing machine can complete the dynamic adjustment of the corresponding process parameters before the specified tooth reaches the defect location.

[0010] Based on the above technical solution, the present invention can be further improved as follows.

[0011] Furthermore, the knife-tooth phase-locked multimodal in-situ sensing module specifically includes:

[0012] S11. A triaxial piezoelectric micro-chatter sensor array is attached to the circumferential region near the end of the hob holder to collect the micro-chatter spectrum of the meshing area between the hob teeth and the workpiece blank at the tooth passing frequency.

[0013] S12. High-resolution linear array vision sensing unit, equipped with a macro lens, high-resolution linear array vision sensor sets scan lines along the direction perpendicular to the chip movement, used to start line scanning acquisition with the tooth phase reference signal as an external trigger signal to obtain a continuous grayscale image sequence of the chip root.

[0014] S13. Wideband acoustic emission sensing unit, which is attached to the hob cutter bar via an acoustic coupling medium, is used to collect the high-frequency friction energy sequence generated by the friction between the hob tooth rake face and the machined surface of the workpiece, as well as the characteristic acoustic emission sequence generated by the phase change of the workpiece material surface induced by the high temperature and high pressure zone of the hob machining.

[0015] S14. The key phase marker and photoelectric speed sensor are installed at the end of the hob spindle. The output end is connected to the synchronous trigger end of the triaxial piezoelectric micro-chatter sensor array, the high-resolution linear array visual sensor unit and the broadband acoustic emission sensor unit, respectively, to output the tooth phase reference signal as the synchronous latching trigger source of the multi-source sensor signal.

[0016] Furthermore, the specific workflow of the high-resolution linear array visual sensing unit includes:

[0017] S121. A directional backlight illuminates the chip root area in the chip discharge path at a preset angle, so that the edge of the serrated shear band at the chip root forms a high-contrast light and dark boundary line.

[0018] S122. A high-resolution linear array vision sensor equipped with a macro lens sets up scanning lines perpendicular to the chip movement direction, and uses the tooth phase reference signal as an external trigger signal to start line scanning acquisition to obtain a continuous grayscale image sequence of the chip root.

[0019] S123. Perform real-time edge detection and morphological segmentation on the acquired image sequence to extract the serrated shear band morphology sequence, the spacing fluctuation sequence between adjacent shear bands, and the pulsation sequence of the chip root thickness along the movement direction.

[0020] S124. Align and encapsulate the above three sequence data according to the acquisition timestamp and the blade phase reference signal, and output the micro-dynamic feature data of the chip root for use by subsequent modules.

[0021] Furthermore, the specific workflow of the broadband acoustic emission sensing unit includes:

[0022] S131. A broadband acoustic emission sensor is tightly attached to the surface of the hob bar through an acoustic coupling medium, so that the high-frequency elastic wave generated by the friction between the hob tooth rake face and the machined surface of the workpiece, as well as the transient elastic wave released by the phase change of the workpiece surface, are effectively coupled to the sensor sensitive element.

[0023] S132. The wideband acoustic emission sensor continuously acquires the coupled acoustic emission raw signal at a sampling frequency of not less than 1MHz. After pre-amplification and bandpass filtering, two characteristic frequency bands are separated. The high-frequency friction energy corresponding frequency band is used to characterize the sliding friction intensity between the back face of the cutting tooth and the tooth surface, and the low-frequency phase transition characteristic frequency band is used to capture the characteristic acoustic emission events generated by the phase transition of the workpiece surface.

[0024] S133. Perform envelope detection and energy integration on the two filtered signals respectively to generate a high-frequency friction energy sequence and a characteristic acoustic emission sequence;

[0025] S134. Latch and align the two sequences according to the timestamp of the tooth phase reference signal, and output the tooth surface friction and phase transition state characteristic data for use by subsequent modules.

[0026] Furthermore, the specific working process of the bond phase marker and photoelectric speed sensor includes:

[0027] S141. The key phase mark installed at the end of the hob spindle rotates synchronously with the spindle. Whenever a specified cutter tooth passes through the detection optical path of the photoelectric speed sensor, the key phase mark triggers the sensor to output a corresponding cutter tooth synchronization pulse signal. The cutter tooth synchronization pulse signal serves as the cutter tooth phase reference signal.

[0028] S142. Using the rising edge of the blade synchronization pulse signal as the synchronization latch trigger source, hardware-level synchronization latch acquisition is performed on the micro-chatter spectrum output by the triaxial piezoelectric micro-chatter sensor array, the chip root image sequence output by the high-resolution linear array vision sensor unit, and the characteristic acoustic emission sequence output by the broadband acoustic emission sensor unit.

[0029] S143. Align and encapsulate the latched multi-channel sensor data according to the tooth passing cycle on the time axis to generate a multimodal latent state feature vector with a single tooth as the index unit.

[0030] Furthermore, the tooth-tooth surface latent defect decoupling and tracing module specifically includes:

[0031] S21. Cutting edge passivation evolution unit, with its input end connected to the sawtooth shear band morphology sequence and shear band spacing fluctuation sequence in the multimodal latent state feature vector, used to extract the micro passivation radius features and micro chipping depth features of each cutting edge tooth.

[0032] S22. Tooth surface thermo-mechanical-phase change coupling unit, the input end of which is connected to the high-frequency friction energy sequence and the characteristic acoustic emission sequence in the multimodal latent state feature vector, used to extract the instantaneous thermal accumulation features of the tooth surface creation zone and the tooth surface white layer generation tendency features tooth by tooth;

[0033] S23. Machine tool frequency-varying stiffness phase lag unit, the input end is connected to the micro-chatter spectrum in the multimodal latent state feature vector, used to extract the dynamic stiffness phase lag characteristics and tooth meshing impact non-uniformity characteristics of the hob cutter bar-spindle system in the tooth passing frequency band;

[0034] S24. Defect generation determination network unit, which is connected to the output of the above three units respectively, is used to input the extracted six features into the defect generation determination network of the tooth-tooth surface latent defect decoupling model, and output the confidence distribution of tooth surface latent defect type and the corresponding tooth position coordinates where the defect occurs.

[0035] Furthermore, the specific workflow of the cutting edge passivation evolution unit includes:

[0036] S211. Receive the serrated shear band morphology sequence and shear band spacing fluctuation sequence after the cutting tooth phase lock, perform contour fitting on the shear band morphology image corresponding to each cutting tooth, and extract the curvature radius of the shear band root as the quantitative characterization value of the micro-passivation radius of the cutting edge under constant cutting parameter constraints.

[0037] S212. Perform Fourier transform and envelope demodulation on the shear band spacing fluctuation sequence to separate the high-frequency periodic fluctuation component caused by the micro-chipping of the cutting edge, and calculate the micro-chipping depth characteristics of the cutting teeth tooth by tooth based on the fluctuation amplitude.

[0038] S213. Perform a difference operation between the extracted micro-passivation radius and micro-chipping depth features and the pre-stored initial state parameters of the cutting teeth to obtain the wear increment value of the current cutting teeth relative to the new cutting state.

[0039] S214. Output the wear increment value of each cutting tooth according to the cutting tooth number, so that the defect generation determination network unit can evaluate the confidence of passivation defects.

[0040] Furthermore, the specific workflow of the thermal-mechanical-phase change coupling unit in the tooth surface generation zone includes:

[0041] S221. Receive the high-frequency friction energy sequence and characteristic acoustic emission sequence after the phase lock of the cutting teeth, perform sliding window integration processing on the high-frequency friction energy sequence, and extract the cumulative friction energy value of each cutting tooth within the cycle as the instantaneous thermal accumulation feature of the tooth surface creation area.

[0042] S222. Perform wavelet packet decomposition on the characteristic acoustic emission sequence, extract the characteristic frequency band energy ratio related to the phase transition of the workpiece surface, and map the energy ratio as the tooth surface white layer generation tendency feature based on the pre-stored phase transition calibration curve.

