Intelligent piling underwater noise monitoring method and system

Abstract

The invention discloses an intelligent piling underwater noise monitoring method and system, and the method comprises the steps: receiving and preprocessing original underwater sound pressure signals of a plurality of monitoring stations, and obtaining a plurality of underwater sound pressure signals; based on short-time energy analysis and self-adaptive threshold judgment, underwater sound pressure signals of all monitoring stations are detected and extracted, and candidate pulse signals are obtained; performing spectral analysis on the candidate pulse signals to obtain a time-frequency distribution diagram, a log-mel energy spectrum diagram and frequency domain characteristic parameters; inputting the log-mel energy spectrum diagram into a deep learning model to obtain a piling pulse signal; post-processing the piling pulse signal, and calculating acoustic evaluation parameters of the piling underwater noise; and on the basis of the acoustic evaluation parameters of the multiple monitoring stations and the corresponding monitoring position information, the average sound source level of piling operation is obtained through inversion calculation in combination with an ocean underwater sound propagation model. The underwater noise pollution monitoring system meets the requirement that modern ocean engineering urgently needs quick response and automatic monitoring on underwater noise pollution.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of marine environment monitoring, and particularly relates to an intelligent piling underwater noise monitoring method and system. BACKGROUND

[0002] Traditional underwater noise monitoring methods mostly use fixed acoustic monitoring systems or manually deployed hydrophone arrays, which have significant defects. In terms of data processing, existing methods usually rely on offline processing mode, and the collected acoustic data need to be analyzed by humans or later to extract effective information, which leads to serious lag in data processing and makes it impossible to achieve real-time feedback and rapid warning.

[0003] In addition, due to the lack of efficient automatic identification and extraction algorithms, the existing technology is difficult to accurately distinguish and extract piling pulse signals from complex marine environmental noise, further reducing the accuracy and reliability of underwater noise monitoring.

[0004] Therefore, the existing technology cannot meet the urgent need for rapid response and automatic monitoring of underwater noise pollution in modern marine engineering. SUMMARY

[0005] To solve the above technical problems, the present application provides an intelligent piling underwater noise monitoring method, which comprises the following steps: Receiving and preprocessing the original underwater sound pressure signals of multiple monitoring sites to obtain multiple underwater sound pressure signals; Based on short-time energy analysis and adaptive threshold determination, detecting and extracting the underwater sound pressure signals of each monitoring site to obtain candidate pulse signals; Performing spectral analysis on the candidate pulse signals to obtain time-frequency distribution diagrams, log-mel spectrograms, and frequency domain feature parameters; Inputting the log-mel spectrogram into a deep learning model for identification and classification to obtain piling pulse signals from the candidate pulse signals; Post-processing the piling pulse signals and calculating the acoustic evaluation parameters of piling underwater noise; Based on the acoustic evaluation parameters of multiple monitoring sites and their corresponding monitoring location information, and combined with a marine underwater sound propagation model, the average sound source level of piling operations is calculated by inversion.

[0006] Optionally, based on short-time energy analysis and adaptive threshold determination, detecting and extracting the underwater sound pressure signals of each monitoring site to obtain candidate pulse signals, comprising the following steps: Calculating the short-time energy of the underwater sound pressure signals based on a preset window length and a preset overlap rate; An adaptive threshold is calculated based on the average energy, standard deviation of background noise of each monitoring site, and a preset sensitivity factor; The short-time energy of each monitoring site is compared with the adaptive threshold to identify candidate impulse signals; the short-time energy of the candidate impulse signals is greater than the adaptive threshold; The envelope of the candidate impulse signal is extracted; The envelope is low-pass filtered and smoothed, and the start and end times of the candidate impulse signal are determined based on an envelope detection threshold; Based on the start and end times, the complete waveform of the candidate impulse signal is obtained by extending the preset fill time before and after.

[0007] Optionally, the frequency domain feature parameters include in-band energy distribution, peak frequency and bandwidth parameters; The candidate impulse signal is subjected to spectral analysis to obtain a time-frequency distribution map, a log-mel spectrogram and frequency domain feature parameters, including the following steps: Based on the complete waveform of the candidate impulse signal, fast Fourier transform or short-time Fourier transform is performed to obtain a time-frequency distribution map; The time-frequency distribution map is converted into a log-mel spectrogram, and the in-band energy distribution, peak frequency and bandwidth parameters of each candidate impulse signal are calculated based on the time-frequency distribution map.

