Intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switch

Abstract

The invention relates to the technical field of power equipment fault diagnosis, in particular to an intelligent voiceprint monitoring, diagnosis and analysis system for a high-voltage switch. According to the invention, the information amount index is obtained according to the amplitude change trend of the component signal of the current signal segment and the power difference of different frequencies, and the amplitude increase index is obtained according to the difference of the matching conditions of the adjacent audio signal segments and the component signal of the same frequency of the current signal segment. Acquiring a spectrum distribution diffusion index according to the spectrum change degree of the component signals with the same frequency of the different audio signal segments; and then selecting an effective component signal of the current signal section in combination with an information quantity index, an amplitude increase index and a spectrum distribution diffusion index, reconstructing the effective component signal to obtain a current effective signal section, and diagnosing the working state of the high-voltage switch at the current time by using the current effective signal section. According to the invention, the proportion of the abnormal working sound of the high-voltage switch in the current signal segment is increased, and the accuracy of diagnosing the working state of the high-voltage switch is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment fault diagnosis, and in particular to a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system. Background Art

[0002] High-voltage switches, also known as high-voltage circuit breakers, are one of the most important electrical appliances in high-voltage power supply systems. High-voltage switches can not only cut off and connect the no-load current and load current in high-voltage circuits under normal circumstances, but can also cooperate with protective devices and automatic devices when a system fault occurs to quickly cut off the faulty power supply and ensure the safe operation of the system.

[0003] At present, high-voltage switches in my country's power system are basically in a regular maintenance state. Condition-based maintenance is a preventive operation performed based on the operating status of the high-voltage circuit breaker. By online monitoring of the status of the high-voltage switch and determining the maintenance plan, fault prediction and health management of the high-voltage switch can be achieved.

[0004] Existing methods determine the operating status of a high-voltage switch by analyzing the audio during its closing or opening operations. The audio during a high-voltage switch's operation typically includes the switch's operating sound, abnormal operating noise, and ambient noise. Abnormal operating noise is a key factor in determining the switch's operating status. If the impact of ambient noise on the audio is greater than that of the abnormal operating noise, the accuracy of the high-voltage switch's operating status diagnosis will be low. Summary of the Invention

[0005] To address the technical problem that abnormal working noises from high-voltage switches have little impact on the audio during their operation, resulting in low accuracy in diagnosing the working status of the high-voltage switches, the present invention aims to provide a smart voiceprint monitoring, diagnosis, and analysis system for high-voltage switches. The technical solutions employed are as follows:

[0006] The present invention proposes a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system, which includes:

[0007] a data acquisition module, configured to obtain a sequence of audio signal segments when the high-voltage switch operates continuously, and record the last audio signal segment in the sequence as the current signal segment;

[0008] The information content analysis module is used to decompose each audio signal segment to obtain component signals of different frequencies; based on the amplitude change trend of each component signal of the current signal segment and the power difference of different frequencies, the information content index of each component signal of the current signal segment is obtained;

[0009] an amplitude growth analysis module, configured to match the component signals of the same frequency between the current signal segment and the remaining audio signal segments in the audio signal segment sequence, and obtain an amplitude growth index for each component signal of the current signal segment based on the difference in matching between the component signals of the same frequency between two adjacent audio signal segments and the current signal segment;

[0010] a spectrum analysis module, configured to obtain a spectrum distribution diffusion index of each component signal of a current signal segment based on a degree of spectrum change of component signals of the same frequency in the audio signal segment in the audio signal segment sequence;

[0011] The voiceprint diagnosis module is used to adjust the information amount index according to the amplitude growth index and the spectrum distribution diffusion index, reconstruct the effective component signal filtered from the component signal of the current signal segment using the adjusted information amount index, and obtain the current effective signal segment; based on the current effective signal segment, diagnose the current working state of the high-voltage switch.

[0012] Furthermore, obtaining the information quantity index of each component signal of the current signal segment includes:

[0013] For each component signal of the current signal segment, obtain the envelope of the component signal, the envelope including an upper envelope and a lower envelope; obtain the average of the absolute values ​​of the differences between the amplitudes of all two adjacent data points on each envelope of the component signal, and record it as the local amplitude variation of the corresponding envelope; and use the average of the local amplitude variation of all envelopes of the component signal as the comprehensive amplitude variation of the component signal;

[0014] Performing a Fourier transform on the component signal to obtain spectrum data, recording the frequency corresponding to the maximum power in the spectrum data as the main frequency of the component signal; in the spectrum data, frequency segments are formed by continuous data points with power greater than a preset power threshold, and recording the frequency corresponding to the maximum power in each frequency segment as the analysis frequency of the component signal;

[0015] The product of the absolute value of the difference between each analysis frequency of the component signal and the main frequency and the power of each analysis frequency is used as the local distribution extensiveness of each analysis frequency of the component signal; the cumulative sum of the local distribution extensiveness of all analysis frequencies of the component signal is used as the frequency distribution extensiveness of the component signal;

[0016] An information content index of a component signal is obtained according to the comprehensive amplitude variation and the frequency distribution wideness.

