Method for monitoring running state of micro-particle shot blasting equipment based on multi-source data fusion

Abstract

The invention relates to the technical field of data processing, in particular to a micro-particle shot blasting equipment running state monitoring method based on multi-source data fusion. The method comprises the following steps: fusing the same frequency bands in a plurality of abnormal frequency spectrums in the same abnormal state to obtain a comprehensive frequency spectrum, and connecting the comprehensive frequency spectrums of the plurality of frequency bands to form an abnormal fusion frequency spectrum; acquiring various characteristic signals in the operation process of the equipment in real time and acquiring a real-time frequency spectrum; fusing the frequency bands of the plurality of real-time frequency spectrums to obtain a real-time fused frequency spectrum; comparing the difference between each abnormal fusion spectrum and the real-time fusion spectrum to obtain a frequency domain similarity; and calculating an abnormal score based on the frequency domain similarity, and judging an equipment operation state based on the abnormal score. The method and the device have the effect of improving the detection accuracy of the equipment running state.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method for monitoring the operating status of a microparticle shot peening device based on multi-source data fusion. Background Art

[0002] Microparticle shot peening is similar to the traditional shot peening process. In the microparticle shot peening process, smaller-diameter shot is used to impact the surface of the part at high speed. When a microparticle shot peening device is used to process gears, the surface hardness of the processed workpiece is higher, and at the same time, the plastic deformation on the surface layer of the workpiece is smaller. In addition, the microparticle shot peening technology can also reduce the surface roughness of the parts. The microparticle shot peening technology is mainly used in the processing and manufacturing of some high-precision workpieces, such as gears. The stability of the operation of the microparticle shot peening device is an important condition for the microparticle shot peening process. It is necessary to detect the operating status of the microparticle shot peening device in a timely manner to discover equipment failures and avoid serious damage. During the operation of mechanical equipment, sounds are emitted regularly. When the equipment fails, noises will be generated during the operation of the equipment. Therefore, the operating status of the equipment can be detected based on the sound signal during the operation of the equipment. The main steps of acoustic detection include: collecting the sound signal during the operation of the equipment to be detected through a sound sensor, transmitting the sound signal to a central database through a network, and comparing it with the standard sound signal stored in advance, and judging the operating status of the equipment according to the differences in characteristics such as the frequency or amplitude of the sound signal and the standard sound signal.

[0003] However, the reasons for failures during the operation of the microparticle shot peening device are not single, and there may be various types of failures, such as loose screws, component wear, cavitation explosion, and small cracks in the storage components. Traditional acoustic detection technologies are mainly based on a single acoustic signal, and the sensitive frequency band of the acoustic signal is usually between 50 kHz and 2 MHz; although such methods can capture high-frequency abnormal signals (such as cavitation effects) during actual detection, the detection sensitivity for low-frequency fault signals (such as minor cracks and micro-leaks in valves) is relatively low, and even low-frequency fault signals are ignored. Over time, it will eventually lead to serious damage to the equipment; therefore, there are cases of missed detection and misjudgment in the methods for detecting equipment in related technologies during application. Summary of the Invention

[0004] In order to solve the problem of missed detection and misjudgment in the process of detecting the operation of equipment in related technologies, this application provides a method for monitoring the operating status of a microparticle shot peening device based on multi-source data fusion.

