Feature extraction and simulation method for sound signals of autonomous navigation ship tunnel thruster

Abstract

A feature extraction and simulation method for sound signals of an autonomous navigation ship tunnel thruster. A digital filter is used to process measured signals to remove spike signals, and the characteristics of residual wideband floor noise signals are analyzed. A finite impulse response filter is designed, and the impulse response of the finite impulse response filter is convolved with white Gaussian noise, so that simulated noise is matched with frequency-domain characteristics of floor noise signals. A spike signal amplitude is calculated, and a group of sine wave signals having a corresponding frequency are generated as spike signals. Narrowband spike signals and wideband floor signals are superposed and synthesized into simulated tunnel thruster noise. A feasible solution is provided for research and analysis of sound signals of a tunnel thruster. The sound characteristics of the tunnel thruster are accurately extracted, a predicted sound level is improved, and a tunnel thruster sound source database is established, thus providing valuable data support for signal quality analysis of a ship autonomous navigation system, fault diagnosis and improvement of a control strategy for the tunnel thruster.

Description

Feature extraction and simulation method of tunnel thruster sound signals of autonomous ships Technical Field

[0001] The present invention relates to a signal processing and simulation technology, and in particular to a method for extracting and simulating the characteristics of a sound signal of a tunnel thruster of an autonomous navigation ship. Background Art

[0002] Autonomous navigation technology for unmanned ships is a key technology for intelligent ships. Autonomous ships must navigate and operate in complex marine environments, placing stringent demands on their maneuverability and reliability. Tunnel thrusters are installed in transverse tunnels below the waterline at the bow or stern of a ship, primarily providing lateral thrust to the hull during navigation. Analysis, modeling, and prediction of tunnel thruster acoustic signals provide crucial insights for signal quality analysis, thruster operating condition estimation, and fault diagnosis, ensuring that the ship's autonomous navigation system maintains full and timely monitoring and control of tunnel thruster status.

[0003] The noise generated by tunnel thrusters during operation can be primarily categorized as mechanical noise, hydrodynamic noise, aerodynamic noise, and propeller noise. Propeller noise refers to the noise generated by various mechanisms, including changes in the surrounding flow field and pressure caused by the propeller's rotation. Mechanical noise is primarily caused by the vibration or impact of the thruster's moving parts during operation, caused by periodic changes in gas pressure and inertia. Hydrodynamic noise is the noise generated by water flowing over the thruster surface. Aerodynamic noise includes the excitation vibration noise caused by intake and exhaust. Different types of noise exhibit distinct characteristics in the frequency spectrum. Narrowband spikes in the noise energy spectrum primarily originate from mechanical and propeller noise, and most of these signals manifest as harmonics across the entire frequency band. Broadband floor noise is a collection of other non-narrowband noises, primarily originating from hydrodynamic and aerodynamic noise.

[0004] Most existing sound signal processing methods lack the ability to identify, separate, and model tunnel thruster noise from different sources. The processing algorithms are also complex and require significant computational resources. Furthermore, the simulated sound quality is significantly affected by the accuracy of the analysis, resulting in limitations in analyzing and predicting tunnel thruster sound signals in autonomous navigation systems. Therefore, there is an urgent need to develop an efficient and rapid feature recognition and signal simulation method for tunnel thruster noise, suitable for autonomous ship navigation systems.

[0005] Summary of the Invention

[0006] To address the above problems, a feature extraction and simulation method for the sound signal of the tunnel thruster of an autonomous navigation ship is proposed, which provides the autonomous navigation system with feature recognition, separation, modeling and prediction of the tunnel thruster sound.

[0007] The technical solution of the present invention is: a feature extraction and simulation method for the sound signal of a tunnel thruster of an autonomous navigation ship, which comprises the following steps: performing high-pass filtering on the collected tunnel thruster sound signal, calculating the power spectral density of the filtered signal, performing minimum filtering on the obtained power spectral density vector, and separating the peak signal representing narrowband noise and the broadband base noise from the processed signal; identifying and extracting the peak signal in the processed signal, obtaining the frequency and peak signal amplitude corresponding to the peak signal, and reconstructing a sinusoidal signal to form a narrowband peak signal; performing residual broadband base noise signal characteristic analysis on the processed signal, designing a digital filter to simulate a frequency response matching the power spectral density of the base signal, and then generating white noise and convolving the frequency response of the digital filter to generate a simulated broadband base noise signal to form a broadband base signal; and superimposing the narrowband peak signal and the broadband base signal to synthesize the simulated tunnel thruster noise for autonomous navigation system signal analysis and prediction.

