Marine engine self-adaptive aging prediction method and fault diagnosis model

Abstract

The invention relates to the field of marine engine performance monitoring and optimization, and discloses a marine engine adaptive aging prediction method and a fault diagnosis model.The prediction method comprises the following steps that a digital twin model of a marine engine is constructed, and an engine reference health state is generated; acquiring operation parameters of the marine engine in real time, and inputting the operation parameters into the digital twin model to calculate a reference performance index; identifying the degradation offset of the actuating mechanism by comparing the output of the digital twin model with the operation data of the entity engine; constructing a multi-objective optimization problem based on performance indexes of a digital twin model, and generating a Pareto optimal solution set by using an NSGA-III algorithm; and an optimal solution under a specific working condition is selected through a fuzzy comprehensive evaluation method, and degradation state evaluation and optimal control strategy generation are realized. Compared with the prior art, the method can accurately predict the degradation state of the marine engine, reduces the maintenance cost, improves the operation efficiency and reliability, and meets the dynamic optimization requirements of complex working conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of marine engine performance monitoring and optimization, in particular to a marine engine self-adaptive aging prediction method and a fault diagnosis model. BACKGROUND

[0002] Marine diesel engine, as the core power device of the shipping industry, is the key technical foundation to ensure the safe and efficient operation of ocean transportation. With its outstanding energy conversion efficiency and outstanding economic performance, it dominates the modern ship propulsion system. However, marine diesel engines face many challenges during long-term operation in the marine environment: first, the harsh marine environment, including seawater erosion, salt spray corrosion, etc.; second, the complex and variable working conditions, such as random vibration, load fluctuation, etc.; third, the coupling degradation of mechanical systems. According to the International Maritime Organization (IMO) statistics, diesel engine failures account for about 41.6% of marine accidents caused by mechanical failures, and maintenance costs account for 10% to 20% of the total value of the ship. Therefore, developing advanced predictive diagnosis methods, establishing scientific degradation prediction models, and innovating maintenance strategies have important engineering practical significance. This not only helps to optimize control parameters, predict maintenance time windows, and develop scientific maintenance plans, but also has important significance for improving energy conversion efficiency, enhancing system management capabilities, and promoting the intelligent and sustainable development of ship power systems.

[0003] During the continuous operation of marine diesel engines, the performance degradation of the fuel system and the intake and exhaust system will lead to the deviation of the air-fuel ratio matching relationship in the cylinder, and then affect the combustion characteristics and emission performance. Specifically, the degradation mechanism of the fuel system mainly manifests as: mechanical wear of key components such as needle valves, plungers, and springs in the fuel injector, as well as performance degradation of the fuel pump and elastic elements. In addition, the external fuel coking around the nozzle hole and the internal deposit accumulation in the nozzle needle valve area caused by long-term work will cause blockage of the fuel filter, fuel injector, and high-pressure fuel pump, which will have a significant adverse impact on the normal supply and injection process of the fuel. At the same time, the degradation characteristics of the intake and exhaust valves and cam components in the intake and exhaust system will also affect the optimization of the air-fuel ratio, ultimately reducing the combustion efficiency and the continuous operation performance of the engine. When the degree of engine performance degradation gradually accumulates, the original factory calibration parameters will no longer be suitable for the degraded working state, which will cause the engine to continuously operate in a non-optimized state, increasing the risk of failure.

[0004] Although the degradation phenomenon of the above-mentioned air-fuel ratio optimization system can be observed and analyzed by advanced equipment such as a scanning electron microscope (SEM), an energy dispersive spectrometer (EDS), and an electron backscatter diffraction (EBSD) system, considering the actual limiting factors such as high test cost of the marine diesel engine, difficulty in extreme condition experiment, and complex disassembly process of parts, the degradation monitoring method based on experimental observation is difficult to realize real-time degradation observation and fault prediction during operation. In view of this, a marine engine aging prediction method relying only on a small amount of experimental data obtained by a low-cost sensor is urgently needed. SUMMARY

[0005] In view of the shortcomings of the prior art, the present application provides a marine engine self-adaptive aging prediction method and a fault diagnosis model, which solves the problems of difficult accurate prediction of the degradation state of the traditional marine engine, high maintenance cost and insufficient dynamic optimization capability.

