Ship engine on-condition maintenance method and system

Abstract

The invention provides a ship engine on-condition maintenance method and system, and the method comprises the steps: collecting the original state data of lubricating oil of a ship engine and a plurality of mechanical parts in real time, carrying out the preprocessing, and carrying out the fault diagnosis of the mechanical parts based on the operation logic and fault mechanism of the mechanical parts in combination with thermal parameters and vibration parameters. Carrying out reliability analysis by adopting a genetic algorithm, a maximum likelihood method and three-parameter Weibull distribution, and calculating the remaining use time of the mechanical part after each fault; lubricating oil fault diagnosis is carried out based on the thermal parameters and the oil liquid indexes, reliability analysis is carried out by adopting a genetic algorithm and a BP neural network, and the remaining use time of the lubricating oil is calculated; and finally, generating a mechanical part condition-based maintenance scheme and a lubricating oil condition-based maintenance scheme containing the inspection item list based on the remaining use time of the mechanical part after each fault and the remaining use time of the lubricating oil, and further realizing condition-based maintenance of the ship engine according to the mechanical part condition-based maintenance scheme and the lubricating oil condition-based maintenance scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of ship fault diagnosis, and in particular to a condition-based maintenance method and system for a ship engine. Background Art

[0002] In the shipping industry, the stable operation of ship engines is crucial. However, for a long time, the maintenance strategy for ship engines has mostly relied on scheduled maintenance. This approach has many drawbacks. On the one hand, it can lead to excessive maintenance, unnecessary maintenance costs and waste of resources. On the other hand, due to improper maintenance schedules, potential faults cannot be discovered in a timely manner, causing serious mechanical damage, affecting the normal operation of the ship, and resulting in huge economic losses and safety hazards.

[0003] While current monitoring and maintenance technologies for core marine engine components and lubricating oil have made some progress, they still face numerous shortcomings. 1) For core mechanical components such as turbochargers, lubricating oil pumps, and intake and exhaust valves, thermal parameter monitoring is currently the primary method, generating alarms when a parameter exceeds a threshold. In recent years, some well-known suppliers and solution providers have also adopted various fault diagnosis models for mechanical equipment fault diagnosis, achieving some progress. However, these methods also present several challenges. First, fault diagnosis results are not particularly accurate, often leading to false alarms and missed reports. Second, some strategies are too coarse-grained to pinpoint the engine as a whole or a specific parameter, failing to identify specific faults. Third, when certain engine faults occur, they are not effectively recorded and used as learning experiences. Fourth, they lack reliability analysis capabilities, making it impossible to generate targeted, condition-based maintenance plans and strategies. Consequently, existing monitoring methods for core mechanical components are often inaccurate, making it difficult to accurately predict reliability and the onset of failures. 2) For lubricating oil monitoring, the current approach primarily relies on periodic sampling combined with shore-based testing, which presents several shortcomings. First, the lubricating oil collection cycle is long, and the lubricating oil may have deteriorated rapidly during this cycle; second, the results of multiple lubricating oil analyses are isolated, lacking continuous trend analysis; third, it is impossible to accurately predict and guide the oil change time. Therefore, the current lubricating oil monitoring mainly adopts the strategy of onboard collection and shore analysis. The interval between each collection and analysis is long, which cannot timely and effectively analyze the health status of the lubricating oil, nor can it accurately predict the trend of the lubricating oil.

[0004] Due to these issues, current ship management maintenance strategies are often based on experience and manufacturer-recommended maintenance intervals, or on a post-failure repair approach, lacking flexibility and specificity. Furthermore, maintenance strategies cannot be dynamically adjusted based on the actual operating status and operating environment of the ship's engines, making true condition-based maintenance difficult to achieve. Summary of the Invention

[0005] To address issues such as false reporting, missed reporting, and long lubricating oil analysis intervals in traditional engine fault diagnosis, the present invention provides a condition-based maintenance method for marine engines. This method can effectively improve the accuracy of marine engine fault diagnosis and locate the specific faulty component, ensuring the accuracy of input data for reliability analysis. Through lubricating oil fault diagnosis and reliability analysis, the method can accurately predict the state of the lubricating oil, avoiding the rapid deterioration of the lubricating oil due to long sampling intervals. Furthermore, the method can accurately and continuously analyze the health status trend of the lubricating oil, improving the consistency and accuracy of lubricating oil analysis. Furthermore, the method can generate a clear and easy-to-understand condition-based maintenance plan and checklist, dynamically adjusting maintenance strategies based on the actual operating status and operating environment of the marine engine, thereby providing flexibility and pertinence. The present invention also relates to a condition-based maintenance system for marine engines.

[0006] The technical solutions of the present invention are as follows:

[0007] A method for condition-based maintenance of a ship engine, comprising the following steps:

[0008] Data collection and preprocessing steps: real-time collection of raw state data of the lubricating oil and multiple mechanical components of the ship engine, and preprocessing of the raw state data; the raw state data includes thermal parameters, oil index parameters and vibration parameters;

[0009] Mechanical component fault diagnosis steps: Based on the operation logic and fault mechanism of the mechanical component, the pre-processed thermal parameters and vibration parameters are selected as the diagnostic indicators of the mechanical component, and a fault diagnosis mechanism model is established based on the diagnostic indicators. Then, the diagnostic indicators of the mechanical component under different working conditions when the engine is in a healthy state are statistically analyzed and interpolated and fitted to obtain the benchmark values of the diagnostic indicators of the mechanical component under different working conditions of the engine. The fault diagnosis mechanism model is calibrated by the benchmark values, and the real-time diagnostic indicators of a mechanical component under a certain working condition are input into the calibrated fault diagnosis mechanism model, and the predicted value of the mechanical component under the working condition is output. The difference between the predicted value of the mechanical component and the real-time diagnostic indicator is compared with a preset threshold to determine whether the mechanical component has a fault. If the difference is less than or equal to the preset threshold, it is determined that no fault has occurred; if the difference is greater than the preset threshold, it is determined that the mechanical component has a fault, and the time of each historical fault of the mechanical component is obtained to generate a fault history data set.

[0010] Lubricating oil fault diagnosis steps: selecting the pre-processed thermal parameters and oil indicators as lubricating oil diagnostic indicators, and comparing each indicator in the lubricating oil diagnostic indicators with its corresponding preset lubricating oil failure threshold to determine whether the lubricating oil has failed. If any indicator in the lubricating oil diagnostic indicators is greater than its corresponding preset lubricating oil failure threshold, the lubricating oil is determined to have failed; if all indicators in the lubricating oil diagnostic indicators are less than or equal to their corresponding preset lubricating oil failure thresholds, the lubricating oil is determined to be in a valid state;

[0011] Mechanical component reliability analysis steps: Based on the failure history data set and using the three-parameter Weibull distribution, a reliability function and probability density function for the time when the mechanical component fails are constructed respectively. Then, a log-likelihood function is established based on the probability density function and the maximum likelihood function. A genetic algorithm is used to calculate the optimal solution of the log-likelihood function, and then the optimal scale parameter, optimal shape parameter, and optimal location parameter are obtained. Based on the optimal scale parameter, optimal shape parameter, optimal location parameter, and the preset target reliability, the reliability function is used to calculate the remaining service life of the mechanical component after each failure, which is used as the reliability analysis result of the mechanical component.

[0012] Lubricating oil reliability analysis steps: after determining that the lubricating oil has failed, randomly initialize the weights and bias values of the BP neural network to obtain the initial weights and initial bias values; and use the genetic algorithm to calculate the optimal weights and optimal bias of the BP neural network, and use the optimal weights and optimal bias to update the initial weights and initial bias values in the BP neural network to obtain a prediction model, and input the pre-processed thermal parameters and oil indicators at the current moment as input indicators into the prediction model to predict the viscosity, moisture and particulate matter content increments of the lubricating oil at the next moment after the current moment; and respectively predict the viscosity, moisture and particulate matter content increments at multiple consecutive moments after the next moment, and then calculate the current moment. The average viscosity increment, the average moisture increment, and the average particle content increment at all moments after the moment are calculated; the remaining viscosity usage time is calculated based on the average viscosity increment, the actual viscosity measured at each moment, and a preset viscosity threshold; the remaining moisture usage time is calculated based on the average moisture increment, the actual moisture measured at each moment, and a preset moisture threshold; and the remaining particle content usage time is calculated based on the average particle content increment, the actual particle content measured at each moment, and a preset particle content threshold; and the minimum value of the remaining viscosity, moisture, and particle content usage times is taken as the remaining lubricating oil usage time, which is used as the lubricating oil reliability analysis result;

[0013] Engine condition-based maintenance steps: Generate a mechanical component condition-based maintenance plan and a lubricating oil condition-based maintenance plan containing a checklist of inspection items based on the remaining service life of the mechanical component and the remaining service life of the lubricating oil after each failure, and then implement condition-based maintenance of the ship engine according to the mechanical component condition-based maintenance plan and the lubricating oil condition-based maintenance plan.

[0014] Preferably, the order of the mechanical component fault diagnosis step and the lubricating oil fault diagnosis step is interchangeable;

[0015] And / or, the order of the mechanical component reliability analysis step and the lubricating oil reliability analysis step can be interchanged.

