1. An intelligent monitoring method for natural gas station equipment operation is characterized in that: the method comprises the following steps:

s1, collecting the audio signal of the natural gas station equipment running state by using an acoustic array;

s2, performing outlier detection on the audio of each channel of the array to screen out abnormal audio and corresponding array elements thereof;

s3, performing audio imaging by using a delay-sum beam forming algorithm and acquiring an audio frequency spectrum of the target equipment;

and S4, extracting Mel cepstrum characteristics from the formed frequency spectrum of the wave beam, standardizing, and inputting the Mel cepstrum characteristics into a pre-trained natural gas station equipment state classification model to diagnose the equipment state.

2. The intelligent monitoring method for the operation of the natural gas station equipment according to claim 1, characterized in that: in step S1, the acoustic array includes a microphone, a camera, and a data acquisition card; wherein, the microphones on the array plane are in a multi-arm spiral shape and are used for receiving the audio of the equipment; the receiving direction of the camera is the same as that of the array plane, and the camera is used for shooting field images; the data acquisition card is connected with each microphone through a signal line and is used for synchronously sampling the audio frequency of each channel.

3. The intelligent monitoring method for the operation of the natural gas station equipment according to claim 2, characterized in that: step S2 includes the following steps:

s21, extracting Mel cepstrum characteristics from each channel audio, the process is as follows:

the method comprises the following steps of performing framing processing on audio, calculating the power spectrum of each frame of audio, applying a Mel filter bank to the power spectrum, calculating the energy sum of a filter and taking the logarithm, wherein the calculation formula is as follows:

wherein S is_i(k)Power spectrum of the i-th frame audio, B_m(k) For the frequency response of the mth triangular filter, the frequency band range is [ f ]_m-1,f_m+1]Broadening with increasing m value;

further, applying discrete cosine transform to perform decorrelation processing on the filter bank coefficients, the calculation formula is as follows:

wherein M is the number of filters, and L is the order of the Mel cepstrum coefficient;

further, to obtain the dynamic change information between frames, a first order difference MFCC is calculated, and the calculation formula is as follows:

wherein d is_tThe MFCC first-order difference of the t frame audio is calculated according to the MFCC from the t-N frame to the t + N frame audio, and N is 2;

further, taking the average value of the MFCC and the first-order difference MFCC of each frame of audio to form a feature vector;

s22, calculating the abnormal score of each feature vector through a local abnormal factor algorithm;

and calculating the local reachable density of the point p, wherein the calculation formula is as follows:

wherein N is_k(p) is all points that are not more than the kth distance k-distance (p) from point p, reach-distance (p, o) max { k-distance (o) } d (p, o) is the reachable distance of point p relative to point o;

further, the LOF value of the point p is calculated by the following formula:

s23, removing the audio with the abnormal score higher than the set threshold value and the corresponding array element;

and calculating abnormal scores of the characteristic vectors by using an LOF algorithm, if the scores are larger than an abnormal threshold value, considering the abnormal scores as outliers, screening abnormal audios and corresponding array elements to eliminate the interference of a few outliers, enhancing the robustness of a beam forming algorithm and reducing the noise in the audio formed by the beams.

4. The intelligent monitoring method for the operation of the natural gas station equipment according to claim 3, characterized in that: in step S3, the delay-sum beamforming algorithm delays the signals of each channel, and then adds the signals in phase, and the calculation formula is as follows:

wherein E is the number of array elements, x_i(t) is the signal received by the ith array element, τ_iIs x_i(t) a corresponding time delay compensation,is an array manifold matrix, H represents the conjugate transpose of the vector or matrix, and x (t) is the received signal vector;

further, by different spatial anglesScanning the space to obtain a signal power spectrum related to a space angle, wherein the calculation formula is as follows:

wherein, R is a covariance matrix of the received signal;

further, let D be the distance from the target device to the center of the array, and angle in spaceConversion to coordinates (x) in the plane of the array_c,y_c) The calculation formula is as follows:

according to different coordinates (x)_c,y_c) Drawing a color map by using the power spectrum intensity, and obtaining an audio imaging result after the color map is overlapped with the field image;

further, by the spatial angle at the peak of the power spectrumObtaining the array popular matrix corresponding to the array popular matrixAnd carrying out time delay superposition on each channel signal X (omega) to obtain an audio frequency spectrum of the target equipment, wherein the calculation formula is as follows:

5. the intelligent monitoring method for the operation of the natural gas station equipment according to claim 4, characterized in that: in step S4, the MFCC feature extraction process of the beamformed audio spectrum is as follows:

the method comprises the following steps of performing framing processing on audio, calculating the power spectrum of each frame of audio, applying a Mel filter bank to the power spectrum, calculating the energy sum of a filter and taking the logarithm, wherein the calculation formula is as follows:

wherein S is_i(k) Power spectrum of the i-th frame audio, B_m(k) For the frequency response of the mth triangular filter, the frequency band range is [ f ]_m-1,f_m+1]Broadening with increasing m value;

further, applying discrete cosine transform to perform decorrelation processing on the filter bank coefficients, the calculation formula is as follows:

wherein M is the number of filters, and L is the order of the Mel cepstrum coefficient;

further, to obtain the dynamic change information between frames, a first order difference MFCC is calculated, and the calculation formula is as follows:

wherein d is_tThe MFCC first-order difference of the t frame audio is calculated according to the MFCC from the t-N frame to the t + N frame audio, and N is 2;

further, the MFCC of each frame of audio and the average value of the first-order difference MFCC are taken to form a feature vector.

6. The utility model provides a natural gas station equipment operation intelligent monitoring system which characterized in that: the system comprises an audio signal acquisition module, an audio signal outlier detection module, an audio signal beam forming module and a natural gas station equipment diagnosis module which are electrically connected in sequence;

the audio signal acquisition module is used for acquiring an audio signal and a field image of the running state of the natural gas station equipment and outputting the audio signal and the field image to the audio signal outlier detection module;

the audio signal outlier detection module is used for carrying out outlier detection on the audio of each channel acquired by the array so as to screen out abnormal audio and corresponding array elements thereof and outputting the abnormal audio and the corresponding array elements to the audio signal beam forming module;

the audio signal beam forming module is used for carrying out audio imaging on the site image, acquiring an audio frequency spectrum of the target equipment and outputting the audio frequency spectrum to the natural gas station equipment diagnosis module;

the natural gas station equipment diagnosis module is used for extracting MFCC characteristics from a frequency spectrum formed by the wave beams, carrying out standardization processing, inputting the MFCC characteristics into a pre-trained natural gas station equipment state classification model, diagnosing whether the equipment has a fault or not, and diagnosing the fault type if the equipment has the fault.

Background

At present, the main maintenance mode of the long-distance natural gas station is a mode combining regular inspection, periodic maintenance and preventive maintenance, so that the problems of excessive maintenance or insufficient maintenance exist to a certain extent, and the whole operation efficiency of the station is not improved. The key operation parameters and signal detection of the equipment in the station are relatively perfect, but the direct monitoring of mechanical faults is lacked, most of the key operation parameters and the signal detection are discovered by maintenance by operators on duty, or the mechanical faults are inferred to possibly occur through the abnormal detection of related signals.

Generally, at present, because a monitoring system of a station equipment body is not perfect, most of the monitoring systems need to judge mechanical faults by means of personnel experience through a regular maintenance mode, and the monitoring systems are not efficient and reliable enough for field production operation. Therefore, it is an urgent need to provide a systematic and scientific method and system for monitoring the equipment status of a natural gas station.

Disclosure of Invention

The invention provides an intelligent monitoring method and system for natural gas station equipment operation, aiming at overcoming the defect that the existing natural gas station equipment monitoring system is not complete enough, monitoring the equipment operation state in real time, finding out equipment faults in time, reducing the personnel inspection workload and improving the overall production efficiency and operation reliability of a station.

The invention provides an intelligent monitoring method and system for natural gas station equipment operation, aiming at solving the problem of low overall operation efficiency of a station caused by insufficient monitoring system of the existing natural gas station equipment.

In order to achieve the above object, in one aspect, the present invention provides an intelligent monitoring method for natural gas station equipment operation, including the following steps:

s1, collecting the audio signal of the natural gas station equipment running state by using an acoustic array;

s2, performing outlier detection on the audio of each channel of the array to screen out abnormal audio and corresponding array elements thereof;

s3, performing audio imaging by using a delay-sum beam forming algorithm and acquiring an audio frequency spectrum of the target equipment;

and S4, extracting Mel cepstrum characteristics from the formed frequency spectrum of the wave beam, standardizing, and inputting the Mel cepstrum characteristics into a pre-trained natural gas station equipment state classification model to diagnose the equipment state.

