An online nondestructive monitoring method for micro-grinding based on acoustic emission technology
Through dual-channel acoustic emission technology and time-frequency domain analysis, combined with the Hidden Markov model, high sensitivity monitoring of tool wear and surface roughness in fine grinding is achieved, and the monitoring accuracy and reaction speed in the existing technology are solved. Problem, the online non-destructive monitoring capability of fine grinding processing has been improved.
Patent Information
- Application Number
- CN202211354219.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-11-01
- Publication Date
- 2024-11-29
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art cannot effectively realize online non-destructive monitoring in fine grinding processing, especially in the real-time monitoring of the wear degree of fine grinding tools and the surface roughness of the processed parts.
A fine grinding online lossless monitoring system based on acoustic emission technology adopts a dual-channel connection method. The signal is collected through the first acoustic emission sensor and the second acoustic emission sensor, and wear identification is performed in combination with time-frequency domain EMD processing and hidden Markov model. It is judged that kurtosis and skewness are used as time domain eigenvalues for monitoring.
It realizes accurate identification of the wear degree of fine grinding tools, improves monitoring sensitivity and accuracy, and provides wear status feedback quickly and timely, avoids the problem of insignificant changes in force and vibration indicators, and improves computing efficiency.
Smart Images

Figure CN115586262B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of online nondestructive monitoring of precision micro-grinding, and in particular to an online nondestructive monitoring method of micro-grinding based on acoustic emission technology. Background Art
[0002] In the current way, in the fields of aerospace, mechanical electronics, optics and optoelectronics, "micro grinding" is being used more and more because of its important application value and broad application prospects;
[0003] Specifically, in order to achieve nanometer-level surface roughness and micrometer-level surface accuracy in "micro-grinding", the conventional technical means of the prior art are: online non-destructive monitoring is required during the micro-grinding process to prevent the surface quality of the parts from being affected due to severe wear of the micro-grinding tool;
[0004] To explain the defects of the prior art in more detail, the conventional technical means of the prior art have the following constraints when achieving monitoring: (1) Since the micro-grinding processing area is often in a closed state, equipment such as laser vibrometers cannot directly measure the processing position; (2) The grinding force value in micro-grinding is relatively small, and it is often difficult to monitor the grinding force signal through commonly used force sensors; (3) Non-in-situ detection methods require repeated disassembly of parts, which cannot achieve rapid and timely information feedback.
[0005] In addition, acoustic emission signals are highly resistant to interference and easy to collect, and the frequency of the collected acoustic emission signals is much higher than the natural frequency of the equipment itself, so they are less affected by noise; at the same time, acoustic emission signals are extremely sensitive and can provide real-time feedback to the system; however, even in the existing technical means involving acoustic emission monitoring in the prior art, there are the following constraints: for example (1.1) the commonly used time domain characteristic values such as root mean square RMS and amplitude A have the disadvantages of low sensitivity and insensitivity to pulse events that may occur during the cycle; (1.2) commonly used frequency domain analysis methods such as FFT have the disadvantages of not including the time of event occurrence and being unable to describe the transient signal characteristics in the frequency domain.
[0006] In summary, the urgent problem to be solved is to provide an online non-destructive monitoring system and method for micro-grinding that can ensure sensitivity based on the existing acoustic emission monitoring means. Summary of the invention
[0007] The purpose of the present invention is to solve the technical problem that the wear degree of micro-grinding tools and the surface roughness of processed parts in the existing precision micro-grinding process cannot be effectively monitored online without damage, thereby providing an online non-destructive monitoring method for micro-grinding based on acoustic emission technology.
