A system and method for failure analysis of marine equipment

Abstract

This invention discloses a fault analysis system and method for ship equipment. The system includes: a multi-source sensor synchronization module configured to synchronize the operating data of multiple devices through a hardware phase-locked loop mechanism; a multi-physics modeling module coupled to the multi-source sensor synchronization module, configured to construct a virtual state field of the equipment based on the phase-synchronized operating data of the multiple devices, and generate dynamic parameters characterizing the temporal misalignment of the states of the multiple devices; an adaptive diagnostic analysis module, including a deep learning fault analysis model, wherein the implicit feature extraction parameters and fault evolution temporal modeling parameters of the deep learning fault analysis model are adaptively adjusted in real time according to the dynamic parameters to output system-level fault characterization parameters; and a predictive decision-making module configured to generate decision information including fault development trend prediction and preventive maintenance windows based on the system-level fault characterization parameters. This system enables high-precision correlation diagnosis and predictive maintenance decision-making for early-stage faults of multiple ship equipment.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for equipment, and in particular to a fault analysis system and method for marine equipment. Background Technology

[0002] In the field of ship operation and maintenance, traditional fault analysis methods generally rely on manual inspections or post-shutdown checks, making it difficult to perform real-time and accurate correlation diagnosis of the operating status of multiple devices. While some existing technologies attempt to introduce sensor networks and deep learning models for fault prediction, their data acquisition cannot fully consider the low-frequency vibration interference unique to ships, resulting in raw signals mixed with motion noise such as roll and pitch. Furthermore, fault diagnosis models typically employ fixed parameter structures, failing to dynamically adjust feature extraction and modeling strategies based on equipment state evolution. This leads to the neglect of latent fault characteristics (such as early pump wear and valve micro-jamming), and makes it difficult to effectively model the propagation patterns of coupled faults between multiple devices. Therefore, there is an urgent need for a fault analysis system that can integrate the characteristics of the ship's operating environment, achieve collaborative perception of multi-physics states, and drive adaptive optimization of the model through dynamic parameters, in order to overcome the bottlenecks of existing technologies in early warning capabilities and system-level diagnostic accuracy. Summary of the Invention

[0003] In view of this, the present invention proposes a fault analysis system and method for ship equipment, which can achieve high-precision correlation diagnosis and predictive maintenance decision-making for early-stage faults of multiple ship equipment. The present invention provides the following technical solution:

[0004] A fault analysis system for ship equipment includes:

[0005] A multi-source sensor synchronization module is deployed in ship equipment and configured to synchronize the operating data of multiple devices through a hardware phase-locked loop mechanism.

[0006] The multiphysics modeling module is coupled to the multi-source sensing synchronization module and is configured to construct a virtual state field of devices based on the phase-synchronized multi-device operating data, and generate dynamic parameters characterizing the temporal misalignment of the multi-device states.

[0007] The adaptive diagnostic analysis module includes a deep learning fault analysis model. The implicit feature extraction parameters and fault evolution time series modeling parameters of the deep learning fault analysis model are adaptively adjusted in real time according to the dynamic parameters to output system-level fault characterization parameters.

[0008] The predictive decision module, coupled to the adaptive diagnostic analysis module, is configured to generate decision information that includes fault development trend prediction and preventive maintenance windows based on the system-level fault characterization parameters.

[0009] Optionally, the multi-source sensing synchronization module includes:

[0010] The sensor deployment unit is configured to deploy vibration sensors, pressure sensors, temperature sensors, and flow sensors at nodes in the ship's propulsion system, generator set, hydraulic system, and cooling system.

[0011] A phase measurement unit, coupled to the sensing deployment unit, is configured to detect the phase shift of each sensor signal in the 0.1-2Hz low-frequency vibration band in real time.

[0012] A phase-locked control unit, coupled to the phase measurement unit, is configured to generate a hardware-level synchronous clock signal to dynamically compensate for the phase offset, so that the phase synchronization error of the multi-device operating data is no greater than ±0.1°.

[0013] An environment adaptive unit, coupled to the phase-locked control unit, is configured to dynamically adjust the synchronization accuracy threshold of the phase-locked control unit according to the ship's roll frequency and surge intensity.

[0014] Optionally, the multiphysics modeling module includes:

[0015] A multi-dimensional data fusion unit, coupled to the multi-source sensing synchronization module, is configured to integrate phase-synchronized vibration, pressure, temperature and flow time-series data to generate a multi-dimensional feature matrix of equipment status.

[0016] The virtual field construction unit is coupled to the multidimensional data fusion unit and is configured to construct a coupled virtual state field containing mechanical stress field, thermodynamic field and fluid pressure field based on the physical topology of ship equipment.

[0017] The dynamic parameter generation unit is coupled to the virtual field construction unit and is configured to calculate the distribution offset entropy value of the coupled virtual state field in the temporal dimension, as a dynamic parameter characterizing the temporal misalignment of multiple device states.

