A detection method and system for ultrasonic-assisted machining of holes in brittle materials.

Abstract

This application relates to a detection method and system for ultrasonic-assisted machining of holes in brittle materials. The method involves installing an electrical parameter sensor within the ultrasonic tool drive circuit to collect voltage and current signals, and calculating acoustic signature parameters such as the effective value of the current and the spectral energy of a specific frequency band. The real-time acoustic signature parameters are compared with a pre-established acoustic signature feature library to achieve online determination of micro-damage. When the effective current value exhibits a momentary spike exceeding a first threshold, it is determined that an inlet impact has caused a star-shaped crack at the port. When the effective current value remains above a second threshold and the spectral energy of the specific frequency band shows predetermined fluctuations, it is determined that poor chip removal has caused internal wall scratches. If damage is determined, the feed action of the rotary table is adjusted. The system includes signal acquisition, acoustic signature processing, state decision-making, and motion control modules. This solution solves the problems of difficult detection and control lag in micro-damage, achieving precise and intelligent control of the machining quality of holes in brittle materials such as optical fibers.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of precision and ultra-precision machining, and particularly relates to a detection method and system for brittle material hole ultrasonic auxiliary machining. BACKGROUND

[0002] The rapid development of the fields of optical fiber communication, sensing and laser, etc. puts forward almost rigorous requirements for the machining quality of optical fiber capillary elements. Such elements are usually made of brittle materials such as quartz glass, and micron-level high-depth-diameter ratio channels need to be machined in the interior. The entrance quality, inner wall smoothness and geometric shape consistency of these channels directly determine the transmission loss of optical signals, the packaging precision of devices and the long-term reliability.

[0003] Ultrasonic assisted cutting is one of the key processes for realizing high-quality hole machining of optical fiber capillaries. It reduces the cutting force through high-frequency vibration of the tool and promotes the plastic domain removal of brittle materials. However, when machining capillaries with a diameter often less than 0.5 mm, the process window is extremely narrow, and the problem of micro-damage during machining is particularly prominent. Specifically, the following problems exist:

[0004] Port star crack problem: at the moment when the tool initially contacts the end face of the capillary, due to the small contact area, the stress is highly concentrated. The capillary port structure is fragile, and under the combined action of fixed ultrasonic amplitude and improper initial feed force, the high-frequency impact energy is easy to cause radial expansion of micro-cracks from the edge to the surrounding, forming a "port star crack". This damage can significantly reduce the mechanical strength of the optical fiber joint and become a source of optical signal scattering.

[0005] Inner wall scratch problem: when machining high-depth-diameter ratio holes, the micro-chips generated by cutting are confined in the narrow inner space of the hole, and the grinding fluid is difficult to effectively flush in, update and remove chips. The chips that are not removed in time will adhere to the hole wall or be crushed again by the tool, acting as "abrasive particles" between the tool side and the hole wall, and scratching the hole wall in the axial direction as the tool rotates and feeds. These scratches can greatly increase the transmission loss of optical signals and affect the alignment accuracy of optical fibers and ferrules.

[0006] At present, there are obvious deficiencies in the monitoring of the ultrasonic machining process of optical fiber capillaries. The traditional macro-parameters (such as average spindle current, feed force) monitoring is not sensitive to the above-mentioned micro-damage and has a serious lag. The existing technology mostly adopts "open loop" control based on position, and the system cannot perceive the real-time changes of cutting resistance (such as fluctuations caused by local unevenness of grinding fluid, tool micro-wear, material micro-defects), and can only "stubbornly" execute the preset stroke, resulting in uncontrollable contact force and inducing or aggravating micro-damage. In addition, there is a lack of online sensing features and criteria that can accurately represent the unique damage of optical fiber machining (such as star crack and scratch), making the accuracy and reliability of process monitoring difficult to meet the manufacturing requirements of high-end optical fiber devices.

[0007] Therefore, there is an urgent need in the art for a new method and system that can specifically target the hole processing characteristics of fragile elements such as fiber capillary tubes, and can accurately perceive and distinguish microscopic damage such as port star crack, inner wall pull mark, etc. in real time, and can make instantaneous adaptive control accordingly, to ensure ultra-low loss and ultra-high reliability of fiber elements. SUMMARY

[0008] The present application aims to solve the problems of difficult perception of microscopic damage, control lag and lack of adaptability.

[0009] According to an aspect of the present application, a detection method for ultrasonic assisted machining of holes in fragile materials is provided, which is applied to a cutting system comprising an ultrasonic tool and a rotary lifting platform, the transducer of the ultrasonic tool is connected to the power output end of the ultrasonic generator through a drive circuit, and an electric parameter sensor for collecting voltage signal and current signal is arranged in the drive circuit, the method comprises:

[0010] S10, synchronously collecting voltage signal and current signal of the drive circuit through the electric parameter sensor;

[0011] S20, calculating acoustic fingerprint characteristic parameters based on the voltage signal and the current signal, the acoustic fingerprint characteristic parameters comprising current effective value and spectral energy of a specific frequency band;

[0012] S30, comparing the real-time monitored acoustic fingerprint characteristic parameters with a pre-established acoustic fingerprint characteristic library to determine whether microscopic damage occurs, wherein determining the microscopic damage state comprises:

[0013] when the current effective value appears a transient spike exceeding a first threshold value, determining that the port star crack is caused by inlet impact; or,

[0014] when the current effective value is continuously higher than a second threshold value and the spectral energy of the specific frequency band appears a predetermined mode of fluctuation, determining that the inner wall pull mark is caused by poor chip removal;

[0015] S40, if yes, adjusting the feeding action of the rotary lifting platform.

[0016] Preferably, the acoustic fingerprint characteristic parameters calculated in real time in S20 further comprise equivalent impedance amplitude;

[0017] Correspondingly, the determination of the microscopic damage state in S30 further comprises:

[0018] when the equivalent impedance amplitude sharply decreases from a load state to an unloaded state, determining that the machining penetration occurs.

[0019] Preferably, the calculation of acoustic fingerprint characteristic parameters based on the voltage signal and the current signal in S20 comprises:

[0020] S21, calculating the effective value of the voltage signal and the current signal;

[0021] S22, calculating the phase difference between the voltage and the current, the calculation method including zero-crossing detection method or fast Fourier transform method;

[0022] S23, calculating the equivalent impedance amplitude based on the voltage effective value and the current effective value;

[0023] S24, performing fast Fourier transform on the current signal to extract the spectral energy of the specific frequency band.