[0008] In order to solve the above technical problems, the technical solutions of the present invention are as follows:
[0009] An online nondestructive monitoring method for micro-grinding based on acoustic emission technology, comprising:
[0010] Step 1: Construct an online nondestructive monitoring system for micro-grinding based on acoustic emission technology with dual-channel connection;
[0011] The dual-channel connection method is:
[0012] A first acoustic emission sensor 11 is mounted on the shaft side of the micro-grinding tool 2 to obtain a first acoustic emission signal 100 generated by the micro-grinding tool 2 and the workpiece shaft during the micro-grinding process;
[0013] The first acoustic emission signal is transmitted to the signal acquisition card 3 through the first preamplifier 111, and finally transmitted to the host computer system 4;
[0014] The second acoustic emission sensor 22 is placed at the center of the back side of the machining surface of the workpiece 1;
[0015] Generate a second acoustic emission signal 200 and transmit it to the signal acquisition card 3 and then to the host computer system 4 through the second preamplifier 222;
[0016] The host computer system 4 is used to process the second acoustic emission signal through time-frequency domain EMD, and perform judgment operation through a Markov model module to perform wear identification judgment;
[0017] Step 2: Obtain a first acoustic emission signal 100 through a first acoustic emission sensor 11, and perform time domain processing on the first acoustic emission signal through a time domain processing module 10, and perform time domain recognition and judgment, and the method of the time domain recognition and judgment is:
[0018] A judgment is made based on whether the time domain characteristic value reaches a specified threshold. When the time domain judgment result is that the threshold is not reached, the time domain processing module 10 makes another judgment, and the process repeats itself.
[0019] A judgment is made based on whether the time domain characteristic value reaches a specified threshold, and when the time domain judgment result is reached, the next step is executed;
[0020] Step 3: This step is the next step, and when the time domain judgment result is reached, a second acoustic emission sensor 22 works and obtains the second acoustic emission signal 200;
[0021] Performing EMD calculation on the second acoustic emission signal 200 through an EMD calculation module 20 and obtaining a calculation result;
[0022] The calculation result is used through a hidden Markov model module to perform wear identification and judgment on the acoustic emission signal with the second acoustic emission signal 200 as the calculation result;
[0023] The wear identification judgment executes a preset sequence, which is:
[0024] The first determination is to determine whether the micro-grinding tool has reached slight wear. If the result does not reach slight wear, the hidden Markov model module is returned to perform wear identification and judgment on the subsequently received second acoustic emission signal 200 as the acoustic emission signal of the calculation result;
[0025] When the result reaches slight wear, the upper computer system 4 issues a slight wear prompt;
[0026] Determine again whether the micro-grinding tool has reached normal wear. If the result reaches normal wear, the host computer system will issue a normal wear prompt;
[0027] Three judgments are made to determine whether the micro-grinding tool has reached serious wear. If the result reaches serious wear, the host computer system will issue a serious wear prompt;
[0028] When the second determination or the third determination determines that the wear index has been reached, the processing method is the same as the first determination.
[0029] Preferably, the sampling frequency of the acoustic emission signal is 2 MHz;
[0030] The sampling frequency should be greater than or equal to twice the highest frequency in the analog signal spectrum.
[0031] Preferably, the kurtosis (K) and skewness (S) of the time domain processing module 10 are used as the time domain eigenvalues, and the calculation method thereof is:
[0032] Kurtosis K, calculated as:
[0033]
[0034] Among them, n k is the number of samples in the kurtosis calculation, k2 is the unbiased estimate of the second-order cumulant, i.e., the sample variance, S k is the kurtosis correction coefficient. In this method, S k Take 1.2;
[0035] Skewness (S), calculated as:
[0036]
[0037] Among them, n s is the number of samples in the skewness calculation, k2 is the symmetric unbiased estimate of the second-order cumulant, Ss is the skewness correction coefficient. In this method, S s Take 1.4.
[0038] Preferably, when judging based on the time domain eigenvalue and determining whether the eigenvalue reaches a specified threshold, the threshold determination is performed based on the obtained time domain eigenvalue kurtosis (K) and skewness (S). If the threshold is reached, it is a prerequisite for automatically triggering step three.
[0039] Preferably, the calculation method performed by the EMD calculation module 20 is performed in an optimized calculation method, specifically:
[0040] The second sound wave signal is defined as signal I(ω);
[0041] First, the signal I(ω) is subjected to a short-time Fourier transform, and the signal I(ω) can be decomposed in frequency into:
[0042]
[0043] Among them, the noise in the acoustic emission signal is mainly high-frequency noise;
[0044] The high frequency noise is filtered, and in I(ω), I(ω) is decomposed by EMD as follows:
[0045]
[0046] Wherein, I(ω) represents the input signal, IMF i (ω) represents the eigenmode function, Res n (ω) represents the residual;
[0047] The power of the high frequency noise in the signal is calculated according to the following formula:
[0048]
[0049] Then judge the high-frequency noise signal power intensity according to the following formula:
[0050] SNR = 101g (P s / P n )
[0051] Among them, when SNR≤25, it proves that the high-frequency noise power intensity is sufficient. Remove IMF1 in the EMD decomposition and define the new signal as I'(ω), that is, the signal after the first filtering. Then calculate the high-frequency noise power and SNR of I'(ω) until the new signal I n The SNR in (ω) is ≥ 25.