[0043] S223. The instantaneous thermal accumulation characteristics and white layer formation tendency characteristics are time-aligned and normalized, and the cross-correlation delay coefficient between the two is calculated to determine the strength of the causal relationship between thermal accumulation and phase transition.

[0044] S224. The above two features and causal relationship coefficients are encoded and output according to the tooth sequence number, so as to provide the defect generation determination network unit with confidence assessment of white layer defects.

[0045] Furthermore, the specific workflow of the defect generation determination network unit includes:

[0046] S241. Receive six feature data output from the cutting edge passivation evolution unit, the tooth surface thermo-mechanical-phase change coupling unit, and the machine tool frequency-variable stiffness phase lag unit, and fuse the above six features at the feature level according to the cutting tooth number to form a fixed-dimensional multi-dimensional feature vector.

[0047] S242. Input the multidimensional feature vector into the defect generation determination network, and output the confidence distribution of each defect type through the forward propagation calculation of the network.

[0048] S243. Combining the tooth phase information in the micro-flutter spectrum, the confidence distribution is mapped to the specific tooth position coordinates, and the confidence distribution of the hidden defect type on the tooth surface and the corresponding tooth position coordinates where the defect occurs are output.

[0049] Furthermore, the tool tooth phase orientation process parameter control module specifically includes:

[0050] S31. Control strategy library, used to pre-store the process parameter correction mapping relationship corresponding to different defect types and confidence levels;

[0051] S32. Defect type parsing unit, the input end is connected to the confidence distribution of the hidden defect type of the tooth surface and the coordinates of the position of the defect-occurring cutting tooth, used to parse the main category of the current defect and its confidence level, and output the defect type code and the corresponding cutting tooth number;

[0052] S33. Instruction generation unit, with its input end connected to the defect type parsing unit and the control strategy library, is used to call the corresponding correction instruction template in the strategy library according to the defect type code: when the defect type is micro-passivation of the cutting edge, it generates the hob axial feed speed attenuation coefficient and the hob spindle speed derating coefficient; when the defect type is heat accumulation-induced white layer formation, it generates the cutting fluid injection pressure increment; when the defect type is dynamic stiffness phase mismatch, it generates the hob spindle-workpiece rotation axis synchronous phase compensation coefficient.

[0053] S34. Command issuance interface unit, connected to the command generation unit and the gear hobbing machine CNC system, is used to issue the generated correction command to the gear hobbing machine CNC system in a tooth phase orientation manner, so that the gear hobbing machine can complete the dynamic adjustment of the corresponding process parameters before the specified tooth reaches the defect location.

[0054] The beneficial effects of this invention are:

[0055] 1. By using a tooth phase-locking mechanism, the micro-chatter spectrum, the micro-morphology of the shear band at the chip root, and the dual-frequency signal of acoustic emission are anchored to specific teeth. This enables tooth-by-tooth in-situ sensing of the micro-passivation of the cutting edge from the sawtooth shear band at the chip root and the tendency of white layer formation on the tooth surface to be identified from the low-frequency phase transition band of acoustic emission. This breaks through the technical limitations of traditional gear hobbing machines, which can only perform overall condition monitoring or offline sampling inspection.

[0056] 2. By decoupling the three channels in parallel—the micro-passivation channel of the cutting edge, the thermo-mechanical-phase transformation coupling channel of the tooth surface, and the phase lag channel of the machine tool frequency-varying stiffness—the micro-chipping of the cutting edge, the tendency of white layer formation on the tooth surface, and the attenuation of the dynamic stiffness of the machine tool can be accurately traced to the specific tooth position. This overcomes the technical bottleneck of traditional multi-source error coupling and transmission, which makes it difficult to separate and locate the root cause of defects in the cutting tooth.

[0057] 3. Based on the confidence distribution of defect types and the coordinates of the cutter teeth where defects occur, feed rate attenuation, speed derating, cutting fluid pressure increment, or synchronous phase compensation commands are generated and issued in advance using a cutter tooth phase orientation method. This enables the hobbing machine to complete dynamic adjustments before the specified cutter teeth reach the defect position, achieving closed-loop control of hobbing machining quality that is traceable, directional, and quantifiable at the cutter tooth level. Attached Figure Description

[0058] Figure 1 This is a framework diagram of the system described in Embodiments 1-4 of the present invention;

[0059] Figure 2 This is a flowchart of the system described in this invention;

[0060] Figure 3 This is a schematic diagram of the electronic device in this invention. Detailed Implementation

[0061] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0062] Example 1

[0063] like Figures 1-2 As shown, the present invention provides an online inspection system based on a gear hobbing machine, specifically including the following steps:

[0064] S1. The tooth phase-locking multimodal in-situ sensing module is used to acquire the tooth phase reference signal through the key phase mark during the hobbing process, and simultaneously collect the micro-chatter spectrum, the morphology of the sawtooth shear band at the chip root, the fluctuation of the shear band spacing and thickness pulsation, the high-frequency friction energy and the acoustic emission sequence of the workpiece phase transition characteristics, perform tooth phase-locking alignment, and generate a multimodal latent state feature vector.

[0065] S2. The tooth-tooth surface latent defect decoupling and tracing module is used to input the multimodal latent state feature vector into the pre-trained tooth-tooth surface latent defect decoupling model, extract the tooth micro-passivation radius and micro-chipping depth, tooth surface thermal accumulation and white layer generation tendency, dynamic stiffness phase lag and meshing impact non-uniformity, and output the confidence distribution of tooth surface latent defect type and the coordinates of the tooth where the defect occurs.

[0066] S3. The tooth phase orientation process parameter control module is used to call the control strategy library based on the confidence distribution of the hidden defect type on the tooth surface and the coordinates of the tooth where the defect occurs. It generates corresponding process parameter correction instructions according to the defect type and confidence level, and sends them to the gear hobbing machine CNC system in the tooth phase orientation mode, so that the gear hobbing machine can complete the dynamic adjustment of the corresponding process parameters before the specified tooth reaches the defect location.

[0067] Example 2

[0068] Preferably, in Embodiment 1, the knife-tooth phase-locked multimodal in-situ sensing module is implemented through the following steps:

[0069] S11. Triaxial piezoelectric micro-flutter sensor array, specifically:

[0070] Three evenly distributed measuring points are selected circumferentially near the near end of the hobbing cutter shank. A piezoelectric accelerometer is placed at each measuring point. The three sensors correspond to the radial, axial, and tangential vibration response acquisition directions of the cutter shank, respectively, and together they form a three-dimensional piezoelectric micro-chatter sensor array. Each sensor is attached to the surface of the cutter shank by magnetic base or adhesive method. The sensor sensing element is in direct contact with the metal surface of the cutter shank to shorten the vibration transmission path and improve signal fidelity.

[0071] The signal output terminals of each sensor are connected to the signal conditioning circuit via shielded cables. The signal conditioning circuit performs charge-to-voltage conversion and low-noise amplification on the original charge signal, and the output terminal is connected to a high-speed data acquisition card. The high-speed data acquisition card synchronously samples the three vibration signals at a sampling frequency no less than twice the passing frequency of the blade teeth.

[0072] The acquired time-domain vibration signal is subjected to a fast Fourier transform to convert the time-domain signal into a frequency-domain signal. The spectral components at the tooth passage frequency and its harmonics are extracted from the spectrum. The spectral amplitude, phase angle and harmonic energy distribution at the tooth passage frequency are recorded. The above frequency-domain information is integrated into a micro-flutter spectrum. This micro-flutter spectrum is used as one of the input data for subsequent tooth phase latch alignment.

