Method, system and electronic device for monitoring abnormal noise of ship propulsion system

Abstract

The application provides a ship propulsion system abnormal sound monitoring method, system and electronic equipment. The method comprises: acquiring monitoring data of a propulsion system in a ship, the monitoring data comprising vibration signals of multiple components and a rotation speed signal of the propulsion system; converting the vibration signals to an order domain based on the rotation speed signal to obtain a multi-channel order spectrum; performing source component decomposition on the multi-channel order spectrum based on a target constraint to obtain multiple source components and spatial distribution characteristics of each source component; determining an abnormal sound candidate component and a corresponding fault mechanism type in the multiple source components; performing spatial matching on the spatial distribution characteristics of the abnormal sound candidate component and a spatial attenuation mode of each candidate component to obtain a first matching result; performing mechanism matching on the fault mechanism type of the abnormal sound candidate component and a mechanism compatibility of each candidate component to obtain a second matching result; and obtaining an abnormal sound monitoring result of the propulsion system based on the first matching result and the second matching result.

Description

Technical Field

[0001] This application relates to the field of marine technology, and in particular to a method, system and electronic equipment for monitoring abnormal noises in a ship propulsion system. Background Technology

[0002] During the operation of a ship's propulsion system, multiple components such as gearboxes, bearings, and couplings may produce abnormal noises. However, due to the complex propagation path of the ship's structure and the superposition of strong background noise in the engine room, the abnormal noise signals are superimposed at multiple sensor measurement points, resulting in low accuracy in locating abnormal noises and a high rate of misjudgment. This makes it impossible to achieve precise component-level identification of the source of abnormal noises, leading to problems such as long downtime and high maintenance costs. Summary of the Invention

[0003] In view of this, this application provides a method, system and electronic equipment for monitoring abnormal noises in a ship propulsion system, in order to overcome the shortcomings of the prior art.

[0004] According to a first aspect of this application, a method for monitoring abnormal noise in a ship propulsion system is provided, comprising: acquiring monitoring data of the propulsion system in a ship, the monitoring data including vibration signals of multiple components and rotational speed signals of the propulsion system; converting the vibration signals to the order domain based on the rotational speed signals to obtain a multi-channel order spectrum; performing source component decomposition on the multi-channel order spectrum based on target constraints to obtain multiple source components and spatial distribution characteristics of each source component; the target constraints are used to match the activation modes of the decomposed source components on different sensor channels with the coherence observations between channels; the spatial distribution characteristics characterize the weight distribution on the corresponding sensor channel of each source component; determining candidate components for abnormal noise and their corresponding fault mechanism types among the multiple source components; spatially matching the spatial distribution characteristics of the candidate components for abnormal noise with the spatial attenuation modes of each candidate component to obtain a first matching result; performing mechanism matching between the fault mechanism types of the candidate components for abnormal noise and the mechanism compatibility of each candidate component to obtain a second matching result; and obtaining the abnormal noise monitoring result of the propulsion system based on the first matching result and the second matching result.

[0005] The second aspect of this application provides a ship propulsion system abnormal noise monitoring system for implementing the aforementioned ship propulsion system abnormal noise monitoring method. It includes: a data acquisition module for acquiring monitoring data of the ship's propulsion system, the monitoring data including vibration signals of multiple components and the propulsion system's rotational speed signal; an order conversion module for converting the vibration signals to the order domain based on the rotational speed signal to obtain a multi-channel order spectrum; and a component decomposition module for performing source component decomposition on the multi-channel order spectrum based on target constraints to obtain multiple source components and the spatial distribution characteristics of each source component. The target constraints are used to constrain the temporal characteristics, spatial characteristics, and inter-channel consistency during the source component decomposition process. The spatial distribution characteristics characterize the weight distribution of each source component on each sensor channel; the abnormal noise identification module is used to identify abnormal noise candidate components among multiple source components and determine their corresponding fault mechanism types based on the spectral characteristics of the abnormal noise candidate components; the spatial matching module is used to spatially match the spatial distribution characteristics of the abnormal noise candidate components with the spatial attenuation mode of each candidate component to obtain a first matching result; the mechanism matching module is used to perform mechanism matching between the fault mechanism type of the abnormal noise candidate components and the mechanism compatibility of each candidate component to obtain a second matching result; the result generation module is used to obtain the abnormal noise monitoring results of the propulsion system based on the first matching result and the second matching result.

[0006] A third aspect of this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for monitoring abnormal noises in a ship propulsion system.

[0007] The beneficial effects of this application are at least as follows: By adopting the technical solution of this application, the vibration signal is converted to the order domain based on the rotation speed signal to obtain a multi-channel order spectrum, which eliminates the interference of rotation speed fluctuation on the extraction of abnormal noise features, making the abnormal noise features under different working conditions comparable; then, by performing source component decomposition on the multi-channel order spectrum based on target constraints, multiple source components can be effectively separated from the aliased signals of multiple sensor measurement points. The target constraints make the activation mode of the decomposed source components on different sensor channels match the coherence observation between channels, ensuring the physical rationality of the decomposition results and overcoming the signal aliasing problem caused by the complex propagation path of the ship structure and the strong background noise in the engine room.

[0008] Based on this, a first matching result is obtained by spatially matching the spatial distribution characteristics of the candidate components of abnormal noise with the spatial attenuation mode of each candidate component. The propagation attenuation law of vibration signal in the propulsion system structure is used to determine the location of the abnormal noise source. At the same time, a second matching result is obtained by mechanistically matching the fault mechanism type of the candidate components of abnormal noise with the mechanism compatibility of each candidate component. The difference in fault type of different components is used to assist in the judgment of the abnormal noise source. Finally, the abnormal noise monitoring result is obtained by fusing the first and second matching results. This achieves a comprehensive judgment of the abnormal noise source from two dimensions: spatial evidence and mechanism evidence. It improves the accuracy of abnormal noise location, reduces the false judgment rate, and achieves precise component-level identification of abnormal noise in the propulsion system. This effectively shortens downtime and reduces maintenance costs.

[0009] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0010] The above and other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which: Figure 1 This is a flowchart illustrating a method for monitoring abnormal noises in a ship propulsion system, as provided in an embodiment of this application. Figure 2 This is a schematic diagram of a ship propulsion system abnormal noise monitoring system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0011] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0012] Figure 1 A flowchart illustrating a method for monitoring abnormal noises in a ship propulsion system provided in an embodiment of this application is shown.

[0013] like Figure 1 As shown, the method for monitoring abnormal noises in a ship's propulsion system includes steps S101 to S107.

[0014] Step S101: Acquire monitoring data of the propulsion system in the ship. The monitoring data includes vibration signals of multiple components and rotational speed signals of the propulsion system. Step S102: Based on the rotational speed signal, the vibration signal is converted to the order domain to obtain a multi-channel order spectrum; Step S103: Decompose the multi-channel order spectrum into source components based on the objective constraint to obtain multiple source components and the spatial distribution characteristics of each source component; the objective constraint is used to match the activation modes of the decomposed source components on different sensor channels with the coherence observations between channels; the spatial distribution characteristics characterize the weight distribution on the corresponding sensor channel for each source component. Step S104: Determine the candidate components of abnormal noise and their corresponding fault mechanism types among multiple source components; Step S105: Spatial matching is performed between the spatial distribution characteristics of the candidate components of abnormal noise and the spatial attenuation mode of each candidate component to obtain the first matching result; Step S106: The fault mechanism type of the candidate components of abnormal noise is matched with the mechanism compatibility of each candidate component to obtain the second matching result; Step S107: Obtain the abnormal noise monitoring results of the propulsion system based on the first matching result and the second matching result.

[0015] In step S101, the propulsion system refers to the mechanical system in the ship's power transmission chain, which can be understood as the overall structure that transmits the power output from the main engine to the propeller to drive the ship's navigation.

[0016] For example, the propulsion system may include multiple components such as gearbox, thrust bearing, intermediate bearing, coupling and shaft support, which may produce abnormal noises during operation due to failures such as wear, loosening, and cracks.

[0017] Monitoring data refers to physical quantity measurements that reflect the operating status of the propulsion system. It can be understood as data collected by sensors that characterize the system's working condition.

[0018] For example, the monitoring data includes vibration signals of multiple components and rotational speed signals of the propulsion system. The vibration signals are used to capture the vibration response of each component, and the rotational speed signals are used to establish the correlation between vibration characteristics and rotational conditions. Since the vibration characteristics of the propulsion system are closely related to the rotational speed, the influence of rotational speed fluctuations needs to be eliminated through the rotational speed signals in subsequent processing.