[0052] Preferably, the acoustic emission signal of the calculation result is used for wear identification and judgment in the following manner:
[0053] The acoustic emission signal is modeled based on the hidden Markov model;
[0054] The IMF obtained after EMD decomposition is regarded as the observed state, and the actual wear state of the micro-grinding tool is regarded as the hidden state. The chaos matrix in this process is experimentally verified as follows:
[0055]
[0056] Preferably, the parameters in the hidden Markov model are continuously updated using the Nelder-Mead algorithm in combination with the given chaotic matrix and initial parameters, so as to achieve convergence to the local minimum in a short time without obtaining the derivative function of the multivariate function in advance.
[0057] The present invention has the following beneficial effects:
[0058] Firstly, the present invention uses acoustic emission signals as monitoring signals, which can avoid the problem that relevant indicators such as force and vibration do not change significantly during micro-grinding.
[0059] Secondly, the present invention adopts kurtosis and skewness as technical indicators and compares them with other time domain characteristic values, which has higher sensitivity; compared with the existing technology, the advantage is that precision micro-grinding, which has higher requirements on processing accuracy and tool size than traditional grinding, has more research value.
[0060] Thirdly, the present invention monitors the micro-grinding process by combining the time domain analysis method and the time-frequency domain analysis method. Compared with the prior art, it can not only ensure sufficient monitoring precision and accuracy, but also greatly improve the computing efficiency and avoid waste of resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.
[0062] Figure 1 is a system diagram of the present invention;
[0063] Figure 2 It is a structural block diagram of the present invention;
[0064] Figure 3 The figure is a flowchart of the method of the present invention. DETAILED DESCRIPTION
[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention; it should be noted that in this application, for the convenience of description, the "left side" in the current view is the "first end", the "right side" is the "second end", the "upper side" is the "first end", and the "lower side" is the "second end". The purpose of such description is to clearly express the technical solution, and it should not be understood as an improper limitation on the technical solution of this application.
[0066] The purpose of the present invention is to solve the technical problem that the wear degree of micro-grinding tools and the surface roughness of processed parts in the existing precision micro-grinding process cannot be effectively monitored online without damage, thereby providing a micro-grinding online non-destructive monitoring system and method based on acoustic emission technology.
[0067] See also Figure 3 , and combined with the attached Figure 1 , Attachment Figure 2 As shown, the technical solution is an online nondestructive monitoring method for micro-grinding based on acoustic emission technology, comprising:
[0068] Step 1: Construct an online nondestructive monitoring system for micro-grinding based on acoustic emission technology with dual-channel connection;
[0069] The dual channel connection method is:
[0070] A first acoustic emission sensor 11 is mounted on the shaft side of the micro-grinding tool 2 to obtain a first acoustic emission signal 100 generated by the micro-grinding tool 2 and the workpiece shaft during the micro-grinding process;
[0071] The first acoustic emission signal is transmitted to the signal acquisition card 3 through the first preamplifier 111, and finally transmitted to the host computer system 4;
[0072] The second acoustic emission sensor 22 is placed at the center of the back side of the machining surface of the workpiece 1;
[0073] Generate a second acoustic emission signal 200 and transmit it to the signal acquisition card 3 and then to the host computer system 4 through the second preamplifier 222;
[0074] The host computer system 4 is used to process the second acoustic emission signal through the time-frequency domain EMD, and perform judgment operation through a Markov model module to perform wear identification judgment;
[0075] The method first constructs a detection system, that is, the dual-channel connection method mentioned in this application, and connects the attached Figure 3The first channel in this application specifically refers to:
[0076] The first acoustic emission sensor 11 is placed on the shaft side of the micro-grinding tool, and is used to receive the acoustic emission signal (the first acoustic emission signal 100) through the micro-grinding tool and the shaft. The acoustic emission signal collected by the first acoustic emission sensor 11 enters the signal acquisition card 3 through the first preamplifier 111, and is finally transmitted to the host computer system 4.