[0008] Optionally, after obtaining the candidate impulse signal, the method further includes: Time domain feature extraction is performed on the candidate impulse signal to obtain time domain feature parameters; the time domain feature parameters include peak sound pressure, peak-to-peak sound pressure, pulse duration, rise time, fall time, pulse interval and pulse rate.

[0009] Optionally, the log-mel spectrogram is input into a deep learning model for recognition and classification to obtain a pile driving impulse signal from the candidate impulse signal, including the following steps: The log-mel spectrogram is input into a two-dimensional convolutional neural network model to perform feature learning and classification on the log-mel spectrogram; Based on the feature learning result, a class label and a confidence level corresponding to each candidate impulse signal are output to obtain a pile driving impulse signal from the candidate impulse signal; the class label at least includes pile driving sound and ship engine noise.

[0010] Optionally, the method further includes: Based on the pile driving impulse signal, the average sound pressure level in a period is calculated; The average sound pressure level in the one cycle is compared with a preset frequency band threshold value, and when the average sound pressure level in the one cycle is greater than the frequency band threshold value, a first alarm signal is generated and output.

[0011] Optionally, the method further comprises: Based on the pile driving pulse signal, the sound exposure level of a single event is calculated; the single event is a single pile driving impact; The sound exposure level of the single event is compared with a preset sound exposure level threshold value, and when the sound exposure level of the single event is greater than the sound exposure level threshold value, a second alarm signal is generated and output.

[0012] Optionally, the pile driving pulse signal is post-processed, and an acoustic evaluation parameter of pile driving underwater noise is calculated, including the following steps: The pile driving pulse signal is time-smoothed; Based on the propagation distance of each pile driving pulse signal and the ocean acoustic propagation model, the pile driving pulse signal is corrected for propagation loss; Based on the sound pressure level after propagation loss correction, the acoustic evaluation parameter of pile driving underwater noise is calculated, and the acoustic evaluation parameter includes root mean square sound pressure level, instantaneous peak sound pressure level, single sound exposure level and cumulative sound exposure level.

[0013] Optionally, based on the acoustic evaluation parameters of multiple monitoring sites and their corresponding monitoring position information, the average sound source level of pile driving operation is calculated by inverse calculation combined with the ocean acoustic propagation model, including the following steps: The single sound exposure level of each monitoring site and the corresponding monitoring position information are obtained, and the straight-line distance between each monitoring position and the pile driving sound source is calculated; Based on the spherical or cylindrical expansion ocean acoustic propagation model, the propagation loss corresponding to each monitoring site is calculated; The single sound exposure level of each monitoring site is added to the corresponding propagation loss to calculate the sound source level at 1m from the pile driving operation corresponding to each monitoring site; The sound source levels at 1m from the pile driving operation calculated by all monitoring sites are arithmetically averaged to obtain the average sound source level of the pile driving operation.

[0014] Corresponding to the intelligent pile driving underwater noise monitoring method, the present application provides an intelligent pile driving underwater noise monitoring system, which comprises: A signal acquisition module for acquiring original underwater sound pressure signals of multiple monitoring sites; A preprocessing module for receiving and preprocessing the original underwater sound pressure signals of multiple monitoring sites to obtain multiple underwater sound pressure signals; The signal extraction module is used for detecting and extracting the underwater sound pressure signals of each monitoring site based on short-time energy analysis and adaptive threshold determination to obtain candidate pulse signals; The spectrum analysis module is used for performing spectrum analysis on the candidate pulse signals to obtain a time-frequency distribution diagram, a log-mel spectrogram and frequency domain characteristic parameters; The classification module is used for inputting the log-mel spectrogram into a deep learning model for identification and classification to obtain a pile-driving pulse signal from the candidate pulse signals; The acoustic evaluation parameter calculation module is used for post-processing the pile-driving pulse signal and calculating acoustic evaluation parameters of pile-driving underwater noise. The sound source level inversion module is used for inversely calculating the average sound source level of pile-driving operation based on the acoustic evaluation parameters of multiple monitoring sites and corresponding monitoring position information in combination with an ocean acoustic propagation model.

[0015] Compared with the prior art, the present application has the following beneficial effects: (1) The original underwater sound pressure signals of multiple monitoring sites are received and preprocessed, short-time energy analysis and adaptive threshold determination are combined to realize automatic real-time detection and extraction of pulse signals, the data processing lag problem caused by the dependence of the prior art on offline manual analysis is solved, the deep learning model is used for automatic identification and classification of the log-mel spectrogram, the pile-driving pulse signal and the complex marine environmental noise are effectively distinguished, and the defects of low efficiency and poor accuracy of the automatic identification algorithm in the prior art are overcome, the fusion of the multi-point acoustic evaluation parameters and the ocean acoustic propagation model is used to inversely calculate the sound source level, the whole-process automatic processing from data acquisition, signal identification to sound source intensity evaluation is realized, the demand for rapid response and automatic monitoring of modern marine engineering is met, and the real-time performance, accuracy and reliability of the monitoring are significantly improved.