[0017] Furthermore, obtaining the amplitude growth index of each component signal of the current signal segment includes:

[0018] Arrange the component signals of the same frequency of all audio signal segments in the audio signal segment sequence in sequence to obtain a component signal sequence for each frequency; for each component signal sequence, record the last component signal in the component signal sequence as the current component signal, and record the remaining component signals as target component signals;

[0019] Obtaining envelope analysis indicators of the current component signal and each target component signal based on matching conditions of data points on the envelope line of the current component signal and each target component signal;

[0020] The average of the differences between the envelope analysis indicators of the current component signal and the remaining target component signals except the first target component signal and the previous target component signal is obtained as the amplitude growth indicator of the current component signal.

[0021] Furthermore, the obtaining of envelope analysis indicators of the current component signal and each target component signal includes:

[0022] Arrange the data points on each envelope of each component signal in sequence to obtain an analysis sequence of the corresponding component signal; the analysis sequence includes: an upper analysis sequence and a lower analysis sequence;

[0023] For each target component signal, matching the current component signal with the upper analysis sequence of the target component signal, and selecting a matching data point for each data point in the upper analysis sequence of the current component signal from the upper analysis sequence of the target component signal;

[0024] Obtain the mean of the time intervals between the corresponding moments of the data points and their matching data points in the upper analysis sequence of the current component signal, which is recorded as the time difference index; obtain the mean of the absolute values ​​of the differences between the amplitudes of the data points and their matching data points in the upper analysis sequence of the current component signal, which is recorded as the amplitude difference index;

[0025] Obtaining a local analysis index of an upper analysis sequence of a current component signal and a target component signal according to the time difference index and the amplitude difference index; wherein the time difference index and the local analysis index are negatively correlated, and the amplitude difference index and the local analysis index are positively correlated;

[0026] The average of the local analysis indicators of the upper analysis sequence and the lower analysis sequence of the current component signal and the target component signal is used as the envelope analysis indicator of the current component signal and the target component signal.

[0027] Furthermore, obtaining the spectrum distribution spread index of each component signal of the current signal segment includes:

[0028] Subtracting the spectrum data of each target component signal from the spectrum data of the current component signal to obtain difference spectrum data of each target component signal; arranging the difference spectrum data of all target component signals in sequence to obtain a difference spectrum sequence;

[0029] Normalize the mean of the correlation coefficients of two adjacent difference spectrum data in the difference frequency sequence to obtain the spectrum change consistency value of the current component signal;

[0030] Obtaining the mean of the powers of all data points of each difference frequency spectrum data, which is recorded as the first mean power of the corresponding difference frequency data; normalizing the mean of the first mean power of all difference frequency spectrum data in the difference frequency sequence to obtain the second mean power of the current component signal;

[0031] A spectrum distribution spread index is obtained according to the second average power and the spectrum change consistency value; the second average power and the spectrum change consistency value are both positively correlated with the spectrum distribution spread index.

[0032] Furthermore, the method for obtaining the current valid signal segment includes:

[0033] Obtaining an abnormal information content coefficient of each component signal of the current signal segment according to the amplitude growth index and the spectrum distribution diffusion index; performing weighted processing on the information content index using the abnormal information content coefficient to obtain an effective information index of each component signal of the current signal segment;

[0034] For the valid information indicators of all component signals of the current signal segment, component signals corresponding to a maximum preset number of the valid information indicators are taken as valid component signals; all valid component signals of the current signal segment are reconstructed to obtain the current valid signal segment.

[0035] Furthermore, the diagnosing the current working state of the high-voltage switch based on the current valid signal segment includes:

[0036] Extract features from the current valid signal segment to obtain a feature vector, which is recorded as the current diagnostic vector;

[0037] Obtain the cosine similarity between the current diagnostic vector and the preset normal feature vector. If the cosine similarity is greater than the preset normal threshold, the high-voltage switch is in a normal state in the current operation; if the cosine similarity is less than or equal to the preset normal threshold, the high-voltage switch is in an abnormal state in the current operation.

[0038] Furthermore, the method for decomposing each audio signal segment is an EMD algorithm.

[0039] Furthermore, the method for matching the upper analysis sequence of the current component signal and the target component signal is a DTW algorithm.

[0040] Furthermore, the method for extracting features from the current valid signal segment is an MFCC algorithm.

[0041] The present invention has the following beneficial effects:

[0042] In an embodiment of the present invention, abnormal operating noise from a high-voltage switch and stronger ambient noise can cause significant changes in an audio signal segment during the operation of the high-voltage switch, primarily manifested in changes in the signal amplitude and frequency of the audio signal segment. An information content index is then obtained based on the amplitude change trend of the component signals and the power difference at different frequencies. When an abnormality occurs in the high-voltage switch, the proportion of abnormal noise in the audio signal segment gradually increases, causing the frequency and amplitude of component signals of the same frequency within the audio signal segment sequence to increase. Furthermore, the distribution of the speech signal in the frequency domain gradually spreads as the abnormal noise content increases. Based on these two characteristics, an amplitude growth index and a spectrum distribution diffusion index are sequentially obtained. Both indicators reflect the abnormal noise content of the high-voltage switch in the component signals of the current signal segment. Valid component signals are then screened based on the information content index adjusted using the amplitude growth index and the spectrum distribution diffusion index. In valid component signals, the abnormal noise of the high-voltage switch operation is clearer and has a higher proportion. Reconstruction is performed based on the valid component signals to obtain clearer abnormal noise audio data, namely the current valid signal segment, while reducing the ambient noise in the audio signal segment, thereby improving the accuracy of abnormality diagnosis of the high-voltage switch operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 This is a system structure diagram of a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system provided by one embodiment of the present invention;

[0045] Figure 2 A system structure diagram of a method for obtaining an amplitude growth index provided by one embodiment of the present invention;

[0046] Figure 3 A schematic diagram of a computer device for intelligent voiceprint monitoring, diagnosis and analysis of high-voltage switches provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0047] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of a smart voiceprint monitoring, diagnosis, and analysis system for high-voltage switches proposed in accordance with the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0048] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0049] The specific scheme of the intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switches provided by the present invention is described in detail below with reference to the accompanying drawings.