[0077] The specific working principle and purpose is that the first acoustic emission sensor 11 is used as a rough measurement part; the acoustic emission signal collected by the first acoustic emission sensor 11 is processed in the time domain, and when the time domain characteristic value of the signal reaches a certain proportion, the acoustic emission signal (second acoustic emission signal 200) collected by the second acoustic emission sensor 22 is used for further online monitoring;
[0078] The acoustic emission signal collected by the first acoustic emission sensor 11 will be affected by the noise when the main shaft of the workpiece 1 rotates, but contains more abundant information such as the rotation state of the main shaft of the workpiece 1. S21 is marked as the first processing stage mark S21.
[0079] More specifically, in the second channel of the present application, the second acoustic emission sensor 22 is placed at the center of the back side of the workpiece 1, and is not affected by the grinding chips and other substances, and is used to directly receive the acoustic emission signal from the surface of the workpiece;
[0080] The acoustic emission signal collected by the second acoustic emission sensor 22 (i.e., the second acoustic emission signal 200) enters the signal acquisition card 3 via the second preamplifier 222, and is finally transmitted to the host computer system 4; the acoustic emission signal collected by the second acoustic emission sensor 22 is close to the processing position and is not interfered by other noises.
[0081] The second acoustic emission sensor 22 is used as the precision measurement part; when the specified threshold is reached, the acoustic emission signal collected by the second acoustic emission sensor 22 is processed by time-frequency domain EMD, and the hidden Markov model is used for machine learning, specifically for wear identification and judgment.
[0082] In summary, the use of a dual-channel signal acquisition card can ensure the synchronous acquisition of two signals.
[0083] In addition, as attached Figure 2As shown, the overall system of the method includes an acoustic emission signal acquisition module for acquiring acoustic emission signals collected by a signal acquisition card, a gain adjustment amplifying circuit module, an acoustic emission signal acquisition module, a kurtosis and skewness acquisition module for acquiring the kurtosis (K) and skewness (S) of the acoustic emission signal, a module for setting the comparison relationship between the wear condition of a micro-grinding tool and the kurtosis (K) and skewness (S), an acoustic emission signal EMD acquisition module for acquiring EMD processing of the acoustic emission signal, and a Markov model module for performing machine learning on the acoustic emission signal using a Markov model.
[0084] In an implementable manner, it is more preferred that the acoustic emission signal acquisition module is the W500 acoustic emission sensor of Qingcheng Acoustic Emission Research Co., Ltd. The frequency range of the acoustic emission sensor is 100-1000KHz, the resonance frequency is 500kHz, and the sensitivity peak is >65dB.
[0085] In an implementable manner, it is more preferred that the gain adjustment amplifier circuit module is a PAS preamplifier of Qingcheng Acoustic Emission Research Co., Ltd., with a gain of 40dB and a filter range of 20kHz-1200kHz.
[0086] In an practicable manner, the acoustic emission signal acquisition module is a PCI-1714U synchronous analog input card from ADVANTECH, which uses a 4-channel single-head analog input and a 12-bit converter.
[0087] For further information, please see the attached Figure 3 In the specific mark as the second part, the specific content is:
[0088] Step 2: Obtain a first acoustic emission signal 100 through a first acoustic emission sensor 11, and perform time domain processing on the first acoustic emission signal through a time domain processing module 10, and perform time domain recognition and judgment. The method of time domain recognition and judgment is:
[0089] It is judged whether the time domain characteristic value reaches the specified threshold. If the time domain judgment result is that it does not reach the threshold, the time domain processing module 10 makes another judgment, and so on.