[0024] Preferably, the decision logic in S30 also satisfies the following conditions:

[0025] When the current effective value and the spectral energy of the specific frequency band satisfy the threshold condition of the microscopic damage state, and the state lasts for a preset duration or appears a preset number of consecutive times in consecutive analysis windows, the final output of the microscopic damage decision result is output.

[0026] Preferably, the voiceprint feature parameter further includes an energy ratio R;

[0027] The energy ratio R is obtained by calculating the ratio of the target frequency band signal energy and the reference frequency band signal energy;

[0028] Wherein, the center frequency of the target frequency band is located in the interval [f1, f2], and the bandwidth is located in the interval [B1, B2]; the center frequency of the reference frequency band is located in the interval [f3, f4], and the bandwidth is located in the interval [B3, B4];

[0029] In S30, when the current effective value is continuously higher than the second threshold value and the energy ratio R appears a predetermined mode of fluctuation, it is determined that the internal wall pull is caused by poor chip removal.

[0030] Preferably, the construction and dynamic updating of the voiceprint feature library includes:

[0031] S11, collecting the voiceprint feature parameters of the system in the no-load state and the stable cutting state to generate an initial reference template;

[0032] S12, for the stable cutting stage data of subsequent qualified machining, the exponential weighted moving average algorithm with forgetting factor μ ∈ [μ1, μ2] is used to dynamically update the reference template;

[0033] S13, in the updating process, the abnormal data window deviating from the reference template confidence interval [p1, p2] is removed to prevent it from polluting the template data;

[0034] S14, when the cumulative working time of the tool reaches the interval [H1, H2] or the tool wear estimated by the voiceprint feature enters the interval [W1, W2], a new stage template is generated and switched to, so as to maintain the accuracy of the discriminant reference.

[0035] Preferably, the correspondence between the abnormal voiceprint in the voiceprint feature library and the micro-damage type is established by an offline verification method, and the offline verification method comprises:

[0036] Under the preset abnormal process parameters, the machining is performed, and the voiceprint feature parameters are synchronously collected;

[0037] After the machining is completed, the damage morphology of the workpiece port and the inner wall of the hole is analyzed by using a microscopic observation device, and the type and degree of micro-damage are determined;

[0038] The collected abnormal voiceprint feature parameters are associated and mapped with the micro-damage type determined by observation to form the determination basis in the voiceprint feature library.

[0039] Preferably, the offline verification method further comprises statistical significance evaluation and determination threshold setting of the correspondence:

[0040] Based on the voiceprint feature sample sets obtained under normal and abnormal process parameters respectively, blind test evaluation is performed, and the consistency proportion and the confusion matrix of the determination result are calculated;

[0041] When the consistency proportion reaches the preset confidence level [Q1, Q2], and the key misjudgment rate in the confusion matrix is lower than the preset threshold [E1, E2], the first threshold value, the second threshold value and the parameters of the predetermined mode currently used are determined as the final determination basis;

[0042] If the above conditions are not met, the sampling parameters, the frequency band definition or the calculation strategy of the voiceprint feature are traced back and adjusted, and the offline verification process is repeatedly performed.

[0043] The application also provides a detection system for ultrasonic auxiliary machining of brittle material holes, which is applied to a cutting system comprising an ultrasonic tool and a rotary lifting platform, a transducer of the ultrasonic tool is connected with a power output end of an ultrasonic generator through a driving circuit, and the detection system comprises:

[0044] A signal collection module comprising an electric parameter sensor arranged in the driving circuit, for synchronously collecting voltage signals and current signals;

[0045] A voiceprint processing module in communication connection with the signal collection module, for calculating voiceprint feature parameters based on the voltage signals and the current signals;

[0046] A state decision module, which is embedded with a voiceprint feature library and is in communication connection with the voiceprint processing module, is configured to compare the real-time voiceprint feature parameters with the voiceprint feature library to determine whether micro-damage occurs and to generate corresponding control instructions;

[0047] A motion control module, which is in communication connection with the state decision module and the rotary lifting platform, is configured to adjust the feeding action of the rotary lifting platform according to the control instructions.

[0048] Preferably, the voiceprint feature library is established and updated by an offline verification module, and the offline verification module comprises:

[0049] A process parameter setting unit is configured to set and execute an abnormal machining process that can induce specific micro-damage;

[0050] A microscopic observation unit is configured to collect and analyze the damage morphology of the machined workpiece;

[0051] A data correlation unit is configured to correlate and match the voiceprint feature data under abnormal process parameters with the damage type determined by the microscopic observation unit, and feed the matching result to the voiceprint feature library in the state decision module.

[0052] The present application has the following beneficial effects: precise online perception and identification of micro-damage are achieved. The present method directly collects high-frequency electric signals that best reflect the micro-interaction between the tool and the workpiece through an electric parameter sensor arranged in the drive circuit, and extracts current effective value, specific frequency band spectrum energy and other voiceprint feature parameters therefrom. This enables the system to sensitively capture transient impact and high-frequency friction signals that cannot be reflected by traditional macro-parameters, thereby achieving early identification and precise discrimination of the two typical micro-damages, i.e., “port star crack” and “inner wall pull mark”, and solving the problem of insufficient perception dimension and insensitivity to micro-damage in the prior art.

[0053] The adaptive control leap from “open loop” to “intelligent closed loop” is achieved. The core of the present method is to compare the real-time perceived voiceprint features with the pre-set feature library, and immediately adjust the feeding action of the rotary lifting platform when micro-damage is determined to have occurred. This changes the rigid, position-based open-loop control mode in the prior art and forms an intelligent closed-loop control based on real-time feedback of the machining state. The system can actively respond to dynamic changes in the machining process (such as initial contact impact and deteriorating chip removal state), and through adaptive strategies such as pausing, retreating and chip removal, it actively suppresses or eliminates the occurrence and development of damage, thereby fundamentally solving the problem of uncontrollable damage caused by control lag and lack of adaptive ability.