[0005] This application provides a method for monitoring the operating status of a microparticle shot peening device based on multi-source data fusion, and adopts the following technical solutions: A method for monitoring the operating status of a microparticle peening device based on multi-source data fusion includes the steps of: obtaining various characteristic signals of the device in a normal operating state and various abnormal states, where the characteristic signals include at least two of acoustic signals, pressure signals, vibration signals, and flow signals; obtaining the frequency spectra of the characteristic signals in each abnormal state to obtain abnormal frequency spectra, and obtaining the frequency spectrum of the characteristic signals in the normal state to obtain a normal frequency spectrum; dividing frequency bands, calculating the band differences between the abnormal frequency spectra and the normal frequency spectrum in different frequency bands, and determining the fusion weights of different characteristic signals in each frequency band in each abnormal state based on the band differences. Fusing the same frequency bands in multiple abnormal frequency spectra in the same abnormal state based on the fusion weights to obtain a comprehensive frequency spectrum, and connecting the comprehensive frequency spectra of multiple frequency bands to form an abnormal fusion frequency spectrum; collecting various characteristic signals during the operation of the device in real time and obtaining a real-time frequency spectrum; fusing the frequency bands of multiple real-time frequency spectra to obtain a real-time fusion frequency spectrum; comparing the differences between each abnormal fusion frequency spectrum and the real-time fusion frequency spectrum to obtain the frequency domain similarity; calculating an abnormal score based on the frequency domain similarity, and judging the operating status of the device based on the abnormal score.

[0006] The beneficial effects are as follows: There may be various abnormal states during the operation of the device, and the signals of different abnormal states usually show high-frequency or low-frequency characteristics. Therefore, in this application, various characteristic signals in different abnormal states are obtained. Different characteristic signals are collected by different sensors. By comparing the differences between different characteristic signals and normal characteristic signals in different frequency bands in the same abnormal state, the importance of different characteristic signals in different frequency bands for device abnormality detection can be determined, and then the fusion weights corresponding to each frequency band can be obtained. Fusing the abnormal frequency spectra of multiple characteristic signals collected in the same abnormal state based on the fusion weights to obtain an abnormal fusion frequency spectrum. The operating status of the device can be detected based on the frequency domain similarity between the real-time fusion frequency spectrum of various characteristic signals collected in real time and the abnormal fusion frequency spectrum.

[0007] In this method, the importance levels of different frequency bands in different abnormal states are analyzed. For example, when some mechanical failures occur in the low-frequency band, the difference between the abnormal frequency spectrum and the normal frequency spectrum of a certain characteristic signal in the low-frequency band will be greater. This characteristic signal and this frequency band are more important for device abnormality detection, and it also shows that this characteristic signal is more sensitive to low-frequency changes, reducing the occurrence of missed low-frequency fault information and improving the accuracy of device operation fault detection.

[0008] Optionally, the step of calculating the band differences between the abnormal frequency spectra and the normal frequency spectrum in different frequency bands includes: for each abnormal frequency spectrum in each abnormal state, obtaining the absolute value of the difference between the elements of the same ordinal position in the same frequency band of the abnormal frequency spectrum and the normal frequency spectrum; taking the sum of the absolute values of the differences corresponding to all elements in the frequency band as the band difference of this frequency band.

[0009] The beneficial effects are as follows: Obtain the differences between each element in the abnormal spectrum and the normal spectrum, and accumulate them, so as to obtain the frequency band difference between a certain frequency band in the abnormal spectrum and the corresponding frequency band in the normal spectrum.

[0010] Optionally, the step of determining the fusion weights of different characteristic signals in each frequency band based on the frequency band difference includes: taking the sum of the frequency band differences corresponding to different characteristic signals as the total difference, and for each abnormal state, taking the ratio of the frequency band difference of any characteristic signal in any frequency band to the total difference as the fusion weight of the characteristic signal in that frequency band.

[0011] The beneficial effects are as follows: Normalize the frequency band difference to facilitate subsequent frequency band fusion.

[0012] Optionally, the step of fusing the same frequency bands in multiple abnormal spectra under the same abnormal state based on the fusion weights to obtain a comprehensive spectrum includes: taking the product of the fusion weight corresponding to each frequency band and the elements in each frequency band as local elements, and for each abnormal spectrum under each abnormal state, taking the sum of the local elements with the same ordinal position in the same frequency band in multiple abnormal spectra as the comprehensive element; multiple comprehensive elements constitute the comprehensive spectrum.

