Method and system for monitoring internal damage of transmission part of polishing device and terminal equipment

Abstract

The invention provides a polishing device transmission part internal damage monitoring method and system and terminal equipment, and relates to the field of transmission part internal damage monitoring. The method for monitoring the internal damage of the transmission part of the polishing device comprises the following steps of: 1, receiving an audio signal on a to-be-detected transmission part to obtain an original audio sequence; 2, dividing the original audio sequence into a plurality of audio frames to obtain an audio frame sequence; 3, negation is carried out on the amplitude in the audio frame to generate a reverse frame, and a reverse frame sequence is obtained; 4, overlapping each audio frame in the audio frame sequence with the previous reverse frame in the reverse frame sequence to generate fusion frames, and splicing all the fusion frames to generate a flaw detection audio sequence; and step 5, monitoring the flaw detection audio sequence in real time, and giving out early warning when the audio characteristics representing that the acoustic emission signals are generated on the transmission piece to be detected are extracted. According to the technical scheme provided by the invention, the internal dark damage and crack of the transmission part can be monitored in real time, and early warning can be given out in time.

Description

Technical Field

[0001] The present invention relates to the field of sound analysis technology or other fields, and more specifically, to a method, system, and terminal device for monitoring internal damage of transmission components of a polishing device. Background Art

[0002] As an important part of precision machining equipment, the polishing device is widely used in the surface treatment of fields such as optical elements and semiconductor devices. Its power transmission system consists of core transmission components such as a drive motor, a reduction gearbox, a grinding disc and supporting gears, a transmission shaft, and bearings. The dynamic stability of the transmission components directly determines the rotational speed control accuracy of the grinding disc, which in turn affects the polishing quality. Although assembly and debugging can compensate for some machining errors, under the action of long-term alternating loads, the transmission components are prone to internal stress concentration defects (such as internal cracks, hidden injuries, grain boundary corrosion, etc.) due to factors such as metal fatigue and insufficient lubrication. Such defects initially appear in millimeters or even sub-millimeters. If not detected in time, they will gradually cause structural failures such as transmission shaft bending and gear tooth breakage, resulting in an exponential decay of polishing accuracy and seriously affecting the polishing quality and production efficiency.

[0003] Currently, there are mainly the following three detection means for the damage of transmission components, but each has certain limitations:

[0004] Surface detection method: The surface state of the transmission component is evaluated through visual inspection or contact measurement, and macroscopic defects such as tooth surface wear and shaft deformation can be identified. However, this method is difficult to detect internal micro-cracks, fatigue fractures and other hidden damages, and frequent shutdown and disassembly for detection seriously affect the equipment utilization rate.

[0005] Ultrasonic flaw detection technology: The internal defects are located by using ultrasonic echo signals, but there are the following defects: (a) Strong dependence on positioning: It is necessary to pre-estimate the defect area and perform multi-angle scanning, and the single detection takes a long time. (b) Lack of real-time performance: Online monitoring cannot be implemented under the operating state of the equipment, and shutdown detection results in production capacity loss. (c) Poor anti-interference ability: The working noise of the polishing device will cover up the weak echo signal, and the false alarm rate is relatively high.

[0006] Indirect parameter monitoring method: The transmission state is inferred by analyzing indirect parameters such as motor current and temperature. However, such methods are extremely insensitive to early micro-damages and can only trigger an alarm when the failure deteriorates to the later stage (such as transmission shaft fracture), and preventive maintenance cannot be achieved.

[0007] Therefore, the existing detection means for the damage of transmission components are difficult to meet the core requirements of online monitoring ability, micro-defect sensitivity, and anti-interference performance. Summary of the Invention

[0008] The object of the present invention is to provide a method, system and terminal device for monitoring internal damage of transmission parts of a polishing device in view of the deficiencies or defects of the prior art. Through a high-sensitivity broadband acoustic sensor and an adaptive inverse frame superposition noise reduction algorithm, real-time positioning and quantitative evaluation of internal defects of transmission parts are realized, breaking through the detection blind area of the prior art and ensuring the stability of polishing accuracy.

[0009] To achieve the above object, in the first aspect of the present invention, a method for monitoring internal damage of transmission parts of a polishing device is provided, including:

[0010] Step 1: Receive an audio signal on a transmission part to be measured to obtain an original audio sequence;

[0011] Step 2: Divide the original audio sequence into a plurality of audio frames to obtain an audio frame sequence;

[0012] Step 3: Invert the amplitude in the audio frame to generate an inverse frame to obtain an inverse frame sequence;

[0013] Step 4: Superimpose each audio frame in the audio frame sequence with the previous inverse frame in the inverse frame sequence to generate a fusion frame, and splice all the fusion frames to generate a flaw detection audio sequence;

[0014] Step 5: Monitor the flaw detection audio sequence in real time, and issue a warning when an audio feature indicating the generation of an acoustic emission signal on the transmission part to be measured is extracted.