[0018] An environmental attenuation compensation unit, coupled to the dynamic parameter generation unit, is configured to dynamically attenuate the distribution offset entropy value based on seawater salinity, equipment exposure time, and ship vibration spectrum, so that the dynamic parameters conform to the physical evolution law of the ship's operating environment.

[0019] Optionally, the adaptive diagnostic analysis module includes:

[0020] The model initialization unit is configured as the infrastructure for loading a deep learning fault analysis model, the infrastructure including a fusion structure of a convolutional neural network and a long short-term memory network;

[0021] The ship characteristic parameter tuning unit is coupled to the multiphysics modeling module and the model initialization unit, and is configured to receive the dynamic parameters and adjust the center frequency of the convolution kernel of the convolutional neural network and the depth of the memory unit of the long short-term memory network in real time based on the characteristics of the ship's low-frequency vibration environment.

[0022] The cross-device feature fusion unit is coupled to the ship characteristic parameter tuning unit and is configured to extract the latent fault features of each device in the 0.1-2Hz vibration frequency band and model the spatiotemporal evolution path of multi-device coupled faults in the ship's physical connection chain.

[0023] The system characterization generation unit is coupled to the cross-device feature fusion unit and is configured to fuse the latent fault features with the spatiotemporal evolution path to generate system-level fault characterization parameters that characterize the severity and propagation trend of system-level faults.

[0024] Optionally, the predictive decision module includes:

[0025] The trend quantification unit, coupled to the adaptive diagnostic analysis module, is configured to construct a fault physical evolution model based on the system-level fault characterization parameters and output the remaining safe operating time of the equipment and the fault severity growth rate.

[0026] The ship operation plan fusion unit is coupled to the ship automatic identification system and the trend quantification unit, and is configured to dynamically correct the rate of increase of the fault severity by combining ship berthing plan, navigation stage and sea state data.

[0027] The decision generation unit, coupled to the ship operation plan fusion unit, is configured to generate structured decision information including a fault location topology map, root cause probability distribution, cross-equipment impact propagation chain, and preventive maintenance time window.

[0028] The accuracy feedback unit, coupled to the decision generation unit and the multiphysics modeling module, is configured to dynamically calibrate the parameters of the fault physical evolution model based on actual maintenance verification results.

[0029] Optionally, the calculation process of the dynamic parameters includes:

[0030] A time-series distribution matrix characterizing the multi-physical field state of ship equipment is constructed based on the multi-device operation data after phase synchronization. The multi-physical field includes mechanical vibration field, fluid pressure field and thermodynamic field.

[0031] Obtain the equipment health status baseline distribution and environmental degradation factor, wherein the environmental degradation factor is dynamically updated based on the ship's seawater corrosion intensity and low-frequency vibration spectrum;

[0032] The temporal misalignment between the current temporal distribution matrix and the healthy state baseline distribution after environmental attenuation correction is calculated using the information entropy algorithm, generating dynamic parameters characterizing the degree of system-level state anomalies.

[0033] Optionally, the ship characteristic parameter tuning unit performs the following adaptive adjustment strategy:

[0034] The dynamic parameters are input into a preset mapping function to calculate the center frequency adjustment of the convolutional neural network kernel, so that the convolutional kernel focuses on the characteristic fault frequency of the ship equipment in a low-frequency vibration environment.

[0035] Calculate the slope of change of the dynamic parameters in the temporal dimension, determine the memory unit depth of the long short-term memory network based on the slope of change, and expand the memory unit depth when the slope of change increases to enhance the ability to capture the evolution law of progressive failures.

[0036] Based on the adjusted convolution kernel and memory unit depth, the deep learning fault analysis model is updated to improve the accuracy of extracting latent fault features of ship equipment.

[0037] Optionally, the environmental degradation compensation unit and the ship characteristic parameter tuning unit form a closed-loop optimization mechanism:

[0038] The environmental attenuation compensation unit compensates for the dynamic parameters based on the ship's operating environment parameters and then outputs the compensation to the ship characteristic parameter tuning unit.

[0039] The ship characteristic parameter tuning unit adjusts the deep learning model parameters based on the compensated dynamic parameters and generates system-level fault characterization parameters.

[0040] The accuracy feedback unit feeds back the actual maintenance verification results to the environmental attenuation compensation unit to correct the environmental attenuation factor.

[0041] This invention further discloses a fault analysis method for ship equipment, comprising:

[0042] Phase synchronization of operating data from multiple devices is achieved through a hardware phase-locked loop mechanism.

[0043] A virtual state field for devices is constructed based on the multi-device operation data after phase synchronization, and dynamic parameters characterizing the temporal misscheduling of the multi-device states are generated.

[0044] The implicit feature extraction parameters and fault evolution time series modeling parameters of the fault analysis model are adaptively adjusted in real time according to the dynamic parameters to output system-level fault characterization parameters.

[0045] Based on the system-level fault characterization parameters, decision information including fault development trend prediction and preventive maintenance window is generated.

[0046] The present invention further discloses 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 program to implement the above-described method.