[0019] In one feasible implementation, acceleration sensors can be installed at multiple component locations, such as the gearbox housing, thrust bearing housing, intermediate bearing housing, structure near the coupling, and shaft support, to collect vibration signals. At the same time, speed sensors can be installed on the main shaft or gearbox output shaft to collect speed signals, thereby obtaining multi-channel vibration signals and corresponding speed signals.

[0020] In step S102, because the rotational speed of the ship's propulsion system fluctuates during actual operation, the frequency of abnormal noise characteristics related to rotational speed changes with the speed fluctuation when analyzing the vibration signal in the time or frequency domain. This leads to frequency ambiguity and energy dispersion in the spectrum, affecting the extraction of abnormal noise characteristics. The order domain is the analysis domain based on the number of rotational cycles. The order is defined as the ratio of the vibration frequency to the rotational speed frequency. After converting the vibration signal to the order domain, the abnormal noise characteristics related to rotational speed are represented by fixed order values ​​in the order domain, eliminating the influence of rotational speed fluctuations. This makes the abnormal noise characteristics under different operating conditions consistent and comparable, facilitating subsequent source component decomposition and abnormal noise identification.

[0021] In one feasible implementation, the vibration signal of each channel can be resampled at equal angles based on the rotation speed signal, and the vibration signal uniformly sampled in the time domain can be converted into a vibration signal uniformly sampled in the angle domain. Then, Fourier transform is performed on the angle domain signal to obtain the order spectrum. The above processing is performed on the vibration signals of multiple sensor channels respectively to obtain a multi-channel order spectrum.

[0022] It should be noted that each channel in the multi-channel order spectrum corresponds to the measurement position of a sensor. The horizontal axis of the order spectrum is the order value, and the vertical axis is the vibration amplitude. When the order value is an integer, it corresponds to an integer multiple of the rotational frequency. When the order value is a fraction, it corresponds to a fractional multiple of the rotational frequency.

[0023] For example, the process of calculating the rotational phase of a propulsion system based on the rotational speed signal can be represented as: In the formula, ω(t) is the rotational speed signal; θ(t) is the rotational phase obtained by integrating the rotational speed signal; Each vibration signal is resampled at equal phase intervals to obtain the order domain signal: In the formula, x m (t) represents the vibration signal of the m-th sensor channel; This is the order domain signal corresponding to the m-th sensor channel; For the ℓth isophase sampling point; t( () is the phase point The corresponding time.

[0024] Based on the transformation of the order domain signal, a multi-channel order spectrum is obtained: In the formula, Let be the order spectrum of the m-th sensor channel at time τ at order o; o is the order; τ is the time window position; W(τ) is the analysis window centered at τ; g() is the window function.

[0025] In step S103, the target constraint refers to the constraint conditions applied during the source component decomposition process. It can be understood as the restriction conditions that ensure the decomposition results conform to the laws of physical propagation. It is used to match the activation modes of the source components obtained by decomposition on different sensor channels with the coherence observations between channels, thereby avoiding physically unreasonable source components in the decomposition results and improving the accuracy of decomposition.

[0026] Since the vibration signals measured by various sensors in a ship's propulsion system are formed by the superposition of multiple vibration source signals through different propagation paths, it is difficult to directly identify abnormal noises in the superimposed signals. By source component decomposition, the multi-channel order spectrum can be decomposed into multiple independent source components, and each source component corresponds to a potential vibration source.

[0027] Among them, the spatial distribution characteristics refer to the amplitude weight of each source component on each sensor channel. It can be understood as the energy distribution pattern of the source component in space. It is used to characterize the weight distribution on the corresponding sensor channel of each source component, reflecting the attenuation law when the vibration source signal propagates to each sensor position, and providing spatial evidence for subsequent abnormal noise source localization.

[0028] In one feasible implementation, a decomposition model of the multi-channel order spectrum can be constructed, representing the multi-channel order spectrum as a combination of the order spectra of multiple source components and their corresponding spatial distribution characteristics. An objective function is set to minimize the difference between the reconstructed multi-channel order spectrum and the actual measured multi-channel order spectrum. An objective constraint term is added to the objective function. This objective constraint term calculates the coherence observations between channels and compares them with the activation modes of the source components in different channels to minimize the deviation between the two. Through optimization, multiple source components and the spatial distribution characteristics of each source component are obtained.

[0029] It should be noted that the spatial distribution characteristics can be represented as a vector. The dimension of the vector is equal to the number of sensor channels. Each element in the vector represents the weight value of the source component on the corresponding channel. The magnitude of the weight value reflects the degree of contribution of the source component to the measurement signal of that channel.

[0030] For example, when performing source component decomposition on a multi-channel order spectrum, the order spectrum amplitude corresponding to the m-th sensor channel can be expressed as: In the formula, Let be the reconstructed order spectral amplitude matrix of the m-th sensor channel; K is the number of source components; W(o,k) represents the spectral characteristics of the k-th source component at order o; A(m,k) is the spatial weight of the k-th source component on the m-th sensor channel; H(k,τ) represents the activation intensity of the k-th source component at time τ.

[0031] In step S104, the abnormal noise candidate component refers to the component among multiple source components that is suspected of containing abnormal noise characteristics. It can be understood as the abnormal vibration component screened out from the source components obtained by decomposition, which is used for subsequent location and diagnosis of abnormal noise sources. Since the propulsion system contains both vibration components generated during normal operation and abnormal noise components caused by malfunctions, it is necessary to identify abnormal source components from the multiple source components obtained by decomposition as abnormal noise candidate components.

[0032] Fault mechanism type refers to the category of physical causes that lead to abnormal noise, which can be understood as the classification of the root causes of abnormal vibration.

[0033] For example, the fault mechanism types may include gear wear, bearing damage, component loosening, structural cracks, resonance response, etc., and different fault mechanism types exhibit different characteristic patterns in the order domain.

[0034] In one feasible implementation, features can be extracted from each source component. The extracted features include dominant order value, peak amplitude of the order spectrum, and distribution pattern of the order spectrum. Then, the source component is judged as a candidate component for abnormal noise according to preset anomaly judgment rules. The preset rules include the peak amplitude of the order spectrum exceeding the normal threshold, the appearance of non-integer order peaks, and the difference between the distribution pattern of the order spectrum and the normal pattern. For source components judged as candidates for abnormal noise, the fault mechanism type is further determined according to the correspondence between its dominant order value and the characteristic order of the propulsion system component.

[0035] It should be noted that a candidate component for abnormal noise can correspond to one type of fault mechanism or a combination of multiple fault mechanism types. When there are multiple candidate components for abnormal noise, it indicates that there may be multiple fault sources in the propulsion system.

[0036] In step S105, the spatial attenuation mode refers to the amplitude attenuation law of the vibration signal propagating to each sensor position when a candidate component is assumed to be the vibration source. It can be understood as the spatial propagation characteristics of the vibration signal of the candidate component. Since the vibration signal will attenuate during propagation due to factors such as propagation distance, structural damping, and propagation medium, candidate components at different positions have different spatial attenuation modes.

[0037] Spatial matching refers to comparing the similarity between the spatial distribution characteristics of candidate components of abnormal noise and the spatial attenuation patterns of candidate components. The first matching result refers to the quantitative matching degree obtained based on spatial matching.

[0038] In one feasible implementation, a spatial attenuation mode for each candidate component can be pre-established. The theoretical attenuation coefficient is calculated based on the spatial positional relationship between the candidate component and each sensor and the structural propagation path to form a spatial attenuation mode vector. Then, the similarity between the spatial distribution feature vector of the abnormal noise candidate component and the spatial attenuation mode vector of each candidate component is calculated. The similarity calculation can be performed using methods such as cosine similarity, normalized correlation coefficient, or Euclidean distance. The calculated similarity is used as the first matching result.

[0039] In step S106, mechanism compatibility refers to the degree of fit between the candidate component and the specific failure mechanism type. It can be understood as the strength of the association between the candidate component and the failure mechanism type, and is used to describe the probability of a certain candidate component experiencing a certain failure mechanism. Due to the different structural characteristics and operating properties of different components, the types of failure mechanisms they may generate also differ. For example, gear components are more likely to generate gear wear failures, while bearing components are more likely to generate bearing damage failures.

[0040] Mechanism matching refers to comparing the correspondence between the fault mechanism type of the candidate noise component and the mechanism compatibility of the candidate component. When the candidate component has a higher compatibility with a certain fault mechanism type, and the fault mechanism type is consistent with the fault mechanism type of the candidate noise component, the candidate component is more likely to be the source of the noise. The second matching result refers to the quantitative matching degree obtained based on mechanism matching.