[0016] Preferably, in the mechanical component reliability analysis step, calculating the optimal solution of the log-likelihood function using a genetic algorithm specifically includes:

[0017] S1: Randomly initialize the scale parameter, shape parameter, and location parameter in the log-likelihood function to obtain the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter, respectively, and encode the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter to generate an initial population;

[0018] S2: Use the log-likelihood function as the fitness function of the genetic algorithm, and calculate the fitness value of each individual in the initial population based on the fitness function;

[0019] S3: Select multiple individuals from the initial population as parent individuals based on their fitness values using the roulette wheel selection method;

[0020] S4: Determine whether to perform a crossover operation based on the randomly generated probability value and the preset crossover rate. If so, randomly select two individuals from the parent individuals, then randomly select two time units with relative positions from the two individuals, and exchange the codes of these two time units to generate two offspring individuals; if not, keep the number of parent individuals unchanged, and use all individuals in the parent individuals as offspring individuals;

[0021] S5: Repeat step S4 until the preset number of iterations is reached to obtain multiple offspring individuals;

[0022] S6: Perform mutation operation on each offspring individual to obtain multiple new offspring individuals;

[0023] S7: Calculate the fitness value of each new offspring individual according to the fitness function, and determine whether the preset number of iterations is reached. If so, take the individual with the largest fitness value as the optimal solution of the log-likelihood function; if not, repeat steps S3 to S6 until the preset number of iterations is reached. Take the individual with the largest fitness value when the preset number of iterations is reached as the optimal solution of the log-likelihood function, and then obtain the optimal scale parameter, optimal shape parameter and optimal position parameter.

[0024] Preferably, in the engine condition-based maintenance step, the mechanical component condition-based maintenance plan specifically includes:

[0025] Determine whether the remaining usage time of a mechanical component after a certain failure is less than a preset time threshold. If so, include the mechanical component in the inspection item list for inspection, and determine whether the inspection result of the mechanical component is normal. If the inspection result is normal, reset the remaining usage time of the mechanical equipment to the remaining usage time after the previous failure of the failure; if the inspection result is abnormal, record the current time as the time when the mechanical component fails, and store the current time in the fault history data set to form a new fault history data set. Based on the new fault history data set and continuing in accordance with the working principle of the mechanical component reliability analysis step, recalculate the new remaining usage time of the mechanical component.

[0026] Preferably, in the engine condition-based maintenance step, the lubricating oil condition-based maintenance plan specifically includes:

[0027] Determine whether the remaining service life of the lubricating oil is less than the preset time threshold. If so, include the lubricating oil in the inspection item list for inspection, and after the inspection, record the actual viscosity, actual moisture content and actual particulate matter content measured at the current moment, and continue to recalculate the new remaining service life of the lubricating oil according to the working principle of the lubricating oil reliability analysis step.

[0028] Preferably, in the data acquisition and preprocessing steps, the mechanical components include a combination of an engine turbocharger, an oil pump, a fuel pump, an exhaust valve, an intake valve, a cylinder liner, a cylinder head, and bearings; the thermal parameters include engine exhaust temperature, cylinder liner temperature, cooling water pressure, and cooling water temperature; the oil index parameters include oil temperature, viscosity, total iron content, and non-total iron content; and the vibration parameters include turbocharger vibration parameters and oil pump vibration parameters.

[0029] The preprocessing of the original state data includes: removing outliers from the original state data, and extracting features from the vibration parameters and oil index parameters in the remaining original state data after the removal, extracting the average value, effective value, maximum value, minimum value, standard deviation, kurtosis and FFT global amplitude of each parameter in the vibration parameters; and extracting the average value of each parameter in the oil index parameters.

[0030] A ship engine condition-based maintenance system, characterized by comprising a data acquisition and preprocessing module, a mechanical component fault diagnosis module, a lubricating oil fault diagnosis module, a mechanical component reliability analysis module, a lubricating oil reliability analysis module, and an engine condition-based maintenance module, wherein the data acquisition and preprocessing module is respectively connected to the mechanical component fault diagnosis module and the lubricating oil fault diagnosis module, the mechanical component fault diagnosis module is respectively connected to the mechanical component reliability analysis module, the lubricating oil fault diagnosis module is respectively connected to the lubricating oil reliability analysis module, and the engine condition-based maintenance module is respectively connected to the mechanical component reliability analysis module and the lubricating oil reliability analysis module;

[0031] The data acquisition and preprocessing module collects the original state data of the lubricating oil and multiple mechanical components of the ship engine in real time and preprocesses the original state data; the original state data includes thermal parameters, oil index parameters and vibration parameters;

[0032] The mechanical component fault diagnosis module selects pre-processed thermal parameters and vibration parameters as diagnostic indicators of the mechanical component based on the mechanical component operation logic and fault mechanism, establishes a fault diagnosis mechanism model based on the diagnostic indicators, and then performs statistical analysis and interpolation fitting processing on the diagnostic indicators of the mechanical component under different working conditions when the engine is in a healthy state to obtain benchmark values of the diagnostic indicators of the mechanical component under different working conditions of the engine, calibrates the fault diagnosis mechanism model based on the benchmark values, and inputs the real-time diagnostic indicator of a mechanical component under a certain working condition into the calibrated fault diagnosis mechanism model, outputs a predicted value of the mechanical component under the working condition, and compares the difference between the predicted value of the mechanical component and the real-time diagnostic indicator with a preset threshold to determine whether the mechanical component has a fault. If the difference is less than or equal to the preset threshold, it is determined that no fault has occurred; if the difference is greater than the preset threshold, it is determined that the mechanical component has a fault, and obtains the time of each historical fault of the mechanical component to generate a fault history data set;

[0033] The lubricating oil fault diagnosis module selects the pre-processed thermal parameters and oil indicators as lubricating oil diagnostic indicators, and compares each indicator in the lubricating oil diagnostic indicators with its corresponding preset lubricating oil failure threshold to determine whether the lubricating oil has failed. If any indicator in the lubricating oil diagnostic indicators is greater than its corresponding preset lubricating oil failure threshold, the lubricating oil is determined to have failed; if all indicators in the lubricating oil diagnostic indicators are less than or equal to their corresponding preset lubricating oil failure thresholds, the lubricating oil is determined to be in a valid state.

[0034] The mechanical component reliability analysis module constructs a reliability function and a probability density function for the time when the mechanical component fails based on a fault history data set and a three-parameter Weibull distribution, and then establishes a log-likelihood function based on the probability density function and the maximum likelihood function. A genetic algorithm is used to calculate the optimal solution of the log-likelihood function, thereby obtaining the optimal scale parameter, optimal shape parameter, and optimal location parameter. Based on the optimal scale parameter, optimal shape parameter, and optimal location parameter, as well as a preset target reliability, the reliability function is used to calculate the remaining service life of the mechanical component after each failure, and this is used as the reliability analysis result of the mechanical component.

[0035] The lubricating oil reliability analysis module randomly initializes the weights and bias values of the BP neural network after determining that the lubricating oil has failed, and obtains the initial weights and initial bias values; and uses a genetic algorithm to calculate the optimal weights and optimal bias of the BP neural network, and uses the optimal weights and optimal bias to update the initial weights and initial bias values in the BP neural network to obtain a prediction model, and inputs the pre-processed thermal parameters and oil indicators at the current moment as input indicators into the prediction model to predict the viscosity, moisture and particulate matter content increments of the lubricating oil at the next moment after the current moment; and respectively predicts the viscosity, moisture and particulate matter content increments at multiple consecutive moments after the next moment, and then calculates the current moment. The average viscosity increment, the average moisture increment, and the average particle content increment at all moments after the moment; then calculating the remaining service life of the viscosity based on the average viscosity increment, the actual viscosity measured at each moment, and the preset viscosity threshold; calculating the remaining service life of the moisture based on the average moisture increment, the actual moisture measured at each moment, and the preset moisture threshold; and calculating the remaining service life of the particle content based on the average particle content increment, the actual particle content measured at each moment, and the preset particle content threshold; and taking the minimum value of the remaining service life of the viscosity, moisture, and particle content as the remaining service life of the lubricating oil, which is used as the lubricating oil reliability analysis result;

[0036] The engine condition-based maintenance module generates a mechanical component condition-based maintenance plan and a lubricating oil condition-based maintenance plan containing a checklist of inspection items based on the remaining service life of the mechanical component and the remaining service life of the lubricating oil after each failure, and then implements condition-based maintenance of the ship engine according to the mechanical component condition-based maintenance plan and the lubricating oil condition-based maintenance plan.

