A ship acoustic state intelligent prediction method and system based on a lightweight digital model

Abstract

The application discloses a ship acoustic state intelligent prediction method and system based on a light-weight digital model, and belongs to the field of ship acoustic monitoring. The method is based on light-weight digital model technology, real-time measurement of vibration acceleration of key parts of the ship, complex ship structure modeling and simplification, construction of digital mapping relationship between ship structure and acoustic characteristics, and realization of efficient and rapid prediction and evaluation of the ship acoustic state. The three-dimensional graphics rendering technology is used to dynamically display the vibration monitoring results, main noise sources and equipment operation states and other key information of the ship, realize comprehensive reproduction of the overall vibration state of the ship, and realize intelligent prediction of the acoustic state of the whole ship. The patent provides a highly integrated, intelligent and portable comprehensive solution for ship acoustic state perception, and has significant engineering application value and scientific foresight.

Description

Technical Field

[0001] This invention belongs to the field of ship acoustic monitoring, specifically relating to a method and system for intelligent prediction of ship acoustic status based on a lightweight digital model. Background Technology

[0002] With the development of the shipbuilding industry, the monitoring and control of acoustic conditions during ship operation has become increasingly important. When a ship is sailing, the operation of internal mechanical equipment and the effects of the external environment will generate vibration and noise, which not only affects the comfort of the crew, but also threatens the structural safety of the ship and the normal operation of the equipment.

[0003] Traditional methods for monitoring the acoustic condition of ships suffer from numerous problems. In terms of modeling, there is a lack of effective ways to analyze the acoustic characteristics of complex structures, making it difficult to grasp the relationship between structure and acoustic features, resulting in insufficient forecast accuracy and timeliness. In terms of hardware, the equipment suffers from low integration, limited functionality, and poor collaboration, hindering efficient data collection and real-time status feedback. In field testing, most equipment is limited by size and portability, making flexible deployment difficult, and operation cumbersome and inefficient.

[0004] Foreign countries started earlier in the field of ship acoustic monitoring, with some research institutions and companies adopting advanced sensor technologies and data analysis methods. However, these efforts often focus on monitoring single indicators, lacking comprehensive solutions, and some technologies have high hardware requirements and are expensive. While domestic research has made some progress, it still falls short in terms of lightweight models and the integration of intelligent forecasting algorithms. Most systems have low visualization levels and limited interactive functions. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent prediction of ship acoustic status based on a lightweight digital model, which solves the problems of insufficient accuracy and timeliness of traditional ship radiated noise monitoring methods.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] A method for intelligent prediction of ship acoustic state based on a lightweight digital model includes the following steps:

[0008] Step 1: Construct a 3D model of the ship and perform lightweight processing, then assign material acoustic and physical properties to the structure;

[0009] Step 2: Select typical mechanical equipment and arrange vibration monitoring points in key parts and sensitive locations of the ship structure to ensure that the monitoring points are consistent with the coordinate system of the three-dimensional model and have accurate coordinate values;

[0010] Step 3: Construct the mapping relationship between ship structural vibration and acoustics through numerical simulation or experimental measurement, and generate a standardized dataset of ship structural vibration response and underwater radiated noise;

[0011] Step 4: By deploying vibration acceleration sensors at vibration monitoring locations to measure vibration acceleration in real time, the data is transmitted to the monitoring system. A visualization interface for ship vibration acoustic monitoring is designed to associate and map the monitoring data with the three-dimensional model.

[0012] Step 5: Preprocess the measured vibration acceleration and extract the vibration spectrum data using Fast Fourier Transform; construct a deep learning model, input the preprocessed vibration spectrum data, and output the underwater radiated noise spectrum prediction result; input the standardized dataset obtained in Step 3 as the training set into the deep learning model, and optimize the model's prediction performance by adjusting the network parameters; integrate the lightweight three-dimensional model from Step 1 with the deep learning model to construct an intelligent prediction system, realizing real-time mapping and prediction of ship structure and vibration acoustics;

[0013] Step 6: Intelligent prediction and display of the acoustic status of the entire ship: Based on real-time data and prediction results, abnormal areas are marked with dynamic arrows and flashing icons in a 3D visualization interface. Historical data is traced back through the timeline and future noise distribution trends are predicted. Finally, the prediction results are displayed through the visualization interface.

[0014] Furthermore, the three-dimensional model of the ship in step 1 includes the ship's hull outline, internal cabin layout, structural components, and typical mechanical equipment; the lightweighting process reduces model complexity by removing redundant geometric details, merging adjacent surfaces, and optimizing the model's topology, while retaining key acoustic features; the assignment of material acoustic physical properties to the structure includes classifying and assigning physical properties to beams, plates, and columns in the model, resulting in a 30%-80% reduction in the data volume of the lightweight three-dimensional model;

[0015] The reduction in computational complexity after model lightweighting:

[0016] ΔC=C 原 -C 优化ˉ =k·N 冗余

[0017] Where C is the computational complexity, k is the computational cost of a single geometric detail, and N is the computational complexity. 冗余 This represents the number of redundant geometric quantities.