[0050] Example 1:

[0051] See also Figure 1 , which shows a system block diagram of a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system provided by an embodiment of the present invention. The system includes: a data acquisition module 110, an information analysis module 120, an amplitude growth analysis module 130, a spectrum analysis module 140, and a voiceprint diagnosis module 150.

[0052] The data acquisition module 110 is configured to obtain a sequence of audio signal segments when the high-voltage switch operates for consecutive times, and record the last audio signal segment in the sequence as the current signal segment.

[0053] Specifically, a sound collection device is arranged near the high-voltage switch to be monitored. The arrangement method of the sound collection device is microphone array technology. The sound collection device is started each time the high-voltage switch starts to work. The sound collection device is used to collect the sound generated by the high-voltage switch during each operation process to obtain the audio signal segment of each operation.

[0054] It should be noted that the sound collection device is a microphone, and can also be an acoustic sensor and a voice recorder, etc.; the operation of the high-voltage switch refers to the closing and opening operations of the high-voltage switch; the duration of the audio signal segment of each operation is equal to the duration of each closing or opening operation of the high-voltage switch; the data collection frequency of the sound collection device in this embodiment is once every 0.1 second, and the implementer can set it according to the specific situation.

[0055] The horizontal axis of the audio signal segment represents time, and the vertical axis represents sound intensity or energy. A curve fitting is performed on the sound intensity data points at each moment during each high-voltage switch operation to obtain the audio signal segment for each operation. The curve fitting method uses the least squares method. Both the least squares method and microphone array technology are well known to those skilled in the art and will not be discussed further here.

[0056] The audio signal segments of the high-voltage switch during the current operation and the audio signal segments of the high-voltage switch during the previous operation by a preset reference number of times are arranged in the order of the high-voltage switch operation to obtain an audio signal segment sequence. The last audio signal segment in the audio signal segment sequence is the audio signal segment of the high-voltage switch during the current operation and is recorded as the current signal segment.

[0057] It should be noted that, in the embodiment of the present invention, the preset reference number is an empirical value of 10, and the number of audio signal segments in the audio signal segment sequence is 11, which can be set by the implementer according to specific circumstances.

[0058] The information content analysis module 120 is used to decompose each audio signal segment to obtain component signals of different frequencies; based on the amplitude change trend of each component signal of the current signal segment and the power difference of different frequencies, the information content index of each component signal of the current signal segment is obtained.

[0059] This embodiment uses the Empirical Mode Decomposition (EMD) algorithm to decompose each audio signal segment to obtain component signals of different frequencies. Other embodiments may also use signal decomposition methods such as the Ensemble Empirical Mode Decomposition (EEMD) algorithm and the Variational Mode Decomposition (VMD) algorithm. The EMD algorithm is well known to those skilled in the art and will not be described in detail here.

[0060] Audio signal segments typically contain the sound information of high-voltage switches operating, as well as abnormal operating noise and ambient noise. These abnormal operating noise and the stronger noise in the ambient noise can cause significant variations in the audio signal during operation, primarily manifesting as changes in the audio signal amplitude and frequency. Therefore, we determine the information content of the component signals based on the amplitude variation trends and power differences at different frequencies within the current signal segment.

[0061] The first subunit: obtains the comprehensive amplitude variation of the component signals.

[0062] For each component signal of the current signal segment, the envelope of the component signal is obtained, and the envelope includes: an upper envelope and a lower envelope; the average of the absolute values ​​of the differences between the amplitudes of all two adjacent data points on each envelope of the component signal is obtained, and recorded as the local amplitude variation of the corresponding envelope; the average of the local amplitude variation of all envelopes of the component signal is used as the comprehensive amplitude variation of the component signal.

[0063] The integrated amplitude variation shows the amplitude variation trend of the component signal. Since the data points on the component signal envelope are usually related to the main characteristics and mutation points of the signal and can carry more useful information, the greater the amplitude difference between adjacent data points on the component signal envelope, the more dramatic the component signal change and the greater the integrated amplitude variation.

[0064] It should be noted that the data point on each envelope line of the component signal refers to a local maximum point or a local minimum point in the component signal.

[0065] In a specific implementation of the embodiment of the present invention, the comprehensive amplitude variation of the component signal is expressed as follows:

[0066]

[0067] Where DA is the comprehensive amplitude variation of the component signal; A is the total number of envelopes of the component signal, which is equal to a constant 2 in this embodiment; N a is the total number of data points on the ath envelope of the component signal; F a,n is the amplitude of the nth data point on the ath envelope of the component signal; F a,n+1 is the amplitude of the n+1th data point on the ath envelope of the component signal;

[0068] is the local amplitude variation of the a-th envelope of the component signal; || is the absolute value function.

[0069] The second subunit: obtains the frequency distribution of the component signal.