[0015] Through the above technical solution, an audio signal generated during the operation of the transmission part to be measured is collected to form an original audio sequence; then, an audio feature indicating an acoustic emission event is extracted from the sequence. By capturing the transient elastic wave generated by the rapid energy release when the internal stress of the transmission part to be measured exceeds the yield limit, real-time monitoring of internal hidden damages and cracks of the transmission part is realized, and a warning is issued in a timely manner, effectively ensuring the stability of polishing processing accuracy.

[0016] To accurately extract the acoustic emission signal, in this solution, the audio frame is superimposed with its inverse frame. By the principle of destructive interference, the background noise generated when the transmission part to be measured is operating smoothly is effectively suppressed. This processing method improves the signal-to-noise ratio of the acoustic emission signal, enabling the effective acoustic emission features indicating internal damage of the transmission part to be clearly extracted.

[0017] Further, the length of the audio frame is greater than 200 microseconds.

[0018] In the technical solution of the present application, by setting the length of the audio frame to be greater than 200 microseconds, the probability that the acoustic emission signal is located in two audio frames is reduced. This setting can ensure that in any audio frame and its inverse frame, the acoustic emission signal exists only in one of the frames. In this way, after the audio frame and its inverse frame are superimposed and processed, complete acoustic emission information can be obtained.

[0019] When the polishing device runs smoothly, although the audio signals generated by the vibration of each transmission part remain relatively stable as a whole, due to the inherent dynamic characteristics of the mechanical system, the received audio signals will still have fluctuations. When the current audio frame is superimposed with the previous reverse frame, it is not necessarily possible to eliminate the noise generated when the transmission part vibrates. In view of the above technical problems, this application proposes an improved technical solution:

[0020] Furthermore, the length of the audio frame is an integer multiple of the period of the original audio sequence.

[0021] In this technical solution, by optimizing the length setting of the audio frame and configuring it to be an integer multiple of the period of the original audio sequence when the polishing device runs smoothly, it is ensured that each audio frame contains a complete periodic signal. When the audio frame is superimposed with its reverse frame, in the case of no new sound source interference inside the polishing device, the reverse frame and the audio frame have the characteristics of opposite amplitudes and the same phase, so that the new sound source feature information inside the transmission part to be measured can be extracted more accurately.

[0022] In practical applications, the original audio sequence usually contains signals of multiple frequency components, and different frequency components have different periodic characteristics. Therefore, directly calculating the overall period of the original audio sequence cannot accurately describe the variation characteristics of each frequency component, which will lead to inaccurate matching when the subsequent audio frame is superimposed with the reverse frame, thus introducing additional noise.

[0023] Furthermore, extract the periods of the signals of each frequency component in the original audio sequence, and the length of the audio frame is an integer multiple of the least common multiple of the periods of the signals of each frequency component in each original audio sequence.

[0024] When the audio frame is superimposed with the reverse frame, since the length of the audio frame matches the periodic characteristics of each frequency component, it can cause the audio information representing the same event to interact with each other, so that these audio information can cancel each other out, and finally accurately capture the acoustic emission signal generated by the transmission part to be measured.

[0025] However, during the superimposing process, due to the unpredictability of the background noise, it may cause the period and phase of the original audio sequence to shift. Although the superimposition of the audio frame and the reverse frame can effectively eliminate the background noise generated during the operation of the transmission part to be measured, it is still possible to introduce new noise signals due to the randomness of the noise, thus affecting the monitoring accuracy of the acoustic emission signal.

[0026] Furthermore, extract the frequency component information in the audio frame and the reverse frame, and superimpose the frequency information of the audio frame and the reverse frame with each other.

[0027] In the technical solution provided by this application, the frequency components in the audio frame and the reverse frame are superimposed in sequence. When the audio frame and the reverse frame are decomposed in frequency, the noise signal can be isolated, thereby increasing the matching accuracy between the audio frame and the reverse frame and reducing the noise signal in the fusion frame to accurately capture the acoustic emission signal.

[0028] Aiming at the problem that the acoustic emission signal attenuates during transmission due to the large volume of the transmission part, making it difficult to accurately capture the acoustic emission signal, this application provides the following technical solution:

[0029] Furthermore, at least the audio signals at two positions on the transmission part to be measured are received.

[0030] In this technical solution, the audio signals at at least two positions on the transmission part to be measured are received, so that the audio signals can be received at different positions, effectively avoiding the problem of signal attenuation caused by the long-distance propagation of the acoustic emission signal in the transmission part.

[0031] Aiming at the problem that when the acoustic emission signal is located at the junction of two audio frames, it may be scattered at both ends of the fusion frame after fusion processing, thus affecting the recognition of the acoustic emission signal, this application provides the following technical solution:

[0032] Furthermore, when dividing the audio frame, a partial overlap is set between adjacent audio frames.