[0013] The beneficial effects are as follows: Each abnormal state corresponds to multiple abnormal spectra, and each abnormal spectrum corresponds to a fusion weight and multiple frequency bands. The elements in multiple abnormal spectra under the same abnormal state are weighted and summed through the fusion weights, so as to obtain a comprehensive spectrum containing various characteristic information.

[0014] Optionally, in the step of dividing the frequency bands, the spectrum is divided into three segments.

[0015] The beneficial effects are as follows: Different sensors have high sensitivity to different frequency bands. According to this characteristic, the spectrum is divided into multiple frequency bands, and a fusion weight is set for each frequency band, so as to improve the accuracy of the final detection of the device.

[0016] Optionally, the step of calculating the abnormal score based on the frequency domain similarity includes: obtaining the time domain similarity between the real-time characteristic signal and the abnormal characteristic signal; taking the product of the time domain similarity and the frequency domain similarity as the abnormal score.

[0017] Optionally, the step of calculating the abnormal score based on the frequency domain similarity includes: obtaining the time domain similarity between the real-time characteristic signal and the abnormal characteristic signal; taking the product of the time domain similarity and the frequency domain similarity as the first control value, obtaining the comprehensive spectrum corresponding to the normal characteristic signal; calculating the second control value based on the similarity between the normal fusion spectrum of the normal characteristic signal and the normal fusion spectrum of the real-time characteristic signal, and taking the ratio of the first control value to the second control value as the abnormal score.

[0018] The beneficial effects are as follows: In this method, not only the differences in the frequency domain between the abnormal feature signal and the real-time feature signal are compared, but also the differences between the real-time feature signal and the normal feature signal are compared, thereby further improving the accuracy of equipment detection.

[0019] Optionally, the calculation steps of the frequency-domain similarity include: obtaining the absolute value of the difference between the elements of the same ordinal position in the real-time fusion spectrum and the abnormal fusion spectrum, and taking the sum of the absolute values of the differences corresponding to multiple elements as the adjustment value; taking the result with as the base and the negative value of the adjustment value as the exponent as the frequency-domain similarity.

[0020] Optionally, the time-domain similarity between the real-time feature signal and the abnormal feature signal is the average of the Euclidean distances between multiple real-time feature signals and the corresponding abnormal feature signals.

[0021] Optionally, the steps of judging the operation state of the equipment based on the abnormal score include: setting an abnormal threshold, and issuing an alarm in response to the abnormal score being greater than the abnormal threshold.

[0022] The present application has the following technical effects: On the one hand, multiple feature signals in multiple dimensions are collected by multiple sensors; on the other hand, the fusion weights of the feature signal frequency bands are determined according to the differences between the feature signals and the normal signals in different frequency bands under different abnormal states, and the abnormal feature signals are fused. By comparing the differences in the time domain and frequency domain between the real-time feature signal and the abnormal feature signal, it is determined whether the current equipment state is similar to the signal corresponding to a certain abnormal state, realizing the detection of the equipment state. At the same time, multiple sensors are used to comprehensively analyze the low-frequency and high-frequency fault signals of the equipment, reducing the occurrence of missing low-frequency fault signals and improving the accuracy of monitoring the operation state of the equipment. Description of the Drawings

[0023] Figure 1 is the method flow chart of the method for monitoring the operation state of the multi-source data fusion micro-particle shot peening equipment in the embodiment of the present application. Detailed Embodiments

[0024] The embodiment of the present application discloses a method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion. During the operation of the device, data in multiple dimensions under various abnormal states and data in multiple dimensions under normal operating states are collected; the abnormal frequency spectrum of the data under abnormal states and the normal frequency spectrum under normal operating states are extracted; the frequency band differences between the normal frequency spectrum and the abnormal frequency spectrum are compared, and the influence of the characteristic signals in different dimensions on the frequency band differences under each abnormal state is determined. Furthermore, the fusion weights of different characteristic signals in different frequency bands under each abnormal state can be determined. Based on the fusion weights, the abnormal frequency spectra of the characteristic signals in multiple dimensions under abnormal states are weighted and fused to obtain an abnormal fusion frequency spectrum. Similarly, the operating information of the device is collected in real time, and a real-time fusion frequency spectrum is obtained. The abnormal fusion frequency spectrum is compared with the real-time fusion frequency spectrum, and an abnormal score is calculated to judge the operating state of the device. Under different abnormal states, different characteristic signals have different fusion weights, that is, during the process of judging that the device belongs to a certain abnormal state, the attention to different data is different. Therefore, different characteristic signals can be concerned according to the abnormal state, thereby reducing the situation of ignoring low-frequency signals during the abnormal detection process and improving the accuracy of device detection.