[0070] Perform Fourier transform on the component signal to obtain spectrum data, record the frequency corresponding to the maximum power in the spectrum data as the main frequency of the component signal; in the spectrum data, a frequency segment is formed by continuous data points with power greater than a preset power threshold, and record the frequency corresponding to the maximum power in each frequency segment as the analysis frequency of the component signal; the product of the absolute value of the difference between each analysis frequency of the component signal and the main frequency and the power of each analysis frequency is used as the local distribution spread of each analysis frequency of the component signal; the cumulative sum of the local distribution spreads of all analysis frequencies of the component signal is used as the frequency distribution spread of the component signal. Among them, Fourier transform is a well-known technology for those skilled in the art and will not be described in detail here.

[0071] The data point with the maximum power in the frequency band represents the main local feature of the spectrum data. Selecting the relevant data of the data point with the maximum power in the frequency band for analysis is conducive to simplifying the analysis and highlighting the feature information.

[0072] Frequency distribution broadness indicates the power differences between different frequencies in a component signal. A larger difference between the analysis frequency and the primary frequency indicates that the component signal contains more frequency components. A higher power at the analysis frequency indicates that the analysis frequency component plays a more important role in the component signal. This indicates that the component signal contains multiple incoherent frequency components, making the component signal's spectrum more dispersed and the frequency distribution broadness is greater.

[0073] It should be noted that spectral data refers to a spectrogram, with the horizontal axis representing frequency and the vertical axis representing power. Spectral data is a curve obtained by least squares curve fitting using the power data points of the audio signal segment at each frequency. In this embodiment of the present invention, the preset power threshold is equal to one-tenth of the power of the component signal's dominant frequency; implementers may adjust this threshold based on their specific circumstances.

[0074] In a specific implementation of the embodiment of the present invention, the frequency distribution of the component signal is expressed as follows:

[0075]

[0076] Where SF is the frequency distribution of the component signal; B is the total number of analysis frequencies of the component signal; P b is the power of the bth analysis frequency of the component signal; F b is the bth analysis frequency of the component signal; F_main is the main frequency of the component signal; P b ×|F b -F_main| is the local distribution extensiveness of the b-th analysis frequency of the component signal; || is the absolute value function.

[0077] The third subunit: obtaining information content indicators of component signals.

[0078] A greater degree of integrated amplitude variation indicates a more dramatic change in the component signal, meaning the component signal contains more information, and thus, the information content of the analysis signal is richer. A greater frequency distribution variability indicates a more discrete distribution of the component signal's frequency components, indicating that the component signal can simultaneously transmit information at multiple frequencies, which means that the component signal itself has a stronger expressive power and richer information.

[0079] Therefore, both the comprehensive amplitude variation and the frequency distribution extensiveness are positively correlated with the information content index. In the embodiment of the present invention, the product of the comprehensive amplitude variation and the frequency distribution extensiveness of the component signal is normalized to obtain the information content index of the component signal.

[0080] In the embodiment of the present invention, the correlation between the comprehensive amplitude variation and the frequency distribution extensiveness and information content index can also be constructed through other basic mathematical operations, which will not be limited or elaborated here.

[0081] It should be noted that the Norm function is used for normalization in the embodiment of the present invention. Other normalization methods can also be selected in the embodiment of the present invention, such as function transformation, maximum and minimum normalization, etc., which are not limited here; the normalization processing at other positions in this scheme uses the normalization method here.

[0082] According to the above method, the information quantity index of each component signal of the current signal segment is obtained.

[0083] The amplitude growth analysis module 130 is used to match the component signals of the same frequency of the current signal segment with the remaining audio signal segments in the audio signal segment sequence, and obtain the amplitude growth index of each component signal of the current signal segment based on the difference in the matching conditions between the component signals of the same frequency of two adjacent audio signal segments and the current signal segment.

[0084] The normal sound of a high-voltage switch operating is typically similar across audio signal segments. During operation, the switch's jamming, travel, and vibration can all reflect its operating condition. When the high-voltage switch experiences an abnormality, it produces an abnormal sound. This abnormal sound gradually increases in the audio signal segments over multiple operating periods. The audio characteristics of audio signal segments within a sequence of audio signal segments gradually increase, further demonstrating the abnormal sound of the high-voltage switch.

[0085] The difference in matching between two adjacent audio signal segments and the component signals of the same frequency in the current signal segment shows the growth trend of the amplitude and frequency of the component signals of the same frequency in the two adjacent audio signal segments, and then the amplitude growth index is obtained to measure the abnormal sound content of the high-voltage switch in the component signal of the current signal segment.

[0086] See also Figure 2 , which shows a system structure diagram of a method for obtaining an amplitude growth index provided by an embodiment of the present invention. The system includes: a component definition unit 131, an envelope analysis unit 132, and an amplitude analysis unit 133.

[0087] The component definition unit 131 sequentially arranges the component signals of the same frequency of all audio signal segments in the audio signal segment sequence to obtain a component signal sequence for each frequency; for each component signal sequence, the last component signal in the component signal sequence is recorded as the current component signal, and the remaining component signals are recorded as target component signals.

[0088] Using component signals of the same frequency in different audio signal segments for analysis can reduce noise interference, improve the resolution of the audio signal segments, and enhance the robustness of the analysis.

[0089] The envelope analysis unit 132 obtains envelope analysis indicators of the current component signal and each target component signal according to the matching of data points on the envelope lines of the current component signal and each target component signal.