[0054] This approach enhances the quality consistency and process reliability of fiber optic component manufacturing. Through the combination of precise sensing and intelligent control, this solution effectively avoids star-shaped cracks at the fiber capillary ends and internal wall striations, directly ensuring the mechanical strength and optical performance (such as low transmission loss and high alignment accuracy) of the processed components. This not only improves the success rate of single-processing but also significantly enhances the consistency and repeatability of the process due to the system's precise control over microscopic states, providing a reliable technical guarantee for the mass production of high-performance fiber optic devices. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0056] Figure 1 This is a logic block diagram of a detection method for ultrasonic-assisted machining of pores in brittle materials, as described in one embodiment of this application. Detailed Implementation

[0057] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings. Preferred embodiments of this application are shown in the drawings. However, this application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0059] Please refer to Figure 1 This application provides a detection method for ultrasonic-assisted machining of holes in brittle materials, applied to a cutting system including an ultrasonic tool and a rotary table. The transducer of the ultrasonic tool is connected to the power output terminal of an ultrasonic generator via a drive circuit. An electrical parameter sensor for acquiring voltage and current signals is provided in the drive circuit. The method includes:

[0060] S10. Synchronously acquire the voltage and current signals of the drive circuit using electrical parameter sensors. In this step, it should be noted that the electrical parameter sensors (typically high-voltage differential probes and current probes) are directly mounted on the power transmission path between the ultrasonic generator and the transducer, rather than on the mechanical structure. This arrangement allows for the acquisition of the original high-frequency electrical signals driving the ultrasonic cutter vibration without introducing additional attenuation or hysteresis. "Synchronous acquisition" means that the voltage and current signals are sampled at the same time; the sampling rate is preferably not less than 500kHz to ensure the accuracy of subsequent phase and other relationship calculations, which are fundamental data conditions for accurately analyzing the system's energy state and load characteristics.

[0061] S20. Based on the voltage and current signals, calculate the voiceprint feature parameters, which include the effective value of the current and the spectral energy of a specific frequency band. It should be noted that this step is used for feature extraction and aggregation of the original signal.

[0062] RMS current (I_rms): This is the root mean square value of the driving current over one cycle, reflecting the overall average energy output level and load of the ultrasonic system. As the cutting resistance increases, the energy required to maintain vibration increases, and the RMS current will rise accordingly.

[0063] Spectral energy of a specific frequency band: The acquired current signal is converted from the time domain to the frequency domain by performing a Fast Fourier Transform. Then, the signal energy within one or more pre-selected frequency bands sensitive to specific processing phenomena (e.g., the second to fifth harmonics of the ultrasonic fundamental frequency (i.e., 40kHz–100kHz)) is calculated. This parameter is extremely sensitive to abrupt changes in processing conditions, high-frequency friction, and impacts, and can reveal microscopic dynamics that macroscopic current values ​​cannot reflect.

[0064] S30. Compare the real-time monitored voiceprint feature parameters with the pre-established voiceprint feature database to determine whether microscopic damage has occurred. The determination of microscopic damage includes:

[0065] When the effective current value experiences a momentary spike exceeding the first threshold, it is determined to be a port star-crack caused by an inlet impact. Or,

[0066] When the effective value of the current is continuously higher than the second threshold and the spectral energy of a specific frequency band fluctuates in a predetermined pattern, it is determined that the internal wall scratches are caused by poor chip removal.

[0067] In this step, it should be noted that the acoustic signature feature library is a "knowledge base" that has been established in advance through a large number of process experiments and calibrations. It stores the typical patterns and thresholds of acoustic signature feature parameters under various typical conditions (such as no load, stable cutting, inlet impact, poor chip removal, etc.).

[0068] When the effective value of the current exhibits a momentary spike exceeding a first threshold (e.g., 150% to 200% of the no-load current value), it is determined to be a port star crack caused by an inlet impact. The duration of this spike is typically extremely short, falling within the range of [1 ms, 10 ms].

[0069] Alternatively, when the effective value of the current is continuously higher than the second threshold (e.g., 115% to 130% of the stable cutting current value), and the spectral energy of the specific frequency band shows a predetermined pattern of fluctuation (e.g., the fluctuation amplitude exceeds 20% to 50% of the baseline value and lasts for more than [50ms, 200ms]), it is determined that the internal wall scratches are caused by poor chip removal.

[0070] S40. If so, adjust the feed action of the rotary lifting platform.

[0071] It should be noted that this step is the execution phase of the decision-making process. Once microscopic damage is detected, the system will immediately interrupt the preset fixed processing procedure and trigger the corresponding adaptive control strategy.

[0072] For cases identified as entry impact, control strategies may include: immediately pausing the platform feed and controlling its retraction by 10-20 μm, maintaining ultrasonic vibration for 50-100 ms to utilize the cavitation effect to correct the port, and then slowly resuming the feed at 50%-70% of the original speed.

[0073] For cases where chip removal is deemed to be inadequate, control strategies may include: initiating a chip removal cycle, controlling the platform to retract 80-150μm, while maintaining or increasing the ultrasonic amplitude to 120%-150% and triggering a polishing slurry pulse jet for 150-300ms, followed by refeeding at an extremely slow speed and checking the acoustic signature.

[0074] The technical solution implemented in this embodiment can achieve the following technical effects:

[0075] Achieving early and accurate diagnosis of micro-damage: By directly monitoring the "acoustic" characteristics that reflect the micro-interaction between the tool and the workpiece, the problem of slow or even ineffective perception of micro-damage by traditional macro-parameters (such as average current and feed force) is solved, enabling identification and alarm in the early stages of star crack propagation or scratch intensification.

[0076] Transforming from passive to proactive, achieving adaptive intelligent control: By directly closing the loop between processing status perception and motion control, the system is no longer a rigid "blind man" executing the process, but an intelligent agent capable of "thinking" and "reacting" based on real-time working conditions. By adjusting the feed action in real time, it can proactively suppress and eliminate the causes of damage, shifting from "post-event remediation" to "pre-event prevention and in-event suppression," significantly improving the processing success rate.

[0077] Especially suitable for processing high-performance brittle components such as optical fibers: This solution provides a direct solution to the unique problems of "port star crack" and "inner wall scratch" in the hole processing of brittle materials. It can effectively ensure the port integrity and inner wall smoothness of components such as optical fiber capillaries, thereby directly improving their mechanical strength and optical performance, and meeting the stringent requirements of high-end devices for ultra-low loss and ultra-high reliability.

[0078] Furthermore, the acoustic signature parameters calculated in real time in S20 also include the equivalent impedance amplitude. In this step, it's important to note that the equivalent impedance amplitude (Z) is a key electrical parameter reflecting the overall mechanical load of the ultrasonic vibration system. Its calculation formula is Z = V_rms / I_rms, which is the ratio of the effective voltage value to the effective current value. When the tool contacts the workpiece and experiences cutting resistance, the system's mechanical load increases, and the equivalent impedance amplitude significantly rises from a low level under no-load conditions to a stable load level. This parameter provides a more macroscopic and stable load perspective, distinct from instantaneous current and spectral energy.