[0033] In this technical solution, by setting the overlapping area between adjacent audio frames, it is ensured that the acoustic emission signal located at the junction of two audio frames can be completely represented in at least one audio frame. In this way, the problem of inaccurate recognition caused by the scattering of the acoustic emission signal at both ends of the fusion frame can be effectively avoided.

[0034] Furthermore, step 5 includes the following steps:

[0035] Collect the flaw detection audio sequence in real time. When the amplitude of the flaw detection audio sequence exceeds the preset amplitude threshold, extract the audio information with a preset length before and after the amplitude exceeding the amplitude threshold to obtain the monitoring signal;

[0036] Extract the audio features from the monitoring signal;

[0037] Input the audio features into the pre-trained BP network model to generate the damage type corresponding to the monitoring signal.

[0038] The second aspect of the present invention provides an internal damage monitoring system for the transmission part of a polishing device, including:

[0039] An acoustic signal receiver for receiving audio signals;

[0040] An information processing device is configured to receive an audio signal sent by the sound signal receiver and execute the foregoing internal damage monitoring method for the transmission member of the polishing device.

[0041] The third aspect of the present invention provides a terminal device, including:

[0042] One or more processors;

[0043] A memory for storing one or more computer programs,

[0044] The processor executes the computer program to implement the steps of the foregoing internal damage monitoring method for the transmission member of the polishing device.

[0045] As the fourth aspect of the present application, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the processor is caused to execute the foregoing internal damage monitoring method for the transmission member of the polishing device.

[0046] The fifth aspect of the present invention provides a computer program product, including a computer program, and when the computer program is executed by a processor, the foregoing internal damage monitoring method for the transmission member of the polishing device is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Through the following description of the embodiments of the present invention with reference to the drawings, the above content and other objects, features and advantages of the present invention will become clearer. In the drawings:

[0048] Figure 1 It is a flowchart of the internal damage monitoring method for the transmission member of the polishing device in Embodiment 1;

[0049] Figure 2 It is a schematic diagram of generating a fusion frame from the original audio sequence in Embodiment 1;

[0050] Figure 3 It is a flowchart of generating a fusion frame in Embodiment 1;

[0051] Figure 4 It is a flowchart of Step 5 in Embodiment 1;

[0052] Figure 5 It is a schematic structural diagram of the BP neural network in Embodiment 4;

[0053] Figure 6 It is a schematic structural diagram of the polishing device monitoring system in Embodiment 5;

[0054] Figure 7 It is a structural block diagram of the terminal device in Embodiment 6; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In the following detailed description, for the sake of explanation, numerous specific details are set forth to provide a comprehensive understanding of the embodiments of the present invention. However, it is obvious that one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0056] An acoustic emission signal refers to the acoustic wave signal corresponding to the transient elastic wave generated due to the release of energy from a defect source when the internal stress of a material exceeds the yield limit. The detection of this signal usually indicates that the material may exhibit phenomena such as cracks, bending, or fracture.

[0057] The core of the technical solution of this application lies in: real-time collecting and monitoring the audio signals generated during the operation of the transmission components of the polishing device. By extracting the acoustic emission signal characteristics from the audio signals, the occurrence of potential failures can be prevented.

[0058] In this application, the transmission components mainly refer to the power transmission components of the polishing device, which generally include the following three categories:

[0059] 1. Shaft components: Monitor whether there is a risk of fatigue fracture;

[0060] 2. Gear assemblies: Monitor whether there are damages such as cracks;

[0061] 3. Sealed bearings: Monitor whether their sealing performance fails or whether there is a leakage phenomenon.

[0062] By monitoring the defects of the above key components, the working state of the transmission components can be effectively monitored, and early warning of failures and preventive maintenance can be achieved.

[0063] Embodiment 1:

[0064] Figure 1 Shows a flowchart of the method for monitoring internal damage of the transmission components of the polishing device of the present invention. As Figure 1 shown, the method for monitoring internal damage of the transmission components of the polishing device includes the following steps:

[0065] Step 1: Receive the audio signal on the transmission component to be measured to obtain the original audio sequence.

[0066] Arrange acoustic wave sensors on the transmission component for collecting the audio signals generated by the transmission component. Since the acoustic emission signal has a wide frequency distribution characteristic, a broadband acoustic wave sensor is selected as the detection element, and the effective working frequency range of this sensor is from 100 kHz to 1000 kHz.

[0067] Further, at least audio signals at two positions on the transmission part to be measured are received.

[0068] In practical applications, acoustic wave sensors need to be respectively arranged at two different positions of the transmission part to be measured to collect the audio signals at the corresponding positions. The audio signals of each transmission part will be monitored for acoustic emission signals separately. When at least one sensor detects an acoustic emission signal, it can be preliminarily determined that there may be defects such as cracks, deformations, and twists inside the transmission part.