[0016] (2) The short-time energy is calculated by presetting the window length and the overlap rate, and the adaptive threshold is dynamically generated based on the statistical characteristics of the background noise, so that the pulse detection can automatically adjust the sensitivity according to the actual environmental noise level, and the missing or false detection problem caused by the fixed threshold is avoided, the envelope line is subjected to low-pass filtering, smoothing processing and front and rear extension filling, noise interference is effectively suppressed, the complete waveform boundary of the candidate pulse signal is accurately extracted, and the accuracy and robustness of the candidate pulse signal extraction are improved.

[0017] (3) The candidate pulse signal is converted into a time-frequency distribution diagram by fast Fourier transform or short-time Fourier transform, and a log-mel spectrogram is further generated, the visualization and logarithmic perception scale representation of the signal spectrum characteristics are realized, the in-band energy distribution, peak frequency and bandwidth parameters are calculated, the input data for the deep learning model is provided, and the feature learning efficiency and identification accuracy of the classification model are improved.

[0018] (4) Through the log-mel spectrogram input to the two-dimensional convolutional neural network model for end-to-end feature learning and classification, using the powerful nonlinear feature extraction ability of deep learning, the high-precision automatic differentiation of complex interference sources such as piling sound and ship noise is realized; through the output of the category label and the confidence, a quantitative basis is provided for the screening of the piling pulse signal, and the workload and misjudgment risk of manual discrimination are greatly reduced.

[0019] (5) By calculating the average sound pressure level in a period and comparing it with the frequency band threshold, the first alarm signal can be generated immediately when the noise intensity exceeds the standard, realizing the rapid early warning function of the persistent strong noise, and providing timely decision support for the biological protection measures of the site construction.

[0020] (6) By calculating the sound exposure level of a single piling impact event and comparing it with the sound exposure level threshold, the second alarm signal is generated when the single pulse energy exceeds the standard, realizing the accurate monitoring and hierarchical early warning of high-intensity transient noise events.

[0021] (7) Through time smoothing processing of the piling pulse signal and propagation loss correction based on the marine underwater sound propagation model, the influence of distance and propagation path on sound pressure measurement is eliminated, making the data of different monitoring sites comparable; by calculating the standardized acoustic evaluation parameters such as root mean square sound pressure level, instantaneous peak sound pressure level, single sound exposure level and cumulative sound exposure level, the standardized and standardized quantitative indicators are provided for the objective evaluation and environmental protection of the piling noise.

[0022] (8) Through the fusion of multi-point monitoring data, the single-point measurement error and environmental uncertainty are effectively reduced, and the accuracy and robustness of the sound source level estimation are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0023] The drawings described herein are used to provide further understanding of the present application, and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings: Figure 1 Flowchart of an embodiment of the intelligent piling underwater noise monitoring method of the present application; Figure 2 Frame diagram of an embodiment of the intelligent piling underwater noise monitoring system of the present application. DETAILED DESCRIPTION

[0024] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0025] As shown in the drawings, the intelligent pile driving underwater noise monitoring method of the present application comprises the following steps: Figure 1 Receiving the original underwater sound pressure signals of multiple monitoring sites and pre-processing to obtain multiple underwater sound pressure signals; preferably, the pre-processing includes converting the original underwater sound pressure signals into digital signals, and performing band-pass filtering (10Hz-100kHz), gain adjustment processing, removing direct current, environmental low-frequency and high-frequency noise; Based on short-time energy analysis and adaptive threshold determination, the underwater sound pressure signals of each monitoring site are detected and extracted to obtain candidate pulse signals; Performing spectral analysis on the candidate pulse signals to obtain time-frequency distribution diagram, log-mel spectrogram and frequency domain feature parameters; Inputting the log-mel spectrogram into a deep learning model for identification and classification to obtain pile driving pulse signals from the candidate pulse signals; Post-processing the pile driving pulse signals and calculating the acoustic evaluation parameters of the pile driving underwater noise; Based on the acoustic evaluation parameters of multiple monitoring sites and their corresponding monitoring location information, combined with the ocean acoustic propagation model, the average sound source level of the pile driving operation is calculated by inversion. Preferably, the multiple monitoring sites form an underwater acoustic sensor array, which has high sensitivity, wide frequency band and adjustable cable length (adjustment range is 1-50 meters); the underwater acoustic sensor array is composed of multiple underwater acoustic sensors arranged according to water depth layering (such as 5m, 15m, 30m); the layout points can be adjusted within a radius range of 100-5000m from the pile driving point.