[0079] Correspondingly, the determination of microscopic damage states in S30 also includes: when the equivalent impedance amplitude drops sharply from the loaded state to the unloaded state, it is determined as machining penetration. In this determination logic, it should be noted that "machining penetration" is the physical process where, at the moment hole machining is about to be completed, the tool breaks through the last layer of material on the workpiece, causing the load at its tip to suddenly disappear. Electrically, this process is typically manifested as the equivalent impedance amplitude dropping sharply from a stable, relatively high load value (e.g., 50%~100% higher than the unloaded impedance) to a level close to the unloaded state within a very short time (e.g., less than 10ms) (e.g., the drop exceeds 60%~80% of the original load impedance value). Detecting this characteristic "high-dive" signal allows for a highly reliable determination of the penetration event.

[0080] Implementing the further technical solutions of this embodiment can achieve the following additional technical effects:

[0081] Achieving precise and instantaneous determination of the machining endpoint: Utilizing the abrupt change in equivalent impedance amplitude at the moment of penetration, this provides the most direct and fastest endpoint detection method. This fundamentally avoids the "over-machining" problem (i.e., the tool continues to advance after penetration, impacting the fixture or causing chipping at the exit edge) caused by traditional endpoint determination based on position or time, thus perfectly protecting the workpiece exit quality and the tool itself.

[0082] Constructing a more complete micro-damage monitoring system: Combining the macro-load parameter of "equivalent impedance amplitude" with micro-dynamic parameters such as "current spikes" and "spectral fluctuations" together forms a complete monitoring network from micro-impact and process anomalies to macro-state transitions, which significantly improves the comprehensiveness and reliability of the system's insight into the state of the entire processing process.

[0083] Furthermore, in S20, based on voltage and current signals, the calculation of voiceprint characteristic parameters includes:

[0084] S21. Calculate the effective values ​​of the voltage and current signals.

[0085] S22. Calculate the phase difference between voltage and current. Calculation methods include zero-crossing detection or fast Fourier transform.

[0086] S23. Calculate the equivalent impedance magnitude based on the effective values ​​of voltage and current.

[0087] S24. Perform a fast Fourier transform on the current signal to extract the spectral energy of a specific frequency band.

[0088] In this series of sub-steps, it should be noted that this is the core processing flow that transforms the raw electrical signal into voiceprint feature parameters with clear physical meaning:

[0089] S21. Calculate the RMS values: This step calculates the root mean square (RMS) values ​​of the acquired high-speed alternating voltage and current signals within a time window. This yields DC parameters characterizing the average power level of the signal, namely the RMS voltage value (V_rms) and the RMS current value (I_rms), which form the basis for subsequent impedance calculations and assessments of overall energy consumption.

[0090] S22. Calculate the phase difference: This step aims to obtain the phase difference (Φ) between the voltage and current waveforms. When an ultrasonic vibration system operates near its resonant frequency, it acts as a capacitive load, and the voltage and current should inherently have a phase difference. This phase difference changes as the load changes.

[0091] Zero-crossing detection method: The phase difference is calculated by detecting the time difference when the voltage and current waveforms cross zero from negative to positive (or from positive to negative). The method is direct, has low computational load, and is suitable for scenarios with extremely high real-time requirements.

[0092] Fast Fourier Transform (FFT) method: Perform FFT on voltage and current signals separately, extract the phase angle of their fundamental components in the frequency domain, and then calculate the difference. This method has stronger anti-interference ability, more accurate results, and can simultaneously provide data for spectrum analysis.

[0093] S23. Calculate the equivalent impedance magnitude: Based on V_rms and I_rms obtained in step S21, directly calculate the equivalent impedance magnitude using the formula Z=V_rms / I_rms. This parameter comprehensively reflects the electrical characteristics and mechanical load state of the system, and is a macroscopic and stable state indicator.

[0094] S24. Extracting Spectral Energy of a Specific Frequency Band: This step focuses on the frequency domain characteristics of the signal. A Fast Fourier Transform (FFT) is performed on the high-sampling-rate current signal, mapping it from the time domain to the frequency domain. Subsequently, in the frequency domain, a specific frequency band, pre-defined based on process knowledge and sensitive to specific physical processes (such as friction or impact), is located. The sum of squares (or integrals) of the amplitudes of all frequency components within this band is calculated, yielding the spectral energy of that band. This parameter is exceptionally sensitive to high-frequency, transient events generated during the manufacturing process.

[0095] By implementing a further technical solution of this embodiment, and by clarifying the calculation path of the voiceprint feature parameters, the following additional technical effects can be achieved:

[0096] A multi-dimensional and complementary feature parameter system was constructed: feature information of different dimensions was extracted through multi-angle calculations in the time domain (RMS value), frequency domain (spectral energy), and system characteristic domain (phase difference, impedance). These parameters complement and verify each other, together forming a robust and information-rich state-aware matrix, providing a solid data foundation for accurate discrimination.

[0097] The feasibility and accuracy of the technical solution were ensured: the specific calculation methods and options for key parameters (such as phase difference and spectral energy) were clarified, providing clear guidance for the hardware selection of the system (such as the need to support high sampling rates) and the implementation of software algorithms, and ensuring that the technical solution can be implemented stably and accurately.

[0098] An interface is reserved for continuous system optimization: the clear algorithm steps allow for continuous improvement in the accuracy and response speed of state recognition without changing the overall architecture, by optimizing the FFT window size, frequency band division rules, or phase difference calculation accuracy.

[0099] Furthermore, the decision logic in S30 also satisfies the following conditions:

[0100] The final micro-damage determination result is output only when the effective value of the current and the spectral energy of a specific frequency band meet the threshold condition for the micro-damage state, and the state persists for a preset duration, or when the number of occurrences in multiple consecutive analysis windows reaches a preset number of consecutive occurrences.

[0101] In this decision logic, it should be noted that this condition introduces a confirmation mechanism based on time or frequency persistence. Its purpose is to effectively distinguish between real, persistent damage states and transient, isolated signal interference (e.g., external electromagnetic interference, measurement noise, or atypical tiny debris passing by momentarily), thereby significantly reducing the false alarm rate of the system.

[0102] The preset duration of the state refers to the fact that when the signal characteristics (such as the effective value of the current exceeding the second threshold and the abnormal fluctuation of the spectrum energy) are not fleeting, but can be stably maintained for a period of time (for example, 50ms to 200ms), it is considered a stable abnormal state that requires a response, rather than a transient interference.

[0103] If the number of occurrences in multiple consecutive analysis windows reaches a preset number of consecutive occurrences: Considering that the signal may fluctuate, this condition allows the abnormal state to be less than strictly continuous, but as long as it is repeatedly detected in several consecutive (e.g., 3 to 5) analysis windows (each window length can be 10-50ms), the abnormal pattern is considered to be repetitive and regular and should be confirmed.