[0008] Furthermore, a method for collecting tunnel thruster sound signals is provided: a sound sensor is placed near the tunnel thruster at the bottom of the ship to collect and record the sound signals received at this position; the working conditions of the tunnel thruster are changed, and the position and direction of the sound sensor are adjusted to collect sound signals at different positions, and the audio signal data is recorded; the initial discrete time domain signal obtained is real sound pressure data, and the analog signal is converted into a discrete digital signal through a digital signal converter in the sound sensor, and saved in the form of a discrete digital signal for subsequent signal analysis.

[0009] Furthermore, the power spectral density of the filtered signal is calculated: a sound sample of the tunnel propeller in stable operation is selected from the complete sampling signal. The sample is first truncated into smaller segments, and each segment is weighted using the Hanning window function. The discrete Fourier transform of each segment is then calculated.

[0010] The power spectral density of each segment is where c w is the window function correction coefficient; X k is the discrete Fourier transform of each sound segment; f s is the signal sampling frequency, which is related to the sound sensor parameters; t s is the time length of the signal in each period; finally, the average power spectrum density of all samples is calculated; the power spectrum density of the decibel level is calculated as

[0011] Furthermore, the spike signal in the processed signal is identified and extracted: the power spectrum density vector processed by the minimum value filter is compared with the original power spectrum vector, a difference operation is performed, and a threshold is set to determine the spike signal.

[0012] Furthermore, a digital filter design method is used: the power spectrum density is used to design an FIR filter so that the frequency characteristics of the FIR digital filter at discrete frequency points are equal to or close to the values ​​of the broadband background noise signal spectrum at these frequency points, and the characteristics at other frequencies are well approximated; according to the Nyquist sampling theorem, the bilateral frequency response Y(k) of the signal is calculated as follows: m(k) is the power spectral density after mean filtering in the previous paragraph, with a frequency range of 0 to f s / 2, where f s is the signal sampling frequency; first along f s / 2Calculate the complex conjugate of Y(k) and symmetric to f s / 2 to f s , the frequency range is 0 to f s H(k). At this time, H(k) is along f s / 2 conjugate symmetry, that is, H(Nk)=conj(H(N+k)). For H(k) from 0Hz to f s Performing a discrete inverse Fourier transform, we obtain a finite-length digital sequence impulse response h(n), where n=0, 1, ..., 2N-1, as shown in the following formula:

[0013] Finally, the discrete points after h(N+1) are shifted to the range of k<0 to obtain the final impulse signal response g(k) of the FIR filter, that is, for k<0, g(k)=h(k+2N), and since g(n)=g(-n), the phase of this FIR filter is zero.

[0014] Furthermore, the white noise generation method is: Gaussian distributed random white noise, whose mean is 0 and variance is the sampling frequency of the signal, and the power spectral density of the generated white noise is 1.

[0015] The beneficial effect of this invention lies in the fact that the method for extracting and simulating the acoustic signals of tunnel thrusters used in autonomous navigation ships provides a feasible solution for the research and analysis of tunnel thruster acoustic signals. By accurately extracting tunnel thruster acoustic characteristics and improving the predicted sound level, a tunnel thruster sound source database is established, providing valuable data support for signal quality analysis, fault diagnosis, and improved control strategies for tunnel thrusters in autonomous navigation systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG1 is a schematic diagram of a method for extracting and simulating characteristics of a tunnel thruster noise signal according to the present invention. DETAILED DESCRIPTION

[0017] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0018] A method for feature extraction and simulation of tunnel thruster acoustic signals from autonomous ships is presented. A digital filter is used to process the measured signal, removing spikes and analyzing the characteristics of the residual broadband background noise signal. A finite impulse response filter is designed and its impulse response is convolved with Gaussian white noise to match the frequency domain characteristics of the simulated noise with the background noise signal. The spike signal amplitude is calculated, and a set of sinusoidal signals of corresponding frequencies are generated as spike signals. The superposition of the spike signal and the background signal is used as the simulated and predicted tunnel thruster acoustic signal.

