Noise Source Localization System and Method for Planetary Gear Sets in Underwater Ship Gearboxes

Abstract

This invention relates to the field of marine propulsion system technology and discloses a noise source localization system for planetary gear sets in underwater marine gearboxes. The system includes: an ultrasonic transmitting unit, an ultrasonic radiation power supply module, an acceleration sensor, a data acquisition module, and a data processing module. The data acquisition module includes multiple stress sensor groups, a signal processing circuit, and a main control chip. Stress sensor groups are installed on multiple adjacent tooth roots of each planetary gear, and each stress sensor group is connected to the main control chip via the signal processing circuit. The acceleration sensor is used to collect operating status data of the planetary gear set to determine the meshing moment. The data processing module is used to locate the noise source based on the stress data at the meshing moment and a noise localization neural network model. This invention can improve the accuracy of noise source location and optimize gear design.

Description

Technical Field

[0001] This invention relates to the field of marine propulsion system technology, specifically to a noise source localization system and method for planetary gear sets in underwater gearboxes of ships. Background Technology

[0002] The dynamic stress characteristics of a ship's power transmission system directly affect the performance, lifespan, reliability, and noise level of the gear transmission system. In actual operation, gears are subjected to complex loads, especially at high speeds, under heavy loads, and under drastic changes in operating conditions. Their dynamic stress characteristics significantly impact the system's noise, vibration, and failure risk. Therefore, accurately obtaining and analyzing the dynamic stress distribution of gear sets is crucial for noise monitoring and optimization of gear transmission systems.

[0003] Gear dynamic stress testing mainly includes static and dynamic testing methods. Compared to static testing, dynamic testing methods can more realistically reflect the stress variation of gears under actual operating conditions, providing important data support for studying the response characteristics of gears under complex working conditions such as high-frequency meshing and load fluctuations. Currently, dynamic stress testing methods typically employ strain gauge measurement technology combined with signal processing algorithms to analyze the dynamic stress characteristics of gears in both the time and frequency domains. However, gear dynamic stress testing suffers from problems such as weak signals, severe noise interference, and limited acquisition accuracy. Especially under strong noise and high dynamic load environments, the signal is easily submerged by noise, greatly affecting the accuracy of the test results. Moreover, underwater gear dynamic stress testing also faces challenges such as inconvenient power supply and data transmission.

[0004] To address these issues, it is urgent to optimize existing noise localization methods for planetary gear sets in underwater ship gearboxes. By combining multi-scale decomposition and filtering optimization techniques for stress signals, the adaptability and reliability of dynamic stress testing systems can be further improved, providing more accurate data support for noise localization of planetary gear transmission systems. Summary of the Invention

[0005] This invention overcomes the shortcomings of existing technologies and aims to solve the following technical problem: providing a noise source localization system and method for planetary gear sets in underwater ship gearboxes, so as to acquire stress signals in the service state of the gearbox and locate the noise source through the stress signals, thereby providing a scientific basis and theoretical guidance for reducing the noise of underwater ship gearboxes.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a noise source localization system for planetary gear sets in underwater gearboxes of ships, comprising: an ultrasonic transmitting unit, an ultrasonic radiation power supply module, an acceleration sensor, a data acquisition module and a data processing module;

[0007] The ultrasonic radiation power supply module includes: an ultrasonic receiving unit, a supercapacitor bank, a primary coil, and a secondary coil; the ultrasonic transmitting unit and the ultrasonic receiving unit are respectively disposed on the outer wall of the gearbox and inside the gearbox; the primary coil is connected to the ultrasonic receiving unit, and the secondary coil is connected to the supercapacitor bank; the ultrasonic transmitting unit is used to transmit ultrasonic waves to the ultrasonic receiving unit; the ultrasonic receiving unit is used to generate an alternating current in the primary coil through the received ultrasonic waves; the secondary coil is used to charge the supercapacitor bank according to the magnetic field generated by the primary coil; and the supercapacitor bank is used to power the data acquisition module.

[0008] The data acquisition module includes multiple stress sensor groups, signal processing circuits, and a main control chip; each planetary gear of the planetary gear set is provided with multiple measuring points, which are distributed at multiple adjacent tooth root critical sections of the planetary gear. Each measuring point is equipped with a stress sensor group, and each stress sensor group is connected to the main control chip through the signal processing circuit.

[0009] The accelerometer is used to collect the running status data of the planetary gear and send it to the main control chip. The main control chip is used to determine the meshing time of the first measuring point of the planetary gear based on the output signal of the accelerometer, and send a synchronous acquisition command to control the stress sensor group to collect and store stress data at the meshing time. It is also used to synchronously control the supercapacitor group to enter the high power output mode.