[0037] Preferably, in the mechanical component reliability analysis module, calculating the optimal solution of the log-likelihood function using a genetic algorithm specifically includes:

[0038] S1: Randomly initialize the scale parameter, shape parameter, and location parameter in the log-likelihood function to obtain the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter, respectively, and encode the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter to generate an initial population;

[0039] S2: Use the log-likelihood function as the fitness function of the genetic algorithm, and calculate the fitness value of each individual in the initial population based on the fitness function;

[0040] S3: Select multiple individuals from the initial population as parent individuals based on their fitness values using the roulette wheel selection method;

[0041] S4: Determine whether to perform a crossover operation based on the randomly generated probability value and the preset crossover rate. If so, randomly select two individuals from the parent individuals, then randomly select two time units with relative positions from the two individuals, and exchange the codes of these two time units to generate two offspring individuals; if not, keep the number of parent individuals unchanged, and use all individuals in the parent individuals as offspring individuals;

[0042] S5: Repeat step S4 until the preset number of iterations is reached to obtain multiple offspring individuals;

[0043] S6: Perform mutation operation on each offspring individual to obtain multiple new offspring individuals;

[0044] S7: Calculate the fitness value of each new offspring individual according to the fitness function, and determine whether the preset number of iterations is reached. If so, take the individual with the largest fitness value as the optimal solution of the log-likelihood function; if not, repeat steps S3 to S6 until the preset number of iterations is reached. Take the individual with the largest fitness value when the preset number of iterations is reached as the optimal solution of the log-likelihood function, and then obtain the optimal scale parameter, optimal shape parameter and optimal position parameter.

[0045] Preferably, in the engine condition-based maintenance module, the mechanical component condition-based maintenance plan specifically includes:

[0046] Determine whether the remaining usage time of a mechanical component after a certain failure is less than a preset time threshold. If so, include the mechanical component in the inspection item list for inspection, and determine whether the inspection result of the mechanical component is normal. If the inspection result is normal, reset the remaining usage time of the mechanical equipment to the remaining usage time after the previous failure of the failure; if the inspection result is abnormal, record the current time as the time when the mechanical component fails, and store the current time in the fault history data set to form a new fault history data set. Based on the new fault history data set and continuing in accordance with the working principle of the mechanical component reliability analysis module, recalculate the new remaining usage time of the mechanical component.

[0047] Preferably, in the engine condition-based maintenance module, the lubricating oil condition-based maintenance plan specifically includes:

[0048] Determine whether the remaining service life of the lubricating oil is less than the preset time threshold. If so, include the lubricating oil in the inspection item list for inspection, and after the inspection, record the actual viscosity, actual moisture and actual particulate matter content measured at the current moment, and continue to recalculate the new remaining service life of the lubricating oil according to the working principle of the lubricating oil reliability analysis module.

[0049] The beneficial effects of the present invention are:

[0050] The present invention provides a method for condition-based maintenance of ship engines. First, the original status data of the lubricating oil and multiple mechanical components of the ship engine are collected in real time and preprocessed to improve the quality and availability of the data; for the mechanical components, fault diagnosis and reliability analysis are carried out in sequence: a fault diagnosis mechanism model is established based on the operating logic and fault mechanism of the mechanical components and combined with thermal parameters and vibration parameters as diagnostic indicators, and the real-time diagnostic indicator data is compared with the model prediction value. By using a preset threshold, it is determined whether the mechanical component has failed, and the time of each historical failure of the mechanical component is obtained, and then a fault history data set is generated. By establishing a fault diagnosis mechanism model for fault diagnosis, faults can be diagnosed more accurately, the accuracy of fault diagnosis can be improved, false reporting and missed reporting can be reduced, and the specific faulty mechanical component can be accurately located, rather than just the faulty component. Only the entire engine or a certain parameter is located, which effectively improves the granularity of fault diagnosis; then, based on the fault history data set and using the three-parameter Weibull distribution, reliability analysis is performed, and the reliability function and probability density function of the time when the mechanical component fails are constructed respectively. The log-likelihood function is established based on the probability density function and the maximum likelihood function. The genetic algorithm is then used to calculate the optimal solution of the log-likelihood function, and then the optimal scale parameter, optimal shape parameter and optimal position parameter are obtained. According to the optimal scale parameter, optimal shape parameter and optimal position parameter, as well as the preset target reliability, the reliability function is used to calculate the remaining service life of the mechanical component after each failure. This can accurately predict the remaining service life of the mechanical component, realize the reliability analysis of the mechanical component, avoid replacing components too early or too late, and improve maintenance efficiency. For lubricating oil, fault diagnosis and reliability analysis are carried out in sequence: based on thermal parameters and oil indicators as lubricating oil diagnostic indicators, it is judged whether the lubricating oil has failed to achieve the accuracy of fault diagnosis; then the genetic algorithm is used to optimize the weights and bias values of the BP neural network to obtain a prediction model to improve the prediction accuracy; and the prediction model is used to predict the increments of the viscosity, moisture and particulate matter content of the lubricating oil at multiple consecutive moments in the future. By continuously predicting the state changes of the lubricating oil, trend analysis is performed to avoid the problem of rapid deterioration of the lubricating oil caused by long sampling intervals. Through continuous prediction, the isolation of multiple lubricating oil analysis results is reduced, and the consistency and accuracy of lubricating oil analysis are improved; finally, a specific calculation method is used to calculate the remaining service life of the viscosity, moisture and particulate matter content, and the minimum value of the remaining service life of the viscosity, moisture and particulate matter content is taken as the remaining service life of the lubricating oil. This can accurately calculate the remaining service life of the lubricating oil, realize lubricating oil reliability analysis, and guide the oil change time to avoid oil change too early or too late.Finally, based on the remaining service life of mechanical components and the remaining service life of lubricating oil after each failure, a mechanical component condition-based maintenance plan and a lubricating oil condition-based maintenance plan containing a checklist of inspection items are generated. Then, according to the mechanical component condition-based maintenance plan and the lubricating oil condition-based maintenance plan, the ship engine condition-based maintenance is realized, which improves the accuracy and effectiveness of maintenance. Through condition-based maintenance, unnecessary maintenance operations are avoided, maintenance costs are reduced, and the reliability and service life of the engine are improved.

[0051] The present invention provides a method for condition-based maintenance of marine engines. To address the problems of false reporting, missed reporting, and long lubricating oil analysis intervals in traditional engine fault diagnosis, the present invention adds online vibration monitoring and oil online monitoring solutions, and comprehensively adopts multiple fault diagnosis algorithms based on vibration analysis, oil analysis, thermal parameter analysis, mechanism models, etc., to improve the accuracy of fault diagnosis and locate the specific faulty components. In order to ensure the accuracy of the input data for reliability analysis, a feedback mechanism is added to the identified faults and condition-based maintenance inspection items, and an interface for manual input of fault information is provided. In view of the characteristics of different engine components, different reliability analysis methods are used for mechanical equipment and lubricating oil, etc., which can generate clear and easy-to-understand condition-based maintenance plans and inspection lists.

[0052] The present invention, through the above-mentioned technical means, can effectively solve the problems existing in the prior art, such as inaccurate fault diagnosis, coarse strategy granularity, incomplete fault records, and lack of reliability analysis. At the same time, the application of technologies such as real-time data acquisition and preprocessing, fault diagnosis mechanism models, Weibull distribution, genetic algorithms, and BP neural networks can effectively improve the accuracy of ship engine fault diagnosis and locate the specific faulty components, ensuring the accuracy of input data for reliability analysis, effectively shortening the lubricating oil collection cycle, avoiding the problem of rapid deterioration of lubricating oil due to long sampling intervals, and accurately and continuously analyzing the health status trend of lubricating oil, improving the consistency and accuracy of lubricating oil analysis. At the same time, it can generate clear and easy-to-understand condition-based maintenance plans and checklists, dynamically adjust maintenance strategies according to the actual operating status and usage environment of the ship engine, and have flexibility and pertinence, thereby achieving efficient condition-based maintenance of ship engines.

[0053] The present invention also relates to a ship engine condition-based maintenance system, which corresponds to the above-mentioned ship engine condition-based maintenance method and can be understood as a system for realizing the above-mentioned ship engine condition-based maintenance method, including a data acquisition and preprocessing module, a mechanical component fault diagnosis module, a lubricating oil fault diagnosis module, a mechanical component reliability analysis module, a lubricating oil reliability analysis module and an engine condition-based maintenance module, the data acquisition and preprocessing module is respectively connected to the mechanical component fault diagnosis module and the lubricating oil fault diagnosis module, the mechanical component fault diagnosis module is respectively connected to the mechanical component reliability analysis module, the lubricating oil fault diagnosis module is respectively connected to the lubricating oil reliability analysis module, the engine condition-based maintenance module is respectively connected to the mechanical component reliability analysis module and the lubricating oil reliability analysis module, and the modules work together to realize the real-time data acquisition and preprocessing of the ship engine condition-based maintenance system. The application of technologies such as data collection and preprocessing, fault diagnosis mechanism model, Weibull distribution, genetic algorithm, BP neural network, etc. can improve the accuracy of fault diagnosis, optimize maintenance strategy, accurately predict the remaining service time, effectively improve the accuracy of ship engine fault diagnosis and locate the specific components of the fault, ensure the accuracy of input data for reliability analysis, effectively shorten the lubricating oil collection cycle, avoid the problem of rapid deterioration of lubricating oil due to long sampling intervals, and accurately and continuously analyze the health status trend of lubricating oil, improve the consistency and accuracy of lubricating oil analysis, and generate clear and easy-to-understand condition-based maintenance plans and inspection lists, and dynamically adjust the maintenance strategy according to the actual operating status and usage environment of the ship engine, which is flexible and targeted, thereby realizing efficient condition-based maintenance of ship engines. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The present invention is a flow chart of a method for condition-based maintenance of a marine engine.