[0019] As shown in Figure 1, an embodiment of the present invention provides a feature extraction and simulation method for a tunnel thruster noise signal model, which performs high-pass filtering on the collected tunnel thruster sound signal, calculates the power spectral density of the filtered signal, performs minimum filtering on the obtained power spectral density vector, and then separates the peak signal representing narrowband noise and the broadband base noise from the processed signal; identifies and extracts the peak signal in the processed signal, obtains the frequency and peak signal amplitude corresponding to the peak signal, and uses them to reconstruct a sinusoidal signal to form a narrowband peak signal; performs residual broadband base noise signal characteristic analysis on the processed signal, designs a digital filter to simulate a frequency response that matches the power spectral density of the base signal, and then generates white noise and convolves the frequency response of the digital filter to generate a simulated broadband base noise signal to form a broadband base signal; superimposes the narrowband peak signal and the broadband base signal into a simulated tunnel thruster noise for autonomous navigation system signal analysis and prediction.

[0020] To collect sound signals, a sound sensor is placed near the tunnel thruster on the bottom of the ship. Sound signals received at this location are collected and recorded. By adjusting the tunnel thruster's operating conditions and the position and orientation of the sound sensor, sound signals from different locations are collected and recorded for subsequent model signal analysis. The initial discrete time-domain signal obtained represents actual sound pressure data. The digital signal converter in the sound sensor converts the analog signal into a discrete digital signal x[n]. The sound signal is then stored as a discrete digital signal for subsequent model signal analysis.

[0021] Receive a tunnel thruster audio signal from actual measurement to be analyzed, and perform high-pass filtering on the signal with a cutoff frequency generally less than 20Hz to remove low-frequency signal interference that is inaudible to the human ear in the measurement environment.

[0022] The power spectral density analysis of the signal after the first filtering was performed using a discrete Fourier transform. The specific method involved selecting a stable and representative sample from a complete sampling signal (under stable operating conditions for the tunnel propeller). The power spectral density analysis method first truncated the sample into smaller segments, weighted each segment using a Hanning window function, and then calculated the discrete Fourier transform of each segment.

[0023] The power spectral density of each segment is where c w is the window function correction coefficient; X k is the discrete Fourier transform of each sound segment; f s is the signal sampling frequency, which is related to the sound sensor parameters; t s is the time length of the signal in each period. Finally, the average power spectrum density of all samples is calculated. The power spectrum density of the decibel level is calculated as After calculation, the relationship between signal power spectrum density and frequency is obtained.

[0024] Separate the peak signal representing narrowband noise from the broadband base noise. The power spectral density of the obtained sample (i.e., the power spectral density of the previously calculated decibel level) is regarded as a one-dimensional vector along the frequency domain direction, and a minimum filter is performed to remove the narrowband spike. The moving minimum filter uses a sliding window method to calculate the minimum value. A window of a specified length moves sample by sample on the vector and independently calculates the minimum value of the data in the window to remove the peak. Assuming the window length is 2M+1, the window moves to the nth point for minimum filtering, and the filtered value of the nth point is PL 最小值滤波 (n) = min{PL (n-M) , PL (n-M+1) ,...,PL (n) ,...,PL (n+M-1) , PL (n+M) The minimum filter order is related to the resolution of the power spectral density. The window length should be slightly larger than the bandwidth of the narrow-frequency peak in the spectrum. In actual design, it is necessary to try different filter orders multiple times to obtain a better smoothing effect.

[0025] Compare the power spectrum density vector after minimum filtering with the original power spectrum vector, perform difference calculation, set a threshold to determine the peak signal, and record the size of the frequency difference at the corresponding frequency. Record the frequencies corresponding to a series of peak signals. Generate a sinusoidal function of the corresponding frequency and amplitude based on the calculation results, add a random value phase, and the reconstructed sinusoidal signal constitutes a narrowband spike signal model. Since the Hanning window is used in the calculation of the power spectrum density, the energy of the spike noise is dispersed. Therefore, when calculating the peak noise amplitude, the energy spectrum difference of several discrete points of the frequency response near the peak frequency in the power spectrum density should be accumulated. For example, for a specific frequency f1, assuming that the sum of the difference between the original frequency response and the filtered frequency response of the discrete points near it is A, the expression of the sine function is where Δ is 1 / f s , which is the reciprocal of the signal sampling frequency; n is the serial number of the discrete signal, n = 0, 1, 2...; The random phase of the trigonometric function has a random value in the range of 0-π. For different n in y(n), a phase angle is randomly selected.