[0090] It is judged whether the time domain characteristic value reaches the specified threshold. If the time domain judgment result is reached, the next step is executed;
[0091] Step 3: This step is the next step, and when the time domain judgment result is reached, a second acoustic emission sensor 22 works and obtains a second acoustic emission signal 200;
[0092] Performing EMD calculation on the second acoustic emission signal 200 through an EMD calculation module 20 and obtaining a calculation result;
[0093] The calculation result is used to perform wear identification and judgment on the acoustic emission signal with the second acoustic emission signal 200 as the calculation result through a hidden Markov model module;
[0094] Wear identification and judgment execute a preset sequence, which is:
[0095] The first determination is to determine whether the micro-grinding tool has reached slight wear. If the result does not reach slight wear, the hidden Markov model module is returned to perform wear identification and judgment on the acoustic emission signal of the subsequently received second acoustic emission signal 200 as the calculation result;
[0096] When the result reaches slight wear, the upper computer system 4 issues a slight wear prompt;
[0097] Determine again whether the micro-grinding tool has reached normal wear. If the result reaches normal wear, the host computer system will issue a normal wear prompt;
[0098] Three judgments are made to determine whether the micro-grinding tool has reached serious wear. If the result reaches serious wear, the host computer system will issue a serious wear prompt;
[0099] When the second or third determination is that the wear index has been reached, the processing is the same as the first determination.
[0100] Specifically, in the process of precision micro-grinding, different wear states include slight wear, normal wear, severe wear and other different stages; generally, the more severe the wear of the micro-grinding tool, the smaller the number of effective abrasive particles on the surface, and the fewer peaks of the acoustic emission signal in the time domain. This proves that there is a certain qualitative relationship between the characteristic value of the acoustic emission signal after a certain time domain processing and the wear state of the tool. However, it is restricted by many conditions such as processing and monitoring conditions. Simple analytical means cannot meet the requirements of accurate monitoring in precision micro-grinding, but it can still be used as a simple threshold monitoring method to provide rapid response while saving computing resources.
[0101] In a specific implementation, the sampling frequency of the acoustic emission signal is 2 MHz, and the sampling frequency should be greater than or equal to twice the highest frequency in the spectrum of the analog signal.
[0102] According to the Nyquist sampling theorem, in order to restore the analog signal without distortion, the sampling frequency should be greater than or equal to twice the highest frequency in the analog signal spectrum, that is:
[0103] f s ≥2f max
[0104] It is determined that the sampling frequency of the acoustic emission signal should be higher than 2MHz. In actual operation, it is found that a sampling frequency lower than 2MHz will result in too little information contained in the collected acoustic emission signal, and it is impossible to obtain accurate information. But at the same time, the higher the sampling frequency, the better. If calculated according to the sampling frequency of 2.5MHz, the default data width is 12 bits, and the total amount of dual-channel data used in this method is 60MB. This order of magnitude of data cannot complete real-time calculation and feedback. Combined with the above analysis, it is finally determined that the sampling frequency of the acoustic emission signal is 2MHz.
[0105] To explain in detail, specifically, the processing of acoustic emission signals includes time domain analysis, frequency domain analysis and time-frequency domain analysis; the three methods have their own advantages and disadvantages. The time domain analysis method has the fastest response speed, but the accuracy is low; the frequency domain analysis method is more accurate, but it cannot reflect time information in the analysis; the time-frequency domain analysis method is the most accurate, but the response speed is the slowest. In the second processing stage marked S12, the time domain analysis method is used to quickly process the acoustic emission signal. In other acoustic emission time domain monitoring, the root mean square voltage value RMS is usually used as the time domain characteristic value. The voltage value RMS has the advantages of wide use and simple calculation. However, the biggest disadvantage of the voltage value RMS is that its sensitivity is extremely low, and this disadvantage will be used in micro-grinding. Therefore, in the second processing stage marked S12, the kurtosis (K) and skewness (S) with extremely strong sensitivity are used as time domain characteristic values;
[0106] The time domain processing module 10 uses the kurtosis (K) and skewness (S) as time domain feature values, and the calculation method is as follows:
[0107] Kurtosis K, calculated as:
[0108]
[0109] Among them, n k is the number of samples in the kurtosis calculation, k2 is the unbiased estimate of the second-order cumulant, i.e., the sample variance, S k is the kurtosis correction coefficient. In this method, S k Take 1.2;
[0110] Skewness (S), calculated as:
[0111]
[0112] Among them, n s is the number of samples in the skewness calculation, k2 is the symmetric unbiased estimate of the second-order cumulant, S s is the skewness correction coefficient. In this method, S s Take 1.4.