[0026]

[0027] ​The present application solves the data processing lag problem caused by the dependence of the prior art on offline manual analysis by receiving and preprocessing the original underwater sound pressure signals of multiple monitoring sites, combining short-time energy analysis with adaptive threshold determination to realize automatic real-time detection and extraction of pulse signals; through automatic identification and classification of the log-mel spectrogram by the deep learning model, the pile driving pulse signals and complex marine environmental noise are effectively distinguished, overcoming the defects of low efficiency and poor accuracy of the automatic identification algorithm in the prior art; through the fusion of the multi-point acoustic evaluation parameters and the marine underwater sound propagation model to invert the sound source level, the whole process automation processing from data acquisition, signal identification to sound source intensity evaluation is realized, meeting the demand of modern marine engineering for fast response and automatic monitoring, and significantly improving the real-time, accuracy and reliability of the monitoring.

[0028] In the present embodiment, based on short-time energy analysis and adaptive threshold determination, the underwater sound pressure signals of each monitoring site are detected and extracted to obtain candidate pulse signals, including the following steps: The short-time energy of the underwater sound pressure signal is calculated based on the preset window length and the preset overlap rate; the optimal range of the preset window length is 10-50 ms, and the optimal preset overlap rate is 50%; The corresponding adaptive threshold is calculated based on the energy mean and standard deviation of the background noise of each monitoring site, and the preset sensitivity factor; preferably, the product of the energy standard deviation of the background noise and the sensitivity factor is calculated first, and the adaptive threshold is the sum of the product and the energy mean of the background noise; The short-time energy of each monitoring site is compared with the adaptive threshold to identify the candidate pulse signals; the short-time energy of the candidate pulse signal is greater than the adaptive threshold; The envelope of the candidate pulse signal is extracted; The envelope is subjected to low-pass filtering and smoothing processing, and the start and end times of the candidate pulse signal are determined based on the envelope detection threshold; Based on the start and end times, the complete waveform of the candidate pulse signal is obtained by extending the preset filling time length before and after; preferably, the filling time length before the extension is preferably 50 ms, and the filling time length after the extension is preferably 200 ms, to ensure the capture of the envelope and the residual wave; further, the time start and end points, peak time and duration of each candidate pulse signal are also recorded.

[0029] It should be noted that the calculation of the adaptive threshold depends on the statistical characteristics of the background noise of each monitoring site. In a preferred embodiment, the system will automatically learn the background noise for not less than 30 seconds before formally starting the monitoring, during which signals without pile driving operations are collected, and the (short-time) energy mean and standard deviation are calculated, and the sensitivity factor is preferably valued at 3-6. To further adapt to the changes in environmental noise during long-term monitoring, the system supports dynamic updating of the background noise statistics.

[0030] Preferably, the envelope of the candidate pulse signal is extracted by the following steps: Taking absolute value: the part below the horizontal axis in the candidate pulse signal is "folded" above the horizontal axis to obtain a V-shaped signal that is always positive, which is equivalent to obtaining the "amplitude" of the signal first; Low-pass filtering: the candidate pulse signal after taking the absolute value is subjected to low-pass filtering to filter out high-frequency, rapid fluctuations and burrs, and retain the overall, slow-changing trend to obtain the envelope.

[0031] The present application calculates the short-time energy by presetting the window length and overlap rate, and dynamically generates an adaptive threshold based on the statistical characteristics of the background noise, so that the pulse detection can automatically adjust the sensitivity according to the actual environmental noise level, avoiding the problem of missed detection or false detection caused by fixed threshold; by low-pass filtering, smoothing and front and rear extension filling of the envelope, noise interference is effectively suppressed, the complete waveform boundary of the candidate pulse signal is accurately extracted, and the accuracy and robustness of the candidate pulse signal extraction are improved.

[0032] In this embodiment, the frequency domain feature parameters include in-band energy distribution, peak frequency and bandwidth parameters; The candidate pulse signal is subjected to spectral analysis to obtain a time-frequency distribution diagram, a log-mel spectrum diagram and frequency domain feature parameters, including the following steps: Based on the complete waveform of the candidate pulse signal, fast Fourier transform or short-time Fourier transform is performed to obtain a time-frequency distribution diagram; The time-frequency distribution diagram is converted into a log-mel spectrum diagram, and based on the time-frequency distribution diagram, the in-band energy (low frequency / mid frequency / high frequency band) distribution, peak frequency and bandwidth (-3dB) parameters of each candidate pulse signal are calculated.