Further preferably, in step S1, the acoustic array includes a microphone, a camera, and a data acquisition card; wherein, the microphones on the array plane are in a multi-arm spiral shape and are used for receiving the audio of the equipment; the receiving direction of the camera is the same as that of the array plane, and the camera is used for shooting field images; the data acquisition card is connected with each microphone through a signal line and is used for synchronously sampling the audio frequency of each channel.

Further preferably, step S2 includes the steps of:

s21, extracting Mel cepstrum characteristics from each channel audio, the process is as follows:

the method comprises the following steps of performing framing processing on audio, calculating the power spectrum of each frame of audio, applying a Mel filter bank to the power spectrum, calculating the energy sum of a filter and taking the logarithm, wherein the calculation formula is as follows:

wherein S is_i(k) Power spectrum of the i-th frame audio, B_m(k) For the frequency response of the mth triangular filter, the frequency band range is [ f ]_m-1,f_m+1]Broadening with increasing m value;

further, applying discrete cosine transform to perform decorrelation processing on the filter bank coefficients, the calculation formula is as follows:

wherein M is the number of filters, and L is the order of the Mel cepstrum coefficient;

further, to obtain the dynamic change information between frames, a first order difference MFCC is calculated, and the calculation formula is as follows:

wherein d is_tThe MFCC first-order difference of the t frame audio is calculated according to the MFCC from the t-N frame to the t + N frame audio, and N is 2;

further, taking the average value of the MFCC and the first-order difference MFCC of each frame of audio to form a feature vector;

s22, calculating the abnormal score of each feature vector through a local abnormal factor algorithm;

and calculating the local reachable density of the point p, wherein the calculation formula is as follows:

wherein N is_k(p) is all points that are not more than the kth distance k-distance (p) from point p, reach-distance (p, o) max { k-distance (o) } d (p, o) is the reachable distance of point p relative to point o;

further, the LOF value of the point p is calculated by the following formula:

s23, removing the audio with the abnormal score higher than the set threshold value and the corresponding array element;

and calculating abnormal scores of the characteristic vectors by using an LOF algorithm, if the scores are larger than an abnormal threshold value, considering the abnormal scores as outliers, screening abnormal audios and corresponding array elements to eliminate the interference of a few outliers, enhancing the robustness of a beam forming algorithm and reducing the noise in the audio formed by the beams.

Further preferably, in step S3, the delay-sum beamforming algorithm delays the channel signals, and then adds them in phase, and the calculation formula is as follows:

wherein E is the number of array elements, x_i(t) is the signal received by the ith array element, τ_iIs x_i(t) a corresponding time delay compensation,is an array manifold matrix, H represents the conjugate transpose of the vector or matrix, and x (t) is the received signal vector;

further, by different spatial anglesScanning the space to obtain a signal power spectrum related to a space angle, wherein the calculation formula is as follows:

wherein, R is a covariance matrix of the received signal;

further, let D be the distance from the target device to the center of the array, and angle in spaceConversion to coordinates (x) in the plane of the array_c,y_c) The calculation formula is as follows:

according to different coordinates (x)_c,y_c) Drawing a color map by using the power spectrum intensity, and obtaining an audio imaging result after the color map is overlapped with the field image;

further, by the spatial angle at the peak of the power spectrumObtaining the array popular matrix corresponding to the array popular matrixAnd carrying out time delay superposition on each channel signal X (omega) to obtain an audio frequency spectrum of the target equipment, wherein the calculation formula is as follows:

further preferably, in step S4, the MFCC feature extraction process of the beamformed audio spectrum is as follows:

the method comprises the following steps of performing framing processing on audio, calculating the power spectrum of each frame of audio, applying a Mel filter bank to the power spectrum, calculating the energy sum of a filter and taking the logarithm, wherein the calculation formula is as follows:

wherein S is_i(k) Power spectrum of the i-th frame audio, B_m(k) For the frequency response of the mth triangular filter, the frequency band range is [ f ]_m-1,f_m+1]Broadening with increasing m value;

further, applying discrete cosine transform to perform decorrelation processing on the filter bank coefficients, the calculation formula is as follows:

wherein M is the number of filters, and L is the order of the Mel cepstrum coefficient;

further, to obtain the dynamic change information between frames, a first order difference MFCC is calculated, and the calculation formula is as follows:

wherein d is_tThe MFCC first-order difference of the t frame audio is calculated according to the MFCC from the t-N frame to the t + N frame audio, and N is 2;

further, the MFCC of each frame of audio and the average value of the first-order difference MFCC are taken to form a feature vector.