[0055] Figure 2 It is a schematic diagram of the fault entry and query interface of the present invention.

[0056] Figure 3a and Figure 3b They are respectively the flow charts of reliability analysis of lubricating oil and mechanical components of the present invention.

[0057] Figure 4 It is a schematic diagram of a pump reliability curve of the present invention.

[0058] Figure 5a and Figure 5b They are respectively flow charts of the condition-based maintenance plan for mechanical parts and lubricating oil according to the present invention.

[0059] Figure 6 It is a schematic diagram of the equipment maintenance plan and the checklist for inspection according to the present invention. DETAILED DESCRIPTION

[0060] The present invention will be described below with reference to the accompanying drawings.

[0061] The present invention relates to a condition-based maintenance method for marine engines. The method utilizes a combination of engine thermal parameters, vibration parameters, and thermal parameters. First, the collected parameters undergo data processing, such as outlier elimination and eigenvalue extraction. Fault diagnosis is then performed using a variety of methods, primarily based on vibration analysis, thermal parameter analysis, lubricating oil index analysis, and mechanism models. After a fault is identified, the user is required to provide feedback on the accuracy and details of the fault. When the generated fault is marked as correct, the fault diagnosis result and accompanying information are used as historical fault experience values for reliability analysis. To address the possibility of missed faults, fault-related information can also be manually recorded, improving the accuracy of reliability analysis. Based on the characteristics of different engine components, two different approaches are used for reliability analysis: for mechanical components such as turbochargers, pumps, and intake and exhaust valves, a reliability analysis method combining a genetic algorithm (GA) with a maximum likelihood method and a three-parameter Weibull distribution is employed. Specifically, the genetic algorithm and maximum likelihood method are combined to solve for the three parameters of the Weibull distribution. The component reliability is then calculated, thereby determining the remaining service life of the equipment before the next maintenance. For engine lubricating oil, the genetic algorithm (GA) + BP neural network method is used to comprehensively predict the changing trend of key performance indicators of lubricating oil through engine power, speed and ambient temperature, and then solve the remaining service life of lubricating oil based on the changing trend. Based on the reliability analysis results of the engine's cumulative operating time, mechanical components and lubricating oil, a comprehensive condition-based maintenance plan and checklist are generated to guide users to perform condition-based maintenance on the engine. The flow chart of this method is as follows Figure 1 As shown, the following steps are included in sequence:

[0062] 1. Data collection and preprocessing steps: Real-time collection of raw state data of the ship engine's lubricating oil and multiple mechanical components, and preprocessing of the raw state data; the raw state data includes thermal parameters, oil index parameters, and vibration parameters.

[0063] Specifically, the sensors and some operating information of the ship engine itself are communicated with the engine room monitoring and alarm system through Ethernet to collect the original status data of the ship engine's lubricating oil and multiple mechanical components in real time, such as Figure 1As shown, status data is collected. The raw status data includes thermal parameters, oil parameters (physical, chemical, and non-physical and chemical parameters), and vibration parameters. Vibration analysis is an effective means of monitoring moving component failures. To ensure more accurate fault diagnosis results and reduce false alarms and missed reports, an online acceleration vibration information acquisition device has been installed on core mechanical components such as the engine turbocharger, oil pump, fuel pump, cylinder liner, and cylinder head to collect vibration parameters at an 8kHz frequency. The online lubricating oil monitoring device is an effective means of monitoring lubricating oil quality and engine wear. Physicochemical parameters can indicate the degree of lubricating oil performance degradation, while particulate matter indicators can indicate engine wear, indirectly reflecting the degree of lubricating oil performance degradation. The online lubricating oil monitoring device collects oil parameters and assesses lubricating oil condition using viscosity at 100°C, moisture, total iron content, and non-total iron content. Monitoring is performed every 10 minutes. Preferably, mechanical components include the engine turbocharger, oil pump, fuel pump, exhaust valve, intake valve, cylinder liner, and cylinder head; thermal parameters include engine exhaust temperature, cylinder liner temperature, cooling water pressure, and cooling water temperature; oil index parameters include oil temperature, viscosity, total iron content, and non-total iron content; and vibration parameters include turbocharger vibration parameters and oil pump vibration parameters. The collected raw engine status data is the foundation for the engine condition-based maintenance system.

[0064] Since marine engines operate under complex operating conditions and face numerous interfering factors, data preprocessing is required for data application, primarily including outlier removal and feature extraction. Data acquisition can present issues such as disconnections and over-range conditions, which can significantly interfere with the true data and feature extraction, necessitating the removal of these outliers. Furthermore, vibration parameters are acquired at high frequencies, so feature extraction is typically performed prior to application. Therefore, outliers are first removed from the raw data. Feature extraction is then performed on the remaining vibration parameters and oil index parameters. Vibration characteristic values, such as the mean, effective value, maximum, minimum, standard deviation, kurtosis, and FFT global amplitude, are extracted from the remaining raw data. These indicators can effectively identify equipment faults and resonance conditions. To minimize significant fluctuations in lubricating oil status information, the average of six hourly results for each oil index parameter is used for subsequent reliability analysis.

[0065] 2. Mechanical component fault diagnosis steps: First, based on the mechanical component operation logic and fault mechanism, the pre-processed thermal parameters and vibration parameters are selected as the diagnostic indicators of the mechanical component, and a fault diagnosis mechanism model is established based on the diagnostic indicators. Then, the diagnostic indicators of the mechanical component under different working conditions when the engine is in a healthy state are pre-processed by statistical analysis and feature extraction to obtain the benchmark values of the diagnostic indicators under different engine working conditions. The fault diagnosis mechanism model is calibrated by the benchmark values, and the real-time diagnostic indicators of a mechanical component under a certain working condition are input into the calibrated fault diagnosis mechanism model, and the predicted value of the mechanical component under the working condition is output. The difference between the predicted value of the mechanical component and the real-time diagnostic indicator is compared with the preset threshold to determine whether the mechanical component has a fault. If the difference is less than or equal to the preset threshold (that is, the difference does not exceed the preset threshold), it is determined that no fault has occurred; if the difference is greater than the preset threshold, it is determined that the mechanical component has a fault, and the time of each historical fault of the mechanical component is obtained to generate a fault history data set. The fault entry and query interface example is as follows: Figure 2 As shown in the figure, the fault information for the main lubricating oil pump of engine No. 1 includes the time each failure occurred and the time it was repaired. The user then needs to determine whether the fault diagnosis results are correct. If the user determines that the fault diagnosis is correct, the parameter information at the time of the failure, namely the time each mechanical component failure occurred, will be used as the basis for subsequent reliability analysis.

[0066] 3. Lubricating oil fault diagnosis steps: Select the pre-processed thermal parameters and oil indicators as lubricating oil diagnostic indicators, and compare each indicator in the lubricating oil diagnostic indicators with the corresponding preset lubricating oil failure threshold to determine whether the lubricating oil has failed. If an indicator in the lubricating oil diagnostic indicators is greater than its corresponding preset lubricating oil failure threshold, the lubricating oil is determined to have failed. For example, if the water content exceeds 10,000 ppm, the lubricating oil is determined to have failed. If all indicators in the lubricating oil diagnostic indicators are less than or equal to their corresponding preset lubricating oil failure thresholds, the lubricating oil is determined to be valid.

[0067] It is important to note that the order of the mechanical component fault diagnosis steps and the lubricating oil fault diagnosis steps described above can be reversed; both constitute fault diagnosis. Based on vibration analysis, oil analysis, thermal parameter analysis, and other techniques, combined with various mechanism models and intelligent methods, the system performs fault diagnosis on abnormal engine conditions, locates the fault, and outputs the fault diagnosis results. Of course, the order of the mechanical component reliability analysis steps and the lubricating oil reliability analysis steps described below can also be reversed; both constitute reliability analysis.

[0068] 4. Reliability analysis steps for mechanical components: Based on the fault history data set and using the three-parameter Weibull distribution, the reliability function and probability density function of the time when the mechanical component fails are constructed respectively, and then the log-likelihood function is established based on the probability density function and the maximum likelihood function; the genetic algorithm is used to calculate the optimal solution of the log-likelihood function, and then the optimal scale parameter, optimal shape parameter and optimal location parameter are obtained. According to the optimal scale parameter, optimal shape parameter and optimal location parameter, as well as the preset target reliability, the reliability function is used to calculate the remaining service life of the mechanical component after each failure, which is used as the reliability analysis result of the mechanical component.

[0069] This step can also be understood as a reliability analysis of mechanical components based on a statistical model, using a reliability analysis method that uses a genetic algorithm (GA) + maximum likelihood method + (scale parameter, shape parameter, location parameter) three-parameter Weibull distribution to solve the reliability of the mechanical components and then calculate the remaining service life of the mechanical components after each failure, that is, the remaining service life until the next maintenance. Specifically, if Figure 3b As shown in the figure, assuming that the distribution of fault information conforms to the Weibull distribution, the reliability function and probability density function of the time when the mechanical component fails are constructed based on the fault history data set and by adopting the three-parameter Weibull distribution. The reliability function is shown as follows:

[0070]

[0071] The probability density function is shown below:

[0072]

[0073] Among them, α is the scale parameter, β is the shape parameter, and γ is the location parameter.