[0033] The present application converts the candidate pulse signal into a time-frequency distribution diagram by fast Fourier transform or short-time Fourier transform, and further generates a log-mel spectrum diagram, realizing the visualization and logarithmic perception scale representation of the signal spectrum features; at the same time, the in-band energy distribution, peak frequency and bandwidth parameters are calculated, providing input data for the deep learning model and improving the feature learning efficiency and recognition accuracy of the classification model.

[0034] In this embodiment, the log-mel spectrum diagram is input into a deep learning model for recognition and classification to obtain a pile driving pulse signal from the candidate pulse signal, including the following steps: The log-mel spectrum diagram is input into a two-dimensional convolutional neural network model to perform feature learning and classification on the log-mel spectrum diagram; Based on the feature learning result, a class label and a confidence corresponding to each candidate pulse signal are outputted to obtain a pile driving pulse signal from the candidate pulse signals; the class label at least includes a pile driving sound and a ship machine noise.

[0035] The present application realizes high-precision automatic differentiation of complex interference sources such as pile driving sound and ship machine noise by inputting a log-mel spectrogram into a two-dimensional convolutional neural network model for end-to-end feature learning and classification, using the powerful nonlinear feature extraction capability of deep learning; by outputting a class label and a confidence, a quantitative basis is provided for the screening of pile driving pulse signals, greatly reducing the workload and misjudgment risk of manual discrimination.

[0036] In the present embodiment, after obtaining the candidate pulse signal, further comprising: Time domain feature extraction is performed on the candidate pulse signal to obtain time domain feature parameters; the time domain feature parameters include peak sound pressure, peak-to-peak sound pressure, pulse duration (T90 / T50 / T10), rise time, fall time, pulse interval and pulse rate; Further, in the present embodiment, the method further comprises: displaying the time domain feature extraction result and / or the spectrum analysis result in a chart form.

[0037] In the present embodiment, the method further comprises: Based on the pile driving pulse signal, the average sound pressure level in a cycle is calculated; The average sound pressure level in a cycle is compared with a preset frequency band threshold value, and when the average sound pressure level in a cycle is greater than the frequency band threshold value, a first alarm signal is generated and outputted.

[0038] The present application can generate a first alarm signal immediately when the noise intensity exceeds the standard by calculating the average sound pressure level in a cycle and comparing it with the frequency band threshold value, realizing the rapid early warning function of persistent strong noise, and providing timely decision support for the biological protection measures of the site construction.

[0039] In the present embodiment, the method further comprises: Based on the pile driving pulse signal, the sound exposure level of a single event is calculated; the single event is a single pile driving impact; The sound exposure level of a single event is compared with a preset sound exposure level threshold value, and when the sound exposure level of a single event is greater than the sound exposure level threshold value, a second alarm signal is generated and outputted.

[0040] The present application realizes accurate monitoring and graded early warning of high-intensity transient noise events by calculating the sound exposure level of a single pile driving impact event and comparing it with the sound exposure level threshold value, and generating a second alarm signal when the single pulse energy exceeds the standard.

[0041] Preferably, the first alarm signal and the second alarm signal are sent to the administrator through short message, email or APP push.

[0042] Further, in the embodiment, the method further comprises: recording the alarm event based on the first alarm signal and / or the second alarm signal.

[0043] In the embodiment, the pile driving pulse signals are post-processed, and the acoustic evaluation parameters of the pile driving underwater noise are calculated, including the following steps: The pile driving pulse signals are time-smoothed; Based on the propagation distance of each pile driving pulse signal and the ocean acoustic propagation model, the pile driving pulse signals are corrected for propagation loss; Based on the sound pressure level corrected for propagation loss, the acoustic evaluation parameters of the pile driving underwater noise are calculated, including the root mean square sound pressure level, the instantaneous peak sound pressure level, the single sound exposure level and the cumulative sound exposure level.

[0044] Preferably, the root mean square sound pressure level is calculated according to the following formula: ; wherein, is the root mean square sound pressure, is the reference sound pressure, and =1μPa.

[0045] The instantaneous peak sound pressure level is calculated according to the following formula: ; wherein, is the peak sound pressure.

[0046] The single sound exposure level is calculated according to the following formula: ; wherein, T is the time of the pile driving pulse signal, in seconds (s), is the original underwater sound pressure signal, is the reference integral time (usually 1s).

[0047] The cumulative sound exposure level is calculated according to the following formula: ; wherein, i is the i-th pile driving pulse signal, and N is the total number of pile driving pulse signals.