[0018] Furthermore, the typical equipment in step 2 includes the main engine, the propeller in the propulsion system, auxiliary mechanical equipment, and other equipment that generates significant vibration and noise during ship navigation, and vibration monitoring points are arranged at their bases, bulkheads, or structural connection nodes.

[0019] Furthermore, in step 3, the vibration-acoustic mapping relationship of the ship structure is constructed:

[0020] The frequency domain transfer function of the vibration response is: Where U(ω) is the displacement spectrum and F(ω) is the excitation force spectrum;

[0021] The mapping relationship between structural vibration and acoustics is as follows: Where p is the sound pressure level. It represents displacement velocity;

[0022] The numerical simulation is based on ship structural mechanics, acoustic principles, and vibration theory. A finite element / boundary element model of the ship is established, considering the physical parameters of the ship structure's mass, stiffness, and damping, as well as the excitation effect of typical mechanical equipment. The vibration response of the ship structure at the vibration monitoring location under different working conditions is calculated through simulation. Combining the acoustic characteristics of the ship structure material and the acoustic parameters of the surrounding fluid medium, the propagation process of sound waves inside the ship structure and in the underwater acoustic field is simulated, and the underwater radiated noise of the ship is calculated through simulation.

[0023] The experimental measurements involved conducting a real-ship vibration acoustic test. High-precision vibration sensors and acoustic sensors were installed at the vibration monitoring locations to collect data on the ship's speed, load, and sailing attitude under different operating conditions and times, thereby obtaining data on the structure's vibration response and radiated noise.

[0024] Furthermore, in step 4, the analog signals collected by the sensors are converted into digital signals by a data acquisition card and transmitted to the ship's central monitoring system; a visual interface for ship vibration acoustic monitoring is designed based on front-end application development technology, and the monitoring data is associated and mapped with the three-dimensional model; the vibration amplitude, frequency and acoustic intensity information of typical equipment and structures are intuitively displayed through changes in color, shape and size visualization elements, and alarm signals are triggered based on preset thresholds;

[0025] Vibrational acceleration a(t) is mapped to color, shape, or size; where color is based on the acceleration amplitude:

[0026] C = map(a(t), a min ,a max C green C red )

[0027] Where map: a linear mapping function; a min a max : Acceleration range; C green C red : Color values ​​from green to red;

[0028] The same principle applies to mapping acoustic intensity to a 3D model;

[0029] The ship's condition is determined by the following formula:

[0030]

[0031] Among them, a safe a warning a critical These represent the safe, critical, and excessive threshold values ​​for vibration acceleration monitoring, respectively. safe <a warning <a ial The same applies to sound pressure level.

[0032] Based on the rendering loop mechanism of the selected software library, the scene, which includes the updated ship model, equipment model, and measurement point model, as well as the camera view, will be rendered with high quality.

[0033] The scene update formula is:

[0034] Scene t =update(Model,Data) t (Camera)

[0035] Where: Model represents the ship, equipment, and measurement point model; Data t For monitoring data at time t; Camera represents the camera position and viewing angle;

[0036] Add interactive functionality to the model so that clicking on vibration monitoring point i will bring up a pop-up window displaying detailed information:

[0037] Info i ={ID i ,a i (t),SPL i (t),History i Maintenance i}

[0038] Among them, ID i For the measurement point identification, a i (t) represents the vibration acceleration, SPL i (t) represents the sound pressure level, History represents the historical data curve, and Maintenance i To maintain records, the window displays the equipment's real-time operating parameters, historical monitoring data curves, equipment maintenance records, and other information, providing comprehensive decision support for ship maintenance personnel.

[0039] Furthermore, step 5 extracts vibration spectrum data through Fast Fourier Transform.

[0040]

[0041] Where N is the number of sampling points, t n Here, f represents discrete time points, f is the frequency, and j is the imaginary unit. This serves as the input data for a lightweight digital model-based intelligent forecasting algorithm.

[0042] The deep learning model is designed with an input layer, hidden layers, and an output layer. The input layer receives preprocessed vibration spectrum data, and the output layer outputs the underwater radiated noise spectrum and the predicted vibration results.

[0043] y = W (out) h (L) +b (out)

[0044] Where y is the prediction result, W (out) h is the weight matrix of the output layer. (L) For the output of the last hidden layer, b (out) This is the bias term for the output;

[0045] Using vibration acceleration spectrum and radiated noise spectrum datasets as training data, a deep learning model is trained. The training data is then input into the deep learning model, and the predictive performance of the model is optimized by adjusting the network parameters, including forward propagation, loss calculation, backpropagation, and parameter update.