[0047] According to the technical solution of this invention, by constructing a closed-loop analysis system with dynamic parameters as the core coupling factor, the real-time performance and systematic nature of fault diagnosis for multiple ship equipment are significantly improved. Specifically, the multi-source sensor synchronization module utilizes a hardware phase-locked loop mechanism to achieve high-precision phase synchronization in a low-frequency vibration environment of 0.1–2Hz, effectively suppressing the interference of motion noise such as ship roll and pitch on the original signal, providing a clean data foundation for subsequent analysis. Based on this, the multiphysics modeling module integrates mechanical vibration fields, fluid pressure fields, and thermodynamic fields to generate dynamic parameters characterizing the temporal misalignment of multiple equipment states. These parameters not only reflect anomalies in individual equipment but also capture coupling imbalances caused by physical connections between equipment. The adaptive diagnostic analysis module adjusts the implicit feature extraction and evolutionary modeling capabilities of the deep learning model in real time based on these dynamic parameters, enabling the system to focus on the key fault frequencies under ship-specific operating conditions and accurately track the evolution path of progressive faults. Finally, the predictive decision-making module generates decision information containing development trends and maintenance windows, thereby transforming fault identification from passive response to early warning, significantly reducing the misjudgment rate and operation and maintenance costs. Attached Figure Description

[0048] For illustrative and not limiting purposes, the present invention will now be described in conjunction with embodiments and accompanying drawings, wherein:

[0049] Figure 1 This is a schematic diagram of the component modules of the ship equipment fault analysis system in an embodiment of the present invention;

[0050] Figure 2 This is a flowchart illustrating the fault analysis method for ship equipment in an embodiment of the present invention.

[0051] Figure 3 This is a schematic diagram of the structure of the electronic device in an embodiment of the present invention. Detailed Implementation

[0052] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0053] It should be noted that, where there is no conflict, the embodiments and features of the embodiments in this application can be combined with each other. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0054] refer to Figure 1 This embodiment discloses a fault analysis system for ship equipment, including a multi-source sensing synchronization module 21, a multiphysics modeling module 22, an adaptive diagnostic analysis module 23, and a predictive decision-making module 24, which are described in detail below:

[0055] The multi-source sensor synchronization module 21 is deployed in the ship's equipment and is configured to synchronize the operating data of multiple devices through a hardware phase-locked loop mechanism.

[0056] In this embodiment, the multi-source sensor synchronization module 21 is deployed at the operational nodes of key ship equipment, exemplarily including the propulsion shaft bearing housing, generator stator housing, hydraulic pump inlet and outlet pipelines, and the surface of the cooling system heat exchanger. The multi-source sensor synchronization module 21 achieves high-precision phase synchronization of operating data from multiple devices through a hardware phase-locked loop mechanism, thereby eliminating interference from low-frequency vibrations of 0.1–2Hz caused by roll, pitch, and other movements during ship navigation on the original sensor signals.

[0057] Specifically, the multi-source sensor synchronization module includes a sensor deployment unit 211, a phase measurement unit 212, a phase-locked control unit 213, and an environment adaptation unit 214. The sensor deployment unit 211 is equipped with vibration sensors, pressure sensors, temperature sensors, flow sensors, and ship motion attitude sensors, which are respectively installed at the aforementioned device nodes to synchronously acquire raw signals at a sampling rate of 100Hz. ,in The total number of sensors, It is a continuous-time variable.

[0058] Furthermore, due to the periodic low-frequency disturbances generated when the ship navigates in waves, the physical differences in the physical positions of each sensor cause phase shifts in the acquired signals. Therefore, the phase measurement unit 212 detects the phase shift of each channel signal in real time within the 0.1–2Hz frequency band. Its calculation is based on the dominant frequency phase extracted by the short-time Fourier transform (STFT): ,in, This represents the Fourier transform operator. This indicates taking the complex phase angle. For frequency variables.

[0059] Phase-locked control unit 213 receives all A hardware-level synchronous clock signal is generated, and dynamic phase compensation is applied to each acquisition channel through a digitally controlled oscillator (DCO) to ensure that the compensated signal satisfies the following: ,in, The average phase offset is given by the formula. The 0.1° accuracy threshold is derived from the critical point of the first-order bending mode resonance of the ship's piping system. Simulations show that when the phase error exceeds this value, the stress field reconstruction error will increase sharply, leading to the failure of subsequent fault feature extraction.

[0060] Furthermore, the environment adaptive unit 214 is coupled to the ship's motion attitude sensor to acquire the roll frequency in real time. With surge intensity Based on this, the synchronization accuracy threshold of the phase-locked control unit 213 is dynamically adjusted. : This setting helps avoid accidents in high sea states, such as... or The vibration coupling between equipment is enhanced, and the characteristic frequency aliasing caused by small phase mismatch is also improved. In actual ship tests, the signal-to-noise ratio of the pump body wear characteristic frequency can be improved by 13dB in sea state 6 using a 0.05° threshold, which significantly improves the detectability of latent faults.