[0048] The present application eliminates the influence of distance and propagation path on sound pressure measurement by time smoothing processing of pile driving pulse signal and propagation loss correction based on marine acoustic propagation model, so that the data of different monitoring sites are comparable; and the standardized acoustic evaluation parameters such as root mean square sound pressure level, instantaneous peak sound pressure level, single sound exposure level and cumulative sound exposure level are calculated, so as to provide standardized and normalized quantitative indexes for objective evaluation of pile driving noise and environmental protection.

[0049] In the embodiment, the average sound source level of pile driving operation is obtained by inversion calculation based on the acoustic evaluation parameters of multiple monitoring sites and the corresponding monitoring position information, in combination with the marine acoustic propagation model, including the following steps: The single sound exposure level of each monitoring site and the corresponding monitoring position information are obtained, and the straight-line distance between each monitoring position and the pile driving sound source is calculated; The propagation loss corresponding to each monitoring site is calculated based on the spherical or cylindrical expansion marine acoustic propagation model; the calculation formula of the propagation loss TL is specifically: ; wherein k is the expansion loss factor, r is the straight-line distance between the monitoring position and the pile driving sound source, and the unit is meter (m), is the absorption attenuation coefficient (unit: dB / m); The single sound exposure level of each monitoring site and the corresponding propagation loss are added to calculate the sound source level SL of each monitoring site corresponding to 1m away from the pile driving operation; that is: SL ; wherein is the single sound exposure level measured at the monitoring position r away from the pile driving sound source (unit: dB); the formula can be used to inversely deduce the original intensity of the sound source in the case of known monitoring site sound pressure level, or to predict the noise influence range at different distances; The arithmetic mean value of the sound source level SL of 1m away from the pile driving operation calculated based on all monitoring sites is taken as the average sound source level of the pile driving operation.

[0050] Specifically, if M received sound pressure levels are measured at a distance away from the pile driving sound source, and the propagation loss TL( ) corresponding to the jth monitoring site is estimated, then the average sound source level of the pile driving operation can be obtained by the following formula: ; wherein M is the total number of monitoring sites participating in the inversion calculation; j is the jth monitoring site, is the received sound pressure level, that is, the sound pressure level measured at the jth monitoring site; is the straight-line distance between the jth monitoring site and the pile driving sound source. is the propagation loss of the sound wave from the sound source to the jth monitoring site.

[0051] The present application effectively reduces the single-point measurement error and environmental uncertainty influence through multi-point monitoring data fusion, and significantly improves the accuracy and robustness of the sound source level estimation.

[0052] Further, the method further comprises: in response to a user query instruction, exporting user query data from the historical monitoring data. Preferably, the user query instruction includes time, place and other query conditions.

[0053] The method further comprises: generating a compliance determination report according to the average sound source level of the piling operation, directly corresponding to the environmental limit standard to monitor and generate a compliance determination report, and improving the supervision efficiency.

[0054] As shown in Figure 2 The present application also provides an intelligent piling underwater noise monitoring system, which comprises: A signal acquisition module 10 is used to acquire original underwater sound pressure signals of multiple monitoring sites; preferably, the signal acquisition module 10 is a hydrophone sensor array, which has high sensitivity, wide frequency band, and adjustable cable length (adjustment range is 1-50 meters); the hydrophone sensor array is composed of multiple hydrophone sensors arranged according to water depth layers (such as 5m, 15m, 30m); the layout point can be adjusted within a radius range of 100-5000m from the piling point; A preprocessing module 20 is used to receive and preprocess the original underwater sound pressure signals of multiple monitoring sites to obtain multiple underwater sound pressure signals; preferably, the preprocessing module 20 includes an analog-to-digital converter (ADC) and a microcontroller (MCU), which are used to convert the analog signals collected by the hydrophone into digital signals and perform band-pass filtering and gain adjustment processing; A signal extraction module 30 is used to detect and extract the underwater sound pressure signals of each monitoring site based on short-time energy analysis and adaptive threshold determination to obtain candidate pulse signals; A spectrum analysis module 40 is used to perform spectrum analysis on the candidate pulse signals to obtain time-frequency distribution graphs, log-mel spectrograms and frequency domain feature parameters; A classification module 50 is used to input the log-mel spectrogram into a deep learning model for identification and classification to obtain piling pulse signals from the candidate pulse signals; An acoustic evaluation parameter calculation module 60 is used to post-process the piling pulse signals and calculate the acoustic evaluation parameters of the piling underwater noise; A sound source level inversion module 70 is used to calculate the average sound source level of the piling operation based on the acoustic evaluation parameters of multiple monitoring sites and their corresponding monitoring position information, combined with the ocean acoustic propagation model.

