Method and device for detecting abnormal operation of equipment

Abstract

The present invention relates to the technical field of equipment noise detection, and specifically discloses a method and device for detecting abnormal operation of equipment, including: obtaining normal operation sound data and daily operation sound data, performing signal conversion, and obtaining the corresponding time-frequency graph and spectrum graph respectively; performing comparative calculation to obtain the time-frequency graph density comparison result and the spectrum graph energy comparison result, and judging whether there is abnormal data based on the above time-frequency graph density comparison result and spectrum graph energy comparison result; if abnormal data exists, obtaining the operation detection result of the target equipment based on the abnormal data. The present invention analyzes the time-frequency local properties of normal operation sound data and daily operation sound data through the time-frequency graph density comparison result, comprehensively describes the time-varying characteristics and frequency-varying characteristics, and then performs frequency domain analysis in combination with the spectrum graph energy comparison result, thereby effectively improving the accuracy of abnormal operation detection of industrial equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment noise detection, and in particular to a method and a device for detecting abnormal operation of equipment. Background Art

[0002] Since industrial equipment such as belt rollers are located in a complex and harsh environment, the surrounding noise is extremely high. When performing abnormality detection on industrial equipment, noise removal must be performed first. In the existing technology, traditional spectral subtraction is usually used for noise removal. However, traditional spectral subtraction noise removal cannot accurately and clearly process external noise of a specified frequency, and cannot accurately extract the abnormal operating sound of industrial equipment. There will be residual noise or excessive clipping of the original sound in the ordinary energy spectrum or power spectrum, resulting in distortion of the detected sound and waveform, and thus the original sound effect cannot be well restored, resulting in low accuracy of the results of industrial equipment abnormality detection.

[0003] Furthermore, due to the rapid changes in industrial equipment, the existing method of detecting anomalies by transforming and analyzing waveforms through traditional Fourier transform also has major problems, specifically:

[0004] (1) When facing a stationary signal, the traditional Fourier transform can clearly display the frequency of the signal. However, when the signal is non-stationary, the traditional Fourier transform cannot clearly display the frequency of the signal. The abnormality detection of industrial equipment operation is precisely to obtain the abnormality judgment result by analyzing the non-stationary signal;

[0005] (2) The traditional Fourier transform cannot combine time and frequency to detect anomalies in the sound data after noise removal of industrial equipment. The traditional Fourier transform is a global transform that either processes the signal completely in the time domain or completely in the frequency domain. It cannot express the local time-frequency properties of the signal, which is the most basic and critical property of non-stationary signals, resulting in low accuracy in detecting anomalies of industrial equipment.

[0006] Therefore, providing a method for detecting abnormal operation of industrial equipment with high accuracy has become a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0007] The present invention provides a method and a device for detecting abnormal operation of equipment, which solve the problems existing in the related art such as low accuracy of the results of abnormal operation detection of equipment.

[0008] As a first aspect of the present invention, a method for detecting abnormal operation of a device is provided, comprising:

[0009] Acquire normal operating sound data of the target device obtained after noise reduction and gain processing and daily operating sound data obtained after noise reduction and gain processing;

[0010] Performing signal conversion on the normal operation sound data and the daily operation sound data respectively to obtain a time-frequency graph and a spectrum graph corresponding to the normal operation sound data and a time-frequency graph and a spectrum graph corresponding to the daily operation sound data;

[0011] Comparing the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data, a time-frequency graph density comparison result is obtained; and comparing the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data, a spectrum graph energy comparison result is obtained;

[0012] Determine whether there is abnormal data in the daily operation of the target device based on the density comparison results of the time-frequency graph and the energy comparison results of the spectrum graph;

[0013] If there is abnormal data, the operation detection result of the target device is obtained based on the abnormal data.

[0014] Furthermore, signal conversion is performed on the normal operation sound data and the daily operation sound data to obtain a time-frequency graph and a spectrum graph corresponding to the normal operation sound data and a time-frequency graph and a spectrum graph corresponding to the daily operation sound data, respectively, including:

[0015] Performing signal transformation on the normal operating sound data by Hilbert-Huang transform to obtain a time-frequency graph and a spectrum graph corresponding to the normal operating sound data;

[0016] The daily operation sound data is transformed by Hilbert-Huang transform to obtain the time-frequency diagram and spectrum diagram corresponding to the daily operation sound data.

[0017] Furthermore, a time-frequency graph density comparison result is obtained by comparing the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data, and a spectrum graph energy comparison result is obtained by comparing the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data, including:

[0018] The time-frequency graph density comparison result is obtained by comparing and calculating the frequency density of a single frame of daily operation sound data and the frequency density of a single frame of normal operation sound data at the same time;

[0019] Determine whether there is suspected abnormal data in the daily operation of the target device based on the density comparison results of the time-frequency graph;

[0020] If there is suspected abnormal data, the spectrum energy comparison result of the suspected abnormal data is obtained by comparing the spectrum corresponding to the daily operation sound data with the spectrum corresponding to the normal operation sound data.

[0021] Furthermore, based on the density comparison results of the time-frequency graph, it is determined whether there is any suspected abnormal data in the daily operation of the target device, including:

[0022] Determine whether the density comparison result of the time-frequency graph is within the preset density comparison result range;

[0023] If the density comparison result of the time-frequency graph is within the preset density comparison result range, the single frame of daily operation sound data is marked as suspected abnormal data.

[0024] Furthermore, if there is suspected abnormal data, the spectrum energy comparison result of the suspected abnormal data is obtained by comparing the spectrum corresponding to the daily operation sound data with the spectrum corresponding to the normal operation sound data, including:

[0025] Performing a first comparative calculation based on the sound energy of each frame of normal operation sound data within a preset time period after the suspected abnormal data and the sound energy of the corresponding daily operation sound data to obtain a first spectrum energy comparison result;

[0026] Determining whether the energy comparison result of the first spectrum graph is within a preset energy comparison result range;

[0027] Counting the number of frames of daily operation sound data corresponding to the first spectrum energy comparison result within a preset energy comparison result range;

[0028] If the number of frames of daily operation sound data corresponding to the first spectrum graph energy comparison result within the preset energy comparison result range is greater than the preset number of frames, a second comparison calculation is performed based on the sound energy of the normal operation sound data within the preset time period and the sound energy of the daily operation sound data to obtain the second spectrum graph energy comparison result.