[0113] Specifically, when the third processing stage is marked S13 and the fourth processing stage is marked S14, the time domain eigenvalue is judged to determine whether the eigenvalue reaches the specified threshold value, and the kurtosis (K) and skewness (S) of the time domain eigenvalue are used as the basis for threshold determination. If the threshold value is reached, it is automatically triggered to enter the attached process. Figure 3 The second part shown in FIG. That is, when judging based on the time domain eigenvalue and determining whether the eigenvalue reaches the specified threshold, the threshold is judged based on the kurtosis K and skewness S of the obtained time domain eigenvalue. If the threshold is reached, it is a precondition for automatically triggering step three.
[0114] Specifically, in the preferred method, 1.2 times of the peak value when the micro-grinding tool reaches severe wear is used as the threshold. This threshold can avoid the situation where the monitoring is not accurate enough due to the lack of accuracy of the time domain analysis method itself, and can save time and optimize the amount of calculation. In the experiment, it was tested that when this value is used as the threshold, when the number of sampling points of the signal is 20,000, the calculation time can be reduced by 0.25-0.4s in each cycle.
[0115] In the fifth processing mark S21, it is reflected that after the second part is automatically triggered, the acoustic emission signal is acquired from the acoustic emission sensor 2, and the acoustic emission signal is collected by the acoustic emission sensor 2 arranged at the center position of the back side of the workpiece.
[0116] Specifically, arranging the second acoustic emission sensor 22 at the center of the back side of the workpiece can avoid the influence of grinding chips and other external substances, while ensuring that the received acoustic emission signal is less affected by noise.
[0117] The reason why the two acoustic emission sensors are arranged separately in this method is, firstly, to avoid interference between the two acoustic emission sensors; secondly, since in the actual monitoring process, the first part often occupies a higher proportion of time, the precision micro-grinding system often fails at this stage. Therefore, arranging the acoustic emission sensor 1 at the shaft end can monitor the changes in the acoustic emission signal of the spindle while monitoring the roughness of the micro-grinding tool. Once a peak value or deviation with a large difference appears, the entire system should be stopped immediately to avoid damage to the spindle. The second acoustic emission sensor 22 is arranged at the center of the back of the workpiece for processing, which can also complete the monitoring of the surface roughness of the workpiece.
[0118] In the sixth processing stage S22, the acoustic emission signal received from the second acoustic emission sensor 22 is analyzed by EMD time-frequency domain to calculate the acoustic emission signal. Specifically, EMD is a time-frequency domain that decomposes signals based on the time scale characteristics of the data itself. The biggest advantage is that there is no need to pre-set basis functions, which can greatly save calculation time.
[0119] In other words, this method also proposes an optimized form for the EMD algorithm to solve the problem that the acoustic emission signal is easily contaminated by noise during the process of acoustic emission signal collection.
[0120] The calculation method performed by the EMD calculation module 20 is performed in an optimized calculation method, specifically:
[0121] The second sound wave signal is defined as signal I(ω);
[0122] First, the signal I(ω) is subjected to short-time Fourier transform, and the signal I(ω) can be decomposed in frequency as follows:
[0123]
[0124] Among them, the noise in the acoustic emission signal is mainly high-frequency noise;
[0125] For filtering high-frequency noise, in I(ω), the EMD decomposition of I(ω) is:
[0126]
[0127] Where I(ω) represents the input signal, IMF i (ω) represents the eigenmode function, Res n (ω) represents the residual;
[0128] The power of high-frequency noise in the signal is calculated according to the following formula:
[0129]
[0130] Then judge the high-frequency noise signal power intensity according to the following formula:
[0131] SNR = 101g (P s / P n )
[0132] Among them, when SNR≤25, it proves that the high-frequency noise power intensity is sufficient. Remove IMF1 in the EMD decomposition and define the new signal as I'(ω), that is, the signal after the first filtering. Then calculate the high-frequency noise power and SNR of I'(ω) until the new signal I n The SNR in (ω) is ≥ 25.
[0133] The seventh processing stage is marked S23: using the hidden Markov model to identify the acoustic emission signal, and using the hidden Markov model as a basis to model the acoustic emission signal.
[0134] However, the current hidden Markov model has the following problems: First, the large size of the matrix in the model can easily lead to a decrease in overall computational efficiency. Second, the parameter configuration in the model should change with the change of the matrix.