[0010] The data processing module receives stress data collected by the data acquisition module and performs comprehensive analysis on the stress data of each planetary gear based on the noise localization neural network model to obtain the planetary gear corresponding to the noise source.

[0011] The data processing module includes a rod array antenna and a host computer. The rod array antenna is used for data reception, and the host computer is equipped with a wavelet threshold denoising module and a noise localization module.

[0012] The wavelet threshold denoising module includes a wavelet transform module and an adaptive filtering module. The wavelet transform module is used to reconstruct the received stress signal using a wavelet denoising method, and the adaptive filtering module is used to filter the reconstructed stress signal using an adaptive filtering algorithm. The weight update formula of the adaptive filtering module is:

[0013] ;

[0014] in, Step size factor For error signals, For input signals; Indicates the number of iterations; and These represent the weights obtained in the nth and (n+1)th iterations, respectively.

[0015] The noise localization module is used to perform comprehensive analysis on the filtered stress data of each planetary gear based on the noise localization neural network model to obtain the planetary gear corresponding to the noise source.

[0016] The signal processing module includes: multiple Wheatstone bridges, signal amplification and filtering circuits, and AD conversion circuits. The output of each stress sensor group is connected to a Wheatstone bridge, and the outputs of each Wheatstone bridge are connected to the main control chip in sequence through the signal amplification and filtering circuit and the AD conversion circuit.

[0017] The stress sensor group includes four strain gauges, which, together with the corresponding Wheatstone bridge, form a temperature-compensated full-bridge circuit. Two strain gauges are arranged along the principal stress direction to measure the actual strain, while the other two strain gauges are arranged perpendicular to the principal stress direction to achieve temperature compensation.

[0018] The noise source localization system for planetary gear sets in underwater gearboxes of ships further includes a signal generator, a power driver, and an impedance matching network. The signal generator drives the ultrasonic transmitting unit to emit ultrasonic waves via the power driver and the impedance matching network. The signal generator is also used to control the power supply amplitude of the ultrasonic transmitting unit based on the data collected by the accelerometer, so that the secondary coil receives energy stably.

[0019] The data acquisition module also includes a data transmission module, which is used to send the data acquired by the data acquisition module to the data processing module; the main control chip is also used to control the supercapacitor group to enter a high-power output mode to supply power to the data transmission module during data transmission.

[0020] The noise localization neural network model used in the data processing module adopts a fusion architecture of multi-layer convolutional neural network and gated recurrent unit; it includes multiple convolutional layers, pooling layers, GRU layers, and fully connected layers connected in sequence; the noise localization neural network model uses the cross-entropy loss function to optimize network parameters and adjusts network weights through backpropagation algorithm and gradient descent optimizer.

[0021] Furthermore, this invention also provides a noise source localization method for planetary gear sets in underwater ship gearboxes, implemented based on the aforementioned noise source localization system for planetary gear sets in underwater ship gearboxes, comprising the following steps:

[0022] Step 1: Power is supplied to the inside of the gearbox through the ultrasonic transmitting unit and the ultrasonic radiation power supply module.

[0023] Step 2: Obtain the output signal of the accelerometer through the main control chip, determine the meshing time of the first measuring point of the planetary gear, and send a synchronous acquisition command to control the stress sensor group to collect and store stress data at the meshing time. At the same time, activate the supercapacitor group to enter the high power supply mode.

[0024] Step 3: After the engine stops, prepare to enter the data receiving mode, control the supercapacitor group to enter the high-power energy supply mode, and then start receiving and collecting data;

[0025] Step 4: Perform time-domain alignment on the collected multiple data streams, map them to the same time domain, extract feature parameters, and input them into the noise localization neural network model to locate the noise source.

[0026] In step four, the noise localization neural network model takes the stress peak, rate of change, and waveform characteristics of each planetary gear as input and outputs the probability distribution of the noise source location for each planetary gear as output.

[0027] The method for locating noise sources in planetary gear sets of underwater gearboxes for ships further includes the following steps:

[0028] Step 5: Optimize gear materials, structural parameters, and lubrication conditions based on the probability distribution of noise source locations.

[0029] Compared with the prior art, the present invention has the following advantages:

[0030] (1) This invention provides a noise source localization system and method for planetary gear sets in underwater ship gearboxes. It characterizes the noise characteristics of the planetary gear set by synchronously collecting stress distribution data at multiple locations during gear meshing, and accurately identifies the main noise sources of the planetary gear set using a noise source localization algorithm. This provides a scientific basis and theoretical guidance for optimizing noise reduction in underwater ship gearboxes. Based on stress characteristics, this invention can provide optimized design schemes for gear sets and provide key data support for structural improvements of planetary gear sets within gearboxes, thereby effectively reducing noise in underwater ship gearboxes and improving the overall performance and operational reliability of underwater ship gearboxes.