[0029] Furthermore, based on the density comparison results of the time-frequency graph and the energy comparison results of the spectrum graph, it is determined whether there is abnormal data in the daily operation of the target device, including:

[0030] comparing the energy comparison result of the second spectrum graph with a preset threshold;

[0031] If the energy comparison result of the second spectrum graph is greater than a preset threshold, the corresponding daily operation sound data is determined to be abnormal data.

[0032] Furthermore, obtaining normal operation sound data of the target device obtained after noise reduction and gain processing and daily operation sound data obtained after noise reduction and gain processing includes:

[0033] Acquire normal operating sound raw data and environmental noise data of the target device during normal operation;

[0034] Preprocessing the normal operating sound original data and the ambient noise data respectively according to short-time Fourier transform to obtain normal operating sound preprocessed data and first ambient noise intermediate data;

[0035] performing noise reduction gain processing on the normal operating sound preprocessing data according to the first environmental noise intermediate data to obtain normal operating sound data and second environmental noise intermediate data;

[0036] Obtain the original sound data of the target device during daily operation;

[0037] Preprocessing the daily operation sound original data according to short-time Fourier transform to obtain daily operation sound preprocessed data;

[0038] Performing noise reduction and gain processing on the daily operation sound preprocessing data according to the second environmental noise intermediate data to obtain daily operation sound data;

[0039] Among them, the first environmental noise intermediate data is iteratively updated in the process of performing noise reduction gain processing on the normal operation sound preprocessing data until the second environmental noise intermediate data is obtained, and the noise level represented by the second environmental noise intermediate data is greater than the noise level represented by the first environmental noise intermediate data.

[0040] Furthermore, performing noise reduction gain processing on the normal operating sound preprocessing data according to the first environmental noise intermediate data to obtain normal operating sound data and second environmental noise intermediate data includes:

[0041] Calculating a posterior signal-to-noise ratio of the normal operating sound preprocessing data based on the first environmental noise intermediate data;

[0042] Calculating the priori signal-to-noise ratio of the normal operating sound preprocessing data according to the posterior signal-to-noise ratio of the normal operating sound preprocessing data;

[0043] Calculating a gain ratio of the normal operating sound preprocessing data according to a posterior signal-to-noise ratio of the normal operating sound preprocessing data and a priori signal-to-noise ratio of the normal operating sound preprocessing data;

[0044] Gain processing is performed on the normal operating sound preprocessing data according to the gain ratio of the normal operating sound preprocessing data to obtain normal operating sound data and second environmental noise intermediate data.

[0045] Furthermore, performing noise reduction and gain processing on the daily operation sound preprocessing data according to the second environmental noise intermediate data to obtain daily operation sound data includes:

[0046] Calculating a posterior signal-to-noise ratio of the daily operation sound preprocessing data based on the second environmental noise intermediate data;

[0047] Calculating the priori signal-to-noise ratio of the daily operation sound preprocessing data based on the posterior signal-to-noise ratio of the daily operation sound preprocessing data;

[0048] Calculating a gain ratio of the daily operation sound preprocessing data according to a posterior signal-to-noise ratio of the daily operation sound preprocessing data and a priori signal-to-noise ratio of the daily operation sound preprocessing data;

[0049] Perform gain processing on the daily operation sound preprocessing data according to the gain ratio of the daily operation sound preprocessing data to obtain daily operation sound data.

[0050] As another aspect of the present invention, there is provided an apparatus for detecting abnormal operation of equipment, comprising: a data acquisition module for acquiring normal operation sound data of a target equipment obtained after noise reduction and gain processing and daily operation sound data obtained after noise reduction and gain processing;

[0051] A signal conversion module is used to perform signal conversion on the normal operation sound data and the daily operation sound data respectively, to obtain a time-frequency graph and a spectrum graph corresponding to the normal operation sound data and a time-frequency graph and a spectrum graph corresponding to the daily operation sound data;

[0052] A data comparison module is used to compare and calculate the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data to obtain a time-frequency graph density comparison result, and to compare and calculate the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data to obtain a spectrum graph energy comparison result;

[0053] Anomaly judgment module, used to judge whether there is abnormal data in the daily operation of the target device based on the density comparison results of the time-frequency graph and the energy comparison results of the spectrum graph;

[0054] The device operation detection module is used to obtain the operation detection result of the target device based on the abnormal data when there is abnormal data.

[0055] The present invention performs signal conversion on the normal operating sound data and daily operating sound data obtained after noise reduction gain processing of industrial equipment to obtain time-frequency diagrams and spectrum diagrams of the normal operating sound data, as well as time-frequency diagrams and spectrum diagrams of the daily operating sound data. The time-frequency local properties of the normal operating sound data and the daily operating sound data are analyzed through the density comparison results of the time-frequency diagrams, and the time-varying characteristics and frequency-varying characteristics of the normal operating sound data and the daily operating sound data are comprehensively described. The frequency domain properties are further analyzed separately through the energy comparison results of the spectrum diagrams, which effectively improves the accuracy of abnormal operation detection of industrial equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention.

[0057] Figure 1 This is a flow chart of the device operation abnormality detection method provided by the present invention.

[0058] Figure 2 This is another flow chart of the device operation abnormality detection method provided by the present invention.

[0059] Figure 3 The present invention provides a flow chart for performing signal conversion on normal operation sound data and daily operation sound data.

[0060] Figure 4 The present invention provides a flow chart for obtaining the density comparison results of the time-frequency graph and the energy comparison results of the spectrum graph.

[0061] Figure 5 This is a schematic diagram of the time-frequency graph density comparison results provided by the present invention.

[0062] Figure 6 This is a schematic diagram of the energy comparison results of the spectrum diagram provided by the present invention.

[0063] Figure 7 This is a flowchart for determining suspected abnormal data provided by the present invention.

[0064] Figure 8 The present invention provides a flow chart for obtaining energy comparison results of the spectrum of suspected abnormal data.

[0065] Figure 9 This is a flow chart for determining abnormal data provided by the present invention.

[0066] Figure 10 This is a flow chart of obtaining normal operation sound data and daily operation sound data provided by the present invention.

[0067] Figure 11 A flowchart of obtaining normal operating sound data and second environmental noise intermediate data according to first environmental noise intermediate data provided by the present invention.

[0068] Figure 12 Another flow chart of the present invention for obtaining normal operating sound data and second environmental noise intermediate data according to first environmental noise intermediate data.