[0046] Finally, the lightweight 3D model and the trained deep learning model are integrated:

[0047] y final =w d ·f d (X)+w l ·f l (X)

[0048] Among them, f d (X) and f l (X) represents the outputs of the deep learning model and the lightweight model, respectively, with weights w. d and w l Based on the performance settings of the validation set;

[0049] The ship's position coordinates c i Vibration data from monitoring points a i and working conditions w i Associated, forming the input tuple (a i ,c i ,w i ), forming a joint feature vector x i =[a i ,c i ,w i ]∈R m+3+d Design a data interface to transmit the real-time acquired vibration spectrum data a i The coordinates c of the monitoring point in the lightweight model i Exact match: Represented as a paired dataset Where a iIt was collected from location c i The vibration spectrum;

[0050] The vibration spectrum ai is input into the deep learning model to obtain the radiated noise spectrum y. i =f θ (a i Then, based on the one-to-one correspondence of the positions in the three-dimensional coordinate system, the predicted result y is... i Associating with coordinate c in the 3D model i Generate the predicted value of radiated noise at the noise monitoring point.

[0051] Furthermore, step 6 uses a lightweight digital model-based intelligent forecasting algorithm, taking real-time collected ship vibration acoustic monitoring data as input, to intelligently forecast the acoustic state of the entire ship and display the forecast results through a designed visualization interface. The predicted distribution trend of radiated noise across the entire ship is displayed using color gradients and dynamic arrow indicators. Areas or equipment that may experience acoustic anomalies are identified using flashing icons or special markers. The historical changes and future trends of the predicted locations of major noise sources are displayed in the form of charts, curves, etc. Ship maintenance personnel can judge the changing trend of the ship's acoustic state based on the information displayed on the visualization interface and take corresponding control measures in a timely manner.

[0052] A computer device / equipment / system includes a memory, a processor, and a computer program stored in the memory, characterized in that: the processor executes the computer program to implement the steps of a method for intelligent prediction of ship acoustic conditions based on a lightweight digital model.

[0053] The beneficial effects of this invention are as follows:

[0054] (1) By constructing a lightweight digital model of ships, combining the accurate definition of typical mechanical equipment and vibration monitoring locations, and constructing the relationship between ship structural vibration and acoustic mapping, high-precision vibration sensors are used to collect ship vibration data in real time, which realizes accurate prediction of the acoustic state of ships. The monitoring results are dynamically displayed through a visual interface, allowing users to intuitively understand the changing trend of the acoustic state of ships. At the same time, rich interactive functions are provided to facilitate users to conduct data analysis and decision-making.

[0055] (2) The intelligent prediction method, system and device for ship acoustic status based on lightweight digital model is also equipped with complete ship vibration monitoring hardware and ship acoustic status intelligent prediction module, which can monitor the vibration of the ship in real time, discover potential problems in the operation of the ship in a timely manner, and optimize the control strategy of the ship based on the intelligent prediction results, thereby improving the acoustic performance and vibration performance of the ship.

[0056] The present invention relates to a method and system for intelligent prediction of ship acoustic status based on a lightweight digital model. By measuring the vibration acceleration of key parts through high-precision vibration monitoring hardware, and combining the digital mapping relationship between ship structure and acoustic characteristics constructed by the lightweight digital model, the method can efficiently, quickly and accurately predict and assess the acoustic status of the ship, providing strong support for the safe operation, performance optimization and intelligent management of the ship. Attached Figure Description

[0057] Figure 1 Flowchart of a method for intelligent prediction of ship acoustic state based on lightweight digital models. Detailed Implementation

[0058] The present invention will now be further described with reference to the accompanying drawings.

[0059] This invention provides a method and system for intelligent prediction of ship acoustic status based on a lightweight digital model, deeply integrating front-end application development, 3D graphics rendering technology, high-precision vibration monitoring hardware, and advanced intelligent prediction algorithms for ship radiated noise. The system can utilize mature software tools to construct a 3D digital model of the ship (such as ANSYS, Abaqus, etc.) and display ship acoustic information through a visualization platform. Simultaneously, relying on lightweight digital model technology, it accurately models and efficiently simplifies the complex structure and acoustic characteristics of the ship, and combines this with intelligent prediction algorithms for ship radiated noise to achieve intelligent prediction, comprehensive reproduction, and interactive visualization of the ship's acoustic status. Figure 1 The specific steps are as follows:

[0060] Step 1: Construct a 3D model of the ship and perform lightweight processing:

[0061] We comprehensively collect ship design drawings and data, such as lines drawings, general arrangement drawings, detailed structural drawings, and material property information. We then use professional 3D modeling software to construct a high-precision 3D model of the ship, covering the ship's hull outline, internal compartment layout, structural components (such as beams, plates, and columns), and typical mechanical equipment (such as main engines and propellers).