[0069] It should be noted that in the arrangement of acoustic wave sensors, the corresponding installation positions need to be selected according to the structural characteristics of different types of transmission parts. Specifically:

[0070] 1. For shaft transmission parts, acoustic wave sensors should be symmetrically arranged at both ends of the shaft;

[0071] 2. For gear transmission parts, a circumferential array arrangement method is adopted, that is, two or more acoustic wave sensors are evenly distributed along the circumferential direction on the side of the gear, and each sensor is equidistant from the center of the gear and the spacing is equal;

[0072] 3. For sealed bearings, two or more acoustic wave sensors are symmetrically arranged on the outer peripheral side wall.

[0073] The above sensor arrangement scheme can more accurately capture the characteristics of acoustic emission signals on transmission parts at different positions. Taking the gear as an example, the circumferential array arrangement method can ensure a stable signal acquisition state during the rotation of the gear, thereby improving the reliability of the monitoring results. Step 2: Divide the original audio sequence into several audio frames to obtain an audio frame sequence.

[0074] During the process of the polishing device performing fine polishing operations, since the contact state between the grinding disc and the surface of the workpiece to be processed is stable, when the equipment is running smoothly, the audio signals generated by each transmission component due to vibration are relatively constant. Therefore, the audio data collected by the acoustic wave sensors shows obvious periodic characteristics. When there is no abnormality in the transmission system (that is, no acoustic emission signal is generated), the adjacent signal periods recorded continuously usually remain the same.

[0075] Based on the above characteristics, when dividing the audio frames in this application, the length of the audio frame is set to an integer multiple of the period of the original audio sequence. Under the stable operation state of the polishing device, by taking the inverse of the amplitude of the subsequent audio frame and then superimposing it with the current audio frame, the background noise signals generated during the normal operation of the equipment can be effectively eliminated, thereby enhancing the distinguishability of the target acoustic emission signal.

[0076] Since the period of the original audio sequence is relatively chaotic and it is difficult to accurately find the period of the original audio sequence, for this reason, this application provides the following technical solutions:

[0077] Extract the periods of the signal components of each frequency in the original audio sequence. The length of the audio frame is an integer multiple of the least common multiple of the periods of the signal components of each frequency in each original audio sequence. In this way, within one audio frame, the signals of all different frequency components can complete a full cycle change, facilitating subsequent superposition operations.

[0078] In addition, to prevent the acoustic emission signal from spanning two audio frames and causing the superimposed acoustic emission signal to be scattered at both ends of the fusion frame, the present application sets the audio frame length to be greater than 200 microseconds because the time duration of the acoustic emission signal is usually less than 20 microseconds. This can ensure that the acoustic emission signal is completely within one audio frame in the vast majority of cases.

[0079] Furthermore, when dividing the audio frames, there is a partial overlap between adjacent audio frames.

[0080] For example, the overlap ratio between adjacent audio frames is 50%. In this way, even if the acoustic emission signal is exactly at the boundary position between two audio frames, the signal can be completely captured through the overlapping area, thus avoiding the problem that the acoustic emission signal cannot be recognized due to being scattered at both ends of the fusion frame.

[0081] To accurately extract the period characteristics of the signal components of each frequency, multiple schemes can be adopted, such as Fourier transform, wavelet transform, etc. In this scheme, it is preferably to adopt the method for extracting the periods of each frequency described in Embodiment 2.

[0082] Step 3: Invert the amplitude of the audio frame to generate a reverse frame, obtaining a reverse frame sequence.

[0083] For each audio frame, its length corresponds to several complete change cycles in the original audio sequence. Therefore, the reverse frame obtained by inverting the amplitude of the audio frame is an audio signal with the completely opposite amplitude but the same phase as the original audio frame.

[0084] As Figure 2 and Figure 3 shown: Step 4: Superimpose each audio frame in the audio frame sequence with the previous reverse frame in the reverse frame sequence to generate a fusion frame, and splice all the fusion frames to generate a flaw detection audio sequence.

[0085] When the polishing device is in a stable operating state and the transmission part to be measured does not generate an acoustic emission signal, the audio signals of two adjacent audio frames are basically the same. Therefore, after performing the superimposition operation of the current audio frame with the reverse frame of the previous audio frame, the amplitudes of the key frequencies will cancel each other out, and most of the audio information in the finally generated fusion frame is eliminated, leaving only a small amount of disordered background noise.

[0086] However, if the transmission component to be measured generates an acoustic emission signal, the fusion frame can highlight this sound feature and more accurately capture the acoustic emission signal in the audio data.

[0087] In order to optimize the superposition effect and eliminate the sound signal generated by the vibration of the transmission component during the operation of the polishing device, the present application provides the following technical solutions:

[0088] Extract the frequency component information in the audio frame and the reverse frame, and superimpose the frequency information of the audio frame and the reverse frame on each other. Specifically, step 4 includes the following steps:

[0089] Step 41: Divide the audio frame into several frequency components;

[0090] Step 42: Divide the reverse frame into several frequency components.