[0025] Referring to Figure 1 , the method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion includes steps S1 - S5.

[0026] S1: Obtain various characteristic signals of the device under normal operating states and various abnormal states. The characteristic signals include at least two of acoustic signals, pressure signals, vibration signals, and flow signals.

[0027] When the microparticle shot peening device is in different abnormal states (such as bearing wear, microcracks in the storage tank, micro air leakage in the seal, etc.) and normal states, the sensor is essentially a signal converter. Therefore, in this embodiment, sensors are used to collect characteristic signals in multiple dimensions during the operation of the device. In this embodiment, a flow sensor is used to collect the flow information of the microparticles, a dynamic pressure sensor is used to collect the pressure information at the spray gun; an acoustic sensor is used to collect the sound information during the operation of the device and the vibration information during the operation of the device. In other embodiments, two or three of the above-mentioned multiple characteristic signals can be collected.

[0028] After the data collection is completed, multiple sets of abnormal characteristic signals corresponding to multiple abnormal states and one normal characteristic signal are obtained. Each set of abnormal characteristic signals includes multiple abnormal characteristic signals corresponding to various characteristic signals.

[0029] S2: Obtain the spectrum of the characteristic signals in each abnormal state to get the abnormal spectrum, and obtain the spectrum of the characteristic signals in the normal state to get the normal spectrum; divide the frequency bands, calculate the band differences between the abnormal spectrum and the normal spectrum in different frequency bands, and determine the fusion weights of different characteristic signals in each abnormal state in each frequency band based on the band differences.

[0030] Perform Fourier transform on the abnormal characteristic signals and the normal characteristic signals to obtain their corresponding spectra, and get the abnormal spectrum corresponding to the abnormal characteristic signals and the normal spectrum corresponding to the normal characteristic signals.

[0031] Divide the abnormal spectrum and the normal spectrum based on frequency to obtain multiple frequency bands. In this embodiment, for any abnormal spectrum or normal spectrum, it is divided into three frequency bands, namely the low-frequency band, the middle-frequency band, and the high-frequency band. The frequency corresponding to the low-frequency band is: 1 Hz - 1 kHz; the frequency corresponding to the middle-frequency band is: 1 kHz - 50 kHz; the frequency corresponding to the high-frequency band is: 50 kHz - 2 MHz. The specific number of divided frequency bands and the frequencies corresponding to the frequencies are determined by technicians. Taking the example of dividing the spectrum into three frequency bands in this embodiment, then the various collected characteristic signals should have different sensitivities in different frequency bands. For example, one of the characteristic signals can produce a large change when the device has a low-frequency fault, and another group of characteristic signals can produce a large change when the device has a high-frequency fault, etc.

[0032] For different characteristic signals collected in different abnormal states, there are certain differences between their corresponding abnormal spectra and normal spectra, which can be specifically manifested as differences in the low-frequency band, the middle-frequency band, or the high-frequency band. And the sensitive frequency bands of different sensors are different, so the changes of different characteristic signals in the same frequency band are not consistent. Therefore, the fusion weights of different characteristic signals in different frequency bands can be determined based on the band differences. For example, if the abnormality in a certain abnormal state is mainly manifested in the difference in the low-frequency band, then the low-frequency band should be concerned during the detection process. For example, if the characteristic signal sensitive to the change in the low-frequency band is the pressure signal, then when judging whether the operating state of the device is in this abnormal state, more attention should be paid to the change of the pressure signal in this frequency band.