[0041] In one feasible implementation, a mechanism compatibility matrix between candidate components and fault mechanism types can be pre-established. The rows of the matrix correspond to candidate components, the columns correspond to fault mechanism types, and the matrix elements represent the compatibility scores of the corresponding candidate components for the corresponding fault mechanism types. The compatibility scores can be determined based on component design characteristics, historical fault statistics, or expert knowledge. Then, based on the fault mechanism type of the abnormal noise candidate component, the compatibility score of each candidate component for that fault mechanism type is searched in the mechanism compatibility matrix, and this score is used as the second matching result.

[0042] In step S107, the abnormal noise monitoring results may include information such as identification of the abnormal noise source component, determination of the fault mechanism type, and confidence level assessment, which can comprehensively reflect the abnormal noise status of the propulsion system.

[0043] In one feasible implementation, the first matching result and the second matching result can be weighted and fused to obtain the comprehensive matching degree. For each abnormal noise candidate component and each candidate part, the weight coefficient can be determined according to the actual application scenario and historical diagnostic accuracy. Then, for each abnormal noise candidate component, the candidate part with the highest comprehensive matching degree is selected as the abnormal noise source of the candidate component. The identified abnormal noise source part, the corresponding fault mechanism type and the comprehensive matching degree are output as the abnormal noise monitoring result.

[0044] It should be noted that when there are multiple candidate components for abnormal noise, the source of the abnormal noise can be located for each candidate component separately. The abnormal noise monitoring results can include multiple sources of abnormal noise and their corresponding fault mechanism types, reflecting the situation of multi-source abnormal noise in the propulsion system. In addition, the abnormal noise monitoring results can also include auxiliary diagnostic information such as the severity assessment of abnormal noise at each sensor location, the frequency range of abnormal noise, and the speed range in which the abnormal noise occurs, providing maintenance personnel with a more comprehensive basis for fault diagnosis.

[0045] By adopting the technical solution of this application, the vibration signal is converted to the order domain based on the rotation speed signal to obtain a multi-channel order spectrum, which eliminates the interference of rotation speed fluctuation on the extraction of abnormal noise features, making the abnormal noise features under different working conditions comparable; then, by performing source component decomposition on the multi-channel order spectrum based on target constraints, multiple source components can be effectively separated from the aliased signals of multiple sensor measurement points. The target constraints make the activation mode of the decomposed source components on different sensor channels match the coherence observation between channels, ensuring the physical rationality of the decomposition results and overcoming the signal aliasing problem caused by the complex propagation path of the ship structure and the strong background noise in the engine room.

[0046] Based on this, a first matching result is obtained by spatially matching the spatial distribution characteristics of the candidate components of abnormal noise with the spatial attenuation mode of each candidate component. The propagation attenuation law of vibration signal in the propulsion system structure is used to determine the location of the abnormal noise source. At the same time, a second matching result is obtained by mechanistically matching the fault mechanism type of the candidate components of abnormal noise with the mechanism compatibility of each candidate component. The difference in fault type of different components is used to assist in the judgment of the abnormal noise source. Finally, the abnormal noise monitoring result is obtained by fusing the first and second matching results. This achieves a comprehensive judgment of the abnormal noise source from two dimensions: spatial evidence and mechanism evidence. It improves the accuracy of abnormal noise location, reduces the false judgment rate, and achieves precise component-level identification of abnormal noise in the propulsion system. This effectively shortens downtime and reduces maintenance costs.

[0047] To facilitate understanding of the implementation process of the technical solution in this application, an actual application scenario will be used as an example below.

[0048] During a voyage, the crew of a container ship reported significant abnormal vibrations and noise in the propulsion system of the engine room, with the unusual noises being particularly noticeable at medium to high speeds. The ship's propulsion system includes a main reduction gearbox, thrust bearing, two intermediate bearings, couplings, and several shaft supports. Acceleration sensors were installed in multiple locations, including the gearbox housing, thrust bearing housing, intermediate bearing housing 1, intermediate bearing housing 2, and the stern shaft support. A speed sensor was installed on the gearbox output shaft.

[0049] The system collected monitoring data of the propulsion system under normal navigation conditions. The rotational speed signal showed that the main shaft rotational speed fluctuated during this period. Since rotational speed fluctuations can cause energy dispersion in frequency domain analysis, the system converted the vibration signals of each channel to the order domain based on the rotational speed signal, thus eliminating the influence of rotational speed fluctuations. By performing source component decomposition on the multi-channel order spectrum, the system separated several independent source components. Most of these source components correspond to normal vibration characteristics such as thrust bearing rotational frequency and gearbox meshing frequency. However, one source component exhibited an abnormal peak at a specific non-integer order. This order did not correspond to the characteristic frequency of any normal component, and the vibration amplitude was significantly higher than normal. This component was identified as a candidate for abnormal noise.

[0050] Further analysis revealed that the order of the candidate component of the abnormal noise was close to the order of the rolling element passing frequency characteristic of the intermediate bearing, thus determining the fault mechanism type as bearing rolling element damage. Spatially, the candidate component of the abnormal noise had the highest weight at the No. 1 intermediate bearing housing, followed by the No. 2 intermediate bearing housing, while its weight was smaller at other sensor locations. This spatial distribution pattern highly matches the theoretical attenuation law when the No. 1 intermediate bearing acts as the vibration source. Furthermore, from a mechanism compatibility perspective, the bearing rolling element damage fault type has high compatibility with the mechanisms of the No. 1 and No. 2 intermediate bearings, but low compatibility with the mechanisms of other components such as the gearbox and thrust bearing.

[0051] Based on the combined spatial matching and mechanism matching results, the system identified the No. 1 intermediate bearing as the source of abnormal noise. The output abnormal noise monitoring results included information such as the No. 1 intermediate bearing as the source of abnormal noise and the fault mechanism type as damage to the bearing rolling elements.

[0052] As can be seen from the above examples, the embodiments of this application can successfully and accurately locate abnormal noise sources from the aliased vibration signals of multiple components in the propulsion system, avoiding misjudgment and unnecessary disassembly and inspection of other normal components such as gearboxes and thrust bearings, effectively shortening troubleshooting time and reducing maintenance costs. Especially under the actual operating conditions of complex ship propulsion systems with diverse propagation paths and strong background noise in the engine room, the accuracy of abnormal noise source identification is improved by eliminating the influence of speed fluctuations through order domain transformation, achieving signal separation through source component decomposition, and using a two-dimensional fusion of spatial matching and mechanism matching.

[0053] Based on the above embodiments, as a feasible embodiment, step S103 may further include the following steps.

[0054] Step S201: Establish a source component decomposition model for the multi-channel order spectrum. In the source component decomposition model, the multi-channel order spectrum is represented as a linear combination of multiple source components. The parameters to be solved for each source component include the spectral basis, the time activation coefficient, and the spatial weight vector. Step S202: Construct an objective function containing objective constraints based on the source component decomposition model; Step S203: Optimize the objective function to obtain the values ​​of the spectral basis, temporal activation coefficients and spatial weight vectors for each source component; Step S204: Determine the spatial distribution characteristics of each source component based on the obtained spatial weight vector; In step S201, since the actual observed multi-channel order spectrum is formed by the superposition of multiple vibration sources through different propagation paths, it can be represented as a linear combination of multiple source components by establishing a source component decomposition model.

[0055] The source component decomposition model refers to a model that describes the mathematical relationship between a multi-channel order spectrum and multiple source components, used to decompose aliasing observations into source components. In the source component decomposition model, each source component corresponds to three sets of parameters to be solved: spectral basis, temporal activation coefficients, and spatial weight vector.

[0056] Among them, the spectral basis characterizes the spectral characteristics of the source component in the order domain, describing the energy distribution pattern of the source component at different orders; the time activation coefficient characterizes the activation intensity of the source component over time, reflecting the activity level of the source component at different times; and the spatial weight vector characterizes the energy distribution of the source component on each sensor channel, reflecting the amplitude weight of the source component when it propagates to each sensor position.

[0057] In one feasible implementation, the multi-channel order spectrum can be represented as a product of a spectral basis matrix, a temporal activation coefficient matrix, and a spatial weight matrix. The number of columns in the spectral basis matrix is ​​equal to the number of source components, and the number of rows is equal to the order dimension length. The number of rows in the temporal activation coefficient matrix is ​​equal to the number of source components, and the number of columns is equal to the number of time windows. The number of rows in the spatial weight matrix is ​​equal to the number of sensor channels, and the number of columns is equal to the number of source components. The order spectrum of each sensor channel can be represented as a combination of the spectral basis matrix and the spatial weight and temporal activation coefficient matrix corresponding to that channel.