[0061] It is evident that by employing a hardware phase-locked loop mechanism and an environmental adaptive strategy, strict temporal consistency of multi-source data can be ensured under the unique low-frequency vibration environment of ships, providing high-fidelity input for subsequent multiphysics modeling and adaptive diagnosis. This effectively solves the problems of feature distortion and misjudgment caused by signal loss of synchronization in traditional methods.

[0062] The multiphysics modeling module 22 is coupled to the multi-source sensing synchronization module and is configured to construct a virtual state field of devices based on the multi-device operation data after phase synchronization, and generate dynamic parameters characterizing the temporal misscheduling of the multi-device state.

[0063] In this embodiment, the multiphysics modeling module 22 is coupled to the multi-source sensing synchronization module 21 to receive the phase-synchronized multi-device operating data output by the module. ,in Indicates the first The timing signals of the sensors corresponding to each device are obtained after phase compensation. The multiphysics modeling module 22, based on the physical topology of the ship's equipment, constructs a coupled virtual state field that integrates mechanical vibration, fluid pressure, and thermodynamic fields, and generates dynamic parameters characterizing system-level state timing misalignment. Specifically, the multiphysics modeling module 22 includes a multidimensional data fusion unit 221, a virtual field construction unit 222, a dynamic parameter generation unit 223, and an environmental attenuation compensation unit 224. The multidimensional data fusion unit 221 first... Grouping and processing: Vibration signals Pressure signals Temperature signals and flow signals Energy distribution features are extracted using wavelet packet decomposition to generate a multidimensional feature matrix of the device state. ,in This represents the number of device nodes. For feature dimensions.

[0064] Virtual field construction unit 222 is based on the ship's equipment compartment layout diagram and material properties, and will... Mapped to a virtual state field with three fields coupled:

[0065] Mechanical vibration field : Indicates spatial location Equivalent stress at the location;

[0066] Fluid pressure field : Indicates the static pressure distribution of fluid within the pipeline;

[0067] Thermodynamic field : Represents the temperature gradient field on the surface of the device.

[0068] These three elements together constitute a coupled virtual state field. .

[0069] The dynamic parameter generation unit 223 calculates the temporal misalignment between the current state field and the healthy baseline state field. Let the environmentally calibrated baseline distribution under healthy conditions be... ,in Given the initial environmental attenuation factor, the dynamic parameters are... Defined as: ,in, This is the real-time environmental attenuation factor. Let the logarithm function be base 2, summing over all device node positions. This design effectively quantifies the distribution shift of multiphysics in temporal evolution; for example, when the device is healthy. Therefore When early anomalies occur (such as micro-leakage in pipelines), asynchronous shifts occur in the multi-field distribution. Significantly increased.

[0070] Environmental degradation compensation unit 224 dynamically updates according to the ship's operating environment. Let the salinity of seawater be... The cumulative exposure time of the device is The energy of the main frequency of the ship's vibration spectrum is ,but: ,in , and This represents the material-related attenuation coefficient. This exponential attenuation model originates from accelerated aging experiments on actual ships. Tests have shown that in a salt spray environment, the damping characteristics of cast iron pipelines increase with... and The increase in amplitude and the decrease in attenuation lead to a reduction in the stress response amplitude under the same fault condition. Without compensation, this will result in… underestimate.

[0071] In summary, this implementation constructs a virtual state field with three coupled fields and introduces dynamic parameters that are adaptive to the environment. This design not only captures anomalies in individual devices but also reflects the collaborative imbalance caused by the physical connection of multiple devices. Real-world ship verification shows that on bulk carriers, this design enables early identification of micro-leakage faults to be reduced to 82 hours, and maintains a diagnostic accuracy rate of over 94% even after 12 months of salt spray exposure. This effectively solves the problem of condition assessment bias caused by traditional solutions neglecting environmental degradation and multi-field coupling.

[0072] The adaptive diagnostic analysis module 23 includes a deep learning fault analysis model. The implicit feature extraction parameters and fault evolution time series modeling parameters of the deep learning fault analysis model are adaptively adjusted in real time according to the dynamic parameters to output system-level fault characterization parameters.

[0073] The adaptive diagnostic analysis module 23 is coupled to the multiphysics modeling module 22 and receives the dynamic parameters output by it. The system adjusts the internal structure of the deep learning fault analysis model in real time based on these parameters to accurately extract the latent fault characteristics of ship equipment in complex marine environments and model the evolution of multi-factor coupled faults. The adaptive diagnostic analysis module 23 includes a model initialization unit 231, a ship characteristic parameter tuning unit 232, a cross-equipment feature fusion unit 233, and a system characterization generation unit 234.

[0074] Specifically, model initialization unit 231 loads a pre-trained deep learning fault analysis model, whose basic architecture is a fusion structure of convolutional neural network (CNN) and long short-term memory network (LSTM). The input is a multi-dimensional feature matrix after phase synchronization. Three-dimensional tensors formed by stacking along the time dimension ,in, The time window length, Let be the total number of device nodes, and d be the multiphysics feature dimension of each node. The output is the system-level fault characterization parameters. ,in, These are the dimensions used to represent the severity of the fault, the propagation trend, etc.