[0033] First, calculate the band differences between the abnormal spectrum and the normal spectrum in different frequency bands; the steps for calculating the band differences include: for each abnormal spectrum in each abnormal state, obtain the absolute value of the difference between the elements with the same ordinal position in the same frequency band of the abnormal spectrum and the normal spectrum; take the sum of the absolute values of the differences corresponding to all elements in the frequency band as the band difference of this frequency band.

[0034] Specifically, the calculation formula for the band differences in each abnormal state can be expressed as: ; where represents the The frequency band of an abnormal spectrum and the frequency band of a normal spectrum; indicating the th element in the th frequency band of the th abnormal spectrum; indicating the th element in the th frequency band of the normal spectrum; indicating the number of data in the th frequency band, which can also be understood as the length of the th frequency band.

[0035] Indicating the difference between a certain element in the th frequency band of the abnormal spectrum and the normal spectrum. By traversing all the elements in the frequency band, the overall difference of the frequency band is obtained. The greater this difference, the greater the change in this frequency band of the characteristic signal when the corresponding abnormality occurs during the operation of the device, and further it indicates that more attention should be paid to this frequency band of the characteristic signal when judging whether the device belongs to this abnormal state.

[0036] Normalize the frequency band difference to obtain the fusion weight of the frequency band; specifically, the calculation formula for the fusion weight is: ; where represents the fusion weight of the th frequency band of the th abnormal spectrum under each abnormal state; represents the frequency band difference of the th frequency band of the th abnormal spectrum under each abnormal state; represents the number of types of characteristic signals.

[0037] S3: Based on the fusion weights, fuse the same frequency bands in multiple abnormal spectra under the same abnormal state to obtain a comprehensive spectrum; the comprehensive spectra of multiple frequency bands are connected to form an abnormal fusion spectrum.

[0038] Take the product of the fusion weight corresponding to each frequency band and the elements in each frequency band as local elements. For each abnormal spectrum under each abnormal state, take the sum of the local elements with the same ordinal position in the same frequency band of multiple abnormal spectra as comprehensive elements; multiple comprehensive elements constitute a comprehensive spectrum.

[0039] For any abnormal state, there are multiple corresponding abnormal spectra, and fuse the multiple abnormal spectra. Each abnormal spectrum corresponds to multiple frequency bands, so first fuse the same frequency bands in the multiple abnormal spectra.

[0040] Specifically, the calculation formula for the comprehensive element can be expressed as: ; where represents the -th element in the comprehensive spectrum corresponding to the -th frequency band; represents the number of abnormal spectra in the same abnormal state; represents the -th fusion weight of the -th frequency band of the -th abnormal spectrum in the same abnormal state; represents the -th element in the -th frequency band of the

[0041] -th abnormal spectrum in the same abnormal state.

[0042] S4: Real-time collect various characteristic signals during the operation of the device and obtain the real-time spectrum of the real-time characteristic signals; fuse the frequency bands of multiple real-time spectra to obtain the real-time fusion spectrum.

[0043] During the operation of the device, collect the operation information of the device in real time and obtain multiple characteristic signals. Obtain the spectrum of the real-time collected characteristic signals to get the real-time spectrum, and fuse multiple real-time spectra to obtain the real-time fusion spectrum. The acquisition of the real-time fusion spectrum is the same as that of the abnormal fusion spectrum and will not be elaborated here.

[0044] S5: Compare the differences between each abnormal fusion spectrum and the real-time fusion spectrum to obtain the frequency-domain similarity; calculate the abnormal score based on the frequency-domain similarity, and judge the operation state of the device based on the abnormal score.

[0045] The operation state of the device can be judged according to the difference between the abnormal fusion spectrum and the real-time fusion spectrum.