[0069] Figure 13 This is a flow chart of obtaining daily operation sound data based on the second environmental noise intermediate data provided by the present invention.

[0070] Figure 14 Another flow chart of obtaining daily operation sound data according to the second environmental noise intermediate data provided by the present invention.

[0071] Figure 15 This is a structural block diagram of the device for detecting abnormal operation of equipment provided by the present invention. DETAILED DESCRIPTION

[0072] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0073] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0074] It should be noted that the terms "first," "second," and the like in the specification and claims of the present invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate for the embodiments of the present invention described herein. In addition, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatuses.

[0075] The embodiment of the present invention provides a method for detecting abnormal operation of a device. Figure 1 and Figure 2 Shown, including:

[0076] S100, obtaining normal operating sound data of a target device obtained after noise reduction and gain processing and daily operating sound data obtained after noise reduction and gain processing;

[0077] In the embodiments of the present invention, the raw normal operating sound data before noise reduction and gain processing and the raw daily operating sound data before noise reduction and gain processing can be obtained by sound sensors and audio signal collectors installed in the operating environment of industrial equipment. The industrial equipment here can be belt rollers, DC motors, shield machines, etc. It should be understood that the devices for obtaining the raw normal operating sound data and the raw daily operating sound data can be selected according to actual needs and are not limited in this invention.

[0078] The noise reduction and gain processing of the normal operation sound original data and the daily operation sound original data can be implemented by a processor with a noise reduction and gain function.

[0079] S200, performing signal conversion on the normal operation sound data and the daily operation sound data respectively to obtain a time-frequency graph and a spectrum graph corresponding to the normal operation sound data and a time-frequency graph and a spectrum graph corresponding to the daily operation sound data;

[0080] The spectrum diagram corresponding to the normal operating sound data of the industrial equipment and the spectrum diagram corresponding to the daily operating sound data can respectively characterize the frequency domain properties of the normal operating sound data of the industrial equipment and the frequency domain properties of the daily operating sound data; and the time-frequency diagram corresponding to the normal operating sound data of the industrial equipment can simultaneously characterize the time domain and frequency domain properties of the normal operating sound data of the industrial equipment, that is, the time-frequency local properties of the normal operating sound data of the industrial equipment. Similarly, the time-frequency diagram corresponding to the daily operating sound data of the industrial equipment can characterize the time-frequency local properties of the daily operating sound data of the industrial equipment.

[0081] S300, comparing and calculating the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data to obtain a time-frequency graph density comparison result, and comparing and calculating the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data to obtain a spectrum graph energy comparison result;

[0082] In the actual operation of industrial equipment, there is a situation where the frequency distribution of daily operation sound data and normal operation sound data in a certain period of time is the same, but the corresponding sound energy is different. Therefore, the detection of subsequent equipment operation anomalies requires not only the time-frequency graph density comparison results that can characterize the local time-frequency properties, but also the spectrum graph energy comparison results to complete accurate signal judgment.

[0083] S400, judging whether there is abnormal data in the daily operation of the target device based on the density comparison result of the time-frequency graph and the energy comparison result of the spectrum graph;

[0084] This embodiment analyzes the local time-frequency properties based on the time-frequency graph density comparison results, and completes the abnormality judgment of the daily operation sound data of the target device in combination with the spectrum graph energy comparison results.

[0085] S500: If abnormal data exists, obtain the operation detection result of the target device according to the abnormal data.

[0086] In summary, the embodiment of the present invention performs signal conversion on the normal operating sound data and daily operating sound data after noise reduction gain processing of the acquired industrial equipment to obtain the time-frequency diagram and spectrum diagram of the normal operating sound data, as well as the time-frequency diagram and spectrum diagram of the daily operating sound data. The time-frequency local properties of the normal operating sound data and the daily operating sound data are analyzed through the density comparison results of the time-frequency diagram, and the time-varying characteristics and frequency-varying characteristics of the normal operating sound data and the daily operating sound data are comprehensively described. The frequency domain properties are further analyzed separately through the energy comparison results of the spectrum diagram, which effectively improves the accuracy of abnormal operation detection of industrial equipment.

[0087] Furthermore, if Figure 3 As shown, signal transformation is performed on the normal operation sound data and the daily operation sound data to obtain the time-frequency graph and spectrum graph corresponding to the normal operation sound data and the time-frequency graph and spectrum graph corresponding to the daily operation sound data, including:

[0088] S210, performing signal transformation on the normal operating sound data by Hilbert-Huang transform to obtain a time-frequency graph and a spectrum graph corresponding to the normal operating sound data;

[0089] S220 , performing signal transformation on the daily operation sound data through Hilbert-Huang transform to obtain a time-frequency graph and a spectrum graph corresponding to the daily operation sound data.

[0090] The Hilbert-Huang Transform (HHT) mainly includes empirical mode decomposition (EMD) and Hilbert transform (HT). In the embodiment of the present invention, when the normal operation sound data and daily operation sound data of the non-stationary signal are transformed and processed by the Hilbert-Huang transform, specifically, the results of the empirical mode decomposition can be used to describe the changes and waveform characteristics of the sound signal in the time domain, and the Hilbert spectrum analysis of the sound signal through the Hilbert transform can describe the energy distribution and frequency changes of the sound signal at different frequencies.

[0091] Therefore, in this embodiment, the results obtained by performing signal transformation on the normal operating sound data and the daily operating sound data through the empirical mode decomposition and Hilbert transform in the Hilbert-Huang transform mainly cover the characteristics of the normal operating sound data and the daily operating sound data in the time domain and the frequency domain. By performing time-frequency local property analysis on the normal operating sound data and the daily operating sound data, the time-varying characteristics and frequency-varying characteristics of the normal operating sound data and the daily operating sound data can be fully described. The accuracy of the operation detection results of the target equipment obtained based on the above analysis can be greatly improved.

[0092] Furthermore, if Figure 4As shown, the time-frequency graph density comparison result is obtained by comparing the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data, and the spectrum graph energy comparison result is obtained by comparing the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data, including:

[0093] S310, performing a comparison calculation based on the frequency density of a single frame of daily operation sound data and the frequency density of a single frame of normal operation sound data at the same time to obtain a time-frequency graph density comparison result;

[0094] The calculation of the time-frequency graph density comparison result in this embodiment is mainly based on the integral ratio of the frequency density of a single frame of daily operation sound data and the frequency density of a single frame of normal operation sound data at the same time, as specifically shown in the following calculation formula (1):

[0095]

[0096] Wherein, z represents time, p(z) represents the frequency density of a single frame of daily operation sound data at time z, and g(z) represents the frequency density of a single frame of normal operation sound data at time z.