[0090] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the envelope analysis index includes: sequentially arranging the data points on each envelope line of each component signal to obtain an analysis sequence of the corresponding component signal; the analysis sequence includes: an upper analysis sequence and a lower analysis sequence; for each target component signal, matching the current component signal with the upper analysis sequence of the target component signal, selecting a matching data point for each data point in the upper analysis sequence of the current component signal from the upper analysis sequence of the target component signal; obtaining the time interval between the corresponding moments of the data point in the upper analysis sequence of the current component signal and its matching data point , recorded as the time difference index; obtain the average of the absolute values ​​of the differences between the amplitudes of the data points in the upper analysis sequence of the current component signal and its matching data points, recorded as the amplitude difference index; according to the time difference index and the amplitude difference index, obtain the local analysis index of the upper analysis sequence of the current component signal and the target component signal; the time difference index and the local analysis index are negatively correlated, and the amplitude difference index and the local analysis index are positively correlated; the average of the local analysis indexes of the upper analysis sequence and the lower analysis sequence of the current component signal and the target component signal is used as the envelope analysis index of the current component signal and the target component signal.

[0091] The envelope analysis index shows how well the current component signal matches the target component signal. The time difference index reflects the similarity between the current and target component signals, while the amplitude difference index reflects the degree of amplitude change between the two. If the matching distance between the current and target component signals remains unchanged (i.e., the time difference index is smaller), but the amplitude difference of the envelope curve (i.e., the amplitude difference index) is larger, indicating that the current component signal is more similar to the target component signal and the amplitude is increasing, then the proportion of abnormal noise from the high-voltage switch in the current component signal is greater. Therefore, a negative correlation mapping is performed on the time difference index, and the product of the mapping result and the amplitude difference index is normalized to obtain the local analysis index.

[0092] This embodiment uses an exponential function with a natural constant as the base for negative correlation mapping. The specific method is: first take the opposite number of the time difference index, and use the opposite number as the exponent of the exponential function with a natural constant as the base to achieve negative correlation mapping of the time difference index; and use the Norm function for normalization.

[0093] It should be noted that the data points on the envelope of the component signal are arranged in time sequence to obtain an analysis sequence; the analysis sequence corresponding to the upper envelope of the component signal is the upper analysis sequence, and the analysis sequence corresponding to the lower envelope is the lower analysis sequence. In the embodiment of the present invention, a dynamic time warping (DTW) algorithm is used to match the upper analysis sequence of the current component signal with that of the target component signal. The DTW algorithm is well known to those skilled in the art and will not be described in detail here.

[0094] The method for obtaining the local analysis indicators of the upper analysis sequence and the lower analysis sequence of the current component signal and the target component signal is the same; that is, the upper analysis sequence is replaced by the lower analysis sequence, and other contents remain unchanged.

[0095] Amplitude analysis unit 133: obtains the average of the differences between the envelope analysis indices of the current component signal and the remaining target component signals except the first target component signal and the previous target component signal as the amplitude growth index of the current component signal.

[0096] The local analysis index only considers the relationship between a single target component signal and the current component signal, and cannot describe the growth trend of the amplitude and frequency of the component signals of the same frequency in the audio signal segment. This embodiment presents the above-mentioned growth trend through the difference between the envelope analysis indicators of adjacent target component signals and the current component signal.

[0097] If the differences between the envelope analysis indicators of all target component signals and their adjacent previous target component signals and the current component signal are all greater than a constant 0 and the larger the difference, the amplitude and frequency of the component signals of the same frequency in the audio signal segment will show an increasing trend and the trend will be more obvious.

[0098] In a specific implementation of the embodiment of the present invention, the amplitude growth index of the current component signal is expressed as follows:

[0099]

[0100] Where EA is the amplitude growth index of the current component signal; M is the total number of target component signals in the component signal sequence where the current component signal is located, that is, the difference between the total number of component signals in the component signal sequence where the current component signal is located and the constant 1; SAM M-m SAM is the envelope analysis index of the current component signal and the Mm-th target component signal; M-m-1 It is the envelope analysis index of the current component signal and the Mm-1th target component signal.

[0101] It should be noted that if the amplitude growth index is larger, the component signals in the component signal sequence are more similar and the amplitude and frequency show a more obvious increasing trend, then the abnormal sound content of the high-voltage switch in the current component signal is higher; conversely, the weaker the similarity of the component signals in the component signal sequence, the greater the possibility that the information in the current component signal is environmental noise.

[0102] The amplitude growth index of the component signal of each frequency of the current signal segment is obtained. The amplitude growth index is obtained in the same way as the amplitude growth index of the current component signal.

[0103] The spectrum analysis module 140 is configured to obtain a spectrum distribution diffusion index of each component signal of the current signal segment according to a degree of spectrum variation of component signals of the same frequency in the audio signal segments in the audio signal segment sequence.

[0104] Because the abnormal sound content of the high-voltage switch in the component signal sequence gradually increases, the distribution of the speech signal in the frequency domain gradually spreads as the abnormal sound content increases. Therefore, the more obvious the diffusion trend in the frequency domain distribution of the audio signal segments in which the high-voltage switch operates in adjacent periods, the higher the abnormal sound content in the component signal of the current signal segment. The spectral distribution diffusion index is derived based on the degree of spectral variation of the component signals of the same frequency in the audio signal segments of the audio signal sequence, which measures the abnormal content of the high-voltage switch in the component signal of the current signal segment.