[0062] While ensuring the key acoustic characteristics and mechanical performance of the ship structure, the model's data volume and complexity are significantly reduced by removing redundant geometric details, merging adjacent surfaces and volumes, and optimizing the model's topology, thereby improving the efficiency of model storage, transmission, processing, and 3D display. The influence of different materials on sound wave propagation characteristics is fully considered when assigning physical property values ​​to the ship structure.

[0063] The reduction in computational complexity after model lightweighting:

[0064] ΔC=C 原 -C 优化ˉ =k·N 冗余

[0065] Where C is the computational complexity, k is the computational cost of a single geometric detail, and N is the computational complexity. 冗余 This represents the number of redundant geometric quantities.

[0066] Step 2: Define typical mechanical equipment and vibration monitoring locations:

[0067] Based on ship acoustics theory and statistical analysis of extensive real-ship operational data, typical mechanical equipment that significantly impacts the overall acoustic state of the ship was identified. This includes the main engine in the ship's power system, the propeller in the propulsion system, various auxiliary mechanical equipment (such as generators, air conditioning systems, hydraulic pumps, etc.), and other equipment that generates significant vibration and noise during ship navigation. Vibration monitoring points were placed at key locations of these typical mechanical equipment and sensitive locations in the ship's structure (such as bulkheads and decks). This ensured that the locations of the typical mechanical equipment and monitoring points were in the same coordinate system as the ship's three-dimensional model and had accurate coordinate values. The layout of the monitoring points should adhere to the principle of comprehensively and accurately reflecting the overall vibration and acoustic characteristics of the ship, while also considering operability and ease of installation in the actual ship environment to ensure the validity and reliability of the monitoring data.

[0068] Step 3: Construct the mapping relationship between ship structural vibration and acoustics:

[0069] The frequency domain transfer function of the vibration response is: Where U(ω) is the displacement spectrum and F(ω) is the excitation force spectrum.

[0070] The mapping relationship between structural vibration and acoustics is as follows: Where p is the sound pressure level. This represents displacement velocity.

[0071] There are two ways to obtain the relationship between ship structural vibration and acoustic mapping: numerical simulation and experimental measurement.

[0072] The numerical simulation method is based on ship structural mechanics, acoustic principles, and vibration theory to establish a finite element / boundary element model of the ship.

[0073] Finite Element Model (FEM): The ship structure is discretized into a finite number of elements, each with mass, stiffness, and damping characteristics. The equations of motion for the vibrating system are:

[0074] Mu¨+Cu˙+Ku=F(t)

[0075] Where M is the mass matrix, C is the damping matrix, K is the stiffness matrix, and u, These are displacement, velocity, and acceleration vectors, respectively, and F(t) is the external excitation force.

[0076] The excitation effect of typical mechanical equipment (such as main engine, auxiliary engine, propeller, etc.) can be modeled as a simple harmonic force: F(t)=F0e jwt Where F0 is the excitation force amplitude and w is the excitation frequency.

[0077] The vibration response of the ship structure at the vibration monitoring location determined in step S2 under different working conditions was obtained through simulation calculation.

[0078] u(ω)=(K-ω 2 M+iωC) -1 F(ω)

[0079] By combining the acoustic properties of the ship's structural materials with the acoustic parameters of the surrounding fluid medium (such as seawater and air), the propagation process of sound waves inside the ship's structure and in the underwater acoustic field is simulated, and the underwater radiated noise of the ship is obtained through simulation calculation.

[0080] The experimental measurement method is as follows: Conduct a real ship vibration acoustic test experiment, install high-precision vibration sensors and acoustic sensors at the vibration monitoring positions determined in step S2, and collect vibration response and radiated noise data of the ship structure under different operating conditions, different speeds (such as 5 knots, 10 knots, 15 knots), loads (such as no load, half load, full load), and sailing attitudes (such as straight sailing, turning).

[0081] Based on unified data standards and naming rules, detailed identification information, including name, ID, type, and system affiliation, is set for equipment models and vibration monitoring points to ensure the uniqueness and identifiability of the models in complex data interaction environments. Data obtained from simulation and measurement are tagged, including operating conditions and monitoring point coordinates, to form a complete dataset of ship structural vibration and radiated noise.

[0082] Step 4: Dynamically display the vibration monitoring results of typical ship equipment and structures:

[0083] High-precision vibration acceleration sensors are deployed at the vibration monitoring locations defined in step S2 to measure vibration acceleration a(t) in real time, with units of m / s². 2 The sensor sampling frequency must satisfy the Nyquist sampling theorem, and is typically more than twice the highest frequency of the vibration signal (e.g., 10kHz). The sensor measures triaxial acceleration a. x (t),a y (t),a z The resultant acceleration is: a(t) = a x (t) 2 +a y (t) 2 +a z (t) 2 .