[0097] S320: Determine whether there is suspected abnormal data in the daily operation of the target device based on the time-frequency graph density comparison result;

[0098] Compare the time-frequency graph corresponding to the normal operating sound data with the time-frequency graph corresponding to the daily operating sound data to obtain the time-frequency graph density comparison result of each frame of normal operating sound data and daily operating sound data, that is, the comparison result of the frequency distribution density of the two. Make a preliminary judgment based on the time-frequency graph density comparison result that can characterize the local properties of time and frequency, and calibrate the suspected abnormal data.

[0099] S330: If there is suspected abnormal data, a comparison calculation is performed based on the spectrum corresponding to the daily operation sound data and the spectrum corresponding to the normal operation sound data to obtain a spectrum energy comparison result of the suspected abnormal data.

[0100] The density comparison results of the time-frequency graph provided by the embodiment of the present invention are as follows: Figure 5 As shown in the figure, the energy comparison results of the spectrum are as follows: Figure 6 shown.

[0101] Based on the suspected abnormal data obtained from the comparison results of the frequency distribution density, the energy comparison results of the spectrum diagram of the suspected abnormal data are further analyzed to determine whether the suspected abnormal data is indeed abnormal and obtain the operation detection results of the target equipment.

[0102] In summary, the embodiment of the present invention performs signal conversion on the normal operating sound data and the daily operating sound data after noise reduction gain processing of the industrial equipment to obtain the time-frequency graph and spectrum graph of the normal operating sound data, as well as the time-frequency graph and spectrum graph of the daily operating sound data. The suspected abnormal data is judged according to the density comparison result of the time-frequency graph of the normal operating sound data and the daily operating sound data, and the suspected abnormal data is further analyzed according to the energy comparison result of the spectrum graph of the normal operating sound data and the daily operating sound data, and finally the abnormality detection result of the daily operating status of the industrial equipment is obtained. On the basis of analyzing and calibrating the time-frequency local properties of the normal operating sound data and the daily operating sound data to determine the suspected abnormal data, the frequency domain properties of the suspected abnormal data are further analyzed separately, so as to obtain the target equipment operation detection result, thereby effectively improving the accuracy of the abnormality detection of the industrial equipment operation.

[0103] Furthermore, if Figure 7 As shown, based on the time-frequency graph density comparison results, determine whether there is suspected abnormal data in the daily operation of the target device, including:

[0104] S321, determining whether the density comparison result of the time-frequency graph is within a preset density comparison result range;

[0105] Specifically, in this embodiment, the preset density comparison result range is a density estimation function b, b∈(0,1).

[0106] S322: If the density comparison result of the time-frequency graph is within the preset density comparison result range, mark the single frame of daily operation sound data as suspected abnormal data;

[0107] The specific method for judging suspected abnormal data is as follows:

[0108]

[0109] Where b represents the density estimation function, b∈(0,1).

[0110] When the time-frequency graph density comparison result obtained by comparing the frequency density of a single frame of daily operation sound data and the frequency density of a single frame of normal operation sound data at the same time satisfies the above judgment formula (2), the single frame of daily operation sound data is marked as suspected abnormal data.

[0111] Furthermore, if Figure 8 As shown, if there is suspected abnormal data, the spectrum energy comparison result of the suspected abnormal data is obtained by comparing the spectrum corresponding to the daily operation sound data with the spectrum corresponding to the normal operation sound data, including:

[0112] S331, performing a first comparative calculation based on the sound energy of each frame of normal operation sound data within a preset time period after the suspected abnormal data and the sound energy of the corresponding daily operation sound data to obtain a first spectrum energy comparison result;

[0113] The first comparative calculation of sound energy provided in this embodiment is mainly based on the relative deviation of the sound energy of each frame of target normal operation sound data within a preset time period after the suspected abnormal data relative to the sound energy of the corresponding target daily operation sound data, as shown in the following formula (3):

[0114]

[0115] Among them, z f Indicates the time corresponding to the suspected abnormal data, m(z f ) represents z f The sound energy of a single frame of normal operation sound data at the moment, n(z f ) represents z f The sound energy of a single frame of daily running sound data at a moment.

[0116] S332, determining whether the energy comparison result of the first spectrum graph is within a preset energy comparison result range;

[0117] Specifically, in this embodiment, the preset energy comparison result range is the energy difference ratio value c, c∈(0,1), and the judgment method is shown in the following formula (4):

[0118]

[0119] Where c represents the energy difference ratio value, c∈(0,1).

[0120] S333, counting the number of frames of daily operation sound data corresponding to the first spectrum energy comparison result within a preset energy comparison result range;

[0121] That is, the number of frames q of the corresponding daily operation sound data that satisfies the above formula (4) within the preset time period after the suspected abnormal data is counted.

[0122] S334. If the number of frames of the daily operation sound data corresponding to the first spectrum graph energy comparison result within the preset energy comparison result range is greater than the preset number of frames, a second comparison calculation is performed based on the sound energy of the normal operation sound data within the preset time period and the sound energy of the daily operation sound data to obtain a second spectrum graph energy comparison result.

[0123] After statistics, when the number of frames q of daily operation sound data that meet formula (4) within a preset time period after the suspected abnormal data is greater than the preset number of frames Q, a second comparison calculation is performed. Specifically, the second comparison calculation method for obtaining the energy comparison result of the second spectrum provided in this embodiment is shown in the following formula (5):

[0124]

[0125] Among them, z r Indicates the preset time period, z r =z f +0.1×N * , N * Represents a positive integer, N * The value can be set according to actual needs.

[0126] Furthermore, if Figure 9 As shown, based on the density comparison results of the time-frequency graph and the energy comparison results of the spectrum graph, it is determined whether there is abnormal data in the daily operation of the target device, including:

[0127] S410, comparing the energy comparison result of the second spectrum graph with a preset threshold;

[0128] The preset threshold set in this embodiment is specifically 0.

[0129] S420: If the energy comparison result of the second spectrum graph is greater than a preset threshold, the corresponding daily operation sound data is determined to be abnormal data.