[0074] Then, based on the probability density function and the maximum likelihood function, the log-likelihood function is established. That is, formula (2) is first substituted into the maximum likelihood function to obtain the maximum likelihood function for the three-parameter Weibull distribution, as shown below:

[0075]

[0076] Taking the logarithm of both sides gives the log-likelihood function, as shown below:

[0077]

[0078] Then, a genetic algorithm (GA algorithm) is used to calculate the optimal solution of the log-likelihood function to obtain the optimal three parameters α, β, and γ. The specific steps include:

[0079] S1, initialize the population and encode: perform population initialization assignment for the three parameters α, β, and γ, that is, randomly initialize the scale parameter α, shape parameter β, and location parameter γ in the log-likelihood function, obtain the initial values of the scale parameter, shape parameter, and location parameter, respectively, and encode the initial values of the scale parameter, shape parameter, and location parameter respectively to generate the initial population;

[0080] S2, fitness value calculation: the log-likelihood function, i.e., formula (4), is used as the fitness function of the genetic algorithm (GA), and the fitness value of each individual in the initial population is calculated based on the fitness function;

[0081] S3, selection: select multiple individuals from the initial population as parent individuals based on fitness values and using the roulette wheel selection method;

[0082] S4, Crossover: Determine whether to perform a crossover operation based on the randomly generated probability value and the preset crossover rate. If yes, select the single-point crossover method, randomly select two individuals from the parent individuals, then randomly select two time units with relative positions from the two individuals, and swap the codes of these two time units to generate two offspring individuals; if no, keep the number of parent individuals unchanged, and use all individuals in the parent individuals as offspring individuals;

[0083] S5: Repeat step S4 until the preset number of iterations is reached to obtain multiple offspring individuals;

[0084] S6, mutation: perform mutation operation on each offspring individual to obtain multiple new offspring individuals;

[0085] S7: Calculate the fitness value of each new offspring individual according to the fitness function, and determine whether the preset number of iterations is reached. If so, take the individual with the largest fitness value as the optimal solution of the log-likelihood function; if not, repeat steps S3 to S6 until the preset number of iterations is reached. Take the individual with the largest fitness value when the preset number of iterations is reached as the optimal solution of the log-likelihood function, and then obtain the optimal scale parameter, optimal shape parameter and optimal position parameter.

[0086] After obtaining the optimal scale parameter, optimal shape parameter and optimal position parameter, substitute the optimal scale parameter, optimal shape parameter and optimal position parameter into the reliability function and select multiple operating time points to calculate the corresponding reliability, and then obtain the reliability curve of the mechanical component, that is, bring it into formula (1), and accordingly obtain the relationship curve between the reliability of each mechanical component and the time of each failure. Taking the failure data of a pump shown in Table 1 as an example, the genetic algorithm parameter settings are shown in Table 2. The optimal three parameters obtained according to the above method are α=4815, β=1.75, and γ=286. The corresponding reliability curve is shown as follows Figure 4 As shown in the figure, assuming that the reliability is 80% and the mechanical parts are reliable, the remaining service time to the next maintenance ( Figure 4 The next condition-based maintenance check is due at 2329 hours.

[0087] Table 1

[0088] Serial number 1 2 3 4 5 6 7 8 9 10 Time / h 708 1445 2090 2785 3572 4328 5115 5963 7210 8735

[0089] Table 2

[0090] Genetic Algebra Population size Crossover rate mutation rate 100 50 0.75 0.08

[0091] It should be noted that the reliability analysis, i.e., the remaining service life after each mechanical component failure, is recalculated each time a mechanical component failure is added to the failure history data set. If the failure history data remains constant, the remaining service life (lifespan) derived from the reliability analysis gradually decreases as the engine's operating time increases.

[0092] 5. Lubricating oil reliability analysis steps: After determining that the lubricating oil has failed, randomly initialize the weights and bias values of the BP neural network to obtain the initial weights and initial bias values; and use the genetic algorithm to calculate the optimal weights and optimal bias of the BP neural network, and use the optimal weights and optimal bias to update the initial weights and initial bias values in the BP neural network to obtain a prediction model, and input the pre-processed thermal parameters and oil indicators at the current moment as input indicators into the prediction model to predict the viscosity, moisture and particulate matter content increments of the lubricating oil at the next moment after the current moment; and respectively predict the viscosity, moisture and particulate matter content increments at multiple consecutive moments after the next moment, and then calculate the current The average viscosity increment, the average moisture increment and the average particle content increment at all moments after the moment are calculated; the remaining usage time of the viscosity is calculated based on the average viscosity increment, the actual viscosity measured at each moment and the preset viscosity threshold, the remaining usage time of the moisture is calculated based on the average moisture increment, the actual moisture measured at each moment and the preset moisture threshold, and the remaining usage time of the particle content is calculated based on the average particle content increment, the actual particle content measured at each moment and the preset particle content threshold, and the minimum value of the remaining usage time of the viscosity, moisture and particle content is taken as the remaining usage time of the lubricating oil, which is used as the lubricating oil reliability analysis result.

[0093] This step is also called the lubricating oil reliability analysis method using genetic algorithm (GA) + BP neural network to predict the changing trend of key lubricating oil indicators, and then calculate the remaining service life of the lubricating oil in combination with the failure threshold of the lubricating oil. Specifically, Figure 3aAs shown, the network topology is first determined. Based on the characteristics of lubricating oil failure, the main factors affecting lubricating oil failure are inherent performance degradation, intrusion of external contaminants, and the impact of wear particles. This is reflected in lubricating oil performance primarily through physical and chemical properties and the number of wear particles. Specifically, the increase in viscosity @100°C (i.e., viscosity at 100°C), moisture content, and particulate matter content are selected as predictive indicators. Based on the engine's operating mechanism, the main factors affecting lubricating oil are engine power, engine speed, lubricating oil temperature, exhaust temperature, lubricating oil pressure, and cooling water temperature. These indicators serve as inputs. Furthermore, to improve prediction accuracy and efficiency, the increase in viscosity @100°C, moisture content, and particulate matter content are predicted separately. Since the prediction is not complex, a single hidden layer neural network is sufficient.

[0094] Then, the weights and bias values of the BP neural network are randomly initialized to obtain the initial weights and initial bias values, and the root mean square error RMSE is selected as the fitness function of the genetic algorithm (GA algorithm). Then, the genetic algorithm is used to calculate the optimal weights and optimal bias of the BP neural network, that is, selection, crossover, mutation and optimization are performed to determine the optimal search direction and obtain the optimal solution of the weights and bias values. The above process has been introduced above and will not be repeated here; finally, the optimal weights and optimal bias are used to update the initial weights and initial bias values in the BP neural network to obtain the prediction model.

[0095] After obtaining the prediction model, the pre-processed thermal parameters and oil indicators at the current moment are input into the prediction model as input indicators. The viscosity @100°C, moisture content, and particulate matter content increments of the lubricating oil at the next moment after the current moment are predicted. The viscosity, moisture content, and particulate matter content increments at multiple consecutive moments after the next moment are also predicted. That is, the viscosity, moisture content, and particulate matter content increments are predicted for five consecutive moments from the current moment, and the average values are calculated to obtain the average viscosity increment, the average moisture increment, and the average particulate matter content increment. Based on the threshold values of viscosity @100°C, moisture content, and particulate matter content at which the lubricating oil fails, the remaining service life of the viscosity @100°C, moisture content, and particulate matter content is calculated, and the minimum value is taken as the remaining service life of the lubricating oil. The reliability analysis is performed once every fixed time and the reliability analysis results are obtained.

[0096] Example:

[0097] The GA-BP neural network model parameters are shown in Table 3. The historical data set consists of 100 sets. The oil failure thresholds are: oil viscosity at 100°C: 10 mm² / s; moisture: 3‰; and particulate matter content: 2%. Oil performance indicators and reliability analysis results are updated hourly. The three indicators measured by the system in real time at each moment are: actual viscosity @100℃: 14.6mm2 / s; actual moisture: 1.2‰; actual particulate matter content: 0.8%; the incremental average values predicted by the prediction model are: viscosity @100℃ incremental average value: -0.0073mm2 / s; moisture incremental average value: 0.0023‰; particulate matter content incremental average value: 0.0017%; the remaining service life of the three indicators finally obtained are: viscosity @100℃: 630 hours; moisture: 783 hours; particulate matter content: 706 hours; taking the minimum value of the remaining service time of viscosity, moisture and particulate matter content as the remaining service time of the lubricating oil, the remaining service time (remaining service life) of the lubricating oil is 630 hours.

[0098] Table 3

[0099]

[0100] 6. Engine condition-based maintenance steps: Generate a condition-based maintenance plan for mechanical components and a condition-based maintenance plan for lubricating oil containing a checklist of inspection items based on the remaining service life of each mechanical component and the remaining service life of the lubricating oil after each failure. Then, implement condition-based maintenance of the ship engine based on the condition-based maintenance plan for mechanical components and the condition-based maintenance plan for lubricating oil.