[0031] (2) The present invention uses high-energy ultrasonic radiation to solve the power supply problem in the closed gearbox and stores energy through supercapacitors. It can dynamically open a specified number of capacitor modules to supply energy according to the operating status, thereby fully improving the energy utilization rate.

[0032] (3) In stress measurement, this invention utilizes a stress sensor group consisting of four strain gauges forming a temperature-compensated full-bridge structure, and uses a data acquisition module to synchronously acquire the stress signals at the moment of meshing of each planetary gear, thereby improving the accuracy and real-time performance of the data. Moreover, the acquired data is first stored locally and then transmitted after the engine is stopped, which improves the stability of the data.

[0033] In summary, this invention provides a noise source localization system and method for planetary gear sets in underwater gearboxes of ships, integrating power supply and acquisition schemes and data analysis algorithms. It can significantly improve the accuracy and efficiency of stress monitoring and noise localization, and provide reliable key data support for noise control of underwater equipment. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of a noise source localization system for a planetary gear set in an underwater gearbox of a ship, provided by an embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of the ultrasonic radiation power supply module in an embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the planetary gear set in an embodiment of the present invention;

[0037] Figure 4 This is a schematic diagram of the structure of the noise localization neural network model used in the data processing module in this embodiment of the invention;

[0038] In the diagram: 1-Gear ring, 2-Bearing, 3-Planetary gear, 4-Stress sensor, 5-Planet carrier, 6-Sun gear, 7-Rim mounting hole. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Example 1

[0041] like Figure 1 As shown, Embodiment 1 of the present invention provides a noise source localization system for planetary gear sets in underwater gearboxes of ships, including: an ultrasonic radiation power supply module, an acceleration sensor, a data acquisition module, and a data processing module.

[0042] like Figure 2As shown, in this embodiment, the ultrasonic radiation power supply module includes: an ultrasonic transmitting unit, an ultrasonic receiving unit, a supercapacitor bank, a primary coil, and a secondary coil. The ultrasonic transmitting unit and the ultrasonic receiving unit are respectively disposed on the outer wall of the gearbox and inside the gearbox. The primary coil is connected to the ultrasonic receiving unit, and the secondary coil is connected to the supercapacitor bank. The ultrasonic transmitting unit is used to transmit ultrasonic waves to the ultrasonic receiving unit, and the ultrasonic transmitting unit and the ultrasonic receiving unit form an ultrasonic transducer. The ultrasonic receiving unit is used to generate an alternating current in the primary coil through the received ultrasonic waves. The secondary coil is used to charge the supercapacitor bank according to the magnetic field generated by the primary coil. The supercapacitor bank is used to power the data acquisition module.

[0043] Specifically, such as Figure 1 As shown, in this embodiment, the data acquisition module includes multiple stress sensor groups, a signal processing circuit, and a main control chip. The stress sensor groups are distributed on each planetary gear 3, and multiple measuring points are set on each planetary gear 3. The measuring points are distributed at multiple adjacent tooth root critical sections of the planetary gear, and one stress sensor group is arranged at each measuring point. Each stress sensor group is connected to the main control chip via the signal processing circuit. Specifically, one stress sensor group is set at each corresponding tooth root critical section.

[0044] like Figure 3 The diagram shows the structure of a planetary gear set. A ring gear 1 has a ring fixing hole 7 on its outer side. Five planetary gears 3 are mounted inside the ring gear 1 via bearings 2 on the planet carrier 5. A sun gear 6 is located at the center of the ring gear 1. Each planetary gear 3 meshes with the sun gear 6 and the ring gear 1. Each stress sensor 4 in each stress sensor group is located at the critical section of the tooth root of the corresponding planetary gear 3.

[0045] Specifically, such as Figure 2 As shown, in this embodiment, the accelerometer is used to collect the operating status data of each planetary gear 3 and send it to the main control chip.

[0046] Specifically, the accelerometer is installed on the sidewall of each planetary gear 3 to collect triaxial acceleration data during the rotation of the planetary gear 3 in real time, and sends it to the main control chip inside the gearbox for preliminary analysis and processing. Based on the triaxial acceleration data, the main control chip can accurately determine the current rotational position and angle information of the planetary gear 3 in real time, thereby judging the real-time meshing state of the gears, and determining in advance the position of the gears that are about to enter the meshing state, accurately identifying the meshing moment.