[0130] The specific judgment method is as follows:

[0131]

[0132] When the result obtained based on formula (6) is greater than 0, it means that the sound energy of the corresponding single-frame daily operation sound data changes rapidly and has no downward trend. The corresponding daily operation sound data is determined to be abnormal data. The operation detection result of the target device is obtained based on the abnormal data, and an alarm is issued. On the contrary, when the result obtained based on formula (6) is less than or equal to 0, the corresponding daily operation sound data is determined to be non-abnormal data.

[0133] In summary, the present embodiment provides a method for detecting abnormal operation of equipment. The method first compares the frequency density of normal operation sound data with the frequency density of daily operation sound data based on a time-frequency diagram to obtain suspected abnormal data. Then, the energy of the suspected abnormal data is compared based on a spectrum diagram. If, within a preset time after the suspected abnormal data, the number of frames in which the sound energy of daily operation sound data is greater than the sound energy of normal operation sound data is greater than the preset number of frames, and the sound energy of the daily operation sound data in this frame changes rapidly and has no downward trend, then the target equipment is judged to be abnormal in operation and a corresponding alarm is issued. Based on the analysis of the time-frequency local properties of normal operation sound data and daily operation sound data to locate suspected abnormal data, the embodiment of the present invention further analyzes the frequency domain properties of the suspected abnormal data, thereby effectively improving the accuracy of abnormal operation detection of industrial equipment.

[0134] Furthermore, if Figure 10 and Figure 11 As shown, the normal operating sound data of the target device obtained after the noise reduction and gain processing and the daily operating sound data obtained after the noise reduction and gain processing are obtained, including:

[0135] S110, obtaining normal operating sound original data of the target device during normal operation and environmental noise data of the target device during normal operation;

[0136] Specifically, the normal operating sound original data X when the target device is operating normally and the environmental noise data D when the target device is operating normally are acquired.

[0137] S120, preprocessing the normal operating sound original data and the ambient noise data according to short-time Fourier transform to obtain normal operating sound preprocessed data and first ambient noise intermediate data;

[0138] Short-time Fourier transform is used to preprocess the normal operation sound original data X and the ambient noise data D to obtain the normal operation sound preprocessed data, as shown in the following formula (7):

[0139]

[0140] Wherein, h(t) represents the window function, t represents the center value of the window function, τ represents the time period, ω represents the angular frequency, i represents an imaginary number, and x(τ) represents the original data of the normal operating sound in the τ time period.

[0141] According to the above normal operating sound preprocessing data, the noisy phase angle PhAR of each frame of the normal operating sound preprocessing data is further obtained. x And the amplitude mean AmVR of each frame x , and calculate the feature dimension D of each frame x And the frame length of each frame Tx .

[0142] The environmental noise data D is preprocessed by short-time Fourier transform to obtain the first environmental noise intermediate data, as shown in the following formula (8):

[0143]

[0144] Wherein, h(t) represents the window function, t represents the center value of the window function, τ represents the time period, ω represents the angular frequency, i represents the imaginary number, and d(τ) represents the environmental noise data of the τ time period.

[0145] The first environmental noise intermediate data includes the amplitude mean λ of each frame of the first environmental noise intermediate data obtained further. d1 , and the corresponding sound energy η of each frame is further calculated d1 .

[0146] S130, performing noise reduction and gain processing on the normal operating sound preprocessed data according to the first environmental noise intermediate data to obtain normal operating sound data and second environmental noise intermediate data;

[0147] Normal operating sound data is obtained after noise reduction and gain processing The details are shown in the following formula (9):

[0148]

[0149] Among them, the first environmental noise intermediate data is iteratively updated in the process of performing noise reduction gain processing on the normal operation sound preprocessing data until the second environmental noise intermediate data is obtained, and the noise level represented by the second environmental noise intermediate data is greater than the noise level represented by the first environmental noise intermediate data.

[0150] Specifically, the sound energy η of the first ambient noise intermediate data of the 0th frame obtained by short-time Fourier transform is d1 (0) First, the normal operating sound data of the 0th frame is subjected to noise reduction gain processing to obtain the normal operating sound data of the 0th frame after noise reduction gain processing, and the sound energy η of the first environmental noise intermediate data of the 1st frame is calculated based on the normal operating sound data of the 0th frame after noise reduction gain processing. d1 (1) Update and obtain the sound energy of the first environmental noise intermediate data of the updated first frame The first frame of normal operating sound preprocessing data is used to perform noise reduction gain processing on the first frame of normal operating sound data to obtain the first frame of normal operating sound data. Similarly, the first environmental noise data is iteratively updated in the process of performing noise reduction gain processing on the normal operating sound preprocessing data until the second environmental noise intermediate data is obtained. The second environmental noise intermediate data includes the amplitude mean λ d2, and the corresponding sound energy η d2 , the sound energy of the updated first environmental noise intermediate data Greater than the sound energy η of the first environmental noise intermediate data before updating d1 (k) The noise level represented by the second environmental noise intermediate data is greater than the noise level represented by the first environmental noise intermediate data.

[0151] It should be understood that the method of pre-processing the normal operation sound original data X and the environmental noise data D using short-time Fourier transform can meet the processing requirements of unstable and irregular sound data.

[0152] S140, obtaining original data of daily operation sound of the target device during daily operation;

[0153] Specifically, the original data Y of daily operation sound of the target device during daily operation is obtained.

[0154] S150, pre-processing the daily operation sound original data according to short-time Fourier transform to obtain daily operation sound pre-processed data;

[0155] The daily operation sound original data Y is preprocessed by short-time Fourier transform, and the obtained daily operation sound preprocessed data is specifically shown in the following formula (10):

[0156]

[0157] Wherein, h(t) represents the window function, t represents the center value of the window function, τ represents the time period, ω represents the angular frequency, i represents the imaginary number, and y(τ) represents the original data of daily operation sound in the τ time period.

[0158] According to the above daily operation sound preprocessing data, the noisy phase angle PhAR of each frame of the daily operation sound preprocessing data is further obtained. y And the amplitude mean AmVR of each frame y , and calculate the feature dimension D of each frame y And the frame length of each frame T y .

[0159] S160 : Perform noise reduction and gain processing on the daily operation sound preprocessing data according to the second environmental noise intermediate data to obtain daily operation sound data.