[0091] In step 41 and step 42, it is necessary to decompose the audio signal into multiple frequency components, that is, to separate and process the characteristic signals of different frequencies in the audio signal. In practice, processing methods such as Fourier transform and wavelet transform can be used. The present application preferably adopts the frequency information extraction method provided in Embodiment 3, which can decompose the audio signal into signals of individual frequencies, and then obtain different frequency components of the audio signal.

[0092] Specifically, based on the technical solution of Embodiment 3, the audio frame and the reverse frame can be respectively decomposed into several frequency components to obtain the time-domain information at the corresponding frequencies.

[0093] Step 43: Superimpose the audio frame and the reverse frame on each other at each frequency component to obtain each frequency signal frame.

[0094] The mutual superposition in this solution means: align the audio frame and the reverse frame with each other in the time domain and intersect the amplitudes. Because the amplitudes of the reverse frame and the audio frame are opposite and the phases are the same, basically only the noise signal remains in each frequency signal frame.

[0095] Step 44: Mutually fuse each frequency signal frame to obtain a fusion frame.

[0096] After obtaining the time-domain signals of each frequency of the fusion frame, the inverse Fourier transform can be used to mutually fuse each frequency signal frame, thereby obtaining a fusion frame. This fusion process is a process of fusing frequency signals of different components into one signal. In this solution, the inverse operation of the Fourier transform is used to fuse the signals.

[0097] In this technical solution, the superposition operation of the audio frame and the reverse frame is performed separately for each frequency component. After the superposition process is completed for each frequency component, the results are fused to generate the final fused frame. In this superposition method, the combination accuracy of the audio frame and the reverse frame is higher, and it can more effectively eliminate the background audio information unrelated to the acoustic emission signal.

[0098] Step 5: Monitor the flaw detection audio sequence in real time, and issue a warning when the audio features characterizing the acoustic emission signal generated on the transmission part to be measured are extracted.

[0099] As Figure 4 shown, Step 5 includes the following steps:

[0100] Step 51: Collect the flaw detection audio sequence in real time. When the amplitude of the flaw detection audio sequence exceeds the preset amplitude threshold, extract the audio information with a preset length before and after the amplitude exceeding the amplitude threshold to obtain the monitoring signal.

[0101] To optimize the utilization efficiency of the system computing resources, this solution pre-configures the amplitude threshold parameter. When it is detected that the amplitude of a certain point in the audio sequence exceeds the preset threshold, the system determines that there may be abnormal sound information. At this time, collect the audio data with a preset length before and after this signal point as the monitoring information, and analyze these monitoring signals to determine whether they are acoustic emission signals.

[0102] Step 52: Extract the audio features from the monitoring signal;

[0103] Step 53: Input the audio features into the pre-trained BP network model to generate the damage type corresponding to the monitoring signal.

[0104] In practical applications, the damage degree of the transmission part is positively correlated with the intensity of the acoustic emission signal generated by it. For this reason, the audio features are input into the BP neural network model for processing. This model can judge the damage type according to parameters such as the energy and amplitude of the audio features.

[0105] Embodiment 2:

[0106] Embodiment 2 provides a method for extracting the period of each frequency on the basis of Embodiment 1. In Embodiment 1, when segmenting the audio frame, it is necessary to extract the period of the signal of each frequency component in the original audio sequence. In order to accurately extract the period of the signal of each frequency component, the present invention provides the following implementation manners:

[0107] S1: Sample the original audio sequence, and then divide it into multiple short frames. The time-domain signal of the m-th frame is;

[0108] x m[n] = x[n + nH], where n = 0, 1, 2, …… N - 1, N is the number of N sampling points included in each short frame, H is the number of sampling points of the interval between the starting points of adjacent short frames, H = N / 2, m represents the frame number, starting from 0, n + mH represents the sampling point index of the starting position of the m-th frame in the original signal, and x[n] represents the discrete sampling sequence of the original audio sequence.

[0109] S2: Multiply each short frame by the window function w[n] to obtain a supplemented frame

[0110]

[0111] where w[n] is a Hamming window;

[0112]

[0113] In this solution, through the windowing operation, the two ends of the short frame are smoothly transitioned to zero, avoiding spectral leakage caused by truncation.

[0114] S3: Map the time-domain signal to frequency and extract the frequency components.

[0115]

[0116] where X m [k] represents the complex frequency-domain representation of the m-th short frame, and k represents the frequency index;

[0117] S4: Calculate the energy intensity |X m [k]| of the frequency at the k-th frequency point, obtain the amplitude spectrum of the frequency at the k-th frequency point, and extract the peak point k p ;

[0118]

[0119] S5: Convert the peak point k p to the actual frequency f k , calculate the period T k of the frequency f k ;

[0120]

[0121] where F s is the number of sampling points of the original audio sequence.

[0122] Furthermore, in order to overcome the limitation of frequency resolution and improve the accuracy of period calculation, in S5, for each peak point k p , using the adjacent points k p -1 and k pCalculate the amplitude calculation difference offset δ of +1;

[0123] Then f k =(k p +δ)(F s / N).