[0073] S12. High-resolution linear array visual sensing unit, specifically:

[0074] S121. High-contrast illumination arrangement at the root of the chip using a directional backlight, specifically:

[0075] A directional backlight is fixedly installed on the side of the chip discharge path, with the light-emitting surface of the backlight facing the chip root area. A preset angle is formed between the optical axis of the backlight and the normal of the chip root surface. This preset angle is determined based on the surface roughness of the chip root and the micro-undulation height of the serrated shear band, so that the incident light is directionally reflected at the raised edge of the shear band and forms an optical shadow in the recessed area. The value ranges from 15° to 45°, with a classic value of 30°.

[0076] The backlight uses a high-brightness LED array or fiber optic cold light source, and the emission spectrum avoids the interference band of the oxide color on the chip surface, ensuring that the edge of the serrated shear band at the root of the chip forms a high-contrast light-dark boundary on the imaging surface.

[0077] S122. High-resolution linear array vision sensor with serrated phase-synchronized line scanning acquisition, specifically:

[0078] A high-resolution linear array vision sensor equipped with a macro lens is fixed to the side of the chip discharge path, with the front end of the macro lens aligned with the chip root region. The scanning line direction of the linear array vision sensor is perpendicular to the chip movement direction. The magnification of the macro lens is determined based on the microscopic scale of the serrated shear band, ensuring sufficient spatial resolution of the shear band edge on the sensor pixels. The magnification ranges from 2x to 10x, with a classic value of 5x.

[0079] The trigger input of the linear array vision sensor is connected to the output of the key phase marker and the photoelectric speed sensor, using the cutting tooth phase reference signal as the external trigger signal. Whenever the cutting tooth synchronization pulse signal arrives, the linear array vision sensor starts a line scan acquisition, acquiring the grayscale information of the chip root line by line. The results of multiple scans are stitched together in chronological order to form a continuous grayscale image sequence of the chip root.

[0080] S123. Real-time edge detection and morphological segmentation processing of chip root image sequences, specifically:

[0081] Real-time edge detection is performed frame-by-frame on a continuous grayscale image sequence. An edge detection algorithm based on grayscale gradient changes is used to calculate the grayscale gradient magnitude and direction angle of each pixel in each frame in the direction of chip movement and the vertical direction. A grayscale gradient threshold is adaptively set based on the average grayscale difference between the chip root region and the background region, with a value ranging from 15 to 80 and a classic value of 35. Pixels with gradient magnitudes exceeding this threshold are marked as candidate edge points. Non-maximum suppression processing is applied to the candidate edge points. The gradient magnitudes of adjacent pixels are compared in the gradient direction, and only local gradient maxima are retained, while false edge points caused by surface oxide spots or reflective noise are removed. The retained edge points are tracked and connected point by point according to the eight-neighbor spatial connectivity to form the initial contour line of the zigzag shear band.

[0082] The initial contour line undergoes morphological segmentation. A combination of dilation and erosion operations is used for morphological filtering to fill in minor breaks within the contour line and smooth edge burrs. The dilation and erosion operations use structuring elements of the same size to ensure the overall contour position remains stable. Opening operations remove isolated noise pixels outside the contour, while closing operations connect adjacent but broken edge segments within the contour. The region enclosed by the morphologically processed closed contour is segmented from the background at the root of the cut into an independent foreground region. This foreground region represents the spatial occupancy of the jagged shear band in the image, while the background region is set to a uniform grayscale value to eliminate interference.

[0083] A zigzag shear band morphology sequence is extracted from an independent foreground region. The coordinates of each pixel on the shear band contour line are measured point by point along the chip movement direction, and arranged in spatial order to form a zigzag shear band morphology sequence. The root center point of adjacent shear bands is determined by calculating the geometric center or the center of the minimum bounding rectangle of each foreground region. The Euclidean distance between two adjacent root center points is calculated, and the distance value is recorded point by point along the chip movement direction, forming a spacing fluctuation sequence between adjacent shear bands. In the transition region between adjacent shear bands, the thickness of the chip root is measured perpendicular to the chip movement direction, and the thickness change is recorded point by point along the chip movement direction, forming a pulsating sequence of chip root thickness along the movement direction.

[0084] S124. Timestamp alignment and encapsulation output of microscopic dynamic feature data at the chip root, specifically:

[0085] The morphological sequence of the serrated shear band, the fluctuation sequence of the spacing between adjacent shear bands, and the pulsation sequence of the chip root thickness along the direction of motion are marked with their respective acquisition timestamps. Using the time axis of the blade phase reference signal as a reference, the data points acquired by the same blade within the cycle of the three sequences are aggregated into the same time window.

[0086] The data within the time window is validated to remove abnormal data points caused by chip breakage or occlusion. The validated data is then encapsulated into a data structure indexed by the cutter tooth number, so that each cutter tooth number corresponds to a set of microscopic dynamic feature data of the chip root. The encapsulated data structure is output to subsequent modules as a component of the multimodal latent state feature vector.

[0087] S13. Wideband acoustic emission sensing unit, specifically:

[0088] S131. Selection and bonding process of acoustic coupling medium, specifically:

[0089] An alumina ceramic gasket or a special acoustic conductive grease is selected as the acoustic coupling medium. The acoustic impedance of this medium is between that of the metal material of the hob cutter bar and the piezoelectric crystal of the broadband acoustic emission sensor, ensuring that the transmission loss of elastic waves at the heterogeneous interface is minimized. The end face of the sensitive element of the broadband acoustic emission sensor is tightly attached to the near-end area of ​​the hob cutter bar surface through the acoustic coupling medium. Before attachment, the surface of the cutter bar is cleaned and flattened, and the surface oxide layer and oil residue are removed using a lint-free cloth and anhydrous ethanol.

[0090] During bonding, a preset preload is uniformly applied circumferentially along the sensor end face, forming a uniform and continuous acoustic conduction layer between the sensor and the tool holder using the acoustic coupling medium; the value ranges from 3N to 15N, with a classic value of 8N. This acoustic conduction layer effectively transmits the high-frequency elastic waves generated by the friction between the hob tooth's flank face and the machined surface of the workpiece, as well as the transient elastic waves released during the phase transition of the workpiece surface, to the sensor's sensitive element. This avoids sound wave reflection and energy attenuation caused by air gaps, ensuring the amplitude fidelity and phase consistency of the original acoustic emission signal.

[0091] S132. Continuous acquisition and dual-band separation of broadband acoustic emission signals, specifically:

[0092] The wideband acoustic emission sensor continuously samples the coupled acoustic emission raw signal at a sampling frequency of not less than 1MHz. The sampled raw signal is first amplified by a preamplifier circuit for low noise. The amplification factor is set according to the sensor output sensitivity and the input range of the subsequent analog-to-digital converter to improve the amplitude level of the weak elastic wave signal to the effective quantization range. The gain value corresponding to the amplification factor ranges from 20dB to 60dB, with a classic value of 40dB.

[0093] The amplified signal enters the bandpass filtering stage, where two parallel bandpass filters separate two characteristic frequency bands: the high-frequency band covers the main frequency range of elastic waves generated by the friction between the back face of the cutting tooth and the machined surface of the workpiece, and is defined as the frequency band corresponding to high-frequency friction energy, used to characterize the sliding friction intensity; the low-frequency band covers the frequency band of transient elastic waves released by the phase transition of the workpiece surface, and is defined as the low-frequency phase transition characteristic frequency band, used to capture phase transition events.

[0094] S133. Envelope detection and energy integration processing of dual-channel signals, specifically:

[0095] The filtered high-frequency friction energy signal and the low-frequency phase transition characteristic signal are processed independently and in parallel, with the processing flow for both signals being completely identical. Envelope detection is performed on each signal, and the Hilbert transform method is used to obtain orthogonal components from the original signal. The original signal and the orthogonal components are then combined to form an analytic signal, and the magnitude of the analytic signal is calculated as the instantaneous envelope curve.