[0105] Preferably, in some possible implementation methods of the embodiments of the present invention, the method for obtaining the spectrum distribution diffusion index includes: subtracting the spectrum data of each target component signal from the spectrum data of the current component signal to obtain the difference spectrum data of each target component signal; arranging the difference spectrum data of all target component signals in sequence to obtain a difference spectrum sequence; normalizing the mean of the correlation coefficients of two adjacent difference spectrum data in the difference frequency sequence to obtain the spectrum change consistency value of the current component signal; obtaining the mean of the power of all data points of each difference spectrum data, recorded as the first mean power of the corresponding difference frequency data; normalizing the mean of the first mean power of all difference spectrum data in the difference frequency sequence to obtain the second mean power of the current component signal; obtaining the spectrum distribution diffusion index based on the second mean power and the spectrum change consistency value; the second mean power and the spectrum change consistency value are positively correlated with the spectrum distribution diffusion index.

[0106] It should be noted that the power of each frequency in the difference spectrum data of the target component signal is equal to the difference in power of the same frequency in the spectrum data of the current component signal and the target component signal. The correlation coefficient in this embodiment of the present invention is the Pearson correlation coefficient; in other embodiments, the correlation coefficient may also be the Spearman correlation coefficient, the Kendall rank correlation coefficient, the canonical correlation coefficient, etc.

[0107] The similarity index of two adjacent difference spectrum data points in the difference frequency sequence reflects the similarity of the spectral variations of some component signals in the component signal sequence, while the frequency variation consistency value reflects the consistency of the overall spectral variation of the component signals in the component signal sequence. As the proportion of abnormal sound increases during multiple high-voltage switch operations, the spectral difference variations of the audio signal segments during these multiple high-voltage switch operations become more similar and the spectrum becomes more dispersed. Therefore, the larger the spectral variation consistency value, the more dispersed the spectrum of the current component signal, and the larger the spectral distribution diffusion index.

[0108] To analyze the overall trend of the abnormal proportion of high-voltage switches during repeated operation, the average of the first mean power of all difference spectrum data in the difference frequency sequence is taken for analysis, considering the abnormal proportion of each adjacent high-voltage switch operation. A larger second mean power indicates a more significant increase in the abnormal proportion of the high-voltage switch during repeated operation, greater frequency information variation in the audio band of the high-voltage switch during repeated operation, a wider frequency information variation, and a larger spectrum distribution range, thus increasing the spectrum distribution diffusion index.

[0109] Therefore, the second average power is positively correlated with the spectrum variation consistency value and the spectrum distribution spread index. In this embodiment of the present invention, the product of the second average power of the current component signal and the spectrum variation consistency value is normalized to obtain the spectrum distribution spread index of the current component signal. The larger the spectrum distribution spread index, the greater the proportion of abnormal sound from the high-voltage switch in the current component signal. This embodiment uses the Norm function for normalization.

[0110] In the embodiment of the present invention, the correlation between the second average power and the spectrum change consistency value and the spectrum distribution spread index may also be constructed through other basic mathematical operations, which will not be limited or elaborated herein.

[0111] The method for obtaining the spectrum distribution spread index of each component signal of the current signal segment is the same as the method for obtaining the spectrum distribution spread index of the current component signal.

[0112] The voiceprint diagnosis module 150 is used to adjust the information content index according to the amplitude growth index and the spectrum distribution diffusion index, reconstruct the effective component signal selected from the component signal of the current signal segment using the adjusted information content index, and obtain the current effective signal segment; based on the current effective signal segment, diagnose the current working state of the high-voltage switch.

[0113] Both the amplitude growth index and the spectrum distribution diffusion index are used to measure the abnormal sound content of the high-voltage switch in the component signal of the current signal segment. The information amount index is adaptively adjusted by combining the two, so that the adjusted information amount index can more accurately reflect the abnormal sound information in the component signal of the current signal segment.

[0114] Preferably, in some possible implementation methods of the embodiments of the present invention, the method for obtaining the effective information index includes: obtaining the abnormal information coefficient of each component signal of the current signal segment based on the amplitude growth index and the spectrum distribution diffusion index; using the abnormal information coefficient to weightedly process the information index to obtain the effective information index of each component signal of the current signal segment.

[0115] Based on the data from amplitude growth analysis module 130 and spectrum analysis module 140, it can be seen that the larger the amplitude growth index and the spectrum distribution diffusion index, the higher the abnormal noise content of the high-voltage switch in the component signal of the current signal segment. Therefore, the amplitude growth index, the spectrum distribution diffusion index, and the abnormal information content coefficient are positively correlated.

[0116] In a specific implementation of the embodiment of the present invention, the effective information index of each component signal of the current signal segment is expressed as follows:

[0117] E i =Norm(EP i ×EA i )×X i

[0118] Where, E i is the effective information index of the i-th component signal of the current signal segment; EP i is the amplitude growth index of the i-th component signal of the current signal segment; EA i is the spectrum distribution diffusion index of the i-th component signal of the current signal segment; Norm(EP i ×EA i ) is the abnormal information coefficient of the i-th component signal of the current signal segment; X i is the information content index of the i-th component signal of the current signal segment; Norm is the normalization function.

[0119] For all component signals in the current signal segment, the component signals corresponding to the largest preset number of valid information indicators are used as valid component signals. All valid component signals in the current signal segment are reconstructed to obtain the current valid signal segment. In the current valid signal segment, the abnormal sound information of the high-voltage switch operation is relatively clear and has a large proportion.