[0058] In step S202, the objective function refers to the mathematical expression used to evaluate the fitting effect of the decomposition model and the degree of constraint satisfaction. The objective function typically consists of two parts: a data fitting term and a constraint term. The data fitting term is used to minimize the difference between the reconstructed multi-channel order spectrum and the actual observed multi-channel order spectrum, while the constraint term is used to ensure that the decomposition results meet the objective constraint conditions.

[0059] In one feasible implementation, an objective function can be constructed to include a reconstruction error term and an objective constraint term. The reconstruction error term calculates the deviation between the multi-channel order spectrum reconstructed based on the current parameters and the actual observed multi-channel order spectrum. The deviation can be measured by methods such as Euclidean distance. The objective constraint term constructs a corresponding mathematical expression based on the objective constraints described in step S103, and the weight of the constraint term in the objective function is controlled by the regularization coefficient, thereby achieving a balance between fitting accuracy and constraint satisfaction. For example, let V m Let be the actual order spectral amplitude matrix corresponding to the m-th sensor channel, then the objective function can be expressed as: In the formula, V m This is the actual order spectral amplitude matrix corresponding to the m-th sensor channel; Let W be the reconstructed spectral amplitude matrix corresponding to the m-th sensor channel; W be the spectral basis matrix; H be the temporal activation coefficient matrix; and A be the spatial weight matrix. F It is the Frobenius norm; Let L1 be the norm of the time activation coefficient matrix; Φoh(A, H) represents the total variation of the spatial weight vector of the k-th source component; Φoh(A, H) represents the propagation consistency constraint term; λ1, λ2, and λ3 are the weight coefficients.

[0060] In step S203, optimization refers to the process of adjusting model parameters through numerical optimization algorithms to minimize the objective function value, thereby obtaining source component decomposition results that meet the constraints. Since the objective function is usually a non-convex function and the parameters to be solved have a high dimension, iterative solutions using appropriate optimization algorithms are required.

[0061] In one feasible implementation, optimization algorithms such as the least squares method can be used to solve the objective function. After setting the initial parameter values, the spectral basis, temporal activation coefficients, and spatial weight vectors are updated iteratively until the objective function converges or reaches the preset number of iterations. During the solution process, non-negativity constraints can be applied to the parameters to ensure their physical meaning. Finally, the values ​​of the spectral basis, temporal activation coefficients, and spatial weight vectors of each source component are obtained.

[0062] In step S204, since the spatial weight vector directly represents the amplitude weight of the source component on each sensor channel, the obtained spatial weight vector can be used as the spatial distribution feature of the source component.

[0063] In one feasible implementation, the obtained spatial weight vector can be normalized so that the sum of the vector elements equals 1 or the norm of the vector equals 1. The normalized spatial weight vector serves as the spatial distribution feature of the source component, which facilitates subsequent comparison with the spatial attenuation mode of candidate components. The sensor channel corresponding to the element with a larger value in the vector indicates that the source component has a stronger energy contribution at that location.

[0064] Optionally, a multiple random initialization strategy can be adopted during the optimization process. The optimization solution is performed for each initialization and the solution with the minimum objective function value is retained, thereby reducing the risk of getting trapped in local optima and improving the stability and accuracy of the decomposition results.

[0065] By employing the embodiments of this application, a source component decomposition model is established for multi-channel order spectra, and the parameters to be solved for each source component are decomposed into three sets of independent parameters: spectral basis, temporal activation coefficients, and spatial weight vectors. This enables decoupled representation of the frequency domain, temporal domain, and spatial characteristics of the source components, giving each dimension of the characteristics a clear physical meaning. Furthermore, by introducing objective constraint terms into the objective function and performing optimization, the source components obtained from the decomposition can be ensured to satisfy the physical propagation laws while maintaining reconstruction accuracy.

[0066] Based on the above embodiments, as a possible implementation, the target constraints include at least one or more combinations: Temporal sparsity constraints are used to make the anomalous components exhibit sparse characteristics in the time dimension. Spatial continuity constraints are used to ensure that the spatial weights on adjacent sensor channels change smoothly. Propagation consistency constraints are used to ensure that the joint activation characteristics of each source component across multiple sensor channels are consistent with the coherence characteristics observed between channels.

[0067] In this embodiment, the temporal sparsity constraint refers to the constraint condition that causes the temporal activation coefficient of the source component to be close to zero most of the time.

[0068] Since abnormal noises in propulsion systems usually only occur under specific operating conditions or during specific time periods, while vibration components generated during normal operation are usually continuous, applying time sparsity constraints can enhance the distinguishability of abnormal noise components from normal vibration components in the time domain, which is beneficial for accurately identifying candidate abnormal noise components from the decomposition results.

[0069] In one feasible implementation, an L1 norm regularization term or an L0 norm approximation term of the time activation coefficients can be added to the objective function as a time sparsity constraint. The L1 norm regularization causes some coefficients to tend to zero by penalizing the sum of the absolute values ​​of the time activation coefficients, thereby achieving a sparsity effect. The regularization coefficient is used to control the degree of sparsity, and the larger the regularization coefficient, the stronger the sparsity constraint.

[0070] For example, the time sparsity constraint can be expressed as: In the formula, H(k, τ) is the time activation coefficient of the k-th source component at time τ.

[0071] In this embodiment, spatial continuity constraint refers to the constraint condition that causes the spatial weight values ​​of adjacent sensor channels to transition smoothly. It can be understood as a constraint that avoids drastic changes in spatial weights at adjacent positions.

[0072] Since the propagation of vibration signals in the propulsion system structure follows a continuous attenuation law, the vibration amplitude at adjacent sensor locations usually shows a gradual rather than abrupt change. By applying spatial continuity constraints, the spatial weight vector obtained by decomposition can conform to the physical propagation characteristics, thus avoiding unreasonable spatial distribution patterns.

[0073] In one feasible implementation, a penalty term for the spatial weight difference between adjacent channels can be added to the objective function as a spatial continuity constraint. The sum of squares or absolute values ​​of the differences between adjacent elements in the spatial weight vector are calculated. By minimizing this penalty term, the spatial weights of adjacent sensor channels are kept smooth. The adjacency relationship of the sensor channels can be determined based on the physical location of the sensors in the propulsion system.

[0074] For example, spatial continuity constraints can be expressed as: In the formula, ak is the spatial weight vector of the k-th source component; Let A be the set of adjacency relationships between adjacent sensor channels; A(m,k) and A(n,k) are the spatial weights of the k-th source component on adjacent sensor channels m and n, respectively.

[0075] In this embodiment, the propagation consistency constraint refers to the constraint condition that makes the combination result of the temporal activation mode and spatial distribution mode of the source component conform to the actual observed inter-channel correlation, so that the joint activation characteristics of each source component on multiple sensor channels are consistent with the coherence characteristics observed between channels.

[0076] Since the signals from the same vibration source propagating to different sensor locations are coherent, the propagation consistency constraint guides the decomposition of source components by utilizing the coherence information between channels, so that the decomposed source components can correctly reflect the propagation characteristics of the vibration source.

[0077] In one feasible implementation, the coherence matrix between multi-channel order spectra can be pre-calculated as an observation coherence feature. Then, the theoretical coherence between each pair of channels is calculated based on the spatial weight vector and temporal activation coefficient in the source component decomposition model. A penalty term for the difference between theoretical coherence and observation coherence is added to the objective function as a propagation consistency constraint. The difference measure can be the mean square error. By minimizing this penalty term, the decomposition result is matched with the actual observed channel coherence.

[0078] For example, the coherence between the m-th sensor channel and the n-th sensor channel can be expressed as: γ mn (o, τ) represents the coherence between sensor channel m and sensor channel n at order o and time τ; S mn (o, τ) represents the cross spectrum between sensor channel m and sensor channel n; S mm (o, τ) and S nn (o, τ) are the autospectral values ​​of sensor channel m and sensor channel n, respectively; ε is a constant to prevent the denominator from being zero.

[0079] Based on this, the propagation consistency constraint term can be expressed as: In the formula, Γ mn (τ) is the coherence statistic of sensor channel m and sensor channel n at time τ; A(m,k) A(n,k) H(k,τ) is the joint activation intensity of the k-th source component on sensor channels m and n.