[0096] The extracted envelope curve is processed by energy integration. The cycle of a single cutter tooth is used as the integration time window. The square of the envelope amplitude within the window is accumulated over time to obtain the high-frequency friction energy integration value and the characteristic acoustic emission energy integration value within the cycle of the cutter tooth.

[0097] The high-frequency friction energy integral values ​​of multiple consecutive cutting teeth passing through the cycle are arranged in chronological order to form a high-frequency friction energy sequence; the characteristic acoustic emission energy integral values ​​of multiple consecutive cutting teeth passing through the cycle are arranged in chronological order to form a characteristic acoustic emission sequence; each data point in the sequence corresponds to the energy accumulation level of a cutting tooth passing through the cycle, and the index of the data point corresponds one-to-one with the cutting tooth number.

[0098] S134. Knife-tooth phase-locked alignment and encapsulated output of acoustic emission feature sequences, specifically:

[0099] The generated high-frequency friction energy sequence and characteristic acoustic emission sequence are marked with their respective acquisition timestamps, and the timestamp accuracy is aligned with the pulse time of the blade phase reference signal. Using the time axis of the blade phase reference signal as a reference, the high-frequency friction energy integral value and characteristic acoustic emission energy integral value corresponding to the same blade passing period are collected into the same time window.

[0100] The data within the time window is validated to remove abnormal spikes caused by electromagnetic interference or mechanical collisions. Outliers are identified and replaced using sliding median filtering or threshold methods. The validated two-channel sequence data are indexed and encapsulated according to the tooth number, so that each tooth number corresponds to a set of tooth surface friction and phase transition state feature data. The encapsulated data structure is output to subsequent modules as a component of the multimodal latent state feature vector.

[0101] S14. Key phase marking and photoelectric speed sensor, specifically:

[0102] S141. Installation of the key phase marker and generation of the tool tooth synchronization pulse signal, specifically:

[0103] Key phase marks are fixedly installed at the end of the hob spindle. These key phase marks are made of raised metal or highly reflective material, and their number is consistent with the total number of hob teeth. Each key phase mark establishes a one-to-one spatial relationship with a designated tooth. A photoelectric speed sensor is fixed near the end of the hob spindle, and the sensor's detection optical path is aligned with the rotation trajectory of the key phase mark.

[0104] When the hob spindle drives the key phase mark to rotate synchronously, whenever a key phase mark enters the detection optical path of the photoelectric speed sensor, the light receiving end of the sensor detects a sudden change in light intensity. The internal photoelectric conversion circuit converts this change in light intensity into a level jump and outputs a tooth synchronization pulse signal. The pulse timing of this tooth synchronization pulse signal precisely corresponds to the moment when the specified tooth reaches the phase zero point of the meshing area, and its pulse sequence is the tooth phase reference signal.

[0105] S142. The rising edge of the blade synchronization pulse triggers multi-source hardware-level synchronous latch acquisition. Specifically:

[0106] The blade synchronization pulse signal is distributed via shielded cable to the synchronization trigger terminals of the triaxial piezoelectric micro-tear sensor array, the high-resolution linear array vision sensor unit, and the broadband acoustic emission sensor unit. The synchronization trigger terminals of each sensor unit are configured for rising edge-activated triggering. When the rising edge of the blade synchronization pulse signal is detected, the high-speed data acquisition card of the triaxial piezoelectric micro-tear sensor array immediately latches the micro-tear spectrum data within the current sampling period. The high-resolution linear array vision sensor unit initiates a line scan acquisition and latches the current chip root image sequence. The front-end acquisition circuit of the broadband acoustic emission sensor unit latches the characteristic acoustic emission sequence within the current integration period.

[0107] S143. Time axis alignment and multimodal feature vector encapsulation of multi-channel sensor data, specifically:

[0108] After hardware-level synchronous latching, the multi-channel sensor data are aligned on the time axis according to the cutting tooth passing cycle. Using the synchronization pulse signal of a single cutting tooth as the time reference, the micro-flutter spectrum, chip root image sequence and characteristic acoustic emission sequence latched within the same cutting tooth passing cycle are collected into the same time window.

[0109] The validity of the data within each time window is verified, and abnormal data frames caused by chip breakage or electromagnetic interference are removed. The verified three-way data are indexed and encapsulated according to the tooth number, and a multimodal latent state feature vector is established with a single tooth as the index unit. Each tooth number corresponds to a set of structured features including micro-chatter spectrum, micro-dynamic feature data of chip root, and feature data of tooth surface friction and phase transition state. The results are then output to the subsequent tooth-tooth surface latent defect decoupling and tracing module.

[0110] Example 3

[0111] Preferably, in Example 1, the tooth-tooth surface latent defect decoupling and tracing module is implemented through the following steps:

[0112] S21. Cutting edge passivation evolution unit, specifically:

[0113] S211. Contour fitting and root radius of curvature extraction of the shear band topography image, specifically:

[0114] After receiving the serrated shear band morphology sequence after the blade phase lock, for the shear band morphology image corresponding to a single blade, the contour pixel set of the shear band root is extracted point by point along the chip movement direction; the least squares method or cubic spline curve fitting method is used to approximate the pixel set to generate a smooth contour curve of the shear band root.

[0115] Under constant cutting parameter constraints—that is, maintaining constant hob axial feed rate, spindle speed, and radial feed depth—the geometry of the profile curve is primarily governed by the micro-passivation state of the cutting edge. At the root inflection point of the profile curve, continuous pixels within a local neighborhood are selected, and the radius of the fitted circle within that neighborhood is calculated. This fitted circle radius is used as the root curvature radius of the shear band. This curvature radius exhibits a monotonically corresponding relationship with the micro-passivation radius of the cutting edge, and is used as the quantitative characterization value of the micro-passivation radius of the cutting edge.

[0116] S212. Frequency domain decomposition of shear band spacing fluctuation sequence and estimation of micro-breakage depth, specifically:

[0117] After receiving the shear band spacing fluctuation sequence after the blade phase is locked, a fast Fourier transform is performed on the sequence to convert the time-domain distance fluctuation into a frequency-domain spectrum, and the main frequency component and harmonic components in the spectrum are identified. The micro-chipping of the cutting edge will introduce a high-frequency periodic fluctuation component related to the blade rotation period into the shear band spacing. The frequency of this component corresponds to an integer multiple of the blade passing frequency.

[0118] Envelope demodulation was performed on the high-frequency periodic fluctuation component to extract the amplitude information of its modulation envelope. Based on the pre-stored micro-chipping depth calibration relationship, i.e., the mapping table between envelope amplitude and micro-chipping depth established through offline experiments, the extracted envelope amplitude was converted into micro-chipping depth features of the blade teeth. The mapping table is shown in Table 1 below.

[0119] Table 1: Mapping Table between Envelope Amplitude and Micro-chipping Depth

[0120] 0.05–0.15 0–5 Normal wear, cutting edge intact 0.15–0.30 5–12 Minor chipping, localized damage to the cutting edge 0.30–0.50 12–22 Slight chipping, with the area of ​​minor chipping on the edge expanding. 0.50–0.70 22–35 Moderate chipping, with a significant increase in the effective cutting radius of the cutting edge. 0.70–0.85 35–45 Severe chipping occurs when the blade edge shows continuous chipping bands. 0.85–0.95 45–50 Severe chipping, critical point of functional failure of the cutting edge.

[0121] S213. Calculation of differential wear increment based on micro-passivation and micro-chipping characteristics, specifically:

[0122] The micro-passivation radius extracted by S211 and the micro-chipping depth feature extracted by S212 are respectively differentially calculated with the pre-stored initial state parameters of the cutting teeth. The pre-stored initial state parameters of the cutting teeth include the initial passivation radius reference value and the initial micro-chipping depth reference value of each cutting tooth in the new cutting state. These parameters are obtained by joint calibration through the first cutting experiment after the new cutting tool is clamped and microscopic detection. The initial passivation radius reference value ranges from 2μm to 15μm, with a classic value of 5μm, and the initial micro-chipping depth reference value ranges from 0μm to 3μm, with a classic value of 0μm.