On the other hand, the invention provides an intelligent monitoring system for natural gas station equipment operation, which comprises an audio signal acquisition module, an audio signal outlier detection module, an audio signal beam forming module and a natural gas station equipment diagnosis module which are electrically connected in sequence;

the audio signal acquisition module is used for acquiring an audio signal and a field image of the running state of the natural gas station equipment and outputting the audio signal and the field image to the audio signal outlier detection module;

the audio signal outlier detection module is used for carrying out outlier detection on the audio of each channel acquired by the array so as to screen out abnormal audio and corresponding array elements thereof and outputting the abnormal audio and the corresponding array elements to the audio signal beam forming module;

the audio signal beam forming module is used for carrying out audio imaging on the site image, acquiring an audio frequency spectrum of the target equipment and outputting the audio frequency spectrum to the natural gas station equipment diagnosis module;

the natural gas station equipment diagnosis module is used for extracting MFCC characteristics from a frequency spectrum formed by the wave beams, carrying out standardization processing, inputting the MFCC characteristics into a pre-trained natural gas station equipment state classification model, diagnosing whether the equipment has a fault or not, and diagnosing the fault type if the equipment has the fault.

In general, compared with the prior art, the above technical solution contemplated by the present invention can achieve the following beneficial effects:

1. the invention provides an intelligent monitoring method for operation of natural gas station equipment, which can reliably and intelligently monitor the station equipment, effectively reduce difficulty in fault diagnosis of the station equipment, reduce maintenance cost of the station equipment, reduce the inspection workload of personnel and improve the overall production efficiency and operation reliability of a station;

2. according to the intelligent monitoring method for the operation of the natural gas station equipment, the outlier detection is carried out on the audio frequency of each channel of the array, so that the interference of a few outliers on a beam forming algorithm is reduced, the signal to noise ratio of the audio frequency formed by the beam is improved, and the effects of audio frequency imaging and an equipment state classification model can be improved;

3. according to the intelligent monitoring method for the operation of the natural gas station equipment, the MFCC characteristics are extracted from the frequency spectrum formed by the wave beams, and the first-order difference MFCC representing the dynamic characteristics of the audio is considered, so that the accuracy of outlier detection and equipment fault diagnosis can be effectively improved.

Drawings

FIG. 1 is a flow chart of an intelligent monitoring method for operation of natural gas station equipment provided by the invention;

fig. 2 is a flow chart of outlier detection for each channel audio of the array according to the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

Example 1:

as shown in fig. 1, the invention provides an intelligent monitoring method for natural gas station equipment operation, which comprises the following steps:

s1, collecting the audio signal of the natural gas station equipment running state by using an acoustic array;

s2, performing outlier detection on the audio of each channel of the array to screen out abnormal audio and corresponding array elements thereof;

specifically, as shown in fig. 2, the method includes the following steps:

s21, extracting Mel cepstrum characteristics from each channel audio, the process is as follows:

the method comprises the following steps of performing framing processing on audio, calculating the power spectrum of each frame of audio, applying a Mel filter bank to the power spectrum, calculating the energy sum of a filter and taking the logarithm, wherein the calculation formula is as follows:

wherein S is_i(k) Power spectrum of the i-th frame audio, B_m(k) For the frequency response of the mth triangular filter, the frequency band range is [ f ]_m-1,f_m+1]And broadens as the value of m increases. In this embodiment, the frame length is about 80ms, the frame transfer is about 40ms, and the number of the triangular filters is 26;

further, applying discrete cosine transform to perform decorrelation processing on the filter bank coefficients, the calculation formula is as follows:

wherein M is the number of filters, and L is the Mel cepstrum coefficient (MFCC) order. In the embodiment, L is 26, and the 2 nd to 13 th coefficients of the MFCC are stored;

further, to obtain the dynamic change information between frames, a first order difference MFCC is calculated, and the calculation formula is as follows:

wherein d is_tThe MFCC first-order difference of the t frame audio is calculated according to the MFCC from the t-N frame to the t + N frame audio, wherein N is usually 2;

further, taking the average value of the MFCC and the first-order difference MFCC of each frame of audio to form a 24-dimensional feature vector;

s22, calculating the abnormal score of each feature vector through a local abnormal factor algorithm;

and calculating the local reachable density of the point p, wherein the calculation formula is as follows:

wherein N is_k(p) is all points that are not more distant from point p by the kth distance k-distance (p) of point p, reach-distance (p, o) max { k-distance (o) } is the reachable distance of point p with respect to point o. In the embodiment, k is 10;

further, the LOF value of the point p is calculated according to the following formula:

wherein, the lower the local reachability density of the point p, the higher the local reachability density of the nearest neighbor of the kth distance neighborhood of the point p, and the higher the LOF value of the point p. For most points in a cluster, their LOF value is approximately equal to 1;

s23, removing the audio with the abnormal score higher than the set threshold value and the corresponding array element;

and calculating an abnormal score of each feature vector by using an LOF algorithm, if the score is greater than an abnormal threshold value, considering the feature vector as an outlier, and screening abnormal audio and corresponding array elements thereof to eliminate the interference of few outliers. In order to obtain better detection effect, the anomaly threshold value is set to 2 in the embodiment;

s3, performing audio imaging by using a delay-sum (DAS) beam forming algorithm and acquiring an audio frequency spectrum of the target equipment;

specifically, the DAS beam forming algorithm is used to perform appropriate delay on each channel signal, and then the signals are added in phase, and the calculation formula is as follows:

wherein E is the number of array elements, x_i(t) is the signal received by the ith array element, τ_iIs x_i(t) a corresponding time delay compensation,for an array manifold matrix, H is the vector or conjugate transpose of the matrix, and x (t) is the received signal vector. In the embodiment, 30 is taken as E;

further, by different spatial anglesScanning the space to obtain a signal power spectrum related to a space angle, wherein the calculation formula is as follows:

wherein, R is a covariance matrix of the received signal;

further, let D be the distance from the target device to the center of the array, and angle in spaceConversion to coordinates (x) in the plane of the array_c,y_c) The calculation formula is as follows:

according to different coordinates (x)_c,y_c) Drawing a color map by using the power spectrum intensity, and obtaining an audio imaging result after the color map is overlapped with the field image;

further, by the spatial angle at the peak of the power spectrumObtaining the array popularity vector corresponding to the array popularity vectorAnd carrying out time delay superposition on each channel signal X (omega) to obtain an audio frequency spectrum of the target equipment, wherein the calculation formula is as follows:

and S4, extracting Mel cepstrum characteristics from the formed frequency spectrum of the wave beam, standardizing, and inputting the Mel cepstrum characteristics into a pre-trained natural gas station equipment state classification model to diagnose the equipment state.

Specifically, the MFCC feature extraction process for the beamformed audio spectrum is the same as step S21.

Specifically, the beamformed audio frequency spectrums of the different device states are acquired through steps S1-S3, where the device states include normal states and fault states of the different devices; wherein, equipment includes: the device comprises a separator, a pressure regulating pry, a metering pry, a compressor, a pipeline, a valve and the like; the equipment failure comprises: filter element overrun, membrane damage, ultrasonic probe loosening, turbine blade clamping stagnation, rectifier blockage, bolt loosening and the like.

Further, MFCC features are extracted from frequency spectrums of different equipment states and are subjected to standardization processing, and training data sets are obtained after the MFCC features are in one-to-one correspondence with the corresponding states so as to pre-train the natural gas station equipment state classification model.

Specifically, the standardized MFCC features are input into a pre-trained natural gas station equipment state classification model to diagnose the equipment state, if a fault occurs, an operator on duty can quickly determine the fault equipment according to an audio imaging result on a field image, and corresponding processing measures are taken in time according to the fault type.

Example 2:

the invention provides an intelligent monitoring system for natural gas station equipment operation, which comprises an audio signal acquisition module, an audio signal outlier detection module, an audio signal beam forming module and a natural gas station equipment diagnosis module which are electrically connected in sequence;

the audio signal acquisition module is used for acquiring an audio signal and a field image of the running state of the natural gas station equipment and outputting the audio signal and the field image to the audio signal outlier detection module;

the audio signal outlier detection module is used for carrying out outlier detection on the audio of each channel acquired by the array so as to screen out abnormal audio and corresponding array elements thereof and outputting the abnormal audio and the corresponding array elements to the audio signal beam forming module;

the audio signal beam forming module is used for carrying out audio imaging on the site image, acquiring an audio frequency spectrum of the target equipment and outputting the audio frequency spectrum to the natural gas station equipment diagnosis module;

the natural gas station equipment diagnosis module is used for extracting MFCC characteristics from a frequency spectrum formed by the wave beams, carrying out standardization processing, inputting the MFCC characteristics into a pre-trained natural gas station equipment state classification model, diagnosing whether the equipment has a fault or not, and diagnosing the fault type if the equipment has the fault.

The related technical solution of this embodiment is the same as embodiment 1, and is not described herein again.

It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.