[0084] A data acquisition card is used to convert the analog signals collected by the sensors into digital signals, and data transmission technologies (such as wired networks, wireless networks, etc.) are used to transmit the data to the ship's central monitoring system.

[0085] A visualization interface for ship vibration acoustic monitoring was designed based on front-end application development technology, which associates and maps vibration monitoring data with corresponding locations in a 3D model.

[0086] The changes in visual elements such as color, shape, and size intuitively display information such as vibration amplitude, frequency, and acoustic intensity of typical equipment and structures.

[0087] Vibrational acceleration a(t) is mapped to color, shape, or size. For example, color is based on the acceleration amplitude.

[0088] C = map(a(t), a min ,a max C green C red )

[0089] Where map: a linear mapping function; a min a max : Acceleration range; C green C red : Color values ​​from green to red.

[0090] The same principle applies to mapping acoustic intensity into a 3D model.

[0091] Add rich dynamic interactive functions to the visualization interface to enable users to easily view, analyze and operate the monitoring data.

[0092] Referring to the vibration standards for the entire ship stipulated by the International Maritime Organization (IMO), classification society standards, and national standards regarding vibrations of different ship navigation states, different compartment locations, and different equipment, a data early warning and event triggering mechanism is established. When vibration monitoring data exceeds a preset threshold or an abnormal acoustic event occurs, the system automatically triggers an alarm signal and highlights the equipment area with excessive vibration amplitude in red, while providing detailed alarm information and response suggestions to assist ship management personnel in making timely decisions; equipment areas with critical vibration amplitude are marked with a yellow warning; areas with smaller vibration amplitudes within the safe range are displayed in green, and the specific values ​​of the monitoring data are displayed in real time on the model in the form of dynamic digital labels. The state determination is shown in the following formula.

[0093]

[0094] Among them, a safe a warning a criticalThese represent the safe, critical, and excessive threshold values ​​for vibration acceleration monitoring, respectively. safe <a warning <a ial The same applies to sound pressure level.

[0095] Based on the rendering loop mechanism of the selected software library (such as Three.js), the updated scene, including ship models, equipment models, and measurement point models, as well as the camera view, will be rendered with high quality.

[0096] The scene update formula is:

[0097] Scene t =update(Model,Data) t (Camera)

[0098] Among them: Model: Ship, equipment, and measurement point models; Data t Monitoring data at time t; camera position and angle.

[0099] This allows for a clear and intuitive display of the relationship between equipment and vibration monitoring points in the three-dimensional space of the ship, providing rich and accurate visualization effects.

[0100] Add interactive functionality to the model so that clicking on vibration monitoring point i will bring up a pop-up window displaying detailed information:

[0101] Info i ={ID i ,a i (t),SPL i (t),History i Maintenance i}

[0102] Among them, ID i : Measurement point identifier, a i (t) represents the vibration acceleration, SPL i (t) represents the sound pressure level, History represents the historical data curve, and Maintenance i For record keeping. The window displays the equipment's real-time operating parameters, historical monitoring data curves, equipment maintenance records, and other information, providing comprehensive decision support for ship maintenance personnel.

[0103] The system allows users to set timestamps so they can drag the timeline to view changes in the ship's vibration and acoustic properties at different historical moments. It also provides zoom, pan, and rotate controls for the 3D model, enabling users to freely adjust the viewing angle and examine detailed information about different parts of the ship through mouse dragging and other operations.

[0104] Step 5: Implement intelligent forecasting algorithms based on lightweight digital models:

[0105] First, the ship vibration acceleration a(t) collected by high-precision vibration acceleration sensors deployed at the vibration monitoring locations defined in step 2 is preprocessed. Wavelet denoising or low-pass filtering is applied to filter out high-frequency noise and retain effective vibration information. Then, Fast Fourier Transform (FFT) is used to convert the vibration time-domain signal into a frequency-domain signal. The FFT formula is as follows:

[0106]

[0107] Where N is the number of sampling points, t n Here, f represents discrete time points, f is the frequency, and j is the imaginary unit. This serves as the input data for a lightweight digital model-based intelligent forecasting algorithm.

[0108] The deep learning model's input, hidden, and output layers are redesigned. The input layer receives preprocessed vibration spectrum data, and the output layer outputs the underwater radiated noise spectrum and the predicted vibration results.