[0055] Preferably, the system further comprises: a time domain feature extraction module configured to perform time domain feature extraction on the candidate impulse signal to obtain a time domain feature parameter; an alarm module configured to calculate an average sound pressure level in a period based on the piling impulse signal; compare the average sound pressure level in the period with a preset frequency band threshold; when the average sound pressure level in the period is greater than the frequency band threshold, generate and output a first alarm signal; and / or, calculate a sound exposure level of a single event based on the piling impulse signal; the single event is a single piling impact; compare the sound exposure level of the single event with a preset sound exposure level threshold; when the sound exposure level of the single event is greater than the sound exposure level threshold, generate and output a second alarm signal; a visualization module configured to display the time domain feature extraction result and / or the spectrum analysis result in a chart form; further, the visualization module also supports GIS map superposition to display a noise propagation range; a data query module configured to export user query data from the historical monitoring data in response to a user query instruction; preferably, the user query instruction includes time, place and other query conditions; an alarm event recording module configured to record an alarm event based on the first alarm signal and / or the second alarm signal; a wireless communication module configured to transmit the preprocessed underwater sound pressure signal to the signal extraction module 30 in real time, for example, a 4G / 5G module, a LoRa module; an energy management unit including a rechargeable battery module, an interface module reserving a solar cell panel interface (which needs to be installed according to the actual buoy), to ensure the continuous power supply of the system under long-time operation.

[0056] In the embodiment, the system can be arranged in a buoy / submarine main body, which has good water tightness and stability.

[0057] When the system can be arranged in a buoy main body, the system further comprises: a Beidou position module configured to locate the position of the buoy to obtain a drift trajectory of the buoy; Further, the visualization module is also configured to visually display the drift trajectory of the buoy.

[0058] It should be noted that each of the embodiments in the specification adopts a progressive manner for description, and each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to. For the system embodiment, since it is basically similar to the method embodiment, it is described more simply, and the relevant parts are referred to the part of the method embodiment.

[0059] Also, as used herein, the terms "comprise", "comprising", or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0060] The foregoing description illustrates and describes the preferred embodiments of the present application only. It will be understood by those skilled in the art that the present application is not limited to the particular embodiments described herein but is capable of a variety of modifications without departing from the scope of the present application. The application includes any alterations and modifications in the art within the scope of the attached claims.

Claims

1. An intelligent pile driving underwater noise monitoring method, characterized in that, The method comprises the following steps: receiving and preprocessing original underwater sound pressure signals of multiple monitoring sites to obtain multiple underwater sound pressure signals; detecting and extracting the underwater sound pressure signals of each monitoring site based on short-time energy analysis and adaptive threshold determination to obtain candidate impulse signals; performing spectral analysis on the candidate impulse signals to obtain time-frequency distribution diagrams, log-mel spectrograms, and frequency domain characteristic parameters; inputting the log-mel spectrograms into a deep learning model for identification and classification to obtain pile-driving impulse signals from the candidate impulse signals; post-processing the pile-driving impulse signals and calculating acoustic evaluation parameters of pile-driving underwater noise; based on the acoustic evaluation parameters of multiple monitoring sites and the corresponding monitoring position information, combining a marine underwater sound propagation model to obtain the average sound source level of pile-driving operation through inversion calculation.

2. The intelligent pile driving underwater noise monitoring method according to claim 1, characterized in that, The method comprises the following steps: calculating the short-time energy of the underwater sound pressure signals based on a preset window length and a preset overlap rate; calculating the corresponding adaptive threshold based on the energy mean value, standard deviation of the background noise of each monitoring site, and a preset sensitivity factor; comparing the short-time energy and the adaptive threshold of each monitoring site to identify candidate impulse signals; the short-time energy of the candidate impulse signal is greater than the adaptive threshold; extracting the envelope of the candidate impulse signal; performing low-pass filtering and smoothing processing on the envelope, and determining the start and end time of the candidate impulse signal based on an envelope detection threshold; based on the start and end time, the complete waveform of the candidate impulse signal is obtained by extending the preset filling time length before and after.

3. The intelligent pile driving underwater noise monitoring method according to claim 2, characterized in that, The frequency domain characteristic parameters include in-band energy distribution, peak frequency, and bandwidth parameters. The method comprises the following steps: performing fast Fourier transform or short-time Fourier transform based on the complete waveform of the candidate impulse signal to obtain a time-frequency distribution diagram; converting the time-frequency distribution diagram into a log-mel spectrogram, and calculating the in-band energy distribution, peak frequency, and bandwidth parameters of each candidate impulse signal based on the time-frequency distribution diagram.