[0123] A differential operation is performed on the micro-passivation radius, i.e., the micro-passivation radius of the current cutting tooth is subtracted from the initial passivation radius reference value of the corresponding cutting tooth to obtain the passivation radius increment value. A differential operation is also performed on the micro-chipping depth feature, i.e., the micro-chipping depth of the current cutting tooth is subtracted from the initial micro-chipping depth reference value of the corresponding cutting tooth to obtain the micro-chipping depth increment value. The passivation radius increment value and the micro-chipping depth increment value are weighted and fused to obtain the wear increment value of the current cutting tooth relative to the new cutting tooth state. The weight coefficient of the passivation radius increment value ranges from 0.3 to 0.7, with a classic value of 0.4. The weight coefficient of the micro-chipping depth increment value ranges from 0.3 to 0.7, with a classic value of 0.6. The sum of the two weight coefficients is always 1.0.

[0124] S214. Encode and arrange the wear increment values ​​of each cutting tooth according to the cutting tooth number to establish a one-dimensional feature sequence indexed by the cutting tooth number. Each data element in this sequence corresponds to the wear increment value of one cutting tooth, and the arrangement order of the data elements is consistent with the circumferential number of the hob cutting teeth. Output the encoded wear increment value sequence to the defect generation determination network unit as the tooth-by-tooth input feature of the passivation type defect.

[0125] S22. Tooth surface thermo-mechanical-phase change coupled unit, specifically:

[0126] S221. Sliding window integration and instantaneous thermal accumulation feature extraction of high-frequency friction energy sequences, specifically:

[0127] The high-frequency friction energy sequence after receiving the phase-locked blade is set, and the width of the sliding window is set to be equal to the cycle of a single blade passage, and the window step size is equal to the cycle of a single blade passage, so that the sliding window is strictly synchronized with the cycle of the blade passage.

[0128] The data points of the high-frequency friction energy sequence within each window are arithmetically summed by sliding along the time axis. The cumulative sum of all data points within the window reflects the total frictional heat input to the tooth surface creation zone during the complete cutting process. The cumulative sum calculated for each sliding window is taken as the cumulative frictional energy value of that tooth during its cycle. This cumulative frictional energy value is directly related to the friction intensity and duration of the tooth's back face, and is used as the instantaneous heat accumulation characteristic of the tooth surface creation zone, characterizing the instantaneous heat load level transmitted to the tooth surface creation zone during the cutting process.

[0129] S222. Wavelet packet decomposition of characteristic acoustic emission sequences and feature mapping of white layer generation tendency, specifically:

[0130] After receiving the characteristic acoustic emission sequence after phase locking of the cutting teeth, wavelet packet decomposition is performed on the characteristic acoustic emission sequence. Multi-level decomposition operation is performed using wavelet basis functions that match the center frequency of the low-frequency phase transition characteristic band to decompose the signal into sub-sequences of different frequency bands. Characteristic frequency band sub-sequences related to the phase transition of the workpiece surface are extracted from the decomposition results. The ratio between the sum of the energy of all sub-sequences in the characteristic frequency band and the total energy of the entire frequency band is calculated to obtain the energy proportion of the characteristic frequency band.

[0131] During online inspection, the energy proportion of the characteristic frequency band extracted in real time is substituted into the pre-stored phase transition calibration curve. The corresponding predicted value of the tooth surface white layer thickness is obtained through function calculation. This predicted value is directly used as a quantitative characterization value of the tooth surface white layer formation tendency. The larger the predicted value, the higher the tendency of the workpiece tooth surface to produce white layer phase transition structure under the current cutting conditions. The pre-stored phase transition calibration curve is a mapping function established through offline cutting experiments, characterizing the monotonic correspondence between the energy proportion of the characteristic frequency band and the tooth surface white layer thickness. The establishment process of this calibration curve is as follows:

[0132] Under constant cutting parameter constraints, multiple sets of progressive cutting experiments were conducted using the same hob type, workpiece material, and process parameters as the online testing conditions. After each preset cutting stroke was completed, acoustic emission signal acquisition and tooth surface metallographic inspection were performed simultaneously. The acoustic emission signal was decomposed by wavelet packets to extract the energy proportion of the low-frequency phase transition characteristic band. For tooth surface metallographic inspection, the machined tooth surface area at the corresponding time period was extracted, and after inlay, grinding, polishing, and etching, the thickness of the white layer structure on the tooth surface was measured under a scanning electron microscope. The characteristic frequency band energy proportions obtained from multiple sets of experiments and the corresponding tooth surface white layer thicknesses were used to form a discrete data point set. Least square curve fitting was performed with the characteristic frequency band energy proportion as the abscissa and the tooth surface white layer thickness as the ordinate to obtain a monotonically increasing mapping curve. This mapping curve was stored in the system memory in the form of a power function or a piecewise linear interpolation function, which is the pre-stored phase transition calibration curve.

[0133] S223. Temporal alignment and cross-correlation delay coefficient calculation of thermal accumulation and white layer characteristics, specifically:

[0134] The instantaneous thermal accumulation feature output from S221 and the white layer formation tendency feature output from S222 are received and time-aligned along the time axis so that the instantaneous thermal accumulation value and the white layer formation tendency value corresponding to the same knife tooth number are at the same time reference point. The two aligned features are then normalized, and the range normalization method is used to map each feature value to the interval between 0 and 1, eliminating differences in dimensions and numerical ranges.

[0135] The cross-correlation coefficient between the two normalized features is calculated at multiple time delay points. The step size of the delay point is equal to the cycle of a single cutter tooth, and the delay range covers multiple positive and negative cutter tooth cycles. The maximum value among all cross-correlation coefficients is taken as the cross-correlation delay coefficient. The delay time corresponding to this coefficient is the causal delay between thermal accumulation and phase transition. The magnitude of the cross-correlation delay coefficient reflects the degree of linear correlation between thermal accumulation and white layer formation.

[0136] S224. The instantaneous thermal accumulation characteristics, white layer formation tendency characteristics, and cross-correlation delay coefficients are encoded and arranged according to the tooth number to form a tooth-by-tooth feature data set of white layer defects, and output to the defect generation determination network unit to evaluate the confidence of white layer defects.

[0137] S23. Machine tool frequency-varying stiffness phase lag unit, specifically:

[0138] S231. Receive the micro-chatter spectrum after the tooth phase is locked, extract the spectral components at the tooth passage frequency and its integer multiples from the spectral data, accurately identify the spectral peak at the tooth passage frequency using the peak picking method, record the measured phase angle corresponding to the peak, and use the measured phase angle as the measured phase response of the tool holder-spindle system; the measured phase response reflects the vibration phase characteristics of the current machine tool process chain under actual cutting conditions, including the phase shift information of the tool holder-spindle system caused by bearing clearance, loose connection and material damping changes.

[0139] The pre-stored nominal dynamic stiffness model of the tool holder-spindle system is invoked. This model consists of an equivalent lumped mass block of the tool holder, an equivalent support stiffness spring of the spindle bearing, an equivalent viscous damper of the spindle bearing, and an equivalent contact stiffness spring at the interface between the tool holder and the spindle, connected in series to form a lumped parameter vibration system structure with two degrees of freedom. The equivalent lumped mass block of the tool holder represents the inertial characteristics of the tool holder and its clamping system. The equivalent support stiffness spring of the spindle bearing and the viscous damper connected in parallel represent the elastic support and energy dissipation characteristics of the spindle bearing on the tool holder. The equivalent contact stiffness spring at the interface between the tool holder and the spindle represents the contact elastic deformation characteristics at the conical or end face connection.