[0075] Ship characteristic parameter tuning unit 232 according to dynamic parameters Real-time adjustment of key hyperparameters of the model. First, calculate the adjustment amount of the convolution kernel center frequency. : ,in For the ship's vibration intensity coefficient, The activation threshold for the characteristic frequency. The function is the hyperbolic tangent. The adjusted center frequency of the convolution kernel is... ,in This serves as a reference frequency for typical fault characteristics of the equipment under health conditions. Through this design, when... If the frequency is too high, it indicates an early anomaly in the system, and the CNN should be focused on harmonic components near the reference frequency of typical fault characteristics; otherwise, the frequency band should be suppressed to avoid false triggering of roll interference. Secondly, the slope of the dynamic parameters' change in the time dimension is calculated. : ,in This represents the sliding window step size. Based on... Determine the depth of LSTM memory cells : Therefore, it can be seen that a low slope corresponds to steady-state fluctuations, and a shallow LSTM is sufficient to model short-term dependencies; a high slope reflects progressive faults (such as the exponential growth of micro-leakage apertures), requiring a deep LSTM to capture long-term evolution patterns. In actual tests on bulk carriers, this mechanism reduced the leakage rate prediction error from ±18.3 hours to ±4.1 hours.

[0076] The cross-device feature fusion unit 233 receives the adjusted CNN-LSTM model output and extracts the latent fault feature vectors of each device node in the 0.1–2Hz vibration frequency band. And a spatiotemporal evolution diagram is constructed based on the physical connection topology of ships. , where the set of nodes For N devices, Let be the set of edges, and the edges Indicates device With equipment At any moment State interaction strength, and use edge weights Quantify this intensity, that is Furthermore, a spatiotemporal attention mechanism is used to aggregate neighborhood information and generate fused features. .

[0077] System characterization generation unit 234 will The parameters are concatenated and mapped to system-level fault characterization parameters via a fully connected layer. ,in Characterizes the severity of system failures. Characterize the trend of cross-device influence propagation.

[0078] In summary, this implementation method utilizes dynamic parameters. The adaptive reconstruction of CNN and LSTM enables the model to focus on ship-specific fault frequencies and accurately track evolution patterns. Real-world ship validation shows that in coupled fault scenarios, this design improves root cause localization accuracy from 60.4% to 89.7% and advances the early warning time for micro-leaks to 82 hours, effectively solving the problems of inaccurate feature extraction and temporal modeling failure of traditional fixed-parameter models in dynamic ship environments.

[0079] The predictive decision module 24, coupled to the adaptive diagnostic analysis module 23, is configured to generate decision information including fault development trend prediction and preventive maintenance windows based on the system-level fault characterization parameters. The predictive decision module 24, coupled to the adaptive diagnostic analysis module 23, receives the system-level fault characterization parameters output by the module. It integrates ship operation plans and environmental data to generate structured decision information that includes fault development trend prediction and preventive maintenance windows. The predictive decision module 24 includes a trend quantification unit 241, a ship operation plan fusion unit 242, a decision generation unit 243, and an accuracy feedback unit 244.

[0080] Specifically, the trend quantification unit 241 is based on Construct a physical evolution model of the failure. Assume the remaining safe operating time (RUL) of the equipment is... The rate of increase in fault severity is The two satisfy a differential relationship:

[0081] , ,in This is the preset fault threshold. To solve this integral, the exponential decay assumption is used. Modeling: ,in The initial growth rate, This is the propagation sensitivity coefficient. Since multi-equipment coupled failures in ships often exhibit accelerated deterioration characteristics, such as micro-leakage leading to oil temperature rise, which in turn exacerbates valve group wear, the exponential form of this design can effectively fit such nonlinear evolution patterns.

[0082] The vessel operation planning fusion unit 242 is coupled to the Automatic Identification System (AIS) to obtain berthing plans in real time. (Arrival time at the next port), sailing phase (Transoceanic / Nearshore / Anchorage) and Sea State Data (e.g., wave height, wind speed). Based on this, the rate of increase in fault severity is dynamically adjusted: Among them, the correction factor Defined as: This avoids the equipment being subjected to additional dynamic loads under high sea states, which could accelerate fault evolution; while when the system is at rest, the load is low, and the evolution is slower. By introducing an operational plan, the predicted results are closely coupled with the actual operation and maintenance scenario, avoiding giving unrealistic maintenance suggestions during transoceanic voyages.

[0083] Decision generation unit 243 based on the modified Recalculate the RUL and generate structured decision information:

[0084] Fault location topology diagram: Based on the ship's equipment physical connection diagram, highlighted The largest device node;

[0085] Root cause probability distribution: Location of the pair via backpropagation The original sensor features that contribute the most are the output probability vectors. ;

[0086] Cross-device impact propagation chain: Extracting the spatiotemporal evolution diagram Medium weight The edges form a directed path, where, The preset threshold for the intensity of the influence;

[0087] Preventive maintenance time window: defined as ,in , To ensure that repairs are completed before the fault becomes critical, among other things, The earliest time to recommend starting maintenance preparations. The latest deadline for completing the repair.