[0135] Preferably, in this method, in order to solve the problem 1, the IMF obtained after EMD decomposition in the sixth processing stage S22 is first regarded as an observed state, and the actual wear state of the micro-grinding tool is regarded as a hidden state, and the chaotic matrix in this process is experimentally verified to be as follows:
[0136] The method of using the acoustic emission signal of the calculation result to identify wear is as follows:
[0137] The acoustic emission signal is modeled based on the hidden Markov model;
[0138] The IMF obtained after EMD decomposition is regarded as the observed state, and the actual wear state of the micro-grinding tool is regarded as the hidden state. The chaos matrix in this process is experimentally verified as follows:
[0139]
[0140] In addition, the parameters in the hidden Markov model are continuously updated by combining the given chaotic matrix and initial parameters using the Nelder-Mead algorithm, so that the local minimum can be converged in a short time without obtaining the derivative function of the multivariate function in advance.
[0141] First, determine the initial point ω1, and based on it, generate n alternative points as the initial point, namely ω2...ω n+1 , and calculate the average point ω of the replacement points m And ω n+1 About Omega m The symmetric point ω r ,in
[0142]
[0143] ω r =2ω m -ω n+1
[0144] A judgment is then required.
[0145] If I(ω1)≤I(ω n+1 )<I(ω n ), then enter the next cycle.
[0146] If I(ω n+1 )<I(ω1), then the expansion point ω needs to be calculated s ,in
[0147] ω s =3ω m -2ω n+1
[0148] If I(ω s )<I(ω n+1 ), then let ω n+1 =ω s , and enter the next cycle, otherwise let ω n+1 =ω r , and enter the next cycle.
[0149] If I(ω n )≤I(ω r )<I(ω n+1 ), then there is a midpoint in
[0150]
[0151] if make And enter the next loop, otherwise execute the last step.
[0152] If I(ω n+1 )<I(ω r ), then there is an intermediate point in
[0153]
[0154] if make And enter the next loop, otherwise execute the last step.
[0155] make
[0156]
[0157] And enter the next cycle.
[0158] The eighth processing stage is marked S24, the ninth processing stage is marked S25, and the tenth processing stage is marked S26: determine whether the micro-grinding tool has reached the corresponding degree of wear. The host computer system issues a slight wear prompt, the host computer system issues a normal wear prompt, the host computer system issues a severe wear prompt and shuts down.
[0159] Specifically, based on the judgment results, the host computer system will send a prompt and, if necessary, shut down the machine to protect the processing system. In the case of slight wear and normal wear, the host computer system will issue a prompt, and the user will selectively proceed to the next step based on their own processing requirements and actual working conditions.
[0160] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.
Claims
1. A method for online nondestructive monitoring of micro grinding based on acoustic emission technology, characterized in that: include: Step 1: Construct an online nondestructive monitoring system for micro-grinding based on acoustic emission technology with dual-channel connection; The dual-channel connection method is: A first acoustic emission sensor (11) is mounted and clamped on the shaft side of the micro-grinding tool (2) to obtain a first acoustic emission signal (100) generated by the micro-grinding tool (2) and the workpiece shaft during a micro-grinding process; The first acoustic emission signal is transmitted to the signal acquisition card (3) through a first preamplifier (111), and finally transmitted to a host computer system (4); The second acoustic emission sensor (22) is placed at the center of the back side of the machining surface of the workpiece (1); a second acoustic emission signal (200) is generated and transmitted to the signal acquisition card (3) and then to the host computer system (4) through a second preamplifier (222); The host computer system (4) is used to process the second acoustic emission signal through time-frequency domain EMD, and perform judgment operation through a Markov model module to perform wear identification judgment; Step 2: acquiring a first acoustic emission signal (100) through a first acoustic emission sensor (11), and performing time domain processing on the first acoustic emission signal through a time domain processing module (10), and performing time domain identification and judgment, wherein the time domain identification and judgment is performed in the following manner: A judgment is made based on whether the time domain characteristic value reaches a specified threshold. When the time domain judgment result is that the threshold is not reached, the time domain processing module (10) makes another judgment, and the process repeats itself; A judgment is made based on whether the time domain characteristic value reaches a specified threshold, and when the time domain judgment result is reached, the next step is executed; Step 3: This step is the next step, and when the time domain judgment result is reached, a second acoustic emission sensor (22) works and obtains the second acoustic emission signal (200); Performing EMD calculation on the second