4. The intelligent pile driving underwater noise monitoring method according to claim 1, characterized in that, After obtaining the candidate impulse signals, the method further comprises: performing time domain feature extraction on the candidate impulse signals to obtain time domain characteristic parameters; the time domain characteristic parameters include peak sound pressure, peak-to-peak sound pressure, pulse duration, rise time, fall time, pulse interval, and pulse rate.

5. The intelligent pile driving underwater noise monitoring method according to claim 1, characterized in that, The method comprises the following steps: inputting the log-mel spectrogram into a two-dimensional convolutional neural network model to perform feature learning and classification on the log-mel spectrogram; Based on the feature learning result, a class label and a confidence corresponding to each candidate pulse signal are outputted to obtain a pile-driving pulse signal from the candidate pulse signals; the class label at least includes pile-driving sound and ship machinery noise.

6. The intelligent pile driving underwater noise monitoring method according to claim 5, characterized in that, Also comprising: Based on the pile-driving pulse signal, an average sound pressure level in a cycle is calculated; The average sound pressure level in the cycle is compared with a preset frequency band threshold value, and when the average sound pressure level in the cycle is greater than the frequency band threshold value, a first alarm signal is generated and outputted.

7. The intelligent pile driving underwater noise monitoring method according to claim 5, characterized in that, Also comprising: Based on the pile-driving pulse signal, a single-event sound exposure level is calculated; The single event is a single pile-driving impact; The single-event sound exposure level is compared with a preset sound exposure level threshold value, and when the single-event sound exposure level is greater than the sound exposure level threshold value, a second alarm signal is generated and outputted.

8. The intelligent pile-driving underwater noise monitoring method according to claim 1, characterized in that, The pile-driving pulse signal is post-processed, and an acoustic evaluation parameter of pile-driving underwater noise is calculated, including the following steps: The pile-driving pulse signal is time-smoothed; Based on the propagation distance of each pile-driving pulse signal and a marine underwater acoustic propagation model, the pile-driving pulse signal is corrected for propagation loss; Based on the sound pressure level after propagation loss correction, an acoustic evaluation parameter of pile-driving underwater noise is calculated, and the acoustic evaluation parameter includes a root mean square sound pressure level, an instantaneous peak sound pressure level, a single sound exposure level, and a cumulative sound exposure level.

9. The intelligent pile driving underwater noise monitoring method according to claim 8, characterized in that, Based on the acoustic evaluation parameters of multiple monitoring sites and their corresponding monitoring position information, combined with a marine underwater acoustic propagation model, an average sound source level of pile-driving operation is calculated by inversion, including the following steps: The single sound exposure level of multiple monitoring sites and the corresponding monitoring position information are obtained, and the straight-line distance between each monitoring position and the pile-driving sound source is calculated; Based on a spherical or cylindrical marine underwater acoustic propagation model, the propagation loss corresponding to each monitoring site is calculated; The single sound exposure level of each monitoring site is added to the corresponding propagation loss to calculate the sound source level at 1 m from the pile-driving operation corresponding to each monitoring site; The sound source levels at 1 m from the pile-driving operation calculated by all monitoring sites are taken as the arithmetic mean, and the arithmetic mean is taken as the average sound source level of the pile-driving operation.

10. An intelligent pile driving underwater noise monitoring system, characterized in that, Comprising: A signal acquisition module for acquiring original underwater sound pressure signals of multiple monitoring sites; A preprocessing module for receiving and preprocessing the original underwater sound pressure signals of multiple monitoring sites to obtain multiple underwater sound pressure signals; A signal extraction module for detecting and extracting the underwater sound pressure signals of each monitoring site based on short-time energy analysis and adaptive threshold determination to obtain candidate pulse signals; A spectrum analysis module for performing spectrum analysis on the candidate pulse signals to obtain time-frequency distribution diagrams, log-mel spectrograms, and frequency domain feature parameters; A classification module for inputting the log-mel spectrogram into a deep learning model for recognition and classification to obtain a pile-driving pulse signal from the candidate pulse signals; An acoustic evaluation parameter calculation module for post-processing the pile-driving pulse signal and calculating an acoustic evaluation parameter of pile-driving underwater noise. A sound source level inversion module is configured to calculate the average sound source level of the piling operation based on the acoustic evaluation parameters of the plurality of monitoring sites and the corresponding monitoring position information, in combination with a marine underwater acoustic propagation model.

Images