[0160] Daily operation sound data is obtained after noise reduction and gain processing The details are shown in the following formula (11):

[0161]

[0162] Furthermore, if Figure 12 As shown, performing noise reduction gain processing on the normal operating sound preprocessing data according to the first environmental noise intermediate data to obtain the normal operating sound data and the second environmental noise intermediate data includes:

[0163] S131. Calculating a posteriori signal-to-noise ratio of normal operating sound preprocessing data based on the first environmental noise intermediate data;

[0164] Specifically, the calculation formula of the posterior signal-to-noise ratio of the normal operation sound preprocessing data is shown in the following formula (12):

[0165]

[0166] Among them, γ x (k) represents the posterior signal-to-noise ratio of the normal operation sound preprocessing data of the kth frame, η x (k) represents the sound energy of the kth frame normal operation sound preprocessing data, λ d1 (k) represents the amplitude mean of the intermediate data of the first environmental noise of the kth frame.

[0167] S132, calculating the a priori signal-to-noise ratio of the normal operating sound preprocessing data based on the a posteriori signal-to-noise ratio of the normal operating sound preprocessing data;

[0168] Specifically, the calculation formula of the priori signal-to-noise ratio of the normal operation sound preprocessing data is shown in the following formula (13):

[0169]

[0170] Among them, ξ x (k) represents the prior signal-to-noise ratio of the k-th frame normal operating sound preprocessing data, a represents the gain estimation coefficient, X(k) represents the k-th frame normal operating sound original data, λ d1 (k) represents the amplitude mean of the intermediate data of the first environmental noise in the kth frame, γ x (k) represents the posterior signal-to-noise ratio of the normal operation sound preprocessing data of the kth frame, η x (k-1) represents the sound energy of the k-1th frame of normal operation sound preprocessing data, η d1 (k) represents the sound energy of the first environmental noise intermediate data of the kth frame.

[0171] S133, calculating a gain ratio of the normal operating sound preprocessing data according to the a posteriori signal-to-noise ratio of the normal operating sound preprocessing data and the a priori signal-to-noise ratio of the normal operating sound preprocessing data;

[0172] The specific calculation formula for the gain ratio of normal operation sound preprocessing data is shown in the following formula (14):

[0173]

[0174] Among them, H x (k) represents the gain ratio of the normal operation sound preprocessing data of the kth frame, γ x (k) represents the posterior signal-to-noise ratio of the normal operating sound preprocessing data of the kth frame, ξ x (k) represents the priori signal-to-noise ratio of the normal operation sound preprocessing data of the kth frame, and I0 and I1 both represent Bessel functions.

[0175] S134 , performing gain processing on the normal operating sound preprocessing data according to the gain ratio of the normal operating sound preprocessing data to obtain normal operating sound data and second environmental noise intermediate data.

[0176] The specific calculation formula for gain processing of normal operation sound preprocessing data is shown in the following formula (15):

[0177]

[0178] Among them, λ x (k) represents the amplitude mean of the normal operation sound preprocessing data of the kth frame, H represents the amplitude mean of the normal operation sound preprocessing data of the kth frame obtained after gain processing, x (k) represents the gain ratio of the normal operation sound preprocessing data of the kth frame.

[0179] The gain obtained Record the corresponding frame of the output audio and The noisy phase angle PhAR of the normal sound preprocessing data of the kth frame x (k) are multiplied to obtain the normal operation sound data of the kth frame, as shown in formula (16):

[0180]

[0181] Where Z is a set of integers.

[0182] Next, the signal-to-noise energy for VAD calculation (VoiceActivityDetection) is obtained. The specific calculation formula is shown in the following formula (17):

[0183]

[0184] Then, the amplitude mean of the normal operation sound preprocessing data of the kth frame after gain is calculated. The energy mean η of the normal operation sound preprocessing data of the kth frame x (k) Perform single-frame calculation. The specific calculation formula is shown in the following formula (18):

[0185]

[0186] in, It represents the energy mean of the normal operation sound preprocessing data of the kth frame obtained after single frame calculation.

[0187] according to and SNR x (k) The energy mean η of the first ambient noise intermediate data of the k+1th frame d1 (k+1) is updated to obtain the updated Used for gain calculation of normal operating sound preprocessing data of the k+1th frame.

[0188] Specifically, the updated The expression of is shown in formula (19):

[0189]

[0190] The first environmental noise data is iteratively updated in the process of performing noise reduction gain processing on the normal operation sound preprocessing data until the second environmental noise intermediate data is obtained. The second environmental noise intermediate data includes the amplitude mean λ d2 , and the corresponding sound energy η d2 Specifically, the relationship between the second environmental noise intermediate data and the first environmental noise intermediate data is shown in formula (20):

[0191]

[0192] Furthermore, if Figure 13 and Figure 14 As shown, performing noise reduction and gain processing on the daily operation sound preprocessing data according to the second environmental noise intermediate data to obtain daily operation sound data includes:

[0193] S161. Calculating a posterior signal-to-noise ratio of the daily operation sound preprocessing data based on the second environmental noise intermediate data;

[0194] The specific calculation formula for the posterior signal-to-noise ratio of daily operation sound preprocessing data is shown in the following formula (21):

[0195]

[0196] Among them, γ y (k) represents the posterior signal-to-noise ratio of the k-th frame of daily operation sound preprocessing data, η y (k) represents the energy mean of the k-th frame of daily operation sound preprocessing data, λ d2 (k) represents the amplitude mean of the intermediate data of the second environmental noise in the kth frame.

[0197] S162. Calculating a priori signal-to-noise ratio of the daily operation sound preprocessing data based on the posterior signal-to-noise ratio of the daily operation sound preprocessing data;

[0198] The specific calculation formula for the priori signal-to-noise ratio of daily operation sound preprocessing data is shown in the following formula (22):

[0199]

[0200] Among them, ξ y (k) represents the prior signal-to-noise ratio of the k-th frame of daily operation sound preprocessing data, a is the gain estimation coefficient, Y(k) represents the k-th frame of daily operation sound original data, λ d2 (k) represents the amplitude mean of the second ambient noise intermediate data of the kth frame, γ y (k) represents the posterior signal-to-noise ratio of the k-th frame of daily operation sound preprocessing data, η y (k-1) represents the sound energy of the k-1th frame of daily operation sound preprocessing data, η d2 (k) represents the sound energy of the second environmental noise intermediate data of the kth frame.