[0047] The main control chip is used to determine the meshing moment of each planetary gear 3 based on the output signal of the accelerometer and to send a synchronous acquisition command. It controls multiple stress sensor groups corresponding to each planetary gear 3 to collect and store stress data at the meshing moment. It also synchronously controls the supercapacitor group to enter a high-power output mode to power the acquisition circuit of the data acquisition module. During the data acquisition process, the triaxial acceleration data collected in real time by the accelerometer is analyzed by the main control chip. Once the meshing moment is determined to be imminent, the main control chip sends a synchronous acquisition command in advance, and the stress sensor group enters a short-time high-frequency sampling mode to capture the precise stress characteristics of the planetary gear 3 at the moment of meshing.

[0048] Furthermore, the triaxial acceleration signals collected by the accelerometer mounted on the sidewall of planetary gear 3 can accurately reflect the phase characteristic information of the planetary gear 3's rotation process. The accelerometer can accurately sense the gear's operating state, and when the gear is about to enter the meshing state, the main control chip captures these characteristic changes to quickly and accurately determine the rotation angle of planetary gear 3 and the precise meshing moment.

[0049] The data processing module receives stress data acquired by the data acquisition module, performs wavelet threshold denoising, and then locates the noise source based on a noise localization neural network model. Specifically, in this embodiment, the data processing module includes a rod array antenna and a host computer. The rod array antenna receives data, and the host computer performs wavelet threshold denoising and then uses a noise localization neural network model to locate the noise source based on the received data.

[0050] Specifically, such as Figure 1 As shown in this embodiment, the signal processing module includes: multiple Wheatstone bridges, a signal amplification and filtering circuit, and an AD conversion circuit. The output of each stress sensor group is connected to a Wheatstone bridge. The output of each Wheatstone bridge is connected to the main control chip through the signal amplification and filtering circuit and the AD conversion circuit in sequence, and then the signal is transmitted to the host computer.

[0051] Specifically, in this embodiment, the stress sensor 4 in the stress sensor group is a strain gauge. The strain gauge outputs a weak unbalanced voltage signal through the Wheatstone full-bridge circuit. Since the output signal amplitude of the Wheatstone full-bridge circuit is low and easily affected by environmental noise, a differential amplifier circuit is used as the signal amplification circuit to perform high-precision gain amplification of the signal in order to improve the signal amplitude and anti-interference capability. The processing principle is as follows.

[0052] The strain gauge outputs a weak unbalanced voltage signal through a Wheatstone bridge. The output voltage of the Wheatstone bridge is... :

[0053] ; (1)

[0054] in, This indicates the change in resistance of the strain gauge caused by the deformation of the gear under stress. The resistance value of the bridge arm is... The bridge excitation voltage is used to amplify the signal through a high-precision differential amplifier circuit. The output voltage of the differential amplifier circuit is... :

[0055] ; (2)

[0056] in: This is the gain coefficient. and These are the input voltages of the differential circuit, and their difference is the output voltage of the front-end bridge. .

[0057] The amplified stress signal is then amplified through the gain effect of the differential amplifier circuit. :

[0058] ; (3)

[0059] After the amplified stress signal is converted into a digital signal, it is stored in the local FLASH memory through the main control chip (which can be an FPGA).

[0060] Specifically, in this embodiment, each stress sensor group includes four strain gauges, and the four strain gauges and the corresponding Wheatstone bridge form a temperature-compensated full-bridge circuit; two strain gauges are arranged along the principal stress direction to measure the actual strain, and the other two strain gauges are arranged perpendicular to the principal stress direction to achieve temperature compensation.

[0061] Furthermore, such as Figure 2 As shown, this embodiment also includes a signal generator, a power driver, and an impedance matching network disposed outside the gearbox. The signal generator drives the ultrasonic transmitting unit to emit ultrasonic waves via the power driver and the impedance matching network. Furthermore, a secondary coil can be fixedly mounted on one of the planetary gears 3. The signal generator can then determine the position of the secondary coil based on data collected by the accelerometer, thereby controlling the power supply amplitude of the ultrasonic transmitting unit to stabilize the energy received by the secondary coil, thus achieving stable power supply within the enclosed metal gearbox.

[0062] Furthermore, in a noise source localization system for a planetary gear set in an underwater gearbox of a ship according to this embodiment, the data acquisition module further includes a data transmission module, which is used to send the acquired data to the data processing module; the main control chip is also used to control the supercapacitor group to enter a high-power output mode to supply power to the data transmission module during data transmission.