[0104] By implementing a further technical solution of this embodiment and introducing a stability determination condition, the following additional technical effects can be achieved:

[0105] Significantly improves the system's anti-interference capability and judgment reliability: This mechanism can effectively filter out most transient noise and isolated interference signals, ensuring that the system only triggers a response when it detects a stable or recurring abnormal pattern, greatly reducing the probability of false alarms and making the system more stable and reliable in complex industrial environments.

[0106] Achieving the best balance between sensitivity and stability: This design allows the system to maintain high sensitivity to detect weak anomalous signals (the threshold can be set quite sensitively), while avoiding the risk of misjudgment through continuous and repeatable conditions, thus achieving a unity of high sensitivity and high stability.

[0107] The accuracy and authority of control decisions are enhanced: because the judgment conditions for triggering control actions (such as retraction and chip removal) are more stringent, frequent and unnecessary processing interruptions caused by misjudgment are avoided, making each control intervention more targeted and accurate, thereby ensuring the smoothness and efficiency of the processing flow.

[0108] Furthermore, voiceprint feature parameters also include the energy ratio R.

[0109] The energy ratio R is obtained by calculating the ratio of the target frequency band signal energy to the reference frequency band signal energy.

[0110] The target frequency band has a center frequency in the range [f1, f2] and a bandwidth in the range [B1, B2]. The reference frequency band has a center frequency in the range [f3, f4] and a bandwidth in the range [B3, B4].

[0111] In S30, when the effective value of the current is continuously higher than the second threshold and the energy ratio R fluctuates in a predetermined pattern, it is determined that the inner wall scratches are caused by poor chip removal.

[0112] In this step, it should be noted that the purpose of introducing the energy ratio R is to construct a more robust damage sensitivity index.

[0113] The core idea is that the absolute value of a single spectral energy is easily affected by common factors such as system power fluctuations and sensor gain changes. However, by calculating the ratio of the energies of two different frequency bands, these common interferences can be effectively canceled out, thereby extracting signal features related to specific physical processes (such as friction at a specific frequency caused by poor chip removal) more purely.

[0114] Selection of Target and Reference Frequency Bands: The target frequency band should be selected to be sensitive to target damage (in this case, internal wall scratching). Internal wall scratching is accompanied by continuous and intense friction and chip scraping, which usually excites high-order harmonic vibrations in the system. Therefore, the target frequency band is preferably set near the 3rd to 5th harmonics (60kHz~100kHz) of the ultrasonic fundamental frequency. The reference frequency band should be selected to be insensitive to target damage but to reflect the common background noise and energy level of the system. Usually, a frequency band much higher than the main harmonic components (such as 150kHz~200kHz) is selected as the reference.

[0115] Fluctuations in the predetermined pattern: For poor chip removal, its development is a gradual process. Therefore, the typical pattern is that the energy ratio R shows a stable and continuous upward trend over several consecutive analysis windows (e.g., 5-10 windows), reflecting the increasing degree of friction and scratching. Combining this pattern with a consistently high effective current value (reflecting an increase in overall load) constitutes a strong chain of evidence for determining internal wall scratches.

[0116] By implementing a further technical solution of this embodiment and introducing an energy ratio R, the following additional technical effects can be achieved:

[0117] Significantly improves the anti-interference and specificity of damage identification. The energy ratio R, as a relative quantity, can effectively suppress the effects of overall system energy fluctuations and sensor drift, making the identification of inner wall texture more stable and significantly reducing misjudgments caused by non-damaging factors.

[0118] The system enables early and trend-based identification of the damage evolution process. By monitoring the trend change (continuous increase) of the R value, the system can capture abnormal signs in the early stages when the depth and severity of the scratches are still low and the changes in macroscopic current may not be very obvious. This allows for earlier warning and intervention, keeping the damage to a minimum.

[0119] This enhances the portability of the algorithm across different devices. Because the ratio feature weakens the dependence of the absolute value on specific device parameters, the model trained based on this feature or the set threshold has better adaptability and generalization ability on devices of different models or states.

[0120] In one optional embodiment, the construction and dynamic updating of the voiceprint feature database includes:

[0121] S11. Acquire acoustic signature parameters of the system under no-load and stable cutting conditions to generate an initial benchmark template. This step is the "cold start" of the knowledge base. By collecting a large amount of data under known good conditions (no-load, stable cutting), the statistical distribution (mean, variance) of each acoustic signature parameter (such as the effective value of current, energy ratio R, etc.) is calculated to form an initial, healthy acoustic signature template, which serves as the benchmark for subsequent discrimination.

[0122] S12. For stable cutting stage data that is subsequently deemed "qualified" in machining, an exponentially weighted moving average algorithm with a forgetting factor μ ∈ [μ1, μ2] is used to dynamically update the baseline template. This step achieves "progressive learning" of the knowledge base. For stable cutting data that is also judged as "qualified" in subsequent machining, an exponentially weighted moving average algorithm with a forgetting factor μ (e.g., 0.9) is used to update the baseline template. New template = μ × old template + (1-μ) × new data. This method allows the template to slowly track the normal gradual changes in the system (such as slight changes in the properties of the polishing fluid), while giving higher weight to recent data to maintain the timeliness of the template.

[0123] S13. During the update process, outlier data windows that deviate from the baseline template confidence interval [p1, p2] are removed to prevent them from contaminating the template data. This step is a quality control checkpoint to ensure the "purity" of the knowledge base. Before updating, it is checked whether the new data falls within the confidence interval of the current baseline template (e.g., the range of ±2.5 times the standard deviation of the mean). If there is a significant deviation, it is considered that the data window may have been affected by transient anomalies that were not determined in real time, and it is removed and not included in the template update, thereby preventing "bad data from contaminating a good model".

[0124] S14. When the cumulative working time of the tool reaches the interval [H1, H2], or the tool wear estimated by the acoustic signature enters the interval [W1, W2], a new stage template is generated and switched to maintain the accuracy of the judgment benchmark. This step solves the "lifecycle" management problem of the knowledge base. As the tool wears, the vibration characteristics of the system will drift systematically, and the old health template will no longer be applicable.

[0125] Wear can be estimated indirectly by using the cumulative working time as a direct indicator, or by analyzing the characteristics of the voiceprint itself (such as the trend of increasing energy ratio in the high-frequency band used to monitor wear).

[0126] When wear reaches a certain level (entering a preset range), it indicates that the machining state has entered a new stage. At this point, the system generates a new stage template based on the current state data and immediately switches to this template as the new judgment benchmark. This ensures that the system's definition of "health status" remains accurate throughout the entire tool life.