[0201] S163, calculating a gain ratio of the daily operation sound preprocessing data according to the posterior signal-to-noise ratio of the daily operation sound preprocessing data and the prior signal-to-noise ratio of the daily operation sound preprocessing data;

[0202] The specific calculation formula for the gain ratio of daily operation sound preprocessing data is shown in the following formula (23):

[0203]

[0204] Among them, H y (k) represents the gain ratio of the k-th frame of daily operation sound preprocessing data, ξ y (k) represents the prior signal-to-noise ratio of the k-th frame of daily operation sound preprocessing data, γ y (k) represents the posterior signal-to-noise ratio of the k-th frame of daily operation sound preprocessing data, and I0 and I1 both represent Bessel functions.

[0205] S164 , performing gain processing on the daily operation sound preprocessing data according to the gain ratio of the daily operation sound preprocessing data to obtain daily operation sound data.

[0206] The specific calculation formula for gain processing of daily operation sound preprocessing data is shown in the following formula (24):

[0207]

[0208] Among them, λ y(k) represents the amplitude mean of the k-th frame of daily operation sound preprocessing data, H represents the amplitude mean of the k-th frame of daily operation sound preprocessing data obtained after gain processing, y (k) represents the gain ratio of the k-th frame of daily operation sound preprocessing data.

[0209] The gain obtained Record the corresponding frame of the output audio and The noisy phase angle PhAR of the normal sound preprocessing data of the kth frame y (k) multiply to obtain the k-th frame of daily operation sound data:

[0210]

[0211] The signal-to-noise energy used for VAD calculation is obtained. The specific calculation formula is shown in the following formula (25):

[0212]

[0213] According to the amplitude mean of the k-th frame daily operation sound preprocessing data after gain The energy mean η of the k-th frame daily operation sound preprocessing data y (k) Perform single frame calculation. The specific calculation formula is shown in the following formula (26):

[0214]

[0215] in, Represents the energy mean of the k-th frame of daily operation sound preprocessing data obtained after single-frame calculation.

[0216] according to and SNR y (k) The energy mean η of the second ambient noise intermediate data of the k+1th frame d2 (k+1) is updated to obtain the updated Used for the gain calculation of the k+1 frame daily operation sound preprocessing data. Specifically, the updated The expression of is shown in formula (27):

[0217]

[0218] The device operation abnormality detection method provided by the embodiment of the present invention iteratively updates the first environmental noise intermediate data in the target device operation environment during the process of processing the normal operation sound preprocessing data of the target device to achieve noise reduction and gain processing of the target device normal operation sound preprocessing data, and then iteratively updates the second environmental noise intermediate data during the subsequent processing of the target device daily operation sound preprocessing data using the obtained second environmental noise intermediate data to achieve noise reduction and gain processing of the target device daily operation sound preprocessing data. This method does not require the establishment of too many comparative sound data models, nor does it require too many restrictions on the detection environment, and can adapt to device operation abnormality detection in a variety of environments.

[0219] Another embodiment of the present invention provides a device 010 for detecting abnormal operation of equipment, such as Figure 15 Shown, including:

[0220] The data acquisition module 011 is used to acquire the normal operation sound data of the target device obtained after the noise reduction and gain processing and the daily operation sound data obtained after the noise reduction and gain processing;

[0221] The signal conversion module 012 is used to perform signal conversion on the normal operation sound data and the daily operation sound data respectively, and obtain the time-frequency graph and spectrum graph corresponding to the normal operation sound data and the time-frequency graph and spectrum graph corresponding to the daily operation sound data;

[0222] The data comparison module 013 is used to compare and calculate the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data to obtain a time-frequency graph density comparison result, and to compare and calculate the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data to obtain a spectrum graph energy comparison result;

[0223] Anomaly determination module 014, used to determine whether there is abnormal data in the daily operation of the target device based on the time-frequency graph density comparison result and the spectrum graph energy comparison result;

[0224] The device operation detection module 015 is used to obtain the operation detection result of the target device according to the abnormal data when there is abnormal data.

[0225] The equipment operation abnormality detection device 010 provided in an embodiment of the present invention performs signal conversion on the normal operation sound data and daily operation sound data obtained after noise reduction gain processing of the industrial equipment to obtain the time-frequency diagram and spectrum diagram of the normal operation sound data, as well as the time-frequency diagram and spectrum diagram of the daily operation sound data. On the basis of analyzing the time-frequency local properties of the normal operation sound data and the daily operation sound data through the density comparison results of the time-frequency diagram, the frequency domain properties are further analyzed separately through the energy comparison results of the spectrum diagram, thereby effectively improving the accuracy of industrial equipment operation abnormality detection.

[0226] The specific working principle and effect of the device operation abnormality detection device 010 provided in the embodiment of the present invention can be referred to the description of the device operation abnormality detection method above, and will not be repeated here.

[0227] It will be understood that the above embodiments are merely exemplary embodiments for illustrating the principles of the present invention, and the present invention is not limited thereto. Those skilled in the art will appreciate that various modifications and improvements can be made without departing from the spirit and substance of the present invention, and such modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for detecting abnormal operation of equipment, characterized in that: include: Acquire normal operating sound data of the target device obtained after noise reduction and gain processing and daily operating sound data obtained after noise reduction and gain processing; Performing signal conversion on the normal operation sound data and the daily operation sound data respectively to obtain a time-frequency graph and a spectrum graph corresponding to the normal operation sound data and a time-frequency graph and a spectrum graph corresponding to the daily operation sound data; Comparing the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data to obtain a time-frequency graph density comparison result, and comparing the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data to obtain a spectrum graph energy comparison result; Determine whether there is abnormal data in the daily operation of the target device based on the density comparison result of the time-frequency graph and the energy comparison result of the spectrum graph; If there is abnormal data, obtaining the operation detection result of the target device according to the abnormal data; The comparing and calculating the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data to obtain a time-frequency graph density comparison result, and the comparing and calculating the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data to obtain a spectrum graph energy comparison result, including: The time-frequency graph density comparison result is obtained by comparing and calculating the frequency density of a single frame of daily operation sound data and the frequency density of a single frame of normal operation sound data at the same time; Determine whether there is suspected abnormal data in the daily operation of the target device based on the density comparison result of the time-frequency graph; If there is suspected abnormal data, a comparison calculation is performed based on the spectrum corresponding to the daily operation sound data and the spectrum corresponding to the normal operation sound data to obtain an energy comparison result of the spectrum of the suspected abnormal data.