[0140] S232. Substitute the current cutting tooth passing frequency into the nominal dynamic stiffness model of the tool holder-spindle system, solve for the theoretical phase angle corresponding to this frequency, use this theoretical phase angle as the theoretical phase response, and perform a difference operation between the measured phase response and the theoretical phase response, that is, subtract the theoretical phase angle from the measured phase angle, to obtain the dynamic stiffness phase lag characteristic of the tool holder-spindle system in the cutting tooth passing frequency band; this dynamic stiffness phase lag characteristic reflects the degree of dynamic stiffness attenuation caused by the expansion of bearing clearance, the increase of tool holder material damping, or the loosening of the connection interface under the actual cutting load of the machine tool process chain. The larger the phase lag, the more significant the deviation of the system's dynamic stiffness from the nominal state, and the higher the risk of tooth surface waviness and cumulative tooth pitch error.

[0141] S233. Analyze the amplitude data of the micro-chatter spectrum at the cutting tooth passing frequency within a continuous cycle of multiple cutting teeth passing through, extract the effective amplitude value at this frequency in each cycle, calculate the arithmetic mean of the effective amplitude values ​​of multiple consecutive cycles, and obtain the average amplitude level at the cutting tooth passing frequency. Calculate the absolute deviation between each effective amplitude value and the average amplitude for each cycle, and statistically analyze the root mean square value of this absolute deviation as a characteristic of the non-uniformity of cutting tooth meshing impact. This characteristic of non-uniformity of cutting tooth meshing impact reflects the consistency level of cutting force impact when each cutting tooth meshes with the workpiece blank. The larger the root mean square value, the more significant the difference in meshing impact between cutting teeth, and the more severe the inconsistency of cutting tooth wear state or the uneven distribution of dynamic stiffness of the tool holder.

[0142] S234. The dynamic stiffness phase lag characteristics and the non-uniformity characteristics of the tooth meshing impact are encoded and arranged according to the tooth number to form tooth-by-tooth characteristic data of stiffness phase mismatch defects. The data is then output to the defect generation determination network unit for subsequent confidence assessment of stiffness phase mismatch defects.

[0143] S24. Defect generation determination network unit, specifically:

[0144] S241. Alignment of the blade number of the six features and feature-level fusion, specifically:

[0145] The system receives tooth-by-tooth outputs from the cutting edge passivation evolution unit, which outputs the micro-passivation radius and micro-chipping depth of the cutting edge; tooth-by-tooth outputs from the tooth surface thermo-mechanical-phase change coupling unit, which outputs the instantaneous heat accumulation characteristics of the tooth surface creation zone and the tooth surface white layer formation tendency characteristics; and tooth-by-tooth outputs from the machine tool frequency-varying stiffness phase lag unit, which outputs the dynamic stiffness phase lag characteristics and the non-uniformity of the tooth meshing impact, for a total of six characteristic data.

[0146] The tooth-tooth surface latent defect decoupling model consists of three feature extraction channels and a defect generation determination network. The micro-passivation evolution channel of the cutting edge, the thermo-mechanical-phase transformation coupling channel of the tooth surface, and the machine tool frequency-varying stiffness phase lag channel respectively complete the tooth-by-tooth extraction of the above six features. The defect generation determination network is responsible for fusing the extracted features and determining the defects.

[0147] The above six features are fused at the feature level according to the tooth number. That is, under each tooth number, the six features are concatenated in a fixed order to form a multi-dimensional feature vector of fixed dimensions. The dimension of the vector is equal to the number of the six features. Each element of the vector corresponds to the complete latent state representation of the corresponding tooth in a single pass cycle, so that each tooth number corresponds to an independent feature vector sample.

[0148] S242. The multidimensional feature vector is input into the input layer of the defect generative determination network. The multidimensional feature vector is passed from the input layer to the hidden layer. The hidden layer consists of multiple neurons. Each neuron performs a weighted summation operation on all feature elements from the input layer, multiplying each feature element by its corresponding connection weight and summing the results, and then adding the neuron's bias term to obtain the weighted summation result. This weighted summation result is transformed by a non-linear activation function to generate the activation output value of the neuron. The activation output value of the previous hidden layer is used as the input feature of the next hidden layer, and so on, layer by layer, to complete the multi-layer forward propagation calculation, realizing high-dimensional mapping and cross-dimensional interaction of features.

[0149] The activation values ​​output by the hidden layer are passed to the output layer. The output layer has three nodes, corresponding to three defect types: micro-passivation of the cutting edge, thermal accumulation-induced white layer generation, and dynamic stiffness phase mismatch. Each output layer node performs a weighted summation and bias superposition on the activation values ​​from the hidden layer to generate the original output values ​​for each defect type.

[0150] The three original output values ​​are subjected to exponential operations to calculate the exponential function value of each original output value. The three exponential function values ​​are arithmetically summed to obtain the sum of the exponential function values. The exponential function value of each defect type is divided by the sum of the exponential function values ​​to obtain the normalized probability value. The three normalized probability values ​​constitute the confidence distribution of each defect type. Each probability value is in the interval between 0 and 1, and the sum of the three probability values ​​is equal to 1. The defect type with the highest probability value is the most likely latent defect category of the cutting tooth at present.

[0151] S243. Confidence distribution, knife-tooth phase mapping and coordinate output, specifically:

[0152] By combining the tooth phase information embedded in the micro-flutter spectrum, the confidence distribution output by the defect generation determination network is mapped to the specific tooth position coordinates. The tooth phase information originates from the tooth phase reference signal generated by the bond phase marker and the photoelectric speed sensor. This signal establishes a one-to-one correspondence between each tooth synchronization pulse and the specific tooth number on the hob circumference. The confidence values ​​of each defect type in the confidence distribution are spatially located according to the tooth number, so that each tooth number corresponds to a set of confidence values ​​for defects including micro-passivation of the cutting edge, thermal accumulation-induced white layer formation, and dynamic stiffness phase mismatch.

[0153] The mapped data is encapsulated into a structured output, which includes the confidence distribution of hidden defects on the tooth surface and the corresponding coordinates of the cutting tooth where the defects occur. This output data, indexed by the cutting tooth number, provides the positional basis and defect category basis for the cutting tooth phase orientation process parameter control module, enabling the process parameter correction to accurately target specific cutting teeth and the tooth surface creation area they cut.

[0154] Example 4

[0155] Preferably, in Example 1, the tool tooth phase orientation process parameter control module is implemented through the following steps:

[0156] S31. Regulation strategy library, specifically:

[0157] During the system initialization phase, a control strategy library is established based on experimental data of gear hobbing process and tool wear life curve. This strategy library uses defect type and confidence level as dual indexes and pre-stores multiple sets of process parameter correction mapping relationships. The defect type index includes three main categories: micro-passivation of cutting edge, thermal accumulation-induced white layer formation, and dynamic stiffness phase mismatch. The confidence level index is discretized and graded according to the confidence distribution output by the defect generation determination network, divided into three intervals: low confidence level, medium confidence level, and high confidence level.

[0158] For each combination of defect type and confidence level, a corresponding correction instruction template is pre-stored. This template includes adjustment rules for four types of process parameters: hob axial feed rate attenuation coefficient, hob spindle speed derating coefficient, coolant injection pressure increment, and synchronous phase compensation coefficient between the hob spindle and workpiece rotation axis. The control strategy library is stored in the system's non-volatile memory in the form of a structured data table, allowing the instruction generation unit to query and call it in real time.

[0159] S32. Defect type parsing unit, specifically:

[0160] The input terminal receives the confidence distribution of latent defect types on the tooth surface and the coordinates of the tooth where the defect occurs. It compares the confidence values ​​of the three defect types in the confidence distribution and selects the defect type with the highest confidence value as the main category of the current defect. The confidence value corresponding to the main category is then compared with a preset confidence level threshold. The preset confidence level threshold includes an upper limit for low confidence level and a lower limit for high confidence level. The upper limit for low confidence level ranges from 0.3 to 0.5, with a classic value of 0.4. The lower limit for high confidence level ranges from 0.6 to 0.8, with a classic value of 0.7.