[0127] By implementing the technical solution of this optional embodiment and introducing a dynamic update mechanism for the voiceprint feature database, the following additional technical effects can be achieved:

[0128] It endows the system with long-term stability and adaptability: the system can automatically track normal state drift caused by slow tool wear, component aging, etc., and adjust the judgment benchmark accordingly, avoiding the problem of soaring misjudgment rate in the later stage due to benchmark solidification.

[0129] Ensuring the quality and reliability of knowledge base data: By combining dynamic updates with anomaly removal, the templates are kept up-to-date, and the damage of transient interference to long-term models is effectively prevented, so that the core decision knowledge of the system always maintains high fidelity.

[0130] State-based template maintenance has been implemented: template switching is triggered by monitoring tool status, directly linking the template lifecycle with the tool lifecycle in the physical world, making system management and maintenance more scientific and automated, and reducing reliance on external manual intervention.

[0131] In one optional embodiment, the correspondence between abnormal voiceprints and micro-damage types in the voiceprint feature database is established using an offline verification method, which includes:

[0132] Processing is performed under preset abnormal process parameters, and acoustic signature parameters are collected simultaneously. This step is to actively create faults. By intentionally setting process parameters known to cause specific types of damage (such as excessively fast feed rates to induce port star cracks, or insufficient abrasive slurry supply to induce internal wall streaks), a series of typical abnormal processing states are created in a controlled environment.

[0133] After processing, microscopic observation equipment is used to analyze the damage morphology of the workpiece port and the inner wall of the hole to determine the type and extent of microscopic damage. This is crucial for data correlation. During abnormal processing, all acoustic signature parameters are recorded synchronously and with high fidelity, capturing complete electrical signal images at the time of the anomaly. After processing, high-precision microscopic observation equipment (such as a scanning electron microscope (SEM)) is immediately used to perform non-destructive or micro-destructive testing on the workpiece to obtain clear images and quantitative data of the damage morphology. This constitutes a complete data chain of "cause-signal-result".

[0134] The collected abnormal acoustic signature parameters are correlated and mapped with the observed micro-damage types to form the judgment criteria in the acoustic signature feature library. This step is the "refinement" and "solidification" of knowledge. Through the analysis of a large amount of verification test data, repeatable regular patterns are found in the acoustic signature parameters (such as the instantaneous peak characteristics of the effective current value and the explosive growth of energy in a specific high-frequency band) when each specific type of micro-damage (such as radial cracks) occurs. These patterns are summarized and quantified into specific thresholds and logical judgment conditions, ultimately forming the judgment criteria that can be called upon in the feature library.

[0135] By implementing the technical solution of this optional embodiment and introducing a systematic offline verification method, the following additional technical effects can be achieved:

[0136] It achieves refined modeling of damage patterns, and through microscopic observation, it can accurately distinguish different types of damage (such as distinguishing the level of star cracks and the density of striations), and correlate them with subtle differences in acoustic characteristics. This enables the system not only to determine whether there is damage, but also to identify what kind of damage it is, providing the possibility for subsequent differentiated adaptive control strategies.

[0137] This provides a data foundation and iterative closed loop for continuous model optimization. The offline validation process generates a large amount of well-labeled "abnormal sample-damage result" pairing data. This high-quality data can not only be used to build an initial feature library, but also to train more complex machine learning models in the future, or to optimize and calibrate existing thresholds, forming a virtuous cycle of continuous technological progress.

[0138] Furthermore, the offline verification method also includes statistical significance evaluation and threshold tuning for the correspondence:

[0139] Based on the voiceprint feature sample sets obtained under normal and abnormal process parameters, a blind test evaluation was conducted to calculate the consistency ratio and confusion matrix of the judgment results.

[0140] When the consistency ratio reaches the preset confidence level [Q1,Q2] and the critical false positive rate in the confusion matrix is ​​lower than the preset threshold [E1,E2], the parameters of the currently used first threshold, second threshold and predetermined mode will be determined as the final judgment criteria.

[0141] If the above conditions are not met, the sampling parameters, frequency band definition, or calculation strategy of the voiceprint features are backtracked and adjusted, and the offline verification process is repeated.

[0142] In this step, it should be noted that it is not enough to simply establish a preliminary association between "voiceprint and damage". The statistical significance and discriminative power of this association must be rigorously quantitatively evaluated to ensure that the judgment criteria for the final implantation of the feature library are reliable, accurate and optimal.

[0143] Blind testing and confusion matrix: This is the gold standard for evaluating performance. The current discrimination rule (i.e., the current threshold and pattern) is tested using new samples (including normal and various abnormal states) that were not involved in model building. The confusion matrix clearly shows the detailed results of the judgments, including the number of correct judgments and the number of various false positives (e.g., classifying a good item as a bad item – a false positive, or classifying a bad item as a good item – a false negative).

[0144] Performance acceptance criteria: Two key quantitative acceptance indicators were set:

[0145] Consistency ratio: This refers to the overall accuracy rate, which is required to reach a high level (e.g., ≥95%), ensuring the basic reliability of the judgment.

[0146] Key false alarm rate: In particular, the false alarm rate is more serious than false alarms in some high-risk situations. Therefore, it needs to be strictly controlled to be below a more stringent threshold (e.g., ≤3%).

[0147] Iterative optimization loop: This is the guarantee for achieving performance targets. If the evaluation results do not meet the preset performance indicators, it indicates that the current feature extraction or discrimination logic is insufficient. The system will initiate a backtracking optimization process to guide developers to adjust the preceding parameters, such as:

[0148] Sampling parameters: Try higher sampling rates (e.g., increase from 500kHz to 1MHz) to capture richer signal details.

[0149] Frequency band definition: Re-examine and select target frequency bands and reference frequency bands that are more sensitive to target damage.

[0150] Computational strategy: Try different spectrum calculation methods or phase difference algorithms to extract more discriminative features. Then, re-execute the entire offline verification process based on the new parameters until the performance meets the requirements.

[0151] By implementing a further technical solution of this embodiment, and by introducing statistical evaluation and threshold tuning procedures, the following additional technical effects can be achieved:

[0152] Ensure that the discrimination model meets quantifiable industrial-grade reliability standards: By setting clear quantitative acceptance criteria (consistency ratio, false positive rate), the final deployed voiceprint feature library and discrimination logic are no longer empirical or vague, but have undergone rigorous verification, and their performance can be accurately measured and meet the reliability requirements of specific application scenarios.