[0088] The accuracy feedback unit 244 compares the actual maintenance verification results with the predicted RUL and calculates the relative error. ,in, This refers to the actual failure time, i.e., the actual maintenance verification result. For example, when a hydraulic system experiences a failure due to micro-leakage, it becomes a fault characterization parameter. The time when the preset critical value of 1.0 is reached at 154 hours is defined as the failure time. This indicates that the system has entered a state requiring repair, rather than a point of physical damage resulting in complete equipment failure. When At that time, the environmental degradation factor of the environmental degradation compensation unit is triggered. Correction: ,in The learning rate is used. Therefore, during long-term operation, the material aging model may drift due to deviations in actual operating conditions. Dynamic calibration based on maintenance feedback can maintain the long-term stability of the system's predictive accuracy.

[0089] In summary, this implementation method constructs a system with dynamic parameters. A closed-loop fault analysis system based on coupling factors enables high-precision, system-level predictive diagnosis of early-stage faults in multiple ship equipment. First, the multi-source sensor synchronization module 21 utilizes a hardware phase-locked loop mechanism to achieve ±0.1° phase synchronization in a low-frequency vibration environment of 0.1–2Hz, effectively suppressing ship motion noise. Second, the multiphysics modeling module 22 integrates mechanical, fluid, and thermodynamic fields, introducing an environmental attenuation factor. Dynamically compensate for the effects of seawater corrosion and vibration, and generate dynamic parameters characterizing the system's state misalignment. Furthermore, the adaptive diagnostic analysis module, based on... By adjusting the convolution kernel center frequency and LSTM memory depth of the deep learning model in real time, latent fault features are accurately extracted and the spatiotemporal evolution of coupled faults is modeled. Finally, the predictive decision module combines ship operation plans and maintenance feedback to dynamically correct the fault evolution model and generate decision information including maintenance windows, forming a complete technology chain from data acquisition to closed-loop optimization. Real-ship verification shows that this solution advances the early warning time for micro-leaks and other early faults to more than 82 hours, improves the accuracy of coupled fault root cause localization to 89.7%, and maintains a prediction error of less than 5% in long-term salt spray environments. This effectively avoids the problems of diagnostic lag, low accuracy, and insufficient correlation analysis caused by environmental interference, model rigidity, and data fragmentation.

[0090] refer to Figure 2 This embodiment further discloses a fault analysis method for ship equipment, including the following steps:

[0091] S100: Synchronizes the phase of operating data from multiple devices through a hardware phase-locked loop mechanism.

[0092] Specifically, vibration, pressure, temperature, flow rate, and ship motion attitude sensors are deployed at equipment nodes in the ship's propulsion system, generator set, hydraulic system, and cooling system to collect raw signals at a sampling rate of 100Hz. Since the ship operates in a low-frequency vibration environment (such as roll and pitch) of 0.1–2Hz, phase shifts occur in each sensor due to differences in their physical locations. Therefore, a hardware phase-locked loop (PLL) is used to detect the phase shift of each channel in the 0.1–2Hz frequency band in real time. Furthermore, a synchronous clock signal is generated by a digitally controlled oscillator for dynamic compensation, ensuring that the phase error of all channels meets the requirements. ,in This represents the average phase offset. This synchronization accuracy originates from the resonant critical point of the first-order bending mode of the ship's pipeline, effectively eliminating the interference of motion noise on subsequent analysis.

[0093] S200: Construct a virtual state field for devices based on the multi-device operation data after phase synchronization, and generate dynamic parameters that characterize the temporal misscheduling of the multi-device states.

[0094] Specifically, the phase-synchronized multidimensional data is grouped by device node, and the energy features of vibration, pressure, temperature, and flow are extracted through wavelet packet decomposition to form a multidimensional feature matrix of device status. Based on the physical topology of the ship equipment, a fused mechanical vibration field is constructed. Fluid pressure field With thermodynamic field Coupled virtual state field Introducing an environmental degradation factor For the health baseline field Perform dynamic correction. Calculate the information entropy shift between the current state field and the corrected reference field to generate dynamic parameters: This parameter can effectively quantify the coordinated misscheduling caused by physical connections among multiple devices. A normal value indicates a healthy system, while a significant increase indicates an early abnormality.

[0095] S300: Based on the dynamic parameters, the implicit feature extraction parameters and fault evolution time series modeling parameters of the fault analysis model are adaptively adjusted in real time to output system-level fault characterization parameters.

[0096] Specifically, dynamic parameters Input a pre-trained CNN-LSTM fusion model and adjust its hyperparameters in real time: First, calculate the adjustment amount of the convolution kernel center frequency. This allows CNN to focus on the typical fault characteristics of ship equipment; secondly, it calculates... slope of change ,like The LSTM memory cell depth is then increased from 64 to 128 to enhance its ability to capture the evolution of progressive faults. The model outputs latent fault feature vectors for each device and constructs a spatiotemporal evolution diagram based on the ship's physical connection topology. By aggregating cross-device influences through a graph attention mechanism, system-level fault characterization parameters are ultimately generated. These respectively characterize the severity of the fault and its propagation trend.