[0026] The power spectrum density vector y[n] after minimum filtering is PL 最小值滤波 (n), continue to perform sliding average filtering on the residual broadband base signal to further smooth the power spectrum density curve of the signal. Assuming the window length is 2K+1 (the typical value of the window length is K=1 or 2), the output of the symmetric moving average filter is where b j The sum of the two is 1. For the one-dimensional vector b, either an average or Hanning window distribution can be used. A moving average filter is a low-pass filter that complements pre-filtering to reduce the impact of random interference. The order of a sliding average filter is typically small, ensuring that it removes sample bias while preserving power spectral density details. The specific order requires multiple trials in actual design.

[0027] At this time, a broadband background noise signal is obtained. Use this power spectral density to design an FIR filter so that the frequency characteristics of the FIR digital filter at discrete frequency points are equal to or close to the values ​​of the broadband background noise signal spectrum at these frequency points (broadband background noise signal spectrum, that is, the signal after the previous section has been processed by sliding average filtering. This signal is a power spectrum density-frequency relationship, that is, each discrete frequency point (such as 1Hz, 2Hz, 3Hz...) has a calculated power spectrum density value), and the characteristics at other frequencies have a good approximation. Digital filters with finite duration impulse response have the following main advantages: They have precise linear phase. They are always stable. The design method is usually linear. They can be efficiently implemented in hardware. According to the Nyquist sampling theorem, the bilateral frequency response Y(k) of the signal is calculated as follows: m(k) is the power spectral density after mean filtering in the previous paragraph, with a frequency range of 0 to f s / 2, where f s is the signal sampling frequency. To calculate the impulse response of the FIR filter, first s / 2Calculate the complex conjugate of Y(k) and symmetric to f s / 2 to f s , the frequency range is 0 to f s H(k). At this time, H(k) is along f s / 2 conjugate symmetry, that is, H(Nk)=conj(H(N+k)). For H(k) from 0Hz to f s Performing a discrete inverse Fourier transform, we obtain a finite-length digital sequence impulse response h(n), where n=0, 1, ..., 2N-1, as shown in the following formula:

[0028] Finally, the discrete points after h(N+1) are shifted to the range where k < 0, yielding the final FIR filter impulse signal response g(k). That is, for k < 0, g(k) = h(k+2N). Since g(n) = g(-n), the phase of this FIR filter is zero. After the broadband background noise signal passes through the designed FIR filter, the broadband floor noise frequency response is obtained.

[0029] To generate a simulated broadband background noise signal of a certain length, we first generate a Gaussian-distributed random white noise signal with a mean of 0 and a variance equal to the signal sampling frequency. The power spectral density of this generated white noise is 1. We then convolve this white noise signal with the frequency response of the FIR filter output to artificially generate a simulated broadband background noise signal, which serves as background noise. Because the impulse response remains unchanged after passing through the white noise signal, its power spectral density matches the actual measured broadband background noise signal.

[0030] A set of sinusoidal functions representing narrowband signals and a simulated broadband background noise signal are added to obtain an artificially simulated tunnel thruster audio signal.

[0031] After acquiring simulated tunnel propeller acoustic signals, a tunnel propeller sound source database can be established to evaluate signal quality and improve the onboard operating environment. The relationship between the mechanical structure and acoustic characteristics of the tunnel propeller can be analyzed, and frequency domain relationships can be used to analyze abnormal sounds caused by mechanical failures. This analysis can help detect tunnel propeller failures and implement appropriate repair measures, thereby improving the reliability and operating efficiency of the tunnel propeller.