[0109] y = W (out) h (L) +b (out)

[0110] Where y is the prediction result, W (out) h is the weight matrix of the output layer. (L) For the output of the last hidden layer, b (out) This is the bias term for the output;

[0111] The training data uses vibration acceleration spectrum and radiated noise spectrum datasets. These datasets contain ship structural vibration data, underwater radiated noise data, and corresponding label information. The dataset includes the following information:

[0112] The input feature is X: preprocessed ship structure vibration spectrum data, with a dimension of n×m, where n is the number of monitoring points and m is the frequency point;

[0113] The target output is Y n The corresponding underwater radiated noise data has a dimension of n×c. n , where c n The characteristic quantity of the radiated noise output;

[0114] Training data is input into a deep learning model, and the model's predictive performance is optimized by adjusting network parameters, including forward propagation, loss calculation, backpropagation, and parameter updates.

[0115] Given input data X, after passing through layer l:

[0116] z (l) =W(l) h (l-1) +b (l)

[0117] h (l) =f(z) (l) )

[0118] Among them, W (l) Let b be the weight matrix of the l-th layer. (l) h is the bias vector of the l-th layer. (l-1) is the activation value of the previous layer (X for the input layer), and f is the activation function.

[0119] Final output y pred It is the activation output of the last layer;

[0120] Based on the model output y pred Calculate the loss using the true label y:

[0121]

[0122] Where N is the number of samples, and θ = {W, b} is the set of all network parameters;

[0123] Obtain the gradient by differentiating the loss function with respect to the parameters:

[0124] and

[0125] This process applies the chain rule, propagating the error back from the output layer to the input layer layer by layer;

[0126] Adjusting parameters using gradient descent algorithm:

[0127]

[0128] Where γ is the learning rate, which controls the update step size;

[0129] Cross-validation and regularization are used to prevent overfitting and ensure the model's generalization ability. The dataset is divided into K subsets. Each time, one subset is selected as the validation set, and the remaining K-1 subsets are used as the training set. The training and evaluation process is repeated K times. Finally, the average of the K evaluation results is calculated as an estimate of the model performance. The performance evaluation formula is as follows:

[0130]

[0131] in, This is the value of the loss function calculated in the k-th cross-validation;

[0132] Finally, the lightweight 3D model and the trained deep learning model are integrated:

[0133] yfinal =w d ·f d (X)+w l ·f l (X)

[0134] Among them, f d (X) and f l (X) represents the outputs of the deep learning model and the lightweight model, respectively, with weights w. d and w l Based on the performance settings of the validation set;

[0135] The ship's position coordinates c i Vibration data from monitoring points a i and working conditions w i Associated, forming the input tuple (a i ,c i ,w i ), forming a joint feature vector x i =[a i ,c i ,w i ]∈R m+3+d Design a data interface to transmit the real-time acquired vibration spectrum data a i The coordinates c of the monitoring point in the lightweight model i Exact match: Represented as a paired dataset Where a i It was collected from location c i The vibration spectrum;

[0136] Vibration spectrum a i The radiated noise spectrum y is obtained by inputting it into a deep learning model. i =f θ (a i Then, based on the one-to-one correspondence of the positions in the three-dimensional coordinate system, the predicted result y is... i Associating with coordinate c in the 3D model i Generate the predicted value of radiated noise at the noise monitoring point.

[0137] Ensure accurate matching and synchronization between the input data of the deep learning model and the structural positions in the lightweight model, so as to achieve real-time mapping and prediction of ship structure and vibration acoustics.

[0138] Step 6: Intelligent prediction and display of the ship's acoustic status:

[0139] The intelligent forecasting algorithm based on the lightweight digital model constructed in step 5 uses the ship vibration acoustic monitoring data collected in real time in step 4 as input to intelligently forecast the acoustic state of the entire ship.

[0140] The system features an intuitive and user-friendly visualization interface that displays the shipwide acoustic condition forecast results obtained from the intelligent forecasting model. It uses color gradients and dynamic arrow indicators to show the predicted distribution trend of radiated noise across the ship, identifies areas or equipment that may exhibit acoustic anomalies using flashing icons or special markers, and displays the historical and future trends of the predicted locations of major noise sources using charts and curves.

[0141] Based on the information displayed on the visual interface, ship maintenance personnel can judge the changing trends of the ship's acoustic status and take corresponding control measures in a timely manner, such as adjusting equipment operating parameters and optimizing the ship's navigation route, to reduce the impact of ship noise on personnel health, equipment operation and the surrounding environment, and ensure the safe, comfortable and environmentally friendly operation of the ship.