[0046] Specifically, the calculation formula for the frequency-domain similarity can be expressed as: ; where represents the frequency-domain similarity between the abnormal fusion spectrum and the real-time fusion spectrum of the -th abnormal state; represents the -th element in the real-time fusion spectrum; represents the -th element in the -th abnormal fusion spectrum; represents the number of data in the real-time fusion spectrum, which can also be understood as the length of the real-time fusion spectrum.

[0047] It represents the difference between the elements in the real-time fusion spectrum and the abnormal fusion spectrum. The greater this difference is, the less similar the real-time fusion spectrum and the abnormal fusion spectrum are, and further indicates that there is a large difference between the current operating state of the device and the state corresponding to the abnormal fusion spectrum.

[0048] Calculate the anomaly score based on the frequency-domain similarity between the real-time fusion spectrum and different abnormal fusion spectra: In one embodiment, the steps of calculating the anomaly score based on the frequency-domain similarity include: obtaining the time-domain similarity between the real-time feature signal and the abnormal feature signal; taking the product of the time-domain similarity and the frequency-domain similarity as the anomaly score. In this embodiment, the steps of calculating the time-domain similarity include: obtaining the Euclidean distance between the real-time collected feature signal and the abnormal feature signal, and taking the mean of the Euclidean distances corresponding to multiple feature signals in the same abnormal state as the time-domain similarity.

[0049] Calculate the frequency-domain similarity and the time-domain similarity between the real-time spectrum and the abnormal spectrum, which further improves the accuracy of judging the operating state of the device.

[0050] Multiply the time-domain similarity and the frequency-domain similarity in the same abnormal state to obtain the anomaly score; set the anomaly score threshold, and when the anomaly score is greater than the anomaly threshold, issue an alarm, indicating that the current device operation is highly similar to this abnormal state and there is this type of anomaly.

[0051] In another embodiment: The steps of calculating the anomaly score based on the frequency-domain similarity include: obtaining the time-domain similarity between the real-time feature signal and the abnormal feature signal; taking the product of the time-domain similarity and the frequency-domain similarity as the first control value, obtaining the normal fusion spectrum corresponding to the normal feature signal; taking the similarity between the normal fusion spectrum of the normal feature signal and the real-time fusion spectrum of the real-time feature signal as the second control value, and taking the ratio of the first control value to the second control value as the anomaly score.

[0052] The steps of obtaining the second control value include: obtaining the frequency-domain similarity between the normal fusion spectrum and the real-time fusion spectrum; obtaining the time-domain difference between the normal feature signal and the real-time feature signal in the same abnormal state, and taking the product of the frequency-domain similarity and the time-domain similarity as the second control value.

[0053] The greater the frequency-domain similarity between the normal fusion spectrum and the real-time fusion spectrum is, the more normal the currently real-time collected feature signal is in the frequency domain, and thus the greater the possibility that the current device is in a normal operating state. When the first control value is greater than the second control value, it indicates that the real-time collected feature signal is more similar to the abnormal feature signal, so the current state is more inclined to be abnormal.

[0054] The acquisition of the normal fusion spectrum here is the same as that of the abnormal fusion spectrum, and will not be elaborated here.

[0055] In this method, the normal fusion spectrum during the normal operation of the device is obtained. Considering both the similarity between the real-time collected feature signal and the abnormal feature signal and the similarity between the real-time collected feature signal and the normal feature signal, the accuracy of judging the operating state of the device is further improved.

[0056] Based on the abnormal score, the operating state of the device is judged. For the abnormal score corresponding to any abnormal state, in response to the abnormal score being greater than 1, it is judged that there is such an abnormality in the device operation.

[0057] The above are all the preferred embodiments of the present application, and the protection scope of the present application is not limited hereby. Therefore, all equivalent changes made according to the structure, shape, and principle of the present application shall be covered within the protection scope of the present application.