[0153] A complete "design-verification-optimization" technology development closed loop has been constructed: this method elevates offline verification from a simple "data collection" activity into a systematic, iterative R&D process. It ensures that any performance bottlenecks can be identified in a timely manner and guided to solutions through a scientific methodology, significantly improving the efficiency and quality of technology development outcomes.

[0154] This provides a clear path for the continuous improvement of technical solutions: even if the initial version meets the requirements, the process provides a standardized operating framework for re-optimizing and re-verifying the model when dealing with new materials, new tools or higher precision requirements in the future, thus ensuring the long-term viability of the technology.

[0155] In one specific embodiment, the present invention also provides a detection system for ultrasonic-assisted machining of holes in brittle materials, applied to a cutting system including an ultrasonic tool and a rotary table. The transducer of the ultrasonic tool is connected to the power output terminal of an ultrasonic generator via a drive circuit. The detection system includes:

[0156] The signal acquisition module includes an electrical parameter sensor installed in the drive circuit for synchronously acquiring voltage and current signals.

[0157] The voiceprint processing module communicates with the signal acquisition module and is used to calculate voiceprint feature parameters based on voltage and current signals.

[0158] The state decision module has an embedded voiceprint feature library and communicates with the voiceprint processing module. It is used to compare real-time voiceprint feature parameters with the voiceprint feature library to determine whether microscopic damage has occurred and generate corresponding control commands.

[0159] The motion control module communicates with the state decision module and the rotary lifting platform, and is used to adjust the feed action of the rotary lifting platform according to control commands.

[0160] In this embodiment, it should be noted that the system, through the collaborative work of four core functional modules, materializes the software logic of the aforementioned method embodiment into a runnable hardware and software union:

[0161] Signal Acquisition Module: This module is the system's "sensory nerves." Its core is an electrical parameter sensor directly coupled to the drive circuit, responsible for acquiring, with high fidelity and zero delay, the raw voltage and current signals that best reflect the working state of the ultrasonic tool. This is the data source for all intelligent analysis and decision-making.

[0162] Voiceprint Processing Module: This module is the system's "feature extractor." It receives the raw voltage and current signals and, by executing a series of preset digital signal processing algorithms (such as FFT, RMS calculation, phase detection, etc.), calculates voiceprint feature parameters with clear physical meaning in real time, transforming the raw signal into state information that can be used for decision-making.

[0163] State Decision Module: This module is the system's "intelligent brain." It has an embedded voiceprint feature library storing various states. This module quickly matches and logically judges the real-time feature parameters sent by the voiceprint processing module with the knowledge in the feature library, and finally outputs a conclusion about the processing state (such as normal, port crack, inner wall scratch, processing penetration), and generates control instructions with clear intent based on this conclusion.

[0164] Motion Control Module: This module is the system's "executor arm." It receives control commands from the state decision module and translates them into low-level drive signals that the rotary lifting platform can recognize and execute, thereby precisely adjusting the platform's feed motion and actively intervening in the machining process.

[0165] By implementing the technical solution of this embodiment and constructing a complete perception-decision-execution closed-loop system, the following technical effects can be achieved:

[0166] The method has been engineered and automated: the detection method is encapsulated into an independent, modular system, which can be embedded as a functional unit into existing ultrasonic processing equipment, realizing the transformation from "laboratory method" to "industrial field application" and ensuring the feasibility and automation level of the method.

[0167] The responsibilities and interactions of each component of the system are clearly defined: the clear modular division simplifies the design, development, debugging, and maintenance of the system. Each module has a single function and a clear interface, which facilitates parallel development and later upgrades.

[0168] This system forms a complete control closed loop from signal perception to physical action: it is not just a detection device, but a complete control system. It realizes the connection of the entire chain from the perception of electrical signals at the microscopic state to intelligent algorithm decision-making, and finally to the action of the mechanical actuator, truly achieving state-based intelligent adaptive processing.

[0169] Furthermore, the fingerprint feature database is established and updated through an offline verification module, which includes:

[0170] The process parameter setting unit is used to set and execute abnormal processing techniques that can induce specific microscopic damage.

[0171] The microscopic observation unit is used to collect and analyze the damage morphology of the processed workpiece.

[0172] The data association unit is used to associate and match the acoustic signature data under abnormal process parameters with the damage type determined by the microscopic observation unit, and feed the matching results back to the acoustic signature feature library in the state decision module.

[0173] In this embodiment, it should be noted that the offline verification module is an auxiliary system independent of the online detection system, used to "inject knowledge" into it. Through a systematic engineering process, it provides scientifically validated and reliable judgment rules for the voiceprint feature library in the state decision module.

[0174] The process parameter setting unit allows engineers to precisely set a series of process parameters known to cause specific types of damage in a controlled environment (e.g., setting the feed rate to 150% of the normal value to systematically induce port star cracks), thereby proactively and repeatably producing abnormal processing samples for research.

[0175] The microscopic observation unit utilizes high-precision microscopic observation equipment (such as scanning electron microscopes (SEM) or laser confocal microscopes) to perform non-destructive or micro-destructive testing on the processed workpiece, obtaining qualitative and quantitative data on damage morphology. This step transforms abstract abnormal processing into specific, describable physical damage (such as confirming the existence of radial cracks with a maximum crack length of XX micrometers), providing a real and reliable label for data association.

[0176] The data association unit receives data from the first two units: "acoustic fingerprint data under abnormal processes" and "corresponding damage observation results." Through data mining and statistical analysis algorithms (such as calculating the confusion matrix, consistency ratio, and false positive rate), this unit can identify which acoustic fingerprint feature parameters, at what thresholds or modes, can most accurately and reliably distinguish different damage types. Finally, it packages the refined optimal discrimination rules (e.g., the first threshold should be set to X amperes, and the fluctuation mode of the energy ratio R should be defined as Y) and updates them to the online system's status decision module.

[0177] By implementing a further technical solution of this embodiment, and introducing a dedicated offline verification module, the following additional technical effects can be achieved:

[0178] The scientific and standardized construction of the voiceprint feature database has been achieved: the threshold setting process that relies on expert experience has been transformed into a standardized engineering process based on data-driven and physical verification, which ensures the objectivity and scientific nature of the feature database content and reduces human arbitrariness.

[0179] This forms a complete technical closed loop for the system's self-evolution. When the processing materials or tool models change, this module can be quickly recalibrated, greatly improving the system's adaptability and maintainability.

[0180] The system clearly distinguishes between the development phase and the application operation phase: the offline verification module is used for the development, debugging, and optimization phases; while the online system, consisting of four modules—signal acquisition, voiceprint processing, state decision-making, and motion control—is responsible for daily production operation. This architecture, with its clear responsibilities, facilitates the system's engineering management and lifecycle maintenance.