[0142] Example 1:

[0143] Taking a 5,000-ton bulk carrier as an example, the application of this invention is explained in detail, and the specific steps are as follows:

[0144] Step S1: Construct a 3D model of the ship and perform lightweight processing;

[0145] Comprehensive data collection of the ship's design drawings, including lines drawings, general arrangement drawings, detailed structural drawings, and material properties, was conducted. A 3D model of the ship was constructed using the professional 3D modeling software SolidWorks, covering the hull outline, internal compartment layout, structural components, and typical mechanical equipment. Unnecessary geometric details were removed, adjacent faces and volumes were merged, and the topology was optimized. The influence of materials on sound wave propagation characteristics was fully considered, and physical properties were assigned to the model. This resulted in a 40% reduction in model data size and a significant improvement in storage and transmission efficiency.

[0146] Step S2: Define typical mechanical equipment and vibration monitoring locations;

[0147] Based on ship acoustics theory and the ship's operational data, the main engine, propeller, generator, and air conditioning system were identified as typical equipment. Vibration monitoring points were installed at key locations such as the main engine mounts, propeller shaft, generator base, and air conditioning system inlet and outlet pipes, as well as sensitive locations such as bulkheads and decks, to ensure consistency with the coordinates of the three-dimensional model.

[0148] Step S3: Construct the mapping relationship between ship structural vibration and acoustics;

[0149] Numerical simulation: Based on ship structural mechanics, acoustic principles and vibration theory, a finite element model of the ship is established in Abaqus software. Considering the physical parameters of the ship structure such as mass, stiffness, and damping, as well as the excitation effect of typical mechanical equipment, the vibration response and radiated noise of the ship structure under various working conditions are simulated and calculated.

[0150] Experimental measurement: High-precision sensors were installed at the monitoring location to collect vibration and noise data under different speeds, loads, and sailing attitudes.

[0151] The ship vibration and noise data are tagged to form a complete dataset of ship structural vibration and radiated noise, which is then stored in the database.

[0152] Step S4: Dynamically display the vibration monitoring results of typical ship equipment and structures;

[0153] High-precision vibration acceleration sensors are deployed at vibration monitoring locations to measure vibration acceleration in real time. Analog signals acquired by the sensors are converted into digital signals via a data acquisition card and transmitted to the ship's central monitoring system. A visual interface, developed using Vue architecture and Three.js, maps the monitoring data to a 3D model. An early warning mechanism is implemented according to International Maritime Organization standards; when vibration data exceeds limits, the system automatically alarms, highlights the affected areas, and provides response suggestions.

[0154] Step S5: Intelligent forecasting algorithm based on lightweight digital model.

[0155] The vibration data of the ship structure collected by the vibration acceleration sensor is preprocessed, and the vibration data spectrum is extracted by Fast Fourier Transform (FFT) as input data for the lightweight digital model intelligent prediction algorithm.

[0156] The deep learning model is designed with an input layer, hidden layers, and an output layer. The input layer receives preprocessed vibration spectrum data, and the output layer outputs the predicted underwater radiated noise spectrum. A ship structure vibration and radiated noise dataset is used as training data, containing ship structure vibration data, underwater radiated noise data, and corresponding label information (coordinates and operating conditions). The training data is input into the deep learning model, and the model's prediction performance is optimized by adjusting network parameters. Cross-validation and regularization techniques are used to prevent overfitting.

[0157] By integrating a lightweight digital model with a trained deep learning model, a data interface is developed to associate the ship's position coordinates with monitoring point data. This ensures that the input data of the deep learning model (vibration monitoring point data) is accurately matched with the structural position in the lightweight model, enabling real-time mapping and prediction of the ship's structure and vibration acoustics.

[0158] Step S6: Intelligent prediction and display of the acoustic status of the entire ship.

[0159] A lightweight digital model-based intelligent forecasting algorithm uses real-time collected ship vibration and acoustic monitoring data as input to intelligently predict the overall acoustic status of the ship. An intuitive and user-friendly visualization interface is designed to display the ship's acoustic status forecast results obtained from the intelligent forecasting model. The predicted distribution trend of radiated noise across the ship is displayed using color gradients and dynamic arrow indicators. Areas or equipment that may exhibit acoustic anomalies are marked with flashing icons or special markers. Charts and curves show the historical changes and future trends of the predicted locations of major noise sources. Based on the information displayed on the visualization interface, ship maintenance personnel can determine the changing trends of the ship's acoustic status and take timely control measures.

[0160] The functions of the intelligent prediction system for ship acoustic conditions based on a lightweight digital model in this invention can be described by the aforementioned intelligent prediction method for ship acoustic conditions based on a lightweight digital model, and will not be repeated here.