Claims

1. A method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion, characterized in that, It includes the steps of: obtaining various characteristic signals of the device in the normal operating state and various abnormal states, where the characteristic signals include at least two of acoustic signals, pressure signals, vibration signals, and flow signals; obtaining the spectra of the characteristic signals in each abnormal state to get abnormal spectra, and obtaining the spectrum of the characteristic signals in the normal state to get the normal spectrum; dividing frequency bands, calculating the band differences between the abnormal spectra and the normal spectrum in different frequency bands, and determining the fusion weights of different characteristic signals in each abnormal state in each frequency band based on the band differences. Based on the fusion weights, fusing the same frequency bands in multiple abnormal spectra in the same abnormal state to obtain a comprehensive spectrum, and connecting the comprehensive spectra of multiple frequency bands to form an abnormal fusion spectrum; collecting various characteristic signals during the operation of the device in real time and obtaining the real-time spectrum; fusing the frequency bands of multiple real-time spectra to obtain a real-time fusion spectrum; comparing the differences between each abnormal fusion spectrum and the real-time fusion spectrum to get the frequency-domain similarity; calculating an abnormal score based on the frequency-domain similarity, and judging the operating state of the device based on the abnormal score.

2. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 1, wherein The steps of calculating the band differences between the abnormal spectra and the normal spectrum in different frequency bands include: for each abnormal spectrum in each abnormal state, obtaining the absolute value of the difference between the elements of the same ordinal position in the same frequency band of the abnormal spectrum and the normal spectrum; taking the sum of the absolute values of the differences corresponding to all elements in the frequency band as the band difference of this frequency band.

3. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 2, wherein, The steps of determining the fusion weights of different characteristic signals in each abnormal state in each frequency band based on the band differences include: taking the sum of the band differences corresponding to different characteristic signals as the total difference, and for each abnormal state, taking the ratio of the band difference of any characteristic signal in any frequency band to the total difference as the fusion weight of this characteristic signal in this frequency band.

4. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 3, wherein, The steps of fusing the same frequency bands in multiple abnormal spectra in the same abnormal state based on the fusion weights to obtain a comprehensive spectrum include: taking the product of the fusion weight corresponding to each frequency band and the elements in each frequency band as local elements, and for each abnormal spectrum in each abnormal state, taking the sum of the local elements of the same ordinal position in the same frequency band in multiple abnormal spectra as the comprehensive element; multiple comprehensive elements constitute the comprehensive spectrum.

5. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 1, characterized in that, In the step of dividing frequency bands, the spectrum is divided into three segments.

6. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 1, characterized in that, The steps of calculating the abnormal score based on the frequency-domain similarity include: obtaining the time-domain similarity between the real-time characteristic signal and the abnormal characteristic signal; taking the product of the time-domain similarity and the frequency-domain similarity as the abnormal score.

7. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 1, characterized in that The steps of calculating the abnormal score based on the frequency-domain similarity include: obtaining the time-domain similarity between the real-time characteristic signal and the abnormal characteristic signal; taking the product of the time-domain similarity and the frequency-domain similarity as the first control value, obtaining the comprehensive spectrum corresponding to the normal characteristic signal; calculating the second control value based on the similarity between the normal fusion spectrum of the normal characteristic signal and the normal fusion spectrum of the real-time characteristic signal, and taking the ratio of the first control value to the second control value as the abnormal score.

8. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 6, characterized in that, The calculation steps of the frequency-domain similarity include: obtaining the absolute value of the difference between the elements with the same ordinal position in the real-time fusion spectrum and the abnormal fusion spectrum, and taking the sum of the absolute values of the differences corresponding to multiple elements as the adjustment value; taking the result with as the base and the negative value of the adjustment value as the exponent as the frequency-domain similarity.

9. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 6, characterized in that, The time-domain similarity between the real-time characteristic signal and the abnormal characteristic signal is the mean of the Euclidean distances between multiple real-time characteristic signals and the corresponding abnormal characteristic signals.

10. The method for monitoring the operating state of a microparticle shot peening device based on multi-source data fusion according to claim 1, characterized in that, It includes: The steps of judging the operating state of the device based on the abnormal score include: setting an abnormal threshold, and issuing an alarm in response to the abnormal score being greater than the abnormal threshold.

Images