[0063] Furthermore, the main control chip transmits the pre-processed acceleration data in real time to the signal generator and power driver module located outside the gearbox via the data transmission module. Based on the received acceleration data and the real-time positional changes of the internal ultrasonic receiving unit, the signal generator module dynamically adjusts the transmission power amplitude of the ultrasonic transmitting unit to ensure that the ultrasonic receiving unit stably and efficiently receives ultrasonic energy during the rotation of the gear set.

[0064] Specifically, in this embodiment, the ultrasonic receiving unit is mounted on the planetary carrier 5. Since the relative position of the planetary carrier 5 and each planetary gear 3 is constant, the received energy can be stably transmitted to the data acquisition module mounted on the side wall of each planetary gear 3.

[0065] like Figure 1 As shown, in this embodiment, the host computer is equipped with a wavelet threshold denoising module and a noise source localization module. The wavelet threshold denoising module includes a wavelet transform module and an adaptive filtering module. The wavelet transform module is used to reconstruct the received stress signal using wavelet denoising. Specifically, wavelet denoising reconstructs the stress signal through wavelet transform and inverse transform. The adaptive filtering module is used to filter the reconstructed stress signal using an adaptive filtering algorithm. The noise localization module is used to perform comprehensive analysis on the filtered stress data of each planetary gear 3 according to the noise localization neural network model to obtain the planetary gear 3 corresponding to the noise source.

[0066] Specifically, multiple sets of stress signals for each planetary gear 3 transmitted to the host computer enter the wavelet transform module. The wavelet transform module uses the db4 wavelet basis to perform five-level wavelet decomposition on each stress signal of the planetary gear 3 and performs noise reduction processing. The wavelet decomposition process can be represented as:

[0067] ; (4)

[0068] in: These are the low-frequency approximation coefficients for the fifth layer, reflecting the overall trend of stress signal variation. This represents the position index of the wavelet coefficients in the discrete domain, used to locate the corresponding time sampling point after signal decomposition; This represents the high-frequency detail coefficient of the j-th layer, which includes stress signal details and high-frequency noise.

[0069] The wavelet transform module uses a fixed threshold. And soft thresholding function for high-frequency detail coefficients of each layer Thresholding is performed, and the high-frequency detail coefficients after processing are obtained. Represented as:

[0070] ; (5)

[0071] in: This represents the high-frequency detail coefficients of the j-th layer after thresholding. Here are the high-frequency detail coefficients for the j-th layer; For a symbolic function, it is defined as: when When it is 1, A value of -1 indicates whether the detail change is in the same or opposite direction as the shape of the mother wavelet; The threshold is a fixed value, set based on the noise level, and is independent of the number of layers.

[0072] High-frequency detail coefficients after thresholding The original low-frequency approximation coefficient Perform inverse wavelet transform to reconstruct the wavelet-denoised stress signal. :

[0073] ; (6)

[0074] in: Indicates the first Wavelet basis functions of the layer (high-frequency detail basis) are used for detail signal reconstruction. The scaling function (low-frequency approximation basis) of the 5th layer is used for trend signal reconstruction. Indicates the number of decomposition layers. k Indicates the position index.

[0075] Stress signal after wavelet denoising The signal is transmitted to the adaptive filtering module, which uses the LMS adaptive filtering algorithm to dynamically adjust the filter parameters according to the frequency characteristics of the signal, and filters the residual low-frequency noise and system interference in the signal to optimize the signal quality.

[0076] Specifically, the adaptive filtering module adjusts the filter weights in real time based on the input signal to minimize the mean square error between the output signal and the desired signal. The weight update formula is as follows:

[0077] ; (7)

[0078] in, Step size factor For error signals, Let n be the input signal, and n represent the number of iterations. and These represent the weights obtained in the nth and (n+1)th iterations, respectively.

[0079] The stress signal output after filtering is denoted as It retains the stress signal characteristics relatively smoothly and completely. The stress signal processed by the wavelet threshold denoising module is sent to the noise localization module for analysis, thereby characterizing the gearbox noise features.

[0080] Specifically, in this embodiment of the invention, the wavelet threshold denoising module combines an adaptive filtering algorithm and a dynamic threshold adjustment strategy to optimize the special noise characteristics of the dynamic stress signal of a ship gearbox, thereby significantly improving signal quality and feature extraction accuracy.

[0081] In this embodiment, the noise localization module uses a noise localization neural network model that employs a fusion architecture of a multi-layer convolutional neural network and a gated recurrent unit; such as Figure 4 As shown, it includes multiple convolutional layers, pooling layers, GRU layers, and fully connected layers connected in sequence; the noise localization neural network model uses the cross-entropy loss function to optimize the network parameters and adjusts the network weights through the backpropagation algorithm and gradient descent optimizer.