[0097] S400: Generate decision information including fault development trend prediction and preventive maintenance window based on the system-level fault characterization parameters.

[0098] Specifically, based on Construct a physical evolution model of the failure: Assume the failure severity growth rate Solve for the remaining safe running time by integration. Combining navigation phase information obtained from the Automatic Identification System (AIS) With sea state data Dynamically adjust growth rate Generate structured decision-making information, including a fault location topology map, root cause probability distribution, cross-device impact propagation chains, and preventative maintenance time windows. Simultaneously, the actual maintenance verification results will be fed back to the environmental degradation factor. ,pass Achieve closed-loop optimization to ensure long-term forecast accuracy.

[0099] Figure 3 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as... Figure 3 As shown, the electronic device 50 includes: a processor 501, a memory 502, and a bus 503;

[0100] The processor 501 and the memory 502 communicate with each other via the bus 503; the processor 501 is used to call the program instructions in the memory 502 to execute the methods provided in the above-described embodiments.

[0101] This embodiment provides a non-transitory computer-readable storage medium that stores computer instructions that cause a computer to execute the methods provided in the above-described embodiments.

[0102] Those skilled in the art will understand that all or part of the steps of the above-described method implementation can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above-described method implementation. The aforementioned storage medium includes various storage media capable of storing program code, such as ROM, RAM, magnetic disk, or optical disk.

[0103] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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.

[0104] 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., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of each embodiment or some parts of the embodiments.

[0105] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A fault analysis system for ship equipment, characterized in that, include: A multi-source sensor synchronization module is deployed in ship equipment and configured to synchronize the operating data of multiple devices through a hardware phase-locked loop mechanism. The multi-source sensing synchronization module includes: a sensing deployment unit configured to deploy vibration sensors, pressure sensors, temperature sensors, and flow sensors at nodes in the ship's propulsion system, generator set, hydraulic system, and cooling system; a phase measurement unit coupled to the sensing deployment unit configured to detect the phase offset of each sensor signal in the 0.1-2Hz low-frequency vibration band in real time; a phase-locked loop control unit coupled to the phase measurement unit configured to generate a hardware-level synchronization clock signal and dynamically compensate for the phase offset to ensure that the phase synchronization error of the multi-device operating data is no greater than ±0.1°; and an environmental adaptation unit coupled to the phase-locked loop control unit configured to dynamically adjust the synchronization accuracy threshold of the phase-locked loop control unit according to the ship's roll frequency and surge intensity. A multiphysics modeling module, coupled to the multi-source sensing synchronization module, is configured to construct a virtual state field of equipment based on the phase-synchronized multi-device operating data and generate dynamic parameters characterizing the temporal misalignment of the multi-device states. The multiphysics modeling module includes: a multi-dimensional data fusion unit, coupled to the multi-source sensing synchronization module, configured to integrate phase-synchronized vibration, pressure, temperature, and flow time-series data to generate a multi-dimensional feature matrix of the equipment states; a virtual field construction unit, coupled to the multi-dimensional data fusion unit, configured to construct a coupled virtual state field containing mechanical stress field, thermodynamic field, and fluid pressure field based on the physical topology of the ship's equipment; a dynamic parameter generation unit, coupled to the virtual field construction unit, configured to calculate the distribution offset entropy value of the coupled virtual state field in the temporal dimension as a dynamic parameter characterizing the temporal misalignment of the multi-device states; and an environmental attenuation compensation unit, coupled to the dynamic parameter generation unit, configured to dynamically attenuate and compensate the distribution offset entropy value based on seawater salinity, equipment exposure time, and ship vibration spectrum, so that the dynamic parameters conform to the physical evolution law of the ship's operating environment. The adaptive diagnostic analysis module includes a deep learning fault analysis model. The implicit feature extraction parameters and fault evolution time series modeling parameters of the deep learning fault analysis model are adaptively adjusted in real time according to the dynamic parameters to output system-level fault characterization parameters. The predictive decision module, coupled to the adaptive diagnostic analysis module, is configured to generate decision information that includes fault development trend prediction and preventive maintenance windows based on the system-level fault characterization parameters.

2. The fault analysis system according to claim 1, characterized in that, The adaptive diagnostic analysis module includes: The model initialization unit is configured as the infrastructure for loading a deep learning fault analysis model, the infrastructure including a fusion structure of a convolutional neural network and a long short-term memory network; The ship characteristic parameter tuning unit is coupled to the multiphysics modeling module and the model initialization unit, and is configured to receive the dynamic parameters and adjust the center frequency of the convolution kernel of the convolutional neural network and the depth of the memory unit of the long short-term memory network in real time based on the characteristics of the ship's low-frequency vibration environment. The cross-device feature fusion unit is coupled to the ship characteristic parameter tuning unit and is configured to extract the latent fault features of each device in the 0.1-2Hz vibration frequency band and model the spatiotemporal evolution path of multi-device coupled faults in the ship's physical connection chain. The system characterization generation unit is coupled to the cross-device feature fusion unit and is configured to fuse the latent fault features with the spatiotemporal evolution path to generate system-level fault characterization parameters that characterize the severity and propagation trend of system-level faults.