[0120] It should be noted that, in the embodiment of the present invention, the preset number is equal to the result of rounding up one tenth of the total number of audio signal segments in the audio signal segment sequence, and the implementer can set it according to the specific situation; in the embodiment of the present invention, the EMD algorithm is used to reconstruct the effective component signal.

[0121] This embodiment uses the Mel-Frequency Cepstrum Coefficient (MFCC) algorithm to extract features from the current valid signal segment, obtaining a feature vector, recorded as the current diagnostic vector. The cosine similarity between the current diagnostic vector and a preset normal feature vector is then calculated. If the cosine similarity is greater than a preset normal threshold, the high-voltage switch is operating normally during the current operation. If the cosine similarity is less than or equal to the preset normal threshold, the high-voltage switch is operating abnormally during the current operation. The MACC algorithm is well known to those skilled in the art and will not be described in detail here.

[0122] If the cosine similarity between the current diagnostic vector and the preset normal feature vector is greater, it means that the audio of the high-voltage switch in the current operation is closer to the audio during normal operation, and the state of the high-voltage switch in the current operation is normal; conversely, the greater the difference between the audio of the high-voltage switch in the current operation and the audio during normal operation, the state of the high-voltage switch in the current operation is abnormal.

[0123] In a quiet environment, a sound collection device is used to capture an audio signal segment of a new high-voltage switch during its first operation. The MFCC algorithm is then used to extract the feature vector of this audio signal segment, which is recorded as the preset normal feature vector. It should be noted that a quiet environment avoids the influence of ambient noise. The new high-voltage switch is of the same type as the high-voltage switch to be monitored. The new high-voltage switch is typically in normal working order during its first operation and will not produce abnormal operating noise. Since the cosine similarity value ranges from -1 to 1, the preset normal threshold in this embodiment is an empirical value of 0.3, which can be set by the implementer according to the specific situation.

[0124] By selecting effective component signals for reconstruction, clearer abnormal sound audio data can be obtained while reducing the ambient noise in the audio signal segment. This allows the feature vector converted from the audio data to better reflect the abnormal sound during the working process of the high-voltage switch, making the abnormal diagnosis of the working status of the high-voltage switch more accurate.

[0125] So far, the present invention is completed.

[0126] Example 2:

[0127] Figure 3 A computer device diagram of a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis device provided by one embodiment of the present invention. For example, Figure 3 As shown, the computer device includes: a memory 201, a processor 202, and a computer program 203 stored in the memory 201 and running on the processor 202, wherein when the processor 202 executes the computer program 203, the computer device can execute any one of the high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis systems introduced above.

[0128] In addition, an embodiment of the present application also protects a device, which may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to execute a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system provided by an embodiment of the present application.

[0129] In this embodiment, the device can be divided into functional modules based on the above-described method examples. For example, each functional module can be mapped to a specific functional module, or two or more functions can be integrated into a single processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and represents only a logical functional division. In actual implementation, other division methods may be used.

[0130] It should be understood that the device provided in this embodiment is used to execute the above-mentioned high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system, and thus can achieve the same effect as the above-mentioned implementation method.

[0131] In the case of an integrated unit, the device may include a processing module and a storage module. When the device is applied to a device, the processing module may be used to control and manage the operation of the device. The storage module may be used to support the device in executing mutual program codes, etc.

[0132] The processing module may be a processor or controller that implements or executes various exemplary logic blocks, modules, and circuits disclosed herein. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor (DSP) and a microprocessor, and the storage module may be a memory.

[0133] Example 3:

[0134] This embodiment also provides a computer-readable storage medium, which stores computer program code. When the computer program code runs on a computer, the computer executes the above-mentioned related method steps to implement a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system provided by the above embodiment.

[0135] Among them, the device and computer-readable storage medium provided in this embodiment are used to execute the corresponding system provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding system provided above, and will not be repeated here.

[0136] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0137] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0138] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system, characterized in that: The system includes: A data acquisition module, used to obtain a sequence of audio signal segments when the high-voltage switch works continuously, and record the last audio signal segment in the sequence as the current signal segment; The information volume analysis module is used to decompose each audio signal segment to obtain component signals of different frequencies; according to the amplitude change trend of each component signal of the current signal segment and the power difference of different frequencies, the information volume index of each component signal of the current signal segment is obtained; An amplitude growth analysis module is used to match the component signals of the same frequency of the current signal segment with those of the remaining audio signal segments in the audio signal segment sequence, and obtain an amplitude growth index of each component signal of the current signal segment according to the difference in matching conditions between two adjacent audio signal segments and the component signals of the same frequency of the current signal segment; A spectrum analysis module, used to obtain a spectrum distribution diffusion index of each component signal of the current signal segment according to the spectrum change degree of the component signals of the same frequency of the audio signal segments in the audio signal segment sequence; The voiceprint diagnosis module is used to adjust the information volume index according to the amplitude growth index and the spectrum distribution diffusion index, reconstruct the effective component signal selected from the component signal of the current signal segment using the adjusted information volume index, and obtain the current effective signal segment; based on the current effective signal segment, diagnose the current working state of the high-voltage switch.