[0080] It should be noted that the objective constraint can include only one of the three constraints mentioned above, or it can include a combination of two or three constraints. When using multiple constraints at the same time, different weight coefficients can be set for each constraint to balance the role of each constraint in the objective function. The weight coefficients can be adjusted according to the characteristics of the actual data and prior knowledge.

[0081] Optionally, for time sparsity constraints, the constraint can be applied only to some source components instead of all source components. For example, based on the spectral basis characteristics of the source components, it can be preliminarily determined which source components may correspond to abnormal noise, and time sparsity constraints can be applied only to these suspected abnormal noise source components, thereby avoiding the imposition of unreasonable sparsity requirements on normal vibration components.

[0082] By employing the embodiments of this application, applying temporal sparsity constraints to make abnormal noise components exhibit sparse characteristics in the time dimension, the intermittent occurrence of abnormal noises can be utilized to enhance the distinguishability between abnormal noise components and normal vibration components, thereby improving the accuracy of abnormal noise candidate component identification. Applying spatial continuity constraints to maintain a smooth change in spatial weights on adjacent sensor channels ensures that the spatial distribution characteristics obtained from the decomposition conform to the physical laws of continuous propagation of vibration signals in the structure, avoiding unreasonable spatial distribution patterns and improving the reliability of spatial positioning. Applying propagation consistency constraints to ensure that the joint activation characteristics of the source components are consistent with the coherence characteristics observed between channels, fully utilizing the correlation information between multi-channel signals to guide the decomposition process, enabling the decomposition results to accurately reflect the propagation characteristics of the vibration source.

[0083] Based on the above embodiments, as a feasible embodiment, step S104 may further include the following steps.

[0084] Step S301: Calculate an anomaly score for each source component. The anomaly score is calculated based on at least one of the source component’s temporal sparsity, spectral impact, and deviation from the normal background. Step S302: Source components with abnormal scores greater than or equal to a preset threshold are identified as candidate components for abnormal sounds; Step S303: Match the spectral characteristics of the abnormal noise candidate components with the preset fault mechanism feature library to obtain the fault mechanism type corresponding to the abnormal noise candidate components.

[0085] In step S301, temporal sparsity characterizes the sparsity of the temporal activation coefficients of the source component in the time dimension. It can be measured by calculating the proportion of non-zero elements in the temporal activation coefficients. The higher the temporal sparsity, the more the source component tends to be activated at a few moments, which is more in line with the intermittent characteristics of abnormal sounds.

[0086] Spectral impact characterizes the impact characteristics of the spectral basis of a source component in the frequency domain. It can be measured by calculating the peak factor, waveform factor, or impulse index of the spectral basis. The stronger the spectral impact, the more concentrated the spectral energy of the source component is on a few frequency components with a large amplitude, which is more in line with the impact characteristics of abnormal noise.

[0087] The deviation from the normal background characterizes the degree of difference between the spectral base of the source component and the typical spectral mode under normal operating conditions. It can be calculated by calculating the distance between the spectral base and the normal background spectral template. The greater the deviation, the less the source component conforms to normal vibration characteristics and the more likely it is to be an abnormal noise component.

[0088] In one feasible implementation, three indicators can be calculated for each source component: temporal sparsity, spectral impact, and deviation from the normal background. Each indicator is normalized to make its value range uniform between 0 and 1. Then, a comprehensive anomaly score is calculated by weighted summation. The weight coefficients can be set according to the actual application scenario and empirical knowledge.

[0089] In step S302, by comparing the abnormal score of each source component with a preset threshold, source components with abnormal scores greater than or equal to the preset threshold can be identified as candidate components for abnormal noise, and source components with abnormal scores less than the preset threshold can be identified as normal vibration components.

[0090] In step S303, the fault mechanism feature library refers to a pre-established knowledge base containing spectral feature templates corresponding to various faults, which stores a database of typical spectral patterns for different fault types. Each entry in the fault mechanism feature library corresponds to a fault mechanism type and stores characteristic information such as the typical spectral base mode, characteristic order range, and spectral envelope shape of that fault type.

[0091] By calculating the similarity between the spectral base of the abnormal noise candidate component and the spectral feature templates of each fault type in the fault mechanism feature library, the fault mechanism type that best matches the abnormal noise candidate component can be found.

[0092] In one feasible implementation, the correlation coefficient between the spectral basis of the abnormal noise candidate component and each template in the fault mechanism feature library can be calculated, and the fault mechanism type with the highest similarity can be selected as the fault mechanism type corresponding to the abnormal noise candidate component. When the highest similarity is lower than the preset matching threshold, the abnormal noise candidate component can be marked as an unknown type of abnormal noise.

[0093] Optionally, temporal context information can be introduced when calculating anomaly scores. For example, the rate of change of the temporal activation coefficient of the source component between adjacent time windows can be examined. A larger rate of change indicates that the source component has a sudden characteristic, which can be used as an additional reference for anomaly scoring, thereby improving the sensitivity of transient anomaly identification.

[0094] Optionally, when matching fault mechanism types, the spatial distribution characteristics of the abnormal noise candidate components can be combined, and the location of the component most likely to be the source of the abnormal noise can be determined according to the spatial weight vector. The fault type that may occur at the location of the component can be matched first in the fault mechanism feature library, thereby improving the accuracy and efficiency of matching.

[0095] By adopting the embodiments of this application, an anomaly score based on temporal sparsity, spectral impact, and deviation from normal background is calculated for each source component. This enables a comprehensive evaluation of the anomaly degree of the source component from multiple perspectives, including time domain, frequency domain, and statistical distribution. Compared with a single index judgment method, this improves the accuracy and robustness of abnormal noise identification.

[0096] Based on the above embodiments, as a feasible embodiment, step S105 may further include the following steps.

[0097] Step S401: Obtain the spatial attenuation mode corresponding to each candidate component. The spatial attenuation mode characterizes the energy propagation attenuation law from the candidate component to each sensor channel. Step S402: Calculate the similarity between the spatial distribution characteristics of the candidate components of abnormal noise and the spatial attenuation mode of each candidate component to obtain the spatial matching degree of each candidate component; Step S403: Determine the first matching result based on the spatial matching degree. The first matching result is used as the spatial evidence strength of the abnormal noise source for each candidate component.

[0098] In step S401, the spatial attenuation mode refers to the mode vector that describes the attenuation distribution law of vibration energy propagating from the candidate component location to each sensor location, and is used to establish the mapping relationship between the candidate component location and the sensor observation.

[0099] Because vibration signals attenuate as they propagate through the propulsion system structure due to factors such as propagation distance, propagation path, and structural damping, and because different candidate components have different propagation paths to each sensor, each candidate component has a specific spatial attenuation mode.

[0100] Optionally, the dimension of the spatial decay mode is the same as the number of sensor channels, where each element corresponds to the relative energy value when the vibration of the candidate component propagates to the corresponding sensor location. Typically, the spatial decay mode is normalized to represent a relative decay distribution rather than an absolute amplitude.

[0101] In one feasible implementation, the spatial decay mode can be obtained through finite element simulation. In the finite element model of the propulsion system, a unit excitation is applied to the position of the candidate component, the vibration response amplitude at each sensor position is calculated, and the spatial decay mode of the candidate component is formed by normalizing the response amplitude at each sensor position.

[0102] In step S402, spatial matching degree refers to the similarity measure used to quantify the consistency between the actual spatial distribution of the candidate components of abnormal noise and the theoretical spatial decay mode of the candidate components.

[0103] Since the spatial weight vector of the candidate component of the abnormal noise reflects the actual energy distribution of the abnormal noise component on each sensor channel, if a candidate component is the real source of the abnormal noise, the actual spatial distribution of the abnormal noise should match the theoretical spatial attenuation mode of the candidate component.

[0104] By calculating the similarity between the spatial distribution characteristics of the candidate components of abnormal noise and the spatial attenuation mode of each candidate component, the spatial matching degree of each candidate component can be obtained.

[0105] In one feasible implementation, cosine similarity can be used to calculate spatial matching degree. The cosine value of the angle between two vectors is calculated. The closer the cosine value is to 1, the more consistent the directions of the two vectors are, and the higher the spatial matching degree.

[0106] In step S403, the spatial evidence strength characterizes the sufficiency of evidence supporting a candidate component as the source of abnormal noise based on spatial distribution information. The higher the spatial evidence strength, the more likely the candidate component is to be the true source of abnormal noise from the perspective of spatial distribution.