3. The fault analysis system according to claim 2, characterized in that, The predictive decision-making module includes: The trend quantification unit, coupled to the adaptive diagnostic analysis module, is configured to construct a fault physical evolution model based on the system-level fault characterization parameters and output the remaining safe operating time of the equipment and the fault severity growth rate. The ship operation plan fusion unit is coupled to the ship automatic identification system and the trend quantification unit, and is configured to dynamically correct the rate of increase of the fault severity by combining ship berthing plan, navigation stage and sea state data. The decision generation unit, coupled to the ship operation plan fusion unit, is configured to generate structured decision information including a fault location topology map, root cause probability distribution, cross-equipment impact propagation chain, and preventive maintenance time window. The accuracy feedback unit, coupled to the decision generation unit and the multiphysics modeling module, is configured to dynamically calibrate the parameters of the fault physical evolution model based on actual maintenance verification results.

4. The fault analysis system according to claim 3, characterized in that, The calculation process of the dynamic parameters includes: A time-series distribution matrix characterizing the multi-physical field state of ship equipment is constructed based on the multi-device operation data after phase synchronization. The multi-physical field includes mechanical vibration field, fluid pressure field and thermodynamic field. Obtain the equipment health status baseline distribution and environmental degradation factor, wherein the environmental degradation factor is dynamically updated based on the ship's seawater corrosion intensity and low-frequency vibration spectrum; The temporal misalignment between the current temporal distribution matrix and the healthy state baseline distribution after environmental attenuation correction is calculated using the information entropy algorithm, generating dynamic parameters characterizing the degree of system-level state anomalies.

5. The fault analysis system according to claim 2, characterized in that, The ship characteristic parameter tuning unit executes the following adaptive adjustment strategy: The dynamic parameters are input into a preset mapping function to calculate the center frequency adjustment of the convolutional neural network kernel, so that the convolutional kernel focuses on the characteristic fault frequency of the ship equipment in a low-frequency vibration environment. Calculate the slope of change of the dynamic parameters in the temporal dimension, determine the memory unit depth of the long short-term memory network based on the slope of change, and expand the memory unit depth when the slope of change increases to enhance the ability to capture the evolution law of progressive failures. Based on the adjusted convolution kernel and memory unit depth, the deep learning fault analysis model is updated to improve the accuracy of extracting latent fault features of ship equipment.

6. The fault analysis system according to claim 4, characterized in that, The environmental degradation compensation unit and the ship characteristic parameter tuning unit form a closed-loop optimization mechanism: The environmental attenuation compensation unit compensates for the dynamic parameters based on the ship's operating environment parameters and then outputs the compensation to the ship characteristic parameter tuning unit. The ship characteristic parameter tuning unit adjusts the deep learning model parameters based on the compensated dynamic parameters and generates system-level fault characterization parameters. The accuracy feedback unit feeds back the actual maintenance verification results to the environmental attenuation compensation unit to correct the environmental attenuation factor.

7. A method for fault analysis of ship equipment, characterized in that, include: Phase synchronization of operating data from multiple devices is achieved through a hardware phase-locked loop (PLL) mechanism. This includes: deploying vibration sensors, pressure sensors, temperature sensors, and flow sensors at nodes in the ship's propulsion system, generator set, hydraulic system, and cooling system; real-time detection of the phase offset of each sensor signal within the 0.1-2Hz low-frequency vibration band; generating a hardware-level synchronization clock signal to dynamically compensate for the phase offset, ensuring that the phase synchronization error of the operating data from multiple devices is no greater than ±0.1°; and dynamically adjusting the synchronization accuracy threshold of the PLL control unit based on the ship's roll frequency and surge intensity. A virtual state field for equipment is constructed based on the phase-synchronized multi-device operating data, and dynamic parameters characterizing the temporal misalignment of the multi-device states are generated. This includes: integrating phase-synchronized vibration, pressure, temperature, and flow time-series data to generate a multi-dimensional feature matrix of equipment states; constructing a coupled virtual state field containing mechanical stress field, thermodynamic field, and fluid pressure field based on the physical topology of the ship's equipment; calculating the distribution offset entropy value of the coupled virtual state field in the temporal dimension as a dynamic parameter characterizing the temporal misalignment of the multi-device states; and dynamically attenuating and compensating the distribution offset entropy value according to seawater salinity, equipment exposure time, and ship vibration spectrum to make the dynamic parameters conform to the physical evolution law of the ship's operating environment. The implicit feature extraction parameters and fault evolution time series modeling parameters of the deep learning fault analysis model are adaptively adjusted in real time according to the dynamic parameters to output system-level fault characterization parameters. Based on the system-level fault characterization parameters, decision information including fault development trend prediction and preventive maintenance window is generated.

8. 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 program, it implements the method described in claim 7.

Images