[0101] Specifically, the condition-based maintenance plan is mainly based on the relevant results of reliability analysis, and includes two parts: the condition-based maintenance plan and the inspection item list. The condition-based maintenance plan mainly shows the estimated remaining service life of the equipment. When the remaining time of the equipment is very short, it needs to be included in the inspection item list and needs to be checked urgently. In order to ensure the accuracy of the condition-based maintenance plan and subsequent optimization, feedback on the inspection results is required, and the condition-based maintenance plan will also be adjusted based on the feedback results. The logical flow of the condition-based maintenance plan is as follows: Figure 5a and Figure 5b As shown, in Figure 5a In the inspection, when the mechanical equipment inspection result in the inspection item list is normal, the remaining use time of the mechanical equipment is reset to the time used by the last fault sample. When the inspection result is abnormal, the fault data is used as the new historical fault data and the new remaining use time is recalculated. Figure 5b In the process, no matter whether the lubricating oil result is normal or abnormal, the physical and chemical information of the test will be recorded in the system and the new remaining service time will be recalculated.

[0102] In this embodiment, the information presented is the remaining usage time of the device and the items to be checked, such as Figure 5a As shown, the situation-based maintenance plan for mechanical parts specifically includes: first determine whether the remaining usage time of a certain mechanical part (or mechanical equipment) after a certain failure is less than a preset time threshold (for example, the preset time is 48 hours); if so, include the mechanical part in the inspection item list for inspection, and determine whether the inspection result of the mechanical part is normal; if the inspection result is normal, reset the remaining usage time of the mechanical part (or mechanical equipment) to the remaining usage time after the previous failure; if the inspection result is abnormal, record the current time as the time when the mechanical part fails, and store the current time in the fault history data set to form a new fault history data set; based on the new fault history data set and continuing to perform reliability analysis again according to the working principle of the mechanical part reliability analysis step, recalculate the new remaining usage time of the mechanical part, and the mechanical part maintenance plan and the inspection list examples are shown as follows. Figure 6 As shown, the estimated remaining service life and maintenance recommendations of each mechanical component when the reliability is greater than 80% are displayed.

[0103] like Figure 5b As shown in FIG, the lubricating oil condition-based maintenance plan specifically includes: determining whether the remaining service life of the lubricating oil is less than a preset time threshold (for example, the preset time is 48 hours); if so, the lubricating oil is included in the inspection item list for inspection, and after the inspection, the actual viscosity, actual moisture and actual particulate matter content measured at the current moment are recorded, and the reliability analysis is continued according to the working principle of the lubricating oil reliability analysis step to recalculate the new remaining service life of the lubricating oil.

[0104] The present invention also relates to a ship engine condition-based maintenance system, which corresponds to the above-mentioned ship engine condition-based maintenance method and can be understood as a system for implementing the above-mentioned method. The system includes a data acquisition and preprocessing module, a mechanical component fault diagnosis module, a lubricating oil fault diagnosis module, a mechanical component reliability analysis module, a lubricating oil reliability analysis module, and an engine condition-based maintenance module. The data acquisition and preprocessing module is respectively connected to the mechanical component fault diagnosis module and the lubricating oil fault diagnosis module, the mechanical component fault diagnosis module is connected to the mechanical component reliability analysis module, the lubricating oil fault diagnosis module is connected to the lubricating oil reliability analysis module, and the engine condition-based maintenance module is respectively connected to the mechanical component reliability analysis module and the lubricating oil reliability analysis module. Specifically,

[0105] The data acquisition and preprocessing module collects the original state data of the lubricating oil and multiple mechanical components of the ship engine in real time and preprocesses the original state data; the original state data includes thermal parameters, oil index parameters and vibration parameters;

[0106] The mechanical component fault diagnosis module selects pre-processed thermal parameters and vibration parameters as diagnostic indicators of the mechanical component based on the mechanical component operation logic and fault mechanism, establishes a fault diagnosis mechanism model based on the diagnostic indicators, and then performs statistical analysis and interpolation fitting processing on the diagnostic indicators of the mechanical component under different working conditions when the engine is in a healthy state to obtain benchmark values of the diagnostic indicators of the mechanical component under different working conditions of the engine, calibrates the fault diagnosis mechanism model based on the benchmark values, and inputs the real-time diagnostic indicator of a mechanical component under a certain working condition into the calibrated fault diagnosis mechanism model, outputs a predicted value of the mechanical component under the working condition, and compares the difference between the predicted value of the mechanical component and the real-time diagnostic indicator with a preset threshold to determine whether the mechanical component has a fault. If the difference is less than or equal to the preset threshold, it is determined that no fault has occurred; if the difference is greater than the preset threshold, it is determined that the mechanical component has a fault, and obtains the time of each historical fault of the mechanical component to generate a fault history data set;

[0107] The lubricating oil fault diagnosis module selects the pre-processed thermal parameters and oil indicators as lubricating oil diagnostic indicators, and compares each indicator in the lubricating oil diagnostic indicators with its corresponding preset lubricating oil failure threshold to determine whether the lubricating oil has failed. If any indicator in the lubricating oil diagnostic indicators is greater than its corresponding preset lubricating oil failure threshold, the lubricating oil is determined to have failed; if all indicators in the lubricating oil diagnostic indicators are less than or equal to their corresponding preset lubricating oil failure thresholds, the lubricating oil is determined to be in a valid state.

[0108] The mechanical component reliability analysis module constructs a reliability function and a probability density function for the time when the mechanical component fails based on a fault history data set and a three-parameter Weibull distribution, and then establishes a log-likelihood function based on the probability density function and the maximum likelihood function. A genetic algorithm is used to calculate the optimal solution of the log-likelihood function, thereby obtaining the optimal scale parameter, optimal shape parameter, and optimal location parameter. Based on the optimal scale parameter, optimal shape parameter, and optimal location parameter, as well as a preset target reliability, the reliability function is used to calculate the remaining service life of the mechanical component after each failure, and this is used as the reliability analysis result of the mechanical component.

[0109] The lubricating oil reliability analysis module randomly initializes the weights and bias values of the BP neural network after determining that the lubricating oil has failed, and obtains the initial weights and initial bias values; and uses a genetic algorithm to calculate the optimal weights and optimal bias of the BP neural network, and uses the optimal weights and optimal bias to update the initial weights and initial bias values in the BP neural network to obtain a prediction model, and inputs the pre-processed thermal parameters and oil indicators at the current moment as input indicators into the prediction model to predict the viscosity, moisture and particulate matter content increments of the lubricating oil at the next moment after the current moment; and respectively predicts the viscosity, moisture and particulate matter content increments at multiple consecutive moments after the next moment, and then calculates the current moment. The average viscosity increment, the average moisture increment, and the average particle content increment at all moments after the moment; then calculating the remaining service life of the viscosity based on the average viscosity increment, the actual viscosity measured at each moment, and the preset viscosity threshold; calculating the remaining service life of the moisture based on the average moisture increment, the actual moisture measured at each moment, and the preset moisture threshold; and calculating the remaining service life of the particle content based on the average particle content increment, the actual particle content measured at each moment, and the preset particle content threshold; and taking the minimum value of the remaining service life of the viscosity, moisture, and particle content as the remaining service life of the lubricating oil, which is used as the lubricating oil reliability analysis result;

[0110] The engine condition-based maintenance module generates a mechanical component condition-based maintenance plan and a lubricating oil condition-based maintenance plan containing a checklist of inspection items based on the remaining service life of the mechanical component and the remaining service life of the lubricating oil after each failure, and then implements condition-based maintenance of the ship engine according to the mechanical component condition-based maintenance plan and the lubricating oil condition-based maintenance plan.

[0111] Preferably, in the mechanical component reliability analysis module, using a genetic algorithm to calculate the optimal solution of the log-likelihood function specifically includes:

[0112] S1: Randomly initialize the scale parameter, shape parameter, and location parameter in the log-likelihood function to obtain the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter, respectively, and encode the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter to generate an initial population;

[0113] S2: Use the log-likelihood function as the fitness function of the genetic algorithm, and calculate the fitness value of each individual in the initial population based on the fitness function;

[0114] S3: Select multiple individuals from the initial population as parent individuals based on their fitness values using the roulette wheel selection method;

[0115] S4: Determine whether to perform a crossover operation based on the randomly generated probability value and the preset crossover rate. If so, randomly select two individuals from the parent individuals, then randomly select two time units with relative positions from the two individuals, and exchange the codes of these two time units to generate two offspring individuals; if not, keep the number of parent individuals unchanged, and use all individuals in the parent individuals as offspring individuals;

[0116] S5: Repeat step S4 until the preset number of iterations is reached to obtain multiple offspring individuals;

[0117] S6: Perform mutation operation on each offspring individual to obtain multiple new offspring individuals;

[0118] S7: Calculate the fitness value of each new offspring individual according to the fitness function, and determine whether the preset number of iterations is reached. If so, take the individual with the largest fitness value as the optimal solution of the log-likelihood function; if not, repeat steps S3 to S6 until the preset number of iterations is reached. Take the individual with the largest fitness value when the preset number of iterations is reached as the optimal solution of the log-likelihood function, and then obtain the optimal scale parameter, optimal shape parameter and optimal position parameter.