[0107] In one feasible implementation, the spatial matching degree of each candidate component can be directly used as the spatial evidence strength of that candidate component. The candidate component with the highest spatial matching degree corresponds to the strongest spatial evidence. All candidate components are sorted according to the spatial matching degree, and the sorting result is used as the first matching result.

[0108] It should be noted that when there are multiple abnormal noise candidate components in the propulsion system, steps S402 and S403 need to be executed for each abnormal noise candidate component to independently determine the corresponding first matching result for each abnormal noise candidate component, thereby achieving simultaneous localization of multiple abnormal noise sources.

[0109] Optionally, when obtaining the spatial attenuation mode, the frequency or order correlation of the vibration signal can be considered, and different spatial attenuation modes can be established for different frequency or order ranges. Since the transmission characteristics of the structure are related to the excitation frequency, the propagation attenuation laws of high-frequency vibration and low-frequency vibration may differ. When calculating the spatial matching degree, the corresponding spatial attenuation mode can be selected for matching based on the main frequency or order range of the candidate components of the abnormal noise, which can improve the accuracy of matching.

[0110] By employing the embodiments of this application, and characterizing the energy propagation attenuation law of each candidate component to each sensor through the spatial attenuation mode, a theoretical mapping relationship between the candidate component's location and sensor observations can be established, providing a physical basis for locating abnormal noise sources. Furthermore, by calculating the similarity between the spatial distribution characteristics of the abnormal noise candidate components and the spatial attenuation modes of each candidate component, the spatial matching degree can be obtained, enabling a quantitative assessment of the probability of each candidate component being an abnormal noise source, thus realizing the ranking and screening of candidate components based on spatial evidence.

[0111] Based on the above embodiments, as a feasible embodiment, step S401 may further include the following steps.

[0112] Step S501: Obtain the installation position of each candidate component in the propulsion system and the arrangement position of the corresponding sensors in each sensor channel; Step S502: Based on the structural propagation path of the propulsion system, calculate the structural propagation distance between each candidate component and the corresponding sensor of each sensor channel; Step S503: Based on the structural propagation distance, calculate the attenuation weight from each candidate component to each sensor channel according to the preset distance attenuation function; Step S504: Normalize the attenuation weights to obtain the spatial attenuation mode corresponding to each candidate component. Among them, the structural propagation distance characterizes the effective path length of the vibration signal propagating from the candidate component to the sensor through the structural components of the propulsion system; the distance attenuation function characterizes the attenuation weight as being negatively correlated with the structural propagation distance.

[0113] In step S502, the structural propagation distance refers to the effective path length of the vibration signal propagating from the candidate component to the sensor through the structural components of the propulsion system. It can be understood as the equivalent distance after considering the actual propagation path, which is different from the straight-line distance between the candidate component and the sensor.

[0114] Since vibration signals in the propulsion system mainly propagate through structural components, vibration energy needs to be transferred from the vibration source to the sensor location along the structural connection path. Therefore, it is necessary to calculate the structural propagation distance based on the structural propagation path of the propulsion system.

[0115] In one feasible implementation, a structural propagation path graph can be established based on the structural connection relationship of the propulsion system. The candidate component positions and sensor positions are treated as nodes in the graph, and the structural connection relationship is treated as edges in the graph. The weight of the edge is the length or equivalent length of the corresponding structural component. The shortest path length from the candidate component node to the sensor node is calculated using the shortest path algorithm in graph theory, and this shortest path length is used as the structural propagation distance.

[0116] In step S503, the distance attenuation function is a mathematical function that describes the negative correlation between the attenuation weight and the structural propagation distance. The attenuation weight represents the proportion of vibration energy retained during the propagation from the candidate component to the sensor. The greater the structural propagation distance, the more vibration energy is attenuated, and the smaller the attenuation weight. Therefore, the distance attenuation function indicates that the attenuation weight is negatively correlated with the structural propagation distance.

[0117] In one feasible implementation, an exponential decay function can be used as the distance decay function. The decay weight is equal to an exponential function with the natural constant as the base. The negative exponent is the product of the decay coefficient and the propagation distance of the structure. The positive decay coefficient is used to control the decay rate. The larger the decay coefficient, the faster the vibration energy decays with distance.

[0118] In step S504, the attenuation weights are normalized to eliminate the influence of absolute amplitude and highlight the spatial distribution pattern. The vector obtained after normalization is the spatial attenuation pattern corresponding to the candidate component, which characterizes the relative energy distribution of vibrations emitted from the candidate component at each sensor location.

[0119] Optionally, when calculating the propagation distance of the structure, the differences in propagation characteristics of different structural components can be considered, and different equivalent length coefficients can be set for different types of structural components. For example, the equivalent length coefficient of rigid connection structures is smaller, while the equivalent length coefficient of soft connection or damping connection structures is larger. A more accurate structural propagation distance can be obtained by weighted summation of the equivalent lengths of each structural component.

[0120] By employing the embodiments of this application, and by acquiring the positional information of candidate components and sensors, the spatial geometric basis required for locating abnormal noise sources can be established. Furthermore, by calculating the structural propagation distance based on the structural propagation path of the propagation system, the propagation characteristics of vibration signals in the actual structure can be accurately reflected, which is more consistent with the physical reality of vibration propagation than directly using geometric distance.

[0121] Based on the above embodiments, as a feasible embodiment, step S106 may further include the following steps.

[0122] Step S601: Obtain the mechanism compatibility information for each candidate component. The mechanism compatibility information represents the set of failure mechanism types that occur in the candidate component. Step S602: Determine the compatibility between the fault mechanism type corresponding to the candidate components of abnormal noise and the mechanism compatibility information of each candidate component, and obtain the mechanism matching degree of each candidate component; Step S603: Determine the second matching result based on the mechanism matching degree. The second matching result is used as the strength of mechanism evidence for each candidate component as an abnormal noise source.

[0123] In step S601, mechanism compatibility information refers to information describing the set of failure mechanism types that occur in candidate components.

[0124] Because different components have different structures, working principles and failure modes, each candidate component can only experience a specific type of failure related to its structure and function.

[0125] For example, bearing components may experience failures such as outer ring wear, inner ring pitting, and rolling element spalling, but will not experience gear meshing failures. Gear components may experience failures such as tooth surface wear, tooth cracks, and meshing impacts, but will not experience rolling element spalling failures.

[0126] Optionally, the mechanism compatibility information can be represented as a set or list of failure mechanism types. For each candidate component, its mechanism compatibility information contains identifiers of all possible failure mechanism types that the component may have.

[0127] In one feasible implementation, mechanism compatibility information can be established based on component failure modes recorded in the propulsion system's design data, failure mode and effects analysis documents, or maintenance manuals, and the known failure types of each candidate component can be organized into a set of failure mechanism types. In step S602, the degree of compatibility refers to the degree of matching between the fault mechanism type corresponding to the candidate component of abnormal noise and the mechanism compatibility information of the candidate component. The mechanism matching degree is a quantitative representation of the degree of compatibility.

[0128] By determining the compatibility between the fault mechanism type corresponding to the candidate components of abnormal noise and the mechanism compatibility information of each candidate component, the mechanism matching degree of each candidate component can be obtained.

[0129] In one feasible implementation, a binary judgment method can be used to calculate the mechanism matching degree. If the fault mechanism type corresponding to the candidate component of the abnormal noise belongs to the set of fault mechanism types included in the mechanism compatibility information of the candidate component, then the mechanism matching degree of the candidate component is assigned a value of 1 to indicate complete compatibility; otherwise, the mechanism matching degree is assigned a value of 0 to indicate incompatibility.

[0130] In step S603, the strength of mechanistic evidence characterizes the sufficiency of evidence supporting a candidate component as the source of abnormal noise based on fault mechanism information. The higher the strength of mechanistic evidence, the more likely the candidate component is to be the true source of abnormal noise from the perspective of fault mechanism.

[0131] In one feasible implementation, the mechanism matching degree of each candidate component can be directly used as the strength of the mechanism evidence for that candidate component. The candidate component with a mechanism matching degree of 1 has the strongest mechanism evidence, and the candidate component with a mechanism matching degree of 0 can be excluded from the candidate list. The mechanism matching degree of each candidate component is used as the second matching result.

[0132] It should be noted that the mechanism compatibility information needs to be dynamically updated based on the actual usage of the propulsion system and the failure evolution pattern. When a new failure type not included in the mechanism compatibility information is found in a candidate component, the failure type should be added to the mechanism compatibility information of that component to improve the accuracy of mechanism matching.