[0161] When the confidence value of the main category is lower than the upper threshold of the low confidence level, it is determined to be at a low confidence level; when the confidence value of the main category is higher than the lower threshold of the high confidence level, it is determined to be at a high confidence level; when the confidence value of the main category is between the two thresholds, it is determined to be at a medium confidence level. The determined main category is converted into a defect type code, the cutter tooth number in the cutter tooth position coordinates where the defect occurs is extracted as the corresponding cutter tooth number, and the defect type code and the corresponding cutter tooth number are output to the instruction generation unit.

[0162] S33. Instruction generation unit, specifically:

[0163] The input terminal is connected to the defect type parsing unit and the control strategy library. After receiving the defect type code, it uses the code as an index to query the control strategy library and calls the correction instruction template corresponding to the defect type and confidence level.

[0164] When the defect type code indicates micro-passivation of the cutting edge, the instruction generation unit extracts the hob axial feed rate attenuation coefficient and the hob spindle speed derating coefficient from the correction instruction template. The hob axial feed rate attenuation coefficient is calculated by multiplying the current hob axial feed rate by a preset attenuation ratio, which ranges from 5% to 25%, with a classic value of 15%. This ratio reduces the cutting load on the cutter teeth to delay passivation propagation while maintaining the basic material removal rate. The hob spindle speed derating coefficient is calculated by multiplying the current hob spindle speed by a preset derating ratio, which ranges from 5% to 20%, with a classic value of 10%. This ratio strikes a balance between reducing cutting temperature and friction intensity.

[0165] When the defect type code indicates that heat accumulation induces the formation of white layer, the instruction generation unit extracts the coolant injection pressure increment from the correction instruction template. This increment is calculated by multiplying the current coolant injection pressure by a preset pressure increase ratio. The preset pressure increase ratio ranges from 10% to 50%, with a classic value of 30%. This ratio ensures that the coolant can penetrate the high-temperature cutting zone to suppress heat accumulation on the tooth surface.

[0166] When the defect type code indicates dynamic stiffness phase mismatch, the instruction generation unit extracts the synchronous phase compensation coefficient between the hob spindle and the workpiece rotation axis from the correction instruction template. This coefficient is calculated by superimposing the current synchronous phase angle of the two axes with a preset compensation angle. The preset compensation angle ranges from -0.5° to +0.5°, with a classic value of +0.2°. This small angular offset can offset the periodic meshing impact caused by dynamic stiffness phase lag.

[0167] The generated specific process parameter correction instructions are output to the instruction issuance interface unit.

[0168] S34. Command issuance interface unit, specifically:

[0169] The input end is connected to the instruction generation unit, and the output end is connected to the CNC system of the gear hobbing machine. After receiving the process parameter correction instruction, it reads the real-time speed of the current hob spindle and the tooth phase reference signal output by the key phase mark, calculates the time interval required for the tooth corresponding to the position coordinate of the defective tooth to rotate from the current position to the meshing area; and determines the advance of the correction instruction based on the time interval and the instruction parsing delay of the CNC system.

[0170] The process parameter correction command is sent to the hobbing machine CNC system via industrial Ethernet or fieldbus communication interface in the tooth phase orientation manner. This enables the hobbing machine CNC system to dynamically adjust the hob axial feed speed, hob spindle speed, cutting fluid injection pressure, or synchronous phase angle between the spindle and the workpiece rotation axis at a preset advance time before the specified tooth reaches the defect location, thereby achieving traceable closed-loop control at the tooth level.

[0171] In some embodiments, the online detection system based on a gear hobbing machine of the present invention can be implemented using a combination of hardware and software. As an example, the online detection system based on a gear hobbing machine of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the online detection method based on a gear hobbing machine of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0172] The modules described in the embodiments of the present invention can be implemented in software or hardware. The names of the modules are not, in some cases, intended to limit the functionality of the module itself.

[0173] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned online detection methods based on a gear hobbing machine. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute an online detection method based on a gear hobbing machine as shown in any embodiment of the present invention by calling the computer program.

[0174] In one alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0175] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0176] Bus 4002 may include a path for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus 4002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0177] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0178] The memory 4003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 4001. The processor 4001 executes the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0179] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0180] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0181] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-mentioned online detection methods based on a gear hobbing machine.

[0182] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0183] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned online detection system based on a gear hobbing machine.

[0184] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0185] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation that may be implemented in the methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0186] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0187] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0188] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0189] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0190] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0191] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An online inspection system based on a gear hobbing machine, characterized in that, include: The tooth phase-locking multimodal in-situ sensing module is used to acquire the tooth phase reference signal through key phase marking during the hobbing process, and simultaneously collect the micro-chatter spectrum, the morphology of the sawtooth shear band at the chip root, the fluctuation of the shear band spacing and thickness pulsation, the high-frequency friction energy and the acoustic emission sequence of the workpiece phase transition characteristics, perform tooth phase-locking alignment, and generate a multimodal latent state feature vector. The tooth-tooth surface latent defect decoupling and tracing module is used to input the multimodal latent state feature vector into the pre-trained tooth-tooth surface latent defect decoupling model, extract the tooth micro-passivation radius and micro-chipping depth, tooth surface thermal accumulation and white layer generation tendency, dynamic stiffness phase lag and meshing impact non-uniformity, and output the confidence distribution of tooth surface latent defect type and the coordinates of the tooth where the defect occurs. The tooth phase orientation process parameter control module is used to call the control strategy library according to the confidence distribution of the hidden defect type on the tooth surface and the coordinates of the tooth where the defect occurs, generate the corresponding process parameter correction instruction according to the defect type and confidence level, and send it to the gear hobbing machine CNC system in the tooth phase orientation mode, so that the gear hobbing machine can complete the dynamic adjustment of the corresponding process parameters before the specified tooth reaches the defect location.

2. The online inspection system based on a gear hobbing machine according to claim 1, characterized in that, The blade-tooth phase-locked multimodal in-situ sensing module specifically includes: A triaxial piezoelectric micro-chatter sensor array is attached to the circumferential region near the end of the hob holder to collect the micro-chatter spectrum of the meshing area between the hob teeth and the workpiece blank at the tooth passing frequency. The high-resolution linear array vision sensing unit, equipped with a macro lens, sets up scanning lines along the direction perpendicular to the chip movement. It is used to start line scanning acquisition with the tooth phase reference signal as an external trigger signal to obtain a continuous grayscale image sequence of the chip root. A broadband acoustic emission sensing unit is attached to the hob holder through an acoustic coupling medium to collect the high-frequency friction energy sequence generated by the friction between the hob tooth rake face and the machined surface of the workpiece, as well as the characteristic acoustic emission sequence generated by the phase change of the workpiece material surface induced by the high temperature and high pressure zone of hobbing. The key phase marker and photoelectric speed sensor are installed at the end of the hob spindle to output the tooth phase reference signal as a synchronous latching trigger source for multi-source sensing signals.

3. The online inspection system based on a gear hobbing machine according to claim 2, characterized in that, The specific workflow of the high-resolution linear array vision sensing unit includes: A directional backlight illuminates the chip root area in the chip discharge path at a preset angle, so that the edge of the serrated shear band at the chip root forms a high-contrast light-dark boundary line. A high-resolution linear array vision sensor equipped with a macro lens sets up scanning lines perpendicular to the chip movement direction, and uses the tooth phase reference signal as an external trigger signal to start line scanning acquisition to obtain a continuous grayscale image sequence of the chip root. Real-time edge detection and morphological segmentation are performed on the acquired image sequences to extract the morphological sequence of the serrated shear band, the spacing fluctuation sequence between adjacent shear bands, and the pulsation sequence of the thickness of the chip root along the movement direction. The three sequence data are aligned and encapsulated according to the acquisition timestamp and the blade phase reference signal, and the microscopic dynamic feature data of the chip root is output for use by subsequent modules.