[0119] Preferably, in the engine condition-based maintenance module, the condition-based maintenance plan for mechanical components specifically includes: determining whether the remaining usage time of a certain mechanical component after a certain failure is less than a preset time threshold; if so, listing the mechanical component in the inspection item list for inspection, and determining whether the inspection result of the mechanical component is normal; if the inspection result is normal, resetting the remaining usage time of the mechanical equipment to the remaining usage time after the previous failure; if the inspection result is abnormal, recording the current time as the time when the mechanical component fails, and storing the current time in the fault history data set to form a new fault history data set; based on the new fault history data set and continuing in accordance with the working principle of the mechanical component reliability analysis module, recalculating the new remaining usage time of the mechanical component.

[0120] Preferably, the lubricating oil condition-based maintenance plan specifically includes: determining whether the remaining service life of the lubricating oil is less than a preset time threshold; if so, listing the lubricating oil in the inspection item list for inspection, and recording the actual viscosity, actual moisture and actual particulate matter content measured at the current moment after the inspection, and continuing to recalculate the new remaining service life of the lubricating oil according to the working principle of the lubricating oil reliability analysis module.

[0121] The present invention provides an objective and scientific method and system for condition-based maintenance of ship engines. By applying technologies such as real-time data acquisition and preprocessing, fault diagnosis mechanism models, Weibull distribution, genetic algorithms, and BP neural networks, the method can improve the accuracy of fault diagnosis, optimize maintenance strategies, and accurately predict the remaining service life, thereby achieving efficient condition-based maintenance of ship engines. It can also locate the specific components of the fault, ensure the accuracy of the input data for reliability analysis, and generate clear and easy-to-understand condition-based maintenance plans and checklists.

[0122] It should be noted that the specific embodiments described above can enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although this specification has described the present invention in detail with reference to the drawings and embodiments, those skilled in the art should understand that the present invention can still be modified or replaced with equivalents. In short, all technical solutions and improvements that do not depart from the spirit and scope of the present invention should be included in the scope of protection of the patent for the present invention.

Claims

1. A ship engine condition-based maintenance method, characterized in that: The following steps are involved: Data collection and preprocessing steps: real-time collection of raw status data of the ship's engine's lubricating oil and multiple mechanical components, and preprocessing of the raw status data; The original state data includes thermal parameters, oil index parameters and vibration parameters; Mechanical component fault diagnosis steps: Based on the mechanical component operation logic and fault mechanism, pre-processed thermal parameters and vibration parameters are selected as diagnostic indicators of the mechanical component, and a fault diagnosis mechanism model is established based on the diagnostic indicators. Then, the diagnostic indicators of the mechanical component under different working conditions when the engine is in a healthy state are statistically analyzed and interpolated and fitted to obtain the benchmark values of the diagnostic indicators of the mechanical component under different engine working conditions. The fault diagnosis mechanism model is calibrated using the benchmark values, and the real-time diagnostic indicator of a mechanical component under a certain working condition is input into the calibrated fault diagnosis mechanism model. The predicted value of the mechanical component under the working condition is output, and the difference between the predicted value and the real-time diagnostic indicator of the mechanical component is compared with a preset threshold value to determine whether the mechanical component has a fault. If the difference is less than or equal to the preset threshold value, it is determined that no fault has occurred. If the difference is greater than a preset threshold, it is determined that the mechanical component has failed, and the time of each historical failure of the mechanical component is obtained to generate a failure history data set; Lubricating oil fault diagnosis steps: selecting the pre-processed thermal parameters and oil indicators as lubricating oil diagnostic indicators, and comparing each indicator in the lubricating oil diagnostic indicators with its corresponding preset lubricating oil failure threshold to determine whether the lubricating oil has failed. If any indicator in the lubricating oil diagnostic indicators is greater than its corresponding preset lubricating oil failure threshold, the lubricating oil is determined to have failed; if all indicators in the lubricating oil diagnostic indicators are less than or equal to their corresponding preset lubricating oil failure thresholds, the lubricating oil is determined to be in a valid state; Mechanical component reliability analysis steps: Based on the failure history data set and using the three-parameter Weibull distribution, a reliability function and probability density function for the time when the mechanical component fails are constructed respectively. Then, a log-likelihood function is established based on the probability density function and the maximum likelihood function. A genetic algorithm is used to calculate the optimal solution of the log-likelihood function, and then the optimal scale parameter, optimal shape parameter, and optimal location parameter are obtained. Based on the optimal scale parameter, optimal shape parameter, optimal location parameter, and the preset target reliability, the reliability function is used to calculate the remaining service life of the mechanical component after each failure, which is used as the reliability analysis result of the mechanical component. Lubricating oil reliability analysis steps: after determining that the lubricating oil has failed, randomly initialize the weights and bias values of the BP neural network to obtain the initial weights and initial bias values; and use the genetic algorithm to calculate the optimal weights and optimal bias of the BP neural network, and use the optimal weights and optimal bias to update the initial weights and initial bias values in the BP neural network to obtain a prediction model, and input the pre-processed thermal parameters and oil indicators at the current moment as input indicators into the prediction model to predict the viscosity, moisture and particulate matter content increments of the lubricating oil at the next moment after the current moment; and respectively predict the viscosity, moisture and particulate matter content increments at multiple consecutive moments after the next moment, and then calculate the current moment. The average viscosity increment, the average moisture increment, and the average particle content increment at all moments after the moment are calculated; the remaining viscosity usage time is calculated based on the average viscosity increment, the actual viscosity measured at each moment, and a preset viscosity threshold; the remaining moisture usage time is calculated based on the average moisture increment, the actual moisture measured at each moment, and a preset moisture threshold; and the remaining particle content usage time is calculated based on the average particle content increment, the actual particle content measured at each moment, and a preset particle content threshold; and the minimum value of the remaining viscosity, moisture, and particle content usage times is taken as the remaining lubricating oil usage time, which is used as the lubricating oil reliability analysis result; Engine condition-based maintenance steps: Generate a mechanical component condition-based maintenance plan and a lubricating oil condition-based maintenance plan containing a checklist of inspection items based on the remaining service life of the mechanical component and the remaining service life of the lubricating oil after each failure, and then implement condition-based maintenance of the ship engine according to the mechanical component condition-based maintenance plan and the lubricating oil condition-based maintenance plan.

2. The marine engine condition-based maintenance method according to claim 1, characterized in that: The order of the mechanical component fault diagnosis step and the lubricating oil fault diagnosis step can be interchanged; And / or, the order of the mechanical component reliability analysis step and the lubricating oil reliability analysis step can be interchanged.

3. The ship engine condition-based maintenance method according to claim 1 or 2, characterized in that: In the mechanical component reliability analysis step, the step of calculating the optimal solution of the log-likelihood function using a genetic algorithm specifically includes: S1: Randomly initialize the scale parameter, shape parameter, and location parameter in the log-likelihood function to obtain the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter, respectively, and encode the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter to generate an initial population; S2: Use the log-likelihood function as the fitness function of the genetic algorithm, and calculate the fitness value of each individual in the initial population based on the fitness function; S3: Select multiple individuals from the initial population as parent individuals based on their fitness values using the roulette wheel selection method; S4: Determine whether to perform a crossover operation based on the randomly generated probability value and the preset crossover rate. If so, randomly select two individuals from the parent individuals, then randomly select two time units with relative positions from the two individuals, and exchange the codes of these two time units to generate two offspring individuals; if not, keep the number of parent individuals unchanged, and use all individuals in the parent individuals as offspring individuals; S5: Repeat step S4 until the preset number of iterations is reached to obtain multiple offspring individuals; S6: Perform mutation operation on each offspring individual to obtain multiple new offspring individuals; S7: Calculate the fitness value of each new offspring individual according to the fitness function, and determine whether the preset number of iterations is reached. If so, take the individual with the largest fitness value as the optimal solution of the log-likelihood function; if not, repeat steps S3 to S6 until the preset number of iterations is reached. Take the individual with the largest fitness value when the preset number of iterations is reached as the optimal solution of the log-likelihood function, and then obtain the optimal scale parameter, optimal shape parameter and optimal position parameter.

4. The marine engine condition-based maintenance method according to claim 1, characterized in that: In the engine condition-based maintenance step, the mechanical component condition-based maintenance plan specifically includes: Determine whether the remaining usage time of a mechanical component after a certain failure is less than a preset time threshold. If so, include the mechanical component in the inspection item list for inspection, and determine whether the inspection result of the mechanical component is normal. If the inspection result is normal, reset the remaining usage time of the mechanical equipment to the remaining usage time after the previous failure of the failure; if the inspection result is abnormal, record the current time as the time when the mechanical component fails, and store the current time in the fault history data set to form a new fault history data set. Based on the new fault history data set and continuing in accordance with the working principle of the mechanical component reliability analysis step, recalculate the new remaining usage time of the mechanical component.