[0133] By employing the embodiments of this application, and by obtaining the mechanism compatibility information of each candidate component to characterize the set of fault types it experiences, a mapping relationship between components and fault mechanisms can be established, providing constraints at the fault mechanism dimension for locating abnormal noise sources. Furthermore, by determining the compatibility degree between the fault mechanism type corresponding to the abnormal noise and the mechanism compatibility information of the candidate components, the mechanism matching degree can be obtained. This allows for the screening and sorting of candidate components from a fault mechanism perspective, excluding components that are unlikely to produce that type of fault, thus narrowing the search range for abnormal noise sources.

[0134] Based on the above embodiments, as a feasible embodiment, step S107 may further include the following steps.

[0135] Step S701: Obtain the spatial matching degree of each candidate component in the first matching result; Step S702: Obtain the mechanism matching degree of each candidate component in the second matching result; Step S703: Perform weighted fusion of the spatial matching degree and mechanism matching degree of each candidate component to obtain the target matching degree of each candidate component; Step S704: Based on the target matching degree, at least one candidate component is determined as the abnormal noise source component; Step S705: Generate abnormal noise monitoring results for the abnormal noise source component.

[0136] In step S703, the target matching degree can be calculated using a linear weighted fusion method. The target matching degree is equal to the sum of the spatial matching degree multiplied by the spatial weight coefficient and the mechanism matching degree multiplied by the mechanism weight coefficient. The spatial weight coefficient and the mechanism weight coefficient are preset non-negative weight parameters.

[0137] Optionally, the values ​​of the spatial weight coefficient and the mechanism weight coefficient can be set according to the reliability of spatial information and mechanism information in the actual application scenario. When the sensor layout is relatively complete and the spatial attenuation mode is relatively accurate, a larger spatial weight coefficient can be set. When the fault mechanism knowledge is relatively complete and the mechanism compatibility information is relatively reliable, a larger mechanism weight coefficient can be set.

[0138] In step S704, the candidate component with the highest target matching degree can be determined as the abnormal noise source component. Assuming that each abnormal noise corresponds to a unique abnormal noise source, the candidate component with the highest target matching degree is selected as the final abnormal noise source location result.

[0139] In another feasible implementation, all candidate components whose target matching degree exceeds a preset threshold can be identified as abnormal noise source components. By setting a target matching degree threshold, all candidate components with high probability can be filtered out. The threshold can be set according to the confidence requirements of the positioning. A higher threshold corresponds to stricter screening conditions and a lower false detection rate, while a lower threshold corresponds to more lenient screening conditions and a lower false detection rate.

[0140] In step S705, the abnormal noise monitoring result refers to the output result containing the identification information of the abnormal noise source component and related auxiliary information. It can be understood as the final output product of the abnormal noise source localization analysis, used to present the localization conclusion and related detailed information of the abnormal noise source to the user or downstream system. The abnormal noise monitoring result may include, but is not limited to, the name or identifier of the abnormal noise source component, the target matching degree of the abnormal noise source component, the fault mechanism type corresponding to the abnormal noise source component, the time-frequency characteristic parameters of the abnormal noise candidate components, and the severity assessment of the abnormal noise.

[0141] In one feasible implementation, the abnormal noise monitoring results can be generated in a structured data format, including an abnormal noise source component identification field, a target matching degree field, a spatial matching degree field, a mechanism matching degree field, a fault type field, a detection timestamp field, etc., which facilitates data storage and subsequent processing.

[0142] Optionally, when generating abnormal noise monitoring results, severity assessment information of the abnormal noise can be attached. The severity of the abnormal noise can be judged based on the characteristics of the candidate components of the abnormal noise, such as energy level, duration, and development speed. The severity of the abnormal noise can be divided into minor, moderate, and severe levels. The severity assessment information can help users determine the urgency of the fault and the priority of maintenance.

[0143] Optionally, the abnormal noise monitoring results can be correlated with historical monitoring data to record information such as the changing trend of the target matching degree of the same abnormal noise source component at different times and the evolution law of abnormal noise characteristic parameters. Through longitudinal comparative analysis, the deterioration rate and remaining life of the abnormal noise source can be assessed, providing a basis for predictive maintenance decisions.

[0144] By employing the embodiments of this application, and by obtaining the spatial matching degree and mechanism matching degree from the first matching result and the second matching result, it is possible to integrate evidence information from two independent dimensions: spatial distribution and fault mechanism. Furthermore, by weighted fusion of the spatial matching degree and mechanism matching degree to obtain the target matching degree, it is possible to comprehensively utilize multi-dimensional information to improve the accuracy of abnormal noise source localization, avoid the limitations and uncertainties of single-dimensional information, and achieve the complementary advantages of multi-source information.

[0145] Figure 2 This application provides a schematic diagram of the structure of a ship propulsion system abnormal noise monitoring system, as shown in the embodiment of the present application. Figure 2As shown, the ship propulsion system abnormal noise monitoring system includes: The data acquisition module is used to acquire monitoring data of the propulsion system in the ship. The monitoring data includes vibration signals of multiple components and speed signals of the propulsion system. The order conversion module is used to convert the vibration signal to the order domain based on the rotation speed signal to obtain a multi-channel order spectrum. The component decomposition module is used to perform source component decomposition on multi-channel order spectra based on target constraints to obtain multiple source components and the spatial distribution characteristics of each source component. The target constraints are used to constrain the temporal characteristics, spatial characteristics, and inter-channel consistency during the source component decomposition process. The spatial distribution characteristics characterize the weight distribution of each source component on each sensor channel. An abnormal noise identification module is used to identify candidate components of abnormal noise from multiple source components and determine the corresponding fault mechanism type based on the spectral characteristics of the candidate components of abnormal noise. The spatial matching module is used to spatially match the spatial distribution characteristics of the candidate components of abnormal noise with the spatial attenuation mode of each candidate component to obtain the first matching result. The mechanism matching module is used to perform mechanism matching between the failure mechanism type of the abnormal noise candidate components and the mechanism compatibility of each candidate component to obtain a second matching result; The result generation module is used to obtain the abnormal noise monitoring results of the propulsion system based on the first matching result and the second matching result.

[0146] According to an embodiment of this application, the component decomposition module is further configured to establish a source component decomposition model for a multi-channel order spectrum. In the source component decomposition model, the multi-channel order spectrum is represented as a linear combination of multiple source components, wherein the parameters to be solved for each source component include a spectral basis, a temporal activation coefficient, and a spatial weight vector. An objective function containing objective constraints is constructed based on the source component decomposition model. The objective function is optimized and solved to obtain the values ​​of the spectral basis, temporal activation coefficient, and spatial weight vector for each source component. The spatial distribution characteristics of each source component are determined based on the solved spatial weight vector. The spectral basis characterizes the spectral characteristics of the source component in the order domain, the temporal activation coefficient characterizes the activation intensity of the source component over time, and the spatial weight vector characterizes the energy distribution of the source component on each sensor channel.

[0147] According to an embodiment of this application, the abnormal noise identification module is further configured to calculate an abnormal score for each source component, wherein the abnormal score is calculated based on at least one of the source component’s temporal sparsity, spectral impact, and deviation from the normal background; source components with abnormal scores greater than or equal to a preset threshold are identified as abnormal noise candidate components; and the spectral characteristics of the abnormal noise candidate components are matched with a preset fault mechanism feature library to obtain the fault mechanism type corresponding to the abnormal noise candidate components.

[0148] According to an embodiment of this application, the spatial matching module is further configured to obtain the spatial attenuation mode corresponding to each candidate component, wherein the spatial attenuation mode characterizes the energy propagation attenuation law from the candidate component to each sensor channel; calculate the similarity between the spatial distribution characteristics of the abnormal noise candidate components and the spatial attenuation mode of each candidate component to obtain the spatial matching degree of each candidate component; determine the first matching result based on the spatial matching degree, wherein the first matching result is used as the spatial evidence strength of the abnormal noise source of each candidate component.

[0149] According to an embodiment of this application, the spatial matching module is further configured to obtain the installation position of each candidate component in the propulsion system and the arrangement position of the corresponding sensor in each sensor channel; calculate the structural propagation distance between each candidate component and the corresponding sensor in each sensor channel based on the structural propagation path of the propulsion system; calculate the attenuation weight of each candidate component to each sensor channel according to a preset distance attenuation function based on the structural propagation distance; and normalize the attenuation weight to obtain the spatial attenuation mode corresponding to each candidate component; wherein, the structural propagation distance characterizes the effective path length of the vibration signal propagating from the candidate component to the sensor through the structural components of the propulsion system; and the distance attenuation function characterizes the negative correlation between the attenuation weight and the structural propagation distance.