[0161] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent prediction of ship acoustic state based on a lightweight digital model, characterized in that: Includes the following steps: Step 1: Construct a 3D model of the ship and perform lightweight processing, remove redundant geometric details, merge adjacent faces, optimize the model topology, reduce model complexity, and retain key acoustic features. Then, assign material acoustic physical properties to the structure. Step 2: Select typical mechanical equipment and arrange vibration monitoring points in key parts and sensitive locations of the ship structure to ensure that the monitoring points are consistent with the coordinate system of the three-dimensional model and have accurate coordinate values; Step 3: Construct a mapping relationship between ship structural vibration and acoustics through numerical simulation or experimental measurement, and generate a standardized dataset of ship structural vibration response and underwater radiated noise; the construction of the mapping relationship between ship structural vibration and acoustics is as follows: The frequency domain transfer function of the vibration response is: ,in, For displacement spectrum, The excitation force spectrum; The mapping relationship between structural vibration and acoustics is as follows: ,in, For sound pressure, It represents displacement velocity; The numerical simulation is based on ship structural mechanics, acoustic principles, and vibration theory. A finite element / boundary element model of the ship is established, considering the physical parameters of the ship structure's mass, stiffness, and damping, as well as the excitation effect of typical mechanical equipment. The vibration response of the ship structure at the vibration monitoring location under different working conditions is calculated through simulation. Combining the acoustic characteristics of the ship structure material and the acoustic parameters of the surrounding fluid medium, the propagation process of sound waves inside the ship structure and in the underwater acoustic field is simulated, and the underwater radiated noise of the ship is calculated through simulation. The experimental measurements included: conducting actual ship vibration acoustic testing experiments, installing high-precision vibration sensors and acoustic sensors at vibration monitoring locations, collecting data on different speeds, loads, and sailing attitudes of the ship under different operating conditions, and obtaining data on the vibration response and radiated noise of the structure. Step 4: By deploying vibration acceleration sensors at vibration monitoring locations to measure vibration acceleration in real time, the data is transmitted to the monitoring system. A visualization interface for ship vibration acoustic monitoring is designed to associate and map the monitoring data with the three-dimensional model. The changes in color, shape, and size of the visualization elements intuitively display the vibration amplitude, frequency, and acoustic intensity information of typical equipment and structures, and trigger alarm signals based on preset thresholds. Step 5: Preprocess the measured vibration acceleration and extract the vibration spectrum data using Fast Fourier Transform; construct a deep learning model, input the preprocessed vibration spectrum data, and output the underwater radiated noise spectrum prediction result; input the standardized dataset obtained in Step 3 as the training set into the deep learning model, and optimize the model's prediction performance by adjusting the network parameters; integrate the association mapping dataset generated in Step 4 based on the lightweight 3D model in Step 1 with the deep learning model to construct an intelligent prediction system, realizing real-time mapping and prediction of ship structure and vibration acoustics; Step 6: Intelligent prediction and display of the acoustic status of the whole ship: Based on real-time data and prediction results, abnormal areas are marked with dynamic arrows and flashing icons in the three-dimensional visualization interface, and historical data is traced back through the time axis to predict future noise distribution trends. Finally, the prediction results are displayed through the visualization interface. Step 5 integrates the dataset generated based on the lightweight 3D model with the deep learning model according to the following formula: in, The predicted output of a deep learning model. The weights are the predicted outputs obtained from a dataset generated based on a lightweight 3D model. and To be configured based on the performance of the validation set; Step 5 will determine the ship's position coordinates. Vibration data from monitoring points and working conditions Associate them to form input tuples To form a joint feature vector Design a data interface to process the real-time collected vibration spectrum data. Coordinates of monitoring points in the lightweight model Exact match: , representing a paired dataset ,in It is collected from location The vibration spectrum; the vibration spectrum The radiated noise spectrum is obtained by inputting it into a deep learning model. Then, based on the one-to-one correspondence of positions in the three-dimensional coordinate system, the prediction results are... Coordinates associated with the 3D model Generate the predicted value of radiated noise at the noise monitoring point.

2. The intelligent prediction method for ship acoustic state based on a lightweight digital model according to claim 1, characterized in that: The typical equipment in step 2 includes the main engine, propellers in the propulsion system, auxiliary mechanical equipment, and other equipment that generates significant vibration and noise during ship navigation, with vibration monitoring points arranged at their bases, bulkheads, or structural connection nodes.

3. The intelligent prediction method for ship acoustic state based on a lightweight digital model according to claim 1, characterized in that: Step 6 is based on a lightweight digital model-based intelligent forecasting algorithm. It uses real-time collected ship vibration acoustic monitoring data as input to intelligently forecast the acoustic state of the entire ship and displays the forecast results through a designed visualization interface. The predicted distribution trend of radiated noise across the entire ship is displayed using color gradients and dynamic arrow indicators. Flashing graphs mark areas or equipment where acoustic anomalies may occur. Charts and curves show the historical changes and future trends of the predicted locations of major noise sources. Based on the information displayed on the visualization interface, ship maintenance personnel can judge the changing trends of the ship's acoustic status and take appropriate control measures in a timely manner.

4. A computer device / equipment / system, comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 3.

Images