4. The online inspection system based on a gear hobbing machine according to claim 3, characterized in that, The specific workflow of the broadband acoustic emission sensing unit includes: By using an acoustic coupling medium, a broadband acoustic emission sensor is tightly attached to the surface of the hob bar, so that the high-frequency elastic wave generated by the friction between the hob tooth rake face and the machined surface of the workpiece, as well as the transient elastic wave released by the phase transition of the workpiece surface, can be effectively coupled to the sensor's sensitive element. The wideband acoustic emission sensor continuously acquires the raw acoustic emission signal after coupling. After pre-amplification and bandpass filtering, two characteristic frequency bands are separated. The high-frequency friction energy corresponding frequency band is used to characterize the sliding friction intensity between the back face of the cutting tooth and the tooth surface, and the low-frequency phase transition characteristic frequency band is used to capture the characteristic acoustic emission events generated by the phase transition of the workpiece surface. The two filtered signals are subjected to envelope detection and energy integration to generate a high-frequency friction energy sequence and a characteristic acoustic emission sequence. The two sequences are latched and aligned according to the timestamp of the tooth phase reference signal, and the tooth surface friction and phase transition state characteristic data are output for use by subsequent modules.

5. The online inspection system based on a gear hobbing machine according to claim 4, characterized in that, The specific working process of the bond phase marker and photoelectric speed sensor includes: The key phase mark installed at the end of the hob spindle rotates synchronously with the spindle. Whenever a specified cutter tooth passes through the detection optical path of the photoelectric speed sensor, the key phase mark triggers the sensor to output a corresponding cutter tooth synchronization pulse signal, which serves as the cutter tooth phase reference signal. Using the rising edge of the blade synchronization pulse signal as the synchronization latch trigger source, hardware-level synchronization latch acquisition is performed on the micro-chatter spectrum, the chip root image sequence, and the characteristic acoustic emission sequence, respectively. The latched multi-channel sensor data is aligned and encapsulated according to the cycle of the blade teeth, generating a multimodal latent state feature vector with a single blade tooth as the index unit.

6. The online inspection system based on a gear hobbing machine according to claim 1, characterized in that, The tool tooth-tooth surface latent defect decoupling and tracing module specifically includes: The cutting edge passivation evolution unit is connected to the sawtooth shear band morphology sequence and shear band spacing fluctuation sequence in the multimodal latent state feature vector at its input end. It is used to extract the micro passivation radius features and micro chipping depth features of each cutting edge tooth by tooth. The tooth surface thermo-mechanical-phase change coupling unit has its input end connected to the high-frequency friction energy sequence and the characteristic acoustic emission sequence in the multimodal latent state feature vector, and is used to extract the instantaneous thermal accumulation features of the tooth surface creation zone and the tooth surface white layer generation tendency features tooth by tooth. The machine tool frequency-varying stiffness phase lag unit has its input end connected to the micro-chatter spectrum in the multimodal latent state feature vector, which is used to extract the dynamic stiffness phase lag characteristics and tooth meshing impact non-uniformity characteristics of the hob cutter bar-spindle system in the tooth passing frequency band. The defect generation determination network unit is connected to the output of the above three units respectively. It is used to input the extracted six features into the defect generation determination network of the tooth-tooth surface latent defect decoupling model, and output the confidence distribution of tooth surface latent defect type and the corresponding tooth position coordinates where the defect occurs.

7. The online inspection system based on a gear hobbing machine according to claim 6, characterized in that, The specific workflow of the cutting edge passivation evolution unit includes: The serrated shear band morphology sequence and shear band spacing fluctuation sequence after receiving the phase-locked cutting teeth are used. Contour fitting is performed on the shear band morphology image corresponding to each cutting tooth. Under constant cutting parameter constraints, the curvature radius of the shear band root is extracted as the quantitative characterization value of the micro-passivation radius of the cutting edge. Fourier transform and envelope demodulation were performed on the shear band spacing fluctuation sequence to separate the high-frequency periodic fluctuation component caused by micro-chipping of the cutting edge. The micro-chipping depth characteristics of the cutting teeth were calculated tooth by tooth based on the fluctuation amplitude. The extracted micro-passivation radius and micro-chipping depth features are compared with the pre-stored initial state parameters of the cutting teeth to obtain the wear increment value of the current cutting teeth relative to the new cutting state. The wear increment value of each cutting tooth is encoded and output according to the cutting tooth number, and used by the defect generation determination network unit to evaluate the confidence of passivation defects.

8. The online inspection system based on a gear hobbing machine according to claim 7, characterized in that, The specific workflow of the thermo-mechanical-phase change coupling unit in the tooth surface generative zone includes: The high-frequency friction energy sequence and characteristic acoustic emission sequence after the phase lock of the cutting teeth are received. The high-frequency friction energy sequence is processed by sliding window integration, and the cumulative friction energy value of each cutting tooth in the passing cycle is extracted as the instantaneous thermal accumulation feature of the tooth surface creation area. Wavelet packet decomposition is performed on the characteristic acoustic emission sequence to extract the characteristic frequency band energy ratio related to the phase transition of the workpiece surface. Based on the pre-stored phase transition calibration curve, the energy ratio is mapped to the tooth surface white layer generation tendency feature. The instantaneous thermal accumulation characteristics and white layer formation tendency characteristics are time-aligned and normalized, and the cross-correlation delay coefficient between the two is calculated to determine the strength of the causal relationship between thermal accumulation and phase transition. The two features and causality coefficients mentioned above are encoded and output according to the tooth sequence number, and used by the defect generation determination network unit to evaluate the confidence of white layer defects.

9. The online inspection system based on a gear hobbing machine according to claim 8, characterized in that, The specific workflow of the defect generation determination network unit includes: The system receives six feature data outputs from the cutting edge passivation evolution unit, the tooth surface thermo-mechanical-phase change coupling unit, and the machine tool frequency-variable stiffness phase lag unit. The system then fuses these six features at the feature level according to the cutting tooth number to form a fixed-dimensional multidimensional feature vector. The multidimensional feature vectors are input into the defect generation determination network. After forward propagation calculation, the confidence distribution of each defect type is output. By combining the tooth phase information in the micro-flutter spectrum, the confidence distribution is mapped to the specific tooth position coordinates, and the confidence distribution of the latent defect type on the tooth surface and the corresponding tooth position coordinates where the defect occurs are output.

10. The online inspection system based on a gear hobbing machine according to claim 1, characterized in that, The tool tooth phase orientation process parameter control module specifically includes: A control strategy library is used to pre-store the mapping relationship between process parameter corrections and different defect types and confidence levels; The defect type parsing unit is connected to the confidence distribution of the latent defect type on the tooth surface and the coordinates of the cutter tooth where the defect occurs. It is used to parse the main category and confidence level of the current defect and output the defect type code and the corresponding cutter tooth number. The instruction generation unit, with its input end connected to the defect type parsing unit and the control strategy library, is used to call the corresponding correction instruction template in the strategy library according to the defect type code: when the defect type is micro-passivation of the cutting edge, it generates the hob axial feed rate attenuation coefficient and the hob spindle speed derating coefficient; when the defect type is thermal accumulation-induced white layer formation, it generates the cutting fluid injection pressure increment; when the defect type is dynamic stiffness phase mismatch, it generates the hob spindle-workpiece rotation axis synchronous phase compensation coefficient. The instruction sending interface unit is connected to the instruction generation unit and the gear hobbing machine CNC system. It is used to send the generated correction instruction to the gear hobbing machine CNC system in the form of tooth phase orientation, so that the gear hobbing machine can complete the dynamic adjustment of the corresponding process parameters before the specified tooth reaches the defect location.

Images