[0124] Through the above technical solution, frequency components with sufficient amplitude energy and their corresponding periods can be extracted from the original audio sequence, and then the least common multiple of the periods of all frequency signals can be calculated. Then, the least common multiple is magnified by an integer multiple and its length is made greater than 200 microseconds.

[0125] When the true frequency of the signal does not fall on the discrete frequency bins, its energy will spread to adjacent bins, resulting in peak detection deviation (i.e., resolution limitation). To solve this problem, this solution realizes the accurate estimation of the frequency position by using the amplitude information of the peak point and its adjacent bins. Embodiment 3 is a frequency information extraction method further optimized on the basis of Embodiment 2 and Embodiment 1. Similar to Embodiment 2, Embodiment 3 also uses the Fourier transform to process the audio signal. Embodiment 2 is mainly used to calculate the least common multiple of each frequency component, while Embodiment 3 is to extract the specific signals of each frequency component from the audio signal. For this reason, this application proposes a frequency component extraction method, including the following steps: SO1: Sample the original audio sequence and then divide it into multiple short frames. The time-domain signal of the m-th frame is expressed as:

[0126] x m [n]=x[n + nH], n = 0, 1, 2,..., N - 1, where N is the N sampling points included in each short frame, H: the number of sampling points of the interval between the starting points of adjacent short frames, H = N / 2, m represents the frame number, starting from 0, n + mH represents the sampling point index of the starting position of the m-th frame in the original signal, and x[n] represents the discrete sampling sequence of the original audio sequence.

[0127] SO2: Multiply each short frame by the window function w[n] to obtain the supplemented frame

[0128]

[0129] where w[n] is a Hamming window;

[0130]

[0131] In this solution, through the windowing operation, the two ends of the short frame are smoothly transitioned to zero, avoiding spectrum leakage caused by truncation.

[0132] SO3: Map the time-domain signal to frequency and extract the frequency components.

[0133]

[0134] Among them, X m [k] represents the complex frequency-domain representation of the m-th short frame, and k represents the frequency index;

[0135] SO4: Calculate the energy intensity |X m [k]| of the frequency at the k-th frequency point, obtain the amplitude spectrum of the frequency at the k-th frequency point, and extract the peak point k from the amplitude spectrum p ;

[0136]

[0137] ReX m [k] represents the real part of X m [k], ImX m [k] represents the imaginary part of X m [k];

[0138] SO5: Based on the peak point K p Set the frequency band [f low , f high , f low is the upper limit of the frequency band, f high is the lower limit of the frequency band, f low = K p - f 0 , f high = K p + f 0 , f 0 is the set frequency resolution value;

[0139] SO4: Construct the frequency-domain mask M[k];

[0140] When f low ≤ k ≤ f high , M[k] = 1, otherwise M[k] = 0;

[0141] SO5: Retain the target frequency components, filter out other components, and obtain the frequency information

[0142]

[0143] SO6: Inverse Fourier transform the frequency information to recover the time-domain signal.

[0144] In this solution, by introducing the frequency-domain mask M[k], it is possible to effectively filter out the components that do not belong to the specific frequency band [f low , f highThe information. Based on this technical means, the finally restored time-domain signal only contains the data within this frequency band. In this way, the time-domain information of each frequency component can be extracted from the original audio sequence, and the audio frames and reverse frames can be superimposed on each frequency component.

[0145] Embodiment 4:

[0146] In Embodiment 1, in order to accurately identify the acoustic emission signal, it is necessary to extract the audio features from the monitoring signal. If the monitoring signal is directly input into the BP neural network model, the interference of a large number of noise signals will reduce the learning ability of the model, and then cause the technical problem that the model only performs well on the training data set, but has insufficient recognition accuracy in the actual monitoring scenario. To solve the above problems, Embodiment 4 of the present invention provides a method for extracting audio features, which specifically includes the following steps:

[0147] Step 1: Use the Hilbert transform to calculate the local frequency f(t) of the current residual r k-1 ;

[0148]

[0149] f(t) represents the loss frequency, describing the local frequency at time t, r k-1 represents the residual signal of the k-1 layer, H{r q-1 (t)} represents the Hilbert transform, arg represents the argument of the complex number, represents the derivative with respect to time;

[0150] Step 2: Design the cut-off frequency f c (t) of the filter according to the local frequency f(t);

[0151] f c (t) = α·f(t), α ∈ (0,1)

[0152] α is the attenuation coefficient, controlling the cut-off frequency of the filter;

[0153] Step 3: Generate the filter H(t,f) based on the cut-off frequency f c (t)

[0154]

[0155] σ is the filter bandwidth, controlling the frequency selectivity of the filter, and H(t,f) is used to describe the gain of the frequency f at time t;

[0156] Step 4: Filter the monitoring signal S(t) with the filter H(t,f) to obtain the low-frequency component L(t);

[0157] L(t) = H(t,f) * S(t);

[0158] Step Five: Subtract the low-frequency component from the monitoring signal to obtain the high-frequency residual r(t);

[0159] r(t) = S(t) - L(t);

[0160] Step Six: Check whether the residual r(t) satisfies the two conditions of IMF: If it satisfies, then IMF q (t) = h q (t), where IMF q (t) represents the q-th intrinsic mode function, and update the residual IMF q (t) = h q (t), where q represents the index of the intrinsic mode function;

[0161] If it does not satisfy, then take h q (t) as the new residual and the new monitoring signal, and repeat Steps Four to Five.