2. According to claim 1, a high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system is characterized in that: The obtaining of the information quantity index of each component signal of the current signal segment includes: For each component signal of the current signal segment, an envelope of the component signal is obtained, wherein the envelope includes an upper envelope and a lower envelope; the average of the absolute values ​​of the difference between the amplitudes of all two adjacent data points on each envelope of the component signal is obtained, and recorded as the local amplitude variation of the corresponding envelope; the average of the local amplitude variation of all envelopes of the component signal is used as the comprehensive amplitude variation of the component signal; Performing Fourier transformation on the component signal to obtain spectrum data, recording the frequency corresponding to the maximum power in the spectrum data as the main frequency of the component signal; in the spectrum data, a frequency segment is formed by continuous data points whose power is greater than a preset power threshold, and recording the frequency corresponding to the maximum power in each frequency segment as the analysis frequency of the component signal; The product of the absolute value of the difference between each analysis frequency of the component signal and the main frequency and the power of each analysis frequency is taken as the local distribution extensiveness of each analysis frequency of the component signal; the cumulative sum of the local distribution extensiveness of all analysis frequencies of the component signal is taken as the frequency distribution extensiveness of the component signal; According to the comprehensive amplitude variation and the frequency distribution extensiveness, an information quantity index of the component signal is obtained.

3. A high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system according to claim 2, characterized in that: The obtaining of the amplitude growth index of each component signal of the current signal segment includes: Arrange the component signals of the same frequency of all audio signal segments in the audio signal segment sequence in order to obtain a component signal sequence of each frequency; for each component signal sequence, record the last component signal in the component signal sequence as the current component signal, and record the remaining component signals as target component signals; Obtaining envelope analysis indicators of the current component signal and each target component signal according to matching conditions of data points on the envelope line of the current component signal and each target component signal; The average of the differences between the envelope analysis indexes of the current component signal and the remaining target component signals except the first target component signal and the previous target component signal thereof is obtained as the amplitude growth index of the current component signal.

4. A high-voltage switch intelligent voiceprint monitoring, diagnosis and analysis system according to claim 3, characterized in that: The step of obtaining envelope analysis indicators of the current component signal and each target component signal includes: The data points on each envelope line of each component signal are arranged in sequence to obtain an analysis sequence of the corresponding component signal; the analysis sequence includes: an upper analysis sequence and a lower analysis sequence; For each target component signal, matching the current component signal with the upper analysis sequence of the target component signal, and selecting a matching data point of each data point in the upper analysis sequence of the current component signal from the upper analysis sequence of the target component signal; Obtain the mean of the time intervals between the data points in the upper analysis sequence of the current component signal and the corresponding moments of the matching data points, recorded as the time difference index; obtain the mean of the absolute values ​​of the differences between the amplitudes of the data points in the upper analysis sequence of the current component signal and the matching data points, recorded as the amplitude difference index; According to the time difference index and the amplitude difference index, obtaining the local analysis index of the upper analysis sequence of the current component signal and the target component signal; the time difference index and the local analysis index are negatively correlated, and the amplitude difference index and the local analysis index are positively correlated; The average of the local analysis indicators of the upper analysis sequence and the lower analysis sequence of the current component signal and the target component signal is used as the envelope analysis indicator of the current component signal and the target component signal.

5. The intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switches according to claim 3 is characterized in that: The obtaining of the spectrum distribution diffusion index of each component signal of the current signal segment includes: Subtracting the spectrum data of each target component signal from the spectrum data of the current component signal to obtain difference spectrum data of each target component signal; arranging the difference spectrum data of all target component signals in sequence to obtain a difference spectrum sequence; Normalize the mean of the correlation coefficients of two adjacent difference spectrum data in the difference frequency sequence to obtain the spectrum change consistency value of the current component signal; Obtain the mean value of the power of all data points of each difference frequency spectrum data, recorded as the first mean power of the corresponding difference frequency data; normalize the mean value of the first mean power of all difference frequency spectrum data in the difference frequency sequence to obtain the second mean power of the current component signal; A spectrum distribution diffusion index is obtained according to the second average power and the spectrum change consistency value; the second average power and the spectrum change consistency value are both positively correlated with the spectrum distribution diffusion index.

6. The intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switches according to claim 1 is characterized in that: The method for acquiring the current valid signal segment includes: According to the amplitude growth index and the spectrum distribution diffusion index, the abnormal information coefficient of each component signal of the current signal segment is obtained; the information index is weighted by using the abnormal information coefficient to obtain the effective information index of each component signal of the current signal segment; For the valid information indicators of all component signals of the current signal segment, component signals corresponding to a maximum preset number of the valid information indicators are taken as valid component signals; all valid component signals of the current signal segment are reconstructed to obtain the current valid signal segment.

7. The intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switches according to claim 1 is characterized in that: The diagnosing the current working state of the high-voltage switch based on the current valid signal segment includes: Extract features from the current valid signal segment to obtain a feature vector, which is recorded as the current diagnostic vector; Obtain the cosine similarity between the current diagnostic vector and the preset normal feature vector. If the cosine similarity is greater than the preset normal threshold, the high-voltage switch is in a normal state in the current operation; if the cosine similarity is less than or equal to the preset normal threshold, the high-voltage switch is in an abnormal state in the current operation.

8. The intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switches according to claim 1 is characterized in that: The method for decomposing each audio signal segment is an EMD algorithm.

9. The intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switches according to claim 4 is characterized in that: The method for matching the upper analysis sequence of the current component signal with the target component signal is a DTW algorithm.

10. The intelligent voiceprint monitoring, diagnosis and analysis system for high-voltage switches according to claim 7 is characterized in that: The method for extracting features from the current valid signal segment is the MFCC algorithm.

Images