2. The method according to claim 1, characterized in that The performing signal conversion on the normal operation sound data and the daily operation sound data to obtain a time-frequency graph and a spectrum graph corresponding to the normal operation sound data and a time-frequency graph and a spectrum graph corresponding to the daily operation sound data respectively includes: Performing signal transformation on the normal operating sound data by Hilbert-Huang transform to obtain a time-frequency graph and a spectrum graph corresponding to the normal operating sound data; The daily operation sound data is subjected to signal transformation through Hilbert-Huang transform to obtain a time-frequency diagram and a spectrum diagram corresponding to the daily operation sound data.

3. The method according to claim 1, characterized in that The determining whether there is suspected abnormal data in the daily operation of the target device according to the time-frequency graph density comparison result includes: Determining whether the density comparison result of the time-frequency graph is within a preset density comparison result range; If the density comparison result of the time-frequency graph is within a preset density comparison result range, the single frame of daily operation sound data is marked as suspected abnormal data.

4. The method according to claim 1, wherein If there is suspected abnormal data, a comparison calculation is performed based on the spectrum corresponding to the daily operation sound data and the spectrum corresponding to the normal operation sound data to obtain a spectrum energy comparison result of the suspected abnormal data, including: Performing a first comparative calculation based on the sound energy of each frame of normal operation sound data within a preset time period after the suspected abnormal data and the sound energy of the corresponding daily operation sound data to obtain a first spectrum energy comparison result; Determining whether the energy comparison result of the first spectrum graph is within a preset energy comparison result range; Counting the number of frames of daily operation sound data corresponding to the energy comparison result of the first spectrum graph within a preset energy comparison result range; If the number of frames of daily operation sound data corresponding to the first spectrum graph energy comparison result within the preset energy comparison result range is greater than the preset number of frames, a second comparison calculation is performed based on the sound energy of the normal operation sound data within the preset time period and the sound energy of the daily operation sound data to obtain the second spectrum graph energy comparison result.

5. The method according to claim 4, characterized in that The determining whether there is abnormal data in the daily operation of the target device according to the density comparison result of the time-frequency graph and the energy comparison result of the spectrum graph includes: Comparing the energy comparison result of the second spectrum graph with a preset threshold; If the energy comparison result of the second spectrum graph is greater than the preset threshold, the corresponding daily operation sound data is determined to be abnormal data.

6. The method according to claim 1, characterized in that The obtaining of normal operating sound data of the target device obtained after noise reduction and gain processing and daily operating sound data obtained after noise reduction and gain processing includes: Acquire normal operating sound raw data and environmental noise data of the target device during normal operation; Preprocessing the normal operating sound original data and the ambient noise data according to short-time Fourier transform to obtain normal operating sound preprocessed data and first ambient noise intermediate data; performing noise reduction gain processing on the normal operating sound preprocessed data according to the first environmental noise intermediate data to obtain normal operating sound data and second environmental noise intermediate data; Obtain the original sound data of the target device during daily operation; Preprocessing the daily operation sound original data according to short-time Fourier transform to obtain daily operation sound preprocessed data; performing noise reduction and gain processing on the daily operation sound preprocessing data according to the second environmental noise intermediate data to obtain daily operation sound data; Among them, the first environmental noise intermediate data is iteratively updated in the process of performing noise reduction gain processing on the normal operation sound preprocessing data until the second environmental noise intermediate data is obtained, and the noise level represented by the second environmental noise intermediate data is greater than the noise level represented by the first environmental noise intermediate data.

7. The method according to claim 6, characterized in that The performing noise reduction gain processing on the normal operating sound preprocessed data according to the first environmental noise intermediate data to obtain normal operating sound data and second environmental noise intermediate data includes: Calculating a posterior signal-to-noise ratio of the normal operating sound preprocessing data according to the first environmental noise intermediate data; Calculating the a priori signal-to-noise ratio of the normal operating sound preprocessing data according to the a posteriori signal-to-noise ratio of the normal operating sound preprocessing data; Calculating a gain ratio of the normal operating sound preprocessing data according to a posterior signal-to-noise ratio of the normal operating sound preprocessing data and a priori signal-to-noise ratio of the normal operating sound preprocessing data; Gain processing is performed on the normal operating sound preprocessing data according to the gain ratio of the normal operating sound preprocessing data to obtain normal operating sound data and second environmental noise intermediate data.

8. The method according to any one of claims 6 or 7, characterized in that: The performing noise reduction and gain processing on the daily operation sound preprocessing data according to the second environmental noise intermediate data to obtain the daily operation sound data includes: Calculating a posterior signal-to-noise ratio of the daily operation sound preprocessing data based on the second environmental noise intermediate data; Calculating the priori signal-to-noise ratio of the daily operation sound preprocessing data according to the posterior signal-to-noise ratio of the daily operation sound preprocessing data; Calculating a gain ratio of the daily operation sound preprocessing data according to a posterior signal-to-noise ratio of the daily operation sound preprocessing data and a priori signal-to-noise ratio of the daily operation sound preprocessing data; Performing gain processing on the daily operation sound preprocessing data according to the gain ratio of the daily operation sound preprocessing data to obtain daily operation sound data.

9. A device for detecting abnormal operation of equipment, used to implement the method for detecting abnormal operation of equipment according to any one of claims 1 to 8, characterized in that: include: A data acquisition module is used to acquire normal operating sound data of the target device obtained after noise reduction and gain processing and daily operating sound data obtained after noise reduction and gain processing; a signal conversion module, configured to perform signal conversion on the normal operation sound data and the daily operation sound data, respectively, to obtain a time-frequency graph and a spectrum graph corresponding to the normal operation sound data and a time-frequency graph and a spectrum graph corresponding to the daily operation sound data; a data comparison module, configured to compare and calculate the time-frequency graph corresponding to the daily operation sound data with the time-frequency graph corresponding to the normal operation sound data to obtain a time-frequency graph density comparison result, and to compare and calculate the spectrum graph corresponding to the daily operation sound data with the spectrum graph corresponding to the normal operation sound data to obtain a spectrum graph energy comparison result; An abnormality judgment module, used to judge whether there is abnormal data in the daily operation of the target device based on the density comparison result of the time-frequency graph and the energy comparison result of the spectrum graph; The device operation detection module is used to obtain the operation detection result of the target device according to the abnormal data when there is abnormal data.

Images