[0082] Specifically, the host computer also includes a feature extraction module. The stress signal obtained after filtering by the wavelet threshold denoising module is extracted by the feature extraction module to obtain the stress features of each planetary gear 3, which are then input into the noise localization neural network model. Based on the stress features of the planetary gears 3, the noise localization neural network model can identify abnormal stress patterns, including peak surges, rate fluctuations, or frequency domain energy concentrations, and output the probability distribution of the noise source location. ,in (i=1,2……m) represents the i-th The probability that planetary gear 3 is a noise source ( ). m represents the number of planetary gears 3. The probability value directly reflects the relative intensity of the noise source. Through this probability distribution, the system can provide a quantitative basis for targeted optimization of the gearbox, such as tooth profile correction and lubrication improvement.

[0083] Specifically, in this embodiment, the dataset used for training the noise localization neural network model consists of multiple sets of synchronously acquired multi-channel stress signals and corresponding noise source location labels. Specifically, the planetary gear system is run under conditions where the noise source locations are known, and data is recorded simultaneously. Figure 3Multiple sets of dynamic stress data for the five planetary gears 3 are shown. These stress signals are paired with actual noise source locations to form calibration samples. Furthermore, different noise source conditions are simulated using numerical simulation methods to generate corresponding multi-channel stress signals and noise source labels, thereby expanding the training set. These labeled stress data are then used to train a noise localization neural network model, enabling it to predict noise source locations from stress characteristics.

[0084] Specifically, after receiving multiple synchronous high-precision stress signals, the host computer first performs preprocessing such as denoising and filtering through a wavelet threshold denoising module. Then, it extracts features from each preprocessed stress signal through a feature extraction module. The extracted features include the stress peak value corresponding to each planetary gear 3. ), rate of change ( ) and waveform characteristics ( ),in Here, i represents the number of planetary gear 3, and m represents the quantity of planetary gear 3. Next, the Dynamic Time Warping (DTW) algorithm is used to map multiple signals to the same time domain, ensuring data comparability and consistency. The feature vector set after feature extraction and time alignment is as follows:

[0085] ; (8)

[0086] This is a set of feature vectors. After feature extraction and time alignment, these multiple sets of feature data are used to construct an input matrix. X :

[0087] ; (9)

[0088] For the input matrix X Standardization process:

[0089] ; (10)

[0090] in: For each feature, the mean vector is used. For each feature standard deviation vector, This is the standardized feature matrix.

[0091] The standardized data is input into a pre-trained noise localization neural network model. This model employs a fusion architecture of multi-layer convolutional neural networks and gated recurrent units to fully capture spatiotemporal features and sequence dependencies. The output of the convolutional layers is represented as follows:

[0092] ; (11)

[0093] in: For activation function, For convolution kernel weights, For bias terms, This is the feature output after passing through the nth convolutional layer.

[0094] After several convolutional and pooling layers, the feature map The vector is flattened into a one-dimensional vector and passed to the GRU layer to process long-term dependencies in the time series:

[0095] ;(12)

[0096] in, Indicates the current time t The hidden state vector, This represents the output features of the convolutional layer at time t. The hidden state is the state at the previous time step. GRU is a gated loop unit function that contains a reset gate, an update gate, and a nonlinear mapping of candidate hidden states.

[0097] The GRU layer uses the following update mechanism to pass information:

[0098] 1. Update Gate: Determines how much stress feature history information is retained in the current hidden state. Its output is:

[0099] ; (13)

[0100] in: To update the gate weight matrix, This indicates the hidden state at the previous moment. The output features of the convolutional layer at time t. It uses the Sigmoid activation function, and its output value is between [0,1]. This indicates an update to the gate offset. This indicates that the output of the gate is being updated.

[0101] 2. Reset Gate: Controls the influence of historical stress characteristics on candidate states; its output is:

[0102] ;(14)

[0103] in: To reset the gate weight matrix, It resets the door offset. This indicates that the output of the reset door is being reset. Each component falls within the interval [0, 1]. The output of the reset gate is used for the calculation of subsequent candidate states, when... When an element is close to 0, the historical state information of the corresponding dimension is reset; when it is close to 1, more historical state information is retained.

[0104] 3. Candidate state: Generates the temporary state at the current moment, represented as:

[0105] ; (15)

[0106] in: This represents the candidate state at the current moment. Let be the candidate state weight matrix, and tanh be the hyperbolic tangent activation function with output values ​​between [-1, 1]. This is used to bias the candidate state.