5. The ship engine condition-based maintenance method according to claim 1, characterized in that: In the engine condition-based maintenance step, the lubricating oil condition-based maintenance plan specifically includes: Determine whether the remaining service life of the lubricating oil is less than the preset time threshold. If so, include the lubricating oil in the inspection item list for inspection, and after the inspection, record the actual viscosity, actual moisture content and actual particulate matter content measured at the current moment, and continue to recalculate the new remaining service life of the lubricating oil according to the working principle of the lubricating oil reliability analysis step.

6. The ship engine condition-based maintenance method according to claim 1, characterized in that: In the data acquisition and preprocessing step, the mechanical components include a combination of an engine turbocharger, an oil pump, a fuel pump, an exhaust valve, an intake valve, a cylinder liner, a cylinder head, and bearings; the thermal parameters include engine exhaust temperature, cylinder liner temperature, cooling water pressure, and cooling water temperature; the oil index parameters include oil temperature, viscosity, total iron content, and non-total iron content; and the vibration parameters include turbocharger vibration parameters and oil pump vibration parameters. The preprocessing of the original state data includes: removing outliers from the original state data, and extracting features from the vibration parameters and oil index parameters in the remaining original state data after the removal, extracting the average value, effective value, maximum value, minimum value, standard deviation, kurtosis and FFT global amplitude of each parameter in the vibration parameters; and extracting the average value of each parameter in the oil index parameters.

7. A ship engine condition-based maintenance system, characterized in that: It includes a data acquisition and preprocessing module, a mechanical component fault diagnosis module, a lubricating oil fault diagnosis module, a mechanical component reliability analysis module, a lubricating oil reliability analysis module, and an engine condition-based maintenance module. The data acquisition and preprocessing module is connected to the mechanical component fault diagnosis module and the lubricating oil fault diagnosis module respectively. The mechanical component fault diagnosis module is connected to the mechanical component reliability analysis module, the lubricating oil fault diagnosis module is connected to the lubricating oil reliability analysis module, and the engine condition-based maintenance module is connected to the mechanical component reliability analysis module and the lubricating oil reliability analysis module respectively. The data acquisition and preprocessing module collects the original status data of the lubricating oil and multiple mechanical components of the ship engine in real time and preprocesses the original status data; The original state data includes thermal parameters, oil index parameters and vibration parameters; The mechanical component fault diagnosis module selects pre-processed thermal parameters and vibration parameters as diagnostic indicators of the mechanical component based on the mechanical component operation logic and fault mechanism, establishes a fault diagnosis mechanism model based on the diagnostic indicators, and then performs statistical analysis and interpolation fitting processing on the diagnostic indicators of the mechanical component under different working conditions when the engine is in a healthy state to obtain benchmark values of the diagnostic indicators of the mechanical component under different working conditions of the engine, calibrates the fault diagnosis mechanism model based on the benchmark values, inputs the real-time diagnostic indicator of a mechanical component under a certain working condition into the calibrated fault diagnosis mechanism model, outputs a predicted value of the mechanical component under the working condition, and compares the difference between the predicted value and the real-time diagnostic indicator of the mechanical component with a preset threshold value to determine whether the mechanical component has a fault, and if the difference is less than or equal to the preset threshold value, determines that no fault has occurred; If the difference is greater than a preset threshold, it is determined that the mechanical component has failed, and the time of each historical failure of the mechanical component is obtained to generate a failure history data set; The lubricating oil fault diagnosis module selects the pre-processed thermal parameters and oil indicators as lubricating oil diagnostic indicators, and compares each indicator in the lubricating oil diagnostic indicators with its corresponding preset lubricating oil failure threshold to determine whether the lubricating oil has failed. If any indicator in the lubricating oil diagnostic indicators is greater than its corresponding preset lubricating oil failure threshold, the lubricating oil is determined to have failed; if all indicators in the lubricating oil diagnostic indicators are less than or equal to their corresponding preset lubricating oil failure thresholds, the lubricating oil is determined to be in a valid state. The mechanical component reliability analysis module constructs a reliability function and a probability density function for the time when the mechanical component fails based on a fault history data set and a three-parameter Weibull distribution, and then establishes a log-likelihood function based on the probability density function and the maximum likelihood function. A genetic algorithm is used to calculate the optimal solution of the log-likelihood function, thereby obtaining the optimal scale parameter, optimal shape parameter, and optimal location parameter. Based on the optimal scale parameter, optimal shape parameter, and optimal location parameter, as well as a preset target reliability, the reliability function is used to calculate the remaining service life of the mechanical component after each failure, and this is used as the reliability analysis result of the mechanical component. The lubricating oil reliability analysis module randomly initializes the weights and bias values of the BP neural network after determining that the lubricating oil has failed, and obtains the initial weights and initial bias values; and uses a genetic algorithm to calculate the optimal weights and optimal bias of the BP neural network, and uses the optimal weights and optimal bias to update the initial weights and initial bias values in the BP neural network to obtain a prediction model, and inputs the pre-processed thermal parameters and oil indicators at the current moment as input indicators into the prediction model to predict the viscosity, moisture and particulate matter content increments of the lubricating oil at the next moment after the current moment; and respectively predicts the viscosity, moisture and particulate matter content increments at multiple consecutive moments after the next moment, and then calculates the current moment. The average viscosity increment, the average moisture increment, and the average particle content increment at all moments after the moment; then calculating the remaining service life of the viscosity based on the average viscosity increment, the actual viscosity measured at each moment, and the preset viscosity threshold; calculating the remaining service life of the moisture based on the average moisture increment, the actual moisture measured at each moment, and the preset moisture threshold; and calculating the remaining service life of the particle content based on the average particle content increment, the actual particle content measured at each moment, and the preset particle content threshold; and taking the minimum value of the remaining service life of the viscosity, moisture, and particle content as the remaining service life of the lubricating oil, which is used as the lubricating oil reliability analysis result; The engine condition-based maintenance module generates a mechanical component condition-based maintenance plan and a lubricating oil condition-based maintenance plan containing a checklist of inspection items based on the remaining service life of the mechanical component and the remaining service life of the lubricating oil after each failure, and then implements condition-based maintenance of the ship engine according to the mechanical component condition-based maintenance plan and the lubricating oil condition-based maintenance plan.

8. The ship engine condition-based maintenance system according to claim 7, characterized in that: In the mechanical component reliability analysis module, the optimal solution of the log-likelihood function is calculated using a genetic algorithm, specifically including: S1: Randomly initialize the scale parameter, shape parameter, and location parameter in the log-likelihood function to obtain the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter, respectively, and encode the initial value of the scale parameter, the initial value of the shape parameter, and the initial value of the location parameter to generate an initial population; S2: Use the log-likelihood function as the fitness function of the genetic algorithm, and calculate the fitness value of each individual in the initial population based on the fitness function; S3: Select multiple individuals from the initial population as parent individuals based on their fitness values using the roulette wheel selection method; S4: Determine whether to perform a crossover operation based on the randomly generated probability value and the preset crossover rate. If so, randomly select two individuals from the parent individuals, then randomly select two time units with relative positions from the two individuals, and exchange the codes of these two time units to generate two offspring individuals; if not, keep the number of parent individuals unchanged, and use all individuals in the parent individuals as offspring individuals; S5: Repeat step S4 until the preset number of iterations is reached to obtain multiple offspring individuals; S6: Perform mutation operation on each offspring individual to obtain multiple new offspring individuals; S7: Calculate the fitness value of each new offspring individual according to the fitness function, and determine whether the preset number of iterations is reached. If so, take the individual with the largest fitness value as the optimal solution of the log-likelihood function; if not, repeat steps S3 to S6 until the preset number of iterations is reached. Take the individual with the largest fitness value when the preset number of iterations is reached as the optimal solution of the log-likelihood function, and then obtain the optimal scale parameter, optimal shape parameter and optimal position parameter.

9. The ship engine condition-based maintenance system according to claim 7 or 8, characterized in that: In the engine condition-based maintenance module, the mechanical component condition-based maintenance plan specifically includes: Determine whether the remaining usage time of a mechanical component after a certain failure is less than a preset time threshold. If so, include the mechanical component in the inspection item list for inspection, and determine whether the inspection result of the mechanical component is normal. If the inspection result is normal, reset the remaining usage time of the mechanical equipment to the remaining usage time after the previous failure of the failure; if the inspection result is abnormal, record the current time as the time when the mechanical component fails, and store the current time in the fault history data set to form a new fault history data set. Based on the new fault history data set and continuing in accordance with the working principle of the mechanical component reliability analysis module, recalculate the new remaining usage time of the mechanical component.

10. The ship engine condition-based maintenance system according to claim 7 or 8, characterized in that: In the engine condition-based maintenance module, the lubricating oil condition-based maintenance plan specifically includes: Determine whether the remaining service life of the lubricating oil is less than the preset time threshold. If so, include the lubricating oil in the inspection item list for inspection, and after the inspection, record the actual viscosity, actual moisture and actual particulate matter content measured at the current moment, and continue to recalculate the new remaining service life of the lubricating oil according to the working principle of the lubricating oil reliability analysis module.

Images