acoustic emission signal (200) through an EMD calculation module (20) and obtaining a calculation result; The calculation result is used through a hidden Markov model module to perform wear identification and judgment on the acoustic emission signal with the second acoustic emission signal (200) as the calculation result; The wear identification judgment executes a preset sequence, which is: Firstly, determine whether the micro-grinding tool has reached slight wear. If the result shows that the micro-grinding tool has not reached slight wear, return to the hidden Markov model module to perform wear identification and judgment on the subsequently received second acoustic emission signal (200) which is the acoustic emission signal of the calculation result; When the result reaches slight wear, the host computer system (4) issues a slight wear prompt; Determine again whether the micro-grinding tool has reached normal wear. If the result reaches normal wear, the host computer system will issue a normal wear prompt; Three judgments are made to determine whether the micro-grinding tool has reached serious wear. If the result reaches serious wear, the host computer system will issue a serious wear prompt; When the second determination or the third determination determines that the wear index has been reached, the processing method is the same as the first determination; The kurtosis K and skewness S of the time domain processing module (10) are used as the time domain characteristic values, and the calculation method is: Kurtosis K, calculated as: Among them, n k is the number of samples in the kurtosis calculation, k2 is the unbiased estimate of the second-order cumulant, that is, the sample variance, S k is the kurtosis correction coefficient. In this method, S k Take 1.2; Skewness S, calculated as: Among them, n s is the number of samples in the skewness calculation, k2 is the symmetric unbiased estimate of the second-order cumulant, S s is the skewness correction coefficient. In this method, S s Take 1.4; The calculation method performed by the EMD calculation module (20) is performed in an optimized calculation method, specifically: The second acoustic emission signal is defined as signal I(ω); First, the signal I(ω) is subjected to a short-time Fourier transform, and the signal I(ω) can be decomposed in frequency into: Among them, the noise in the acoustic emission signal is mainly high-frequency noise; The high frequency noise is filtered, and in I(ω), I(ω) is decomposed by EMD as follows: Wherein, I(ω) represents the input signal, IMF i (ω) represents the eigenmode function, Res n (ω) represents the residual; The power of the high frequency noise in the signal is calculated according to the following formula: Then judge the high-frequency noise signal power intensity according to the following formula: SNR=10lg(P s / P n ) Among them, when SNR≤25, it proves that the high-frequency noise power intensity is sufficient. Remove IMF1 in the EMD decomposition and define the new signal as I'(ω), that is, the signal after the first filtering. Then calculate the high-frequency noise power and SNR of I'(ω) until the new signal I n The SNR in (ω) is ≥ 25.
2. The method for online nondestructive monitoring of micro-grinding based on acoustic emission technology according to claim 1 is characterized in that: The sampling frequency of the acoustic emission signal is 2 MHz; The sampling frequency should be greater than or equal to twice the highest frequency in the analog signal spectrum.
3. The method for online nondestructive monitoring of micro-grinding based on acoustic emission technology as claimed in claim 2 is characterized in that: When judging based on the time domain eigenvalue and determining whether the eigenvalue reaches the specified threshold, the threshold determination is performed based on the obtained time domain eigenvalue kurtosis K and skewness S. If the threshold is reached, it is a prerequisite for automatically triggering the step three.
4. The method for online nondestructive monitoring of micro-grinding based on acoustic emission technology as claimed in claim 3 is characterized in that: The method of using the acoustic emission signal of the calculation result to identify wear is as follows: The acoustic emission signal is modeled based on the hidden Markov model; The IMF obtained after EMD decomposition is regarded as the observed state, and the actual wear state of the micro-grinding tool is regarded as the hidden state. The chaos matrix in this process is experimentally verified as follows:
5. The method for online nondestructive monitoring of micro-grinding based on acoustic emission technology as claimed in claim 4 is characterized in that: In combination with the given chaotic matrix and initial parameters, the parameters in the hidden Markov model are continuously updated using the Nelder-Mead algorithm to achieve convergence to the local minimum in a short time without the need to obtain the derivative function of the multivariate function in advance.
Citation Information
Patent Citations
Neural network-based grinding machining working condition detection method
CN101817163A
Cutter wear monitoring method based on current and acoustic emission compound signals
CN104723171A