[0162] Condition 1: The difference between the number of extreme points and the number of zero-crossing points does not exceed 1.

[0163] Condition 2: The local mean of the upper and lower envelopes is zero.

[0164] Stop decomposition when any of the following indicators is satisfied:

[0165] Indicator 1: The number of IMFs reaches the preset value.

[0166] Indicator 2: When the energy of the current residual r q (t) is less than the energy threshold with respect to the total energy of the original signal S(t), then stop decomposition.

[0167] Step Four: Repeat Steps Two to Three to obtain the IMF signals, and get the audio feature IMF:

[0168]

[0169] In the technical solution provided by the present application, the extraction of audio features uses the Empirical Mode Decomposition (EMD) algorithm to decompose the monitoring signal, generating several Intrinsic Mode Functions (IMFs). This method can accurately capture the characteristic information of different frequency parts in the monitoring signal. Since the acoustic emission signals generated inside the metal have a wide range of acoustic wave characteristics, and the characteristic positions corresponding to different damage types are located at different frequencies, after decomposing the monitoring signal into multiple IMF features, it is possible to effectively analyze the characteristic information in different frequency bands.

[0170] However, traditional IMF feature decomposition algorithms have certain limitations when dealing with acoustic emission signals. Due to the short time period, strong instantaneousness, and large frequency mutation ability of acoustic emission signals, traditional methods are difficult to accurately decompose them into different frequency components. This insufficient decomposition can lead to susceptibility to noise interference in subsequent analysis, thus affecting the accuracy of the analysis results.

[0171] To solve the above problems, this solution proposes an improved method, that is, to adjust the cut-off frequency in real time according to the instantaneous cut-off frequency of the signal and only retain the current dominant frequency component. This improvement significantly reduces the modal mixing interference (i.e., the "mode aliasing" phenomenon), thus better adapting to the decomposition of non-stationary signals. In this way, the effective feature information in the acoustic emission signal can be extracted more accurately, improving the accuracy and reliability of the overall analysis.

[0172] After obtaining the audio feature IMF, it is necessary to use a BP neural network model for training to judge the corresponding damage type. For each different type of transmission part to be measured, its corresponding BP neural network model needs to be trained separately. However, since they are all used to identify the acoustic emission signals generated inside metal parts due to the material yield exceeding the critical point, the structures of each BP neural network model are the same. Based on this, this embodiment only provides a training method for the BP neural network model of a transmission part.

[0173] Reference Figure 5 , the BP neural network model includes an input layer, a hidden layer, and a pooling layer output layer:

[0174] Input layer: used to receive audio features. The input layer contains multiple input neurons with the same number as the number of independent components (IMF) in the audio features, and each independent component is input into one input neuron respectively.

[0175] Hidden layer: transforms the input audio features through a non-linear activation function to extract key feature information.

[0176] Pooling layer, downsamples the key features, reduces the amount of data to be processed, and speeds up the training speed of the network.

[0177] Output layer: regresses the extracted key features to the corresponding labels and outputs the prediction results.

[0178] During the training process, the BP neural network model needs to use training samples containing audio features and their corresponding labels. The types of labels include the following:

[0179] Fracture: refers to the rupture phenomenon of the transmission part caused by stress yield, thus triggering an acoustic emission signal.

[0180] Twist: It refers to the situation where the transmission component cannot recover after being subjected to torsional force, resulting in the rupture of the metal material and the generation of acoustic emission signals.

[0181] Tear: It refers to the acoustic emission signal of internal tearing in the transmission component when it bears external forces such as torsional force and tension.

[0182] Noise: It refers to non-target acoustic emission signals that are not caused by the damage of the transmission component.

[0183] In practical applications, by inputting training samples into the BP neural network model, the model can learn the correlation between audio features and labels. After training, inputting new audio features into the model can automatically determine the fault type.

[0184] Therefore, in practice, when the number of faults of the transmission component exceeds the preset threshold, the system can trigger an alarm mechanism; or only record the number of faults of the transmission component for the reference of maintenance staff.

[0185] For this reason, the BP neural network model can accurately determine the fault type of the transmission component based on the characteristics of the acoustic emission signal and provide reliable support for equipment maintenance.

[0186] Embodiment 5

[0187] Reference Figure 6 , the internal damage system of the transmission component of the polishing device includes:

[0188] An acoustic signal receiver for receiving audio signals;

[0189] An information processing device configured to receive the audio signals sent by the acoustic signal receiver and execute the method for monitoring the internal damage of the transmission component of the polishing device described in Embodiment 1.