[0181] The embodiments described above are merely illustrative of several implementations of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the appended claims.

Claims

1. A detection method for ultrasonic assisted machining of brittle material holes, applied to a cutting system comprising an ultrasonic tool and a rotary lifting platform, the transducer of the ultrasonic tool being connected to the power output end of an ultrasonic generator through a drive circuit, characterized in that, An electric parameter sensor for collecting voltage signal and current signal is arranged in the drive circuit, and the method comprises: S10, synchronously collecting voltage signal and current signal of the drive circuit through the electric parameter sensor; S20, calculating voiceprint feature parameters based on the voltage signal and current signal, the voiceprint feature parameters comprising current effective value and spectrum energy of specific frequency band; S30, comparing the real-time monitored voiceprint feature parameters with a pre-established voiceprint feature library to determine whether micro-damage occurs, wherein determining micro-damage state comprises: when the current effective value appears transient sharp peak exceeding a first threshold value, determining that the sharp peak is caused by port star crack due to inlet impact; when the current effective value continuously exceeds a second threshold value and the spectrum energy of the specific frequency band appears fluctuation of predetermined mode, and the state continuously lasts for a preset duration or appears for a preset continuous number of times in continuous multiple analysis windows, determining that the fluctuation is caused by internal wall pull due to poor chip removal; S40, if yes, adjusting the feeding action of the rotary lifting platform.

2. The detection method for brittle material hole type ultrasonic assisted machining according to claim 1, characterized in that, The real-time calculated voiceprint feature parameters in S20 further comprise equivalent impedance amplitude; Correspondingly, determining micro-damage state in S30 further comprises: when the equivalent impedance amplitude sharply decreases from load state to no-load state, determining that the decrease is caused by machining penetration.

3. The detection method for brittle material hole type ultrasonic assisted machining according to claim 2, characterized in that, The calculation of voiceprint feature parameters based on the voltage signal and current signal in S20 comprises: S21, calculating effective value of the voltage signal and current signal; S22, calculating phase difference between voltage and current, the calculation method comprising zero-crossing detection method or fast Fourier transform method; S23, calculating equivalent impedance amplitude based on voltage effective value and current effective value; S24, performing fast Fourier transform on the current signal to extract spectrum energy of the specific frequency band.

4. The method for detection of brittle material hole type ultrasonic assisted machining according to claim 1, characterized in that, The voiceprint feature parameters further comprise energy ratio R; The energy ratio R is obtained by calculating ratio of target frequency band signal energy and reference frequency band signal energy; wherein, center frequency of the target frequency band is located in [f1, f2] interval, and bandwidth is located in [B1, B2] interval; center frequency of the reference frequency band is located in [f3, f4] interval, and bandwidth is located in [B3, B4] interval; In S30, when the current effective value continuously exceeds the second threshold value and the energy ratio R appears fluctuation of predetermined mode, determining that the fluctuation is caused by internal wall pull due to poor chip removal.

5. The method for detection of brittle material hole type ultrasonic assisted machining according to claim 1, characterized in that, Construction and dynamic update of the voiceprint feature library comprises: S11, collecting voiceprint feature parameters of the system in no-load state and stable cutting state to generate initial reference template; S12, for subsequent stable cutting stage data of qualified machining, using exponential weighted moving average algorithm with forgetting factor μ∈[μ1, μ2] to dynamically update the reference template; S13, in the update process, removing abnormal data window deviating from the reference template confidence interval [p1, p2] to prevent pollution of template data; S14, when the cumulative working time of the tool reaches the interval [H1, H2] or the tool wear estimated by the voiceprint feature enters the interval [W1, W2], a new stage template is generated and switched to, so as to maintain the accuracy of the discrimination reference.

6. The method for detection of brittle material hole type ultrasonic assisted machining according to claim 1, characterized in that, The correspondence between the abnormal voiceprint in the voiceprint feature library and the micro-damage type is established by an offline verification method, and the offline verification method comprises: Processing under a preset abnormal process parameter, and synchronously collecting voiceprint feature parameters; After processing, the damage morphology of the workpiece port and the inner wall of the hole is analyzed by using a microscopic observation device to determine the type and degree of micro-damage; The collected abnormal voiceprint feature parameters are associated and mapped with the micro-damage type determined by observation to form the determination basis in the voiceprint feature library.

7. The detection method for brittle material hole type ultrasonic assisted machining according to claim 6, characterized in that, The offline verification method further comprises statistical significance evaluation and discrimination threshold setting of the correspondence: Based on the voiceprint feature sample sets obtained under normal and abnormal process parameters respectively, blind test evaluation is performed, and the consistency ratio and confusion matrix of the determination result are calculated; When the consistency ratio reaches the preset confidence level [Q1, Q2] and the key misjudgment rate in the confusion matrix is lower than the preset threshold [E1, E2], the first threshold, the second threshold and the parameters of the predetermined mode currently used are determined as the final discrimination basis; If the above conditions are not met, the sampling parameters, frequency band definition or calculation strategy of the voiceprint feature are adjusted, and the offline verification process is repeatedly executed.

8. A detection system for ultrasonic assisted machining of brittle material holes, applied to the detection method for ultrasonic assisted machining of brittle material holes according to any one of claims 1-7, applied to a cutting system comprising an ultrasonic tool and a rotating lifting table, the transducer of the ultrasonic tool being connected to the power output of an ultrasonic generator through a drive circuit, characterized in that, The detection system comprises: A signal acquisition module comprising an electrical parameter sensor arranged in the drive circuit for synchronously collecting voltage signals and current signals; A voiceprint processing module in communication connection with the signal acquisition module for calculating voiceprint feature parameters based on the voltage signals and current signals; A state decision module embedded with a voiceprint feature library and in communication connection with the voiceprint processing module for comparing real-time voiceprint feature parameters with the voiceprint feature library to determine whether micro-damage occurs and generating corresponding control instructions; A motion control module in communication connection with the state decision module and the rotary lifting platform for adjusting the feeding action of the rotary lifting platform according to the control instructions.

9. The detection system for ultrasonic assisted machining of brittle material orifice according to claim 8, characterized in that, The voiceprint feature library is established and updated by an offline verification module, and the offline verification module comprises: A process parameter setting unit for setting and executing abnormal machining processes that can induce specific micro-damage; A microscopic observation unit for collecting and analyzing the damage morphology of the machined workpiece; A data association unit for associating and matching the voiceprint feature data under abnormal process parameters with the damage type determined by the microscopic observation unit, and feeding back the matching result to the voiceprint feature library in the state decision module.

Images