[0032] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A method for extracting and simulating the sound signal of a tunnel thruster of an autonomous ship, characterized in that: The collected tunnel thruster sound signal is subjected to high-pass filtering, and the power spectral density of the filtered signal is calculated. After the obtained power spectral density vector is subjected to minimum filtering, the peak signal representing narrowband noise and the broadband base noise are partially separated from the processed signal; the peak signal in the processed signal is identified and extracted, and the frequency and peak signal amplitude corresponding to the peak signal are obtained to reconstruct the sinusoidal signal to form a narrowband spike signal; the residual broadband base noise signal characteristics of the processed signal are analyzed, and a digital filter is designed to simulate a frequency response that matches the power spectral density of the base signal, and then white noise is generated and convolved with the frequency response of the digital filter to generate a simulated broadband base noise signal to form a broadband base signal; the narrowband peak signal and the broadband base signal are superimposed and synthesized into a simulated tunnel thruster noise for signal analysis and prediction of the autonomous navigation system.

2. The feature extraction and simulation method of the sound signal of the tunnel thruster of an autonomous navigation ship according to claim 1 is characterized in that: The method for collecting the sound signal of the tunnel thruster is as follows: placing the sound sensor near the tunnel thruster at the bottom of the ship, collecting and recording the sound signal received at this position; changing the working condition of the tunnel thruster, collecting the sound signals at different positions by adjusting the position and direction of the sound sensor, and recording the audio signal data; The initial discrete time domain signal obtained is the real sound pressure data. The analog signal is converted into a discrete digital signal through the digital signal converter in the sound sensor and saved in the form of a discrete digital signal for subsequent signal analysis.

3. The feature extraction and simulation method of the sound signal of the tunnel thruster of an autonomous navigation ship according to claim 1 is characterized in that: The power spectrum density of the filtered signal is calculated: a sound sample of the tunnel thruster in stable operation is selected from a complete sampling signal, the sample is first truncated into segments of smaller length, and each segment is weighted, the weighting function is the Hanning window function, and then the discrete Fourier transform of each segment is calculated; The power spectral density of each segment is where c w is the window function correction coefficient; X k is the discrete Fourier transform of each sound segment; f s is the signal sampling frequency, which is related to the sound sensor parameters; t s is the time length of the signal in each period of time; finally, the average power spectral density of all samples is calculated; the power spectral density of the decibel level is calculated as follows:

4. The feature extraction and simulation method of the sound signal of the tunnel thruster of an autonomous navigation ship according to claim 1 is characterized in that: Identify and extract spike signals in the processed signal: compare the power spectrum density vector processed by minimum value filtering with the original power spectrum vector, perform difference operation, and set a threshold to determine the spike signal.

5. The feature extraction and simulation method of the sound signal of the tunnel thruster of an autonomous navigation ship according to claim 3 is characterized in that: Design method of digital filter: Use this power spectrum density to design FIR filter, so that the value of the frequency characteristic of FIR digital filter at discrete frequency points is equal to or close to the value of the spectrum of broadband background noise signal at these frequency points, and the characteristics at other frequencies have good approximation; according to Nyquist sampling theorem, the bilateral frequency response Y(k) of the signal is calculated as follows: m(k) is the power spectral density after mean filtering in the previous paragraph, with a frequency range of 0 to f s / 2, where f s is the signal sampling frequency; first along f s / 2Calculate the complex conjugate of Y(k) and symmetric to f s / 2 to f s , and the frequency range is 0 to f s H(k). At this time, H(k) is along f s / 2 conjugate symmetry, that is, H(Nk)=conj(H(N+k)). For H(k) from 0Hz to f s Perform a discrete inverse Fourier transform to obtain a finite length digital sequence impulse response h(n), n = 0, 1, ..., 2N-1, as shown below: Finally, the discrete points after h(N+1) are shifted to the range of k<0 to obtain the final pulse signal response g(k) of the FIR filter, that is, for k<0, g(k)=h(k+2N), and since g(n)=g(-n), the phase of this FIR filter is zero.

6. The method for extracting and simulating the sound signal of the tunnel thruster of an autonomous ship according to claim 5 is characterized in that: White noise generation method: Gaussian distributed random white noise, whose mean is 0 and variance is the sampling frequency of the signal, and the power spectral density of the generated white noise is 1.

Images