[0150] According to an embodiment of this application, the mechanism matching module is further configured to obtain mechanism compatibility information for each candidate component, wherein the mechanism compatibility information characterizes the set of fault mechanism types that occur in the candidate component; determine the degree of compatibility between the fault mechanism type corresponding to the abnormal noise candidate component and the mechanism compatibility information of each candidate component, thereby obtaining the mechanism matching degree of each candidate component; and determine a second matching result based on the mechanism matching degree, wherein the second matching result is used as the strength of mechanism evidence that each candidate component is an abnormal noise source.

[0151] According to an embodiment of this application, the result generation module is further configured to obtain the spatial matching degree of each candidate component in the first matching result; obtain the mechanistic matching degree of each candidate component in the second matching result; perform weighted fusion on the spatial matching degree and mechanistic matching degree of each candidate component to obtain the target matching degree of each candidate component; determine at least one candidate component as an abnormal noise source component based on the target matching degree; and generate abnormal noise monitoring results of the abnormal noise source component.

[0152] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application, such as... Figure 3As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340. The processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions from the memory 330 to execute a method for monitoring abnormal noises in the ship's propulsion system.

[0153] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0154] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the ship propulsion system abnormal noise monitoring method provided by the above methods.

[0155] In another aspect, this application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the ship propulsion system abnormal noise monitoring methods provided by the above methods.

[0156] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0157] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for monitoring abnormal noises in a ship propulsion system, characterized in that, include: Acquire monitoring data of the propulsion system in a ship, the monitoring data including vibration signals of multiple components and rotational speed signals of the propulsion system; Based on the rotational speed signal, the vibration signal is converted to the order domain to obtain a multi-channel order spectrum; Based on the objective constraint, the multi-channel order spectrum is decomposed into source components to obtain multiple source components and the spatial distribution characteristics of each source component. The target constraint is used to match the activation modes of the decomposed source components on different sensor channels with the coherent observations between channels; the spatial distribution feature characterizes the weight distribution on the corresponding sensor channel for each source component; Among the multiple source components, candidate components for abnormal noise and their corresponding fault mechanism types are determined; The spatial distribution characteristics of the candidate components of the abnormal noise are spatially matched with the spatial attenuation mode of each candidate component to obtain the first matching result; The failure mechanism type of the candidate components of the abnormal noise is matched with the mechanism compatibility of each candidate component to obtain a second matching result; The abnormal noise monitoring results of the propulsion system are obtained based on the first matching result and the second matching result.

2. The method according to claim 1, characterized in that, The source component decomposition of the multi-channel order spectrum based on target constraints to obtain multiple source components and the spatial distribution characteristics of each source component includes: A source component decomposition model is established for the multi-channel order spectrum. In the source component decomposition model, the multi-channel order spectrum is represented as a linear combination of multiple source components. The parameters to be solved for each source component include spectral basis, temporal activation coefficient and spatial weight vector. Construct an objective function containing the objective constraints based on the source component decomposition model; The objective function is optimized and solved to obtain the values ​​of the spectral basis, temporal activation coefficients, and spatial weight vectors for each source component; The spatial distribution characteristics of each source component are determined based on the spatial weight vector obtained from the solution. Wherein, the spectral basis characterizes the spectral characteristics of the source component in the order domain, the time activation coefficient characterizes the activation intensity of the source component over time, and the spatial weight vector characterizes the energy distribution of the source component on each sensor channel.

3. The method according to claim 1 or 2, characterized in that, The objective constraints include at least one or a combination of more: Temporal sparsity constraints are used to make the anomalous components exhibit sparse characteristics in the time dimension. Spatial continuity constraints are used to ensure that the spatial weights on adjacent sensor channels change smoothly. Propagation consistency constraints are used to ensure that the joint activation characteristics of each source component across multiple sensor channels are consistent with the coherence characteristics observed between channels.

4. The method according to claim 1, characterized in that, The step of determining the candidate components of abnormal noise and their corresponding fault mechanism types among the multiple source components includes: An anomaly score is calculated for each source component, and the anomaly score is calculated based on at least one of the source component’s temporal sparsity, spectral impact, and deviation from the normal background. Source components with abnormal scores greater than or equal to a preset threshold are identified as candidate components for abnormal sounds. The spectral characteristics of the abnormal noise candidate components are matched with a preset fault mechanism feature library to obtain the fault mechanism type corresponding to the abnormal noise candidate components.

5. The method according to claim 1, characterized in that, The step of spatially matching the spatial distribution characteristics of the candidate noise components with the spatial attenuation mode of each candidate component to obtain a first matching result includes: Obtain the spatial attenuation mode corresponding to each candidate component, wherein the spatial attenuation mode characterizes the energy propagation attenuation law from the candidate component to each sensor channel; Calculate the similarity between the spatial distribution characteristics of the candidate components of the abnormal noise and the spatial attenuation mode of each candidate component to obtain the spatial matching degree of each candidate component; The first matching result is determined based on the spatial matching degree, and the first matching result is used as the spatial evidence strength of the abnormal noise source for each candidate component.

6. The method according to claim 5, characterized in that, The step of obtaining the spatial attenuation mode corresponding to each candidate component includes: Obtain the installation position of each candidate component in the propulsion system and the arrangement position of the corresponding sensors in each sensor channel; Based on the structural propagation path of the propulsion system, calculate the structural propagation distance between each candidate component and the corresponding sensor in each sensor channel; Based on the propagation distance of the structure, the attenuation weight from each candidate component to each sensor channel is calculated according to a preset distance attenuation function; The attenuation weights are normalized to obtain the spatial attenuation mode corresponding to each candidate component; Wherein, the structural propagation distance characterizes the effective path length of the vibration signal propagating from the candidate component to the sensor through the structural components of the propulsion system; the distance attenuation function characterizes the attenuation weight as being negatively correlated with the structural propagation distance.

7. The method according to claim 1, characterized in that, The step of matching the failure mechanism type of the candidate noise components with the mechanism compatibility of each candidate component to obtain a second matching result includes: Obtain the mechanism compatibility information for each candidate component, wherein the mechanism compatibility information characterizes the set of failure mechanism types that occur in the candidate component; Determine the compatibility between the fault mechanism type corresponding to the candidate components of the abnormal noise and the mechanism compatibility information of each candidate component to obtain the mechanism matching degree of each candidate component; The second matching result is determined based on the mechanism matching degree, and the second matching result is used as the strength of mechanism evidence for each candidate component as an abnormal noise source.

8. The method according to claim 1, characterized in that, The step of obtaining the abnormal noise monitoring results of the propulsion system based on the first matching result and the second matching result includes: Obtain the spatial matching degree of each candidate component in the first matching result; Obtain the mechanistic matching degree of each candidate component in the second matching result; The spatial matching degree and mechanistic matching degree of each candidate component are weighted and fused to obtain the target matching degree of each candidate component; Based on the target matching degree, at least one candidate component is identified as the source of the abnormal noise. Generate abnormal noise monitoring results for the aforementioned abnormal noise source component.

9. A monitoring system for abnormal noise in a ship propulsion system, characterized in that, include: The data acquisition module is used to acquire monitoring data of the propulsion system in the ship, including vibration signals of multiple components and rotational speed signals of the propulsion system. The order conversion module is used to convert the vibration signal to the order domain based on the rotation speed signal to obtain a multi-channel order spectrum. The component decomposition module is used to perform source component decomposition on the multi-channel order spectrum based on target constraints to obtain multiple source components and the spatial distribution characteristics of each source component. The target constraint is used to constrain the temporal characteristics, spatial characteristics, and inter-channel consistency during the source component decomposition process; the spatial distribution feature characterizes the weight distribution of each source component on each sensor channel. An abnormal noise identification module is used to identify candidate components for abnormal noise among the multiple source components and determine the corresponding fault mechanism type based on the spectral characteristics of the candidate components for abnormal noise. The spatial matching module is used to spatially match the spatial distribution characteristics of the candidate components of the abnormal noise with the spatial attenuation mode of each candidate component to obtain a first matching result. The mechanism matching module is used to perform mechanism matching between the fault mechanism type of the candidate components of abnormal noise and the mechanism compatibility of each candidate component to obtain a second matching result; The result generation module is used to obtain the abnormal noise monitoring results of the propulsion system based on the first matching result and the second matching result.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for monitoring abnormal noises in a ship propulsion system as described in any one of claims 1-8.

Images