[0107] 4. The formula for calculating the final state is:

[0108] ; (16)

[0109] This represents the current hidden state and is the final output of the GRU layer. The noise source is passed to the fully connected layer, and the specific location and intensity distribution of the noise source are output through the Softmax activation function. The calculation formula for the fully connected layer can be expressed as:

[0110] ; (17)

[0111] Where O represents the output vector of the fully connected layer. This indicates a hidden state at the last moment. It is the weight matrix of the fully connected layer. It's a bias, the output vector. Each dimension corresponds to a measurement point on planetary gear 3, and the value represents the probability that the location is a noise source. The system ultimately determines the location of the noise source based on the index of the highest probability value.

[0112] During model training, the cross-entropy loss function is used to optimize the network parameters:

[0113] ; (18)

[0114] in, Represents the loss function; Encoding of the true label: If the k-th measurement point is a true noise source, then ,otherwise K represents the number of measurement points for suspected noise sources in the planetary gear set; This indicates the probability that the model predicts that the point is a noise source.

[0115] The noisy localization neural network model continuously adjusts the network weights through backpropagation and the gradient descent optimizer (Adam).

[0116] ; (19)

[0117] in: For learning rate, For the loss gradient, and These represent the network weights at times t+1 and t, respectively.

[0118] After training, the noise localization neural network model can effectively identify the stress state differences of different planetary gears 3 at the same time through feature correlation analysis and pattern recognition, thereby achieving accurate localization of specific noise sources in the gear set inside the underwater gearbox.

[0119] Example 2

[0120] Embodiment 2 of the present invention provides a noise source localization method for planetary gear sets in underwater gearboxes of ships, which is based on the noise source localization system for planetary gear sets in underwater gearboxes of ships described in Embodiment 1, and includes the following steps:

[0121] Step 1: Power is supplied to the inside of the gearbox through the ultrasonic transmitting unit and the ultrasonic radiation power supply module.

[0122] Specifically, the position of the planetary gear 3 can be determined based on the data collected by the accelerometer, and then the position of the secondary coil can be determined. The power supply amplitude of the ultrasonic transmitting unit can be dynamically adjusted to stabilize the energy received by the secondary coil.

[0123] Step 2: Obtain the output signal of the accelerometer through the main control chip, determine the meshing time of planetary gear 3, and send a synchronous acquisition command at the meshing time to control the stress sensor group to collect and store stress data at the meshing time, while activating the supercapacitor group to enter the high power supply mode.

[0124] Step 3: After the engine stops, prepare to enter the data receiving mode, control the supercapacitor group to enter the high-power energy supply mode, and then start receiving and collecting data.

[0125] Step 4: After denoising the collected multi-channel data, perform temporal alignment to map them to the same time domain, extract feature parameters, and input them into the noise localization neural network model to locate the noise source.

[0126] In step four, the feature parameters input to the noise localization neural network model include the stress peak value corresponding to each planetary gear 3. ), rate of change ( ) and waveform characteristics ( The output is the probability distribution of the noise source location corresponding to each planetary gear 3.

[0127] The method for locating noise sources in planetary gear sets of underwater gearboxes for ships further includes the following steps:

[0128] Step 5: Optimize gear materials, structural parameters, and lubrication conditions based on the probability distribution of noise source locations.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A noise source localization system for planetary gear sets in underwater gearboxes of ships, characterized in that, include: Ultrasonic transmitting unit, ultrasonic radiation power supply module, accelerometer, data acquisition module and data processing module; The ultrasonic radiation power supply module includes: an ultrasonic receiving unit, a supercapacitor bank, a primary coil, and a secondary coil; the ultrasonic transmitting unit and the ultrasonic receiving unit are respectively disposed on the outer wall of the gearbox and inside the gearbox; the primary coil is connected to the ultrasonic receiving unit, and the secondary coil is connected to the supercapacitor bank; the ultrasonic transmitting unit is used to transmit ultrasonic waves to the ultrasonic receiving unit; the ultrasonic receiving unit is used to generate an alternating current in the primary coil through the received ultrasonic waves; the secondary coil is used to charge the supercapacitor bank according to the magnetic field generated by the primary coil; and the supercapacitor bank is used to power the data acquisition module. The data acquisition module includes multiple stress sensor groups, signal processing circuits, and a main control chip; each planetary gear of the planetary gear set is provided with multiple measuring points, which are distributed at multiple adjacent tooth root critical sections of the planetary gear. Each measuring point is equipped with a stress sensor group, and each stress sensor group is connected to the main control chip through the signal processing circuit. The accelerometer is used to collect the running status data of the planetary gear and send it to the main control chip. The main control chip is used to determine the meshing time of the first measuring point of the planetary gear based on the output signal of the accelerometer, and send a synchronous acquisition command to control the stress sensor group to collect and store stress data at the meshing time. It is also used to synchronously control the supercapacitor group to enter the high power output mode. The data processing module is used to receive stress data collected by the data acquisition module, and to perform comprehensive analysis on the stress data of each planetary gear based on the noise localization neural network model to obtain the planetary gear corresponding to the noise source; the data processing module includes a rod array antenna and a host computer, the rod array antenna is used for data reception, and the host computer is equipped with a wavelet threshold denoising module and a noise localization module. The wavelet threshold denoising module includes a wavelet transform module and an adaptive filtering module. The wavelet transform module is used to reconstruct the received stress signal using a wavelet denoising method, and the adaptive filtering module is used to filter the reconstructed stress signal using an adaptive filtering algorithm. The weight update formula of the adaptive filtering module is: ； in, Step size factor For error signals, For input signals; Indicates the number of iterations; and These represent the weights obtained in the nth and (n+1)th iterations, respectively. The noise localization module is used to perform comprehensive analysis on the filtered stress data of each planetary gear based on the noise localization neural network model to obtain the planetary gear corresponding to the noise source.