[0190] The information processing device includes a preamplifier, a signal processor, a processor, and a display. Among them, the preamplifier is used to amplify the acoustic emission signal. The frequency range of the preamplifier is 20k~1200kHz.

[0191] The signal processor is used to convert the received sound signal into a digital signal, and the signal processor is an A / D converter. The processor is used to execute the method for monitoring the internal damage of the transmission component of the polishing device described in Embodiment 1. The display is used to display the final result generated by the processor.

[0192] Reference Figure 7 , Embodiment 6: FIG. 7 schematically shows a block diagram of a terminal device suitable for implementing the method for monitoring the internal damage of the transmission component of the polishing device according to an embodiment of the present invention.

[0193] As Figure 7As shown, the terminal device according to an embodiment of the present invention includes a processor, which performs various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage section into a random access memory (RAM).

[0194] The processor includes a general microprocessor, an instruction set processor, a related chipset, and a dedicated microprocessor.

[0195] In the RAM, various programs and data required for the operation of the terminal device are stored. The processor, ROM, and RAM are interconnected via a bus.

[0196] The processor executes various operations of the method flow according to an embodiment of the present invention by executing the program in the ROM and / or RAM.

[0197] It should be noted that the program can also be stored in one or more memories other than the ROM and RAM.

[0198] The terminal device further includes an (I / O) interface, and the (I / O) interface is connected to the bus. The terminal device further includes one or more of the following components connected to the I / O interface: an input section such as a keyboard and a mouse; an output section including a display and a speaker, etc.; a communication section including a network interface card such as a LAN card and a modem.

[0199] The processor can also execute the internal damage monitoring method of the transmission member of the polishing device according to Embodiments 1-4 of the present invention by executing the program stored in one or more memories.

[0200] Embodiment 7:

[0201] A computer-readable storage medium, which can be included in the polishing device monitoring system described in Embodiment 5 or the terminal device described in Embodiment 6 above, or can exist separately without being assembled into the device / device / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the internal damage monitoring method of the transmission member of the polishing device according to Embodiment 1 of the present invention is implemented.

[0202] Embodiment 8:

[0203] A computer program product, which includes a computer program, and the computer program contains program codes for executing the internal damage monitoring method of the transmission member of the polishing device described in Embodiment 1.

[0204] Embodiments of the present invention have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although the embodiments have been described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. The scope of the present invention is defined by the appended claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art can make various substitutions and modifications, and these substitutions and modifications should fall within the scope of the present invention.

Claims

1. A method for monitoring internal damage of a transmission part of a polishing device, characterized in that: The steps include: Step 1: receiving an audio signal from a transmission component to be tested to obtain an original audio sequence; Step 2: Divide the original audio sequence into a number of audio frames to obtain an audio frame sequence; Step 3: Invert the amplitude in the audio frame to generate an inverse frame, thereby obtaining an inverse frame sequence; Step 4: superimpose each audio frame in the audio frame sequence with the previous reverse frame in the reverse frame sequence to generate a fused frame, and splice all the fused frames to generate a flaw detection audio sequence; Step 5: Monitor the flaw detection audio sequence in real time and issue an early warning when the audio feature representing the acoustic emission signal generated on the transmission part to be tested is extracted.

2. The method according to claim 1, characterized in that The length of the audio frame is greater than 200 microseconds.

3. The method according to claim 1, characterized in that The length of an audio frame is an integer multiple of the period of the original audio sequence.

4. The method according to claim 3, characterized in that The period of each frequency component signal in the original audio sequence is extracted, and the length of the audio frame is an integer multiple of the least common multiple of the period of each frequency component signal in the original audio sequence.

5. The method according to claim 4, characterized in that The frequency component information in the audio frame and the reverse frame is extracted, and the frequency information in the audio frame and the reverse frame are superimposed on each other.

6. The method according to claim 1, characterized in that At least two audio signals are received at the positions on the transmission member to be tested.

7. The method according to claim 2, characterized in that: When dividing the audio frames, it is set that adjacent audio frames have a partial overlap.

8. The method according to claim 1, characterized in that Step 5 includes the following steps: Collect the flaw detection audio sequence in real time. When the amplitude of the flaw detection audio sequence exceeds a preset amplitude threshold, extract the audio information of a preset length before and after the amplitude exceeds the amplitude threshold to obtain a monitoring signal. Extracting audio features from monitoring signals; The audio features are input into the pre-trained BP network model to generate the damage type corresponding to the monitoring signal.

9. A polishing device transmission part internal damage monitoring system, characterized in that: include: An acoustic signal receiver, used for receiving an audio signal; An information processing device is configured to receive the audio signal sent by the acoustic signal receiver and execute the method according to any one of claims 1 to 8.

10. A terminal device, comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the processor implements the steps of the method according to any one of claims 1 to 8 when executing the computer program.

Images