2. The noise source localization system for planetary gear sets in underwater gearboxes of ships according to claim 1, characterized in that, The signal processing module includes: multiple Wheatstone bridges, signal amplification and filtering circuits, and AD conversion circuits. The output of each stress sensor group is connected to a Wheatstone bridge, and the outputs of each Wheatstone bridge are connected to the main control chip in sequence through the signal amplification and filtering circuit and the AD conversion circuit.

3. A noise source localization system for planetary gear sets in underwater gearboxes of ships according to claim 1, characterized in that, The stress sensor group includes four strain gauges, which, together with the corresponding Wheatstone bridge, form a temperature-compensated full-bridge circuit. Two strain gauges are arranged along the principal stress direction to measure the actual strain, while the other two strain gauges are arranged perpendicular to the principal stress direction to achieve temperature compensation.

4. A noise source localization system for planetary gear sets in underwater gearboxes of ships according to claim 1, characterized in that, It also includes a signal generator, a power driver, and an impedance matching network. The signal generator drives the ultrasonic transmitting unit to emit ultrasonic waves via the power driver and the impedance matching network. The signal generator is also used to control the power supply amplitude of the ultrasonic transmitting unit based on the data collected by the accelerometer, so that the secondary coil receives energy stably.

5. A noise source localization system for planetary gear sets in underwater gearboxes of ships according to claim 1, characterized in that, The data acquisition module also includes a data transmission module, which is used to send the data acquired by the data acquisition module to the data processing module; the main control chip is also used to control the supercapacitor group to enter a high-power output mode to supply power to the data transmission module during data transmission.

6. A noise source localization system for planetary gear sets in underwater gearboxes of ships according to claim 1, characterized in that, The noise localization neural network model used in the data processing module adopts a fusion architecture of multi-layer convolutional neural network and gated recurrent unit; it includes multiple convolutional layers, pooling layers, GRU layers, and fully connected layers connected in sequence; the noise localization neural network model uses the cross-entropy loss function to optimize network parameters and adjusts network weights through backpropagation algorithm and gradient descent optimizer.

7. A method for locating noise sources in planetary gear sets of underwater gearboxes for ships, implemented based on the noise source location system for planetary gear sets of underwater gearboxes for ships as described in any one of claims 1 to 6, characterized in that, Includes the following steps: Step 1: Power is supplied to the inside of the gearbox through the ultrasonic transmitting unit and the ultrasonic radiation power supply module. Step 2: Obtain the output signal of the accelerometer through the main control chip, determine the meshing time of the first measuring point of the planetary gear, and send a synchronous acquisition command to control the stress sensor group to collect and store stress data at the meshing time. At the same time, activate the supercapacitor group to enter the high power supply mode. Step 3: After the engine stops, prepare to enter the data receiving mode, control the supercapacitor group to enter the high-power energy supply mode, and then start receiving and collecting data; Step 4: Perform time-domain alignment on the collected multiple data streams, map them to the same time domain, extract feature parameters, and input them into the noise localization neural network model to locate the noise source.

8. A method for locating noise sources in planetary gear sets of underwater gearboxes for ships according to claim 7, characterized in that, In step four, the input to the noise neural network model is the stress peak value, rate of change, and waveform characteristics corresponding to each planetary gear, and the output is the probability distribution of the noise source location corresponding to each planetary gear.

9. A method for locating noise sources in a planetary gear set of an underwater gearbox for ships according to claim 7, characterized in that, It also includes the following steps: Step 5: Optimize gear materials, structural parameters, and lubrication conditions based on the probability distribution of noise source locations.

Images