[0006] To achieve the above object, the present application is implemented by the following technical scheme: a marine engine self-adaptive aging prediction method, comprising the following steps: A digital twin model of the marine engine is constructed, which is based on experimental data under typical working conditions and a high-precision simulation model, and is used to simulate the baseline health state of the marine engine; By comparing the output of the digital twin model with the actual operating parameters of the physical engine, the degradation offset of the actuator is identified; Based on the multiple performance indicators output by the digital twin model, a multi-objective optimization algorithm is used for optimization to obtain a multi-objective Pareto optimal solution set; In the Pareto solution set, the optimal prediction solution under a specific working condition is determined by a comprehensive evaluation method, and the prediction of the engine degradation state is realized.

[0007] Preferably, the step of constructing the digital twin model of the marine engine comprises: The baseline working condition parameters of the marine engine are determined, including injection quantity, injection timing, intake valve relative opening time, speed, throttle opening degree and fuel pressure; The digital twin model is constructed based on a one-dimensional engine simulation tool, and is integrated according to the physical characteristics of the marine engine; The digital twin model is preliminarily calibrated, and the key parameters of the model are adjusted by using the experimental data collected by the actual engine under the baseline working condition; The digital twin model is verified under typical working conditions, so that the simulation error of the performance indicators is controlled within a preset range; Based on the verified digital twin model, dynamic running simulation is carried out under high load, medium load and low load working conditions to generate baseline model output data covering multiple actual operating states; constructing a complete digital twin model for comparative evaluation with actual operating parameters.

[0008] Preferably, the step of identifying the actuator degradation offset by comparing the digital twin model output with the actual operating parameters of the physical engine comprises: collecting real-time operating parameters of the marine engine, including fuel injection quantity, fuel injection timing, intake valve relative opening time, fuel pressure, speed and throttle opening; inputting the collected real-time operating parameters into the digital twin model, and calculating the baseline health status output of the engine based on the digital twin model; collecting actual output performance parameters of the physical engine under the same operating conditions at the same time, including power, fuel consumption rate, nitrogen oxide emission, air-fuel ratio and combustion chamber peak pressure; comparing the output performance parameters of the digital twin model with the actual output performance parameters of the physical engine, and calculating the difference value between the key performance parameters; based on the difference value, analyzing the degradation characteristics of the actuator, and identifying the degradation characteristics; quantitative analysis of the degradation offset, combined with the operating condition parameters to generate a degradation trend curve, and predict the degradation development state of the actuator.

[0009] Preferably, the degradation offset includes the following single parameter or combination of multiple parameters: fuel injection timing offset; reduction of fuel supply per cycle; intake valve relative opening time lag; compound offset of fuel injection timing offset and fuel supply reduction; compound offset of fuel injection timing offset and intake valve relative opening time lag; compound offset of fuel supply reduction and intake valve relative opening time lag.

[0010] Preferably, the step of using multi-objective optimization algorithm to optimize the multiple performance indicators based on the digital twin model output to obtain a multi-objective Pareto optimal solution set comprises: using the performance indicators and related operating parameters output by the digital twin model as input data to construct the objective function of the multi-objective optimization problem; generate an initial solution set based on the multi-objective optimization algorithm, calculate each solution according to the optimization objective, and form a preliminary solution set; In each iteration, update the solution set based on the principle of non-dominated sorting, eliminate non-Pareto optimal solutions, and retain the Pareto optimal solution set; A diversity preservation mechanism is introduced to control the uniformity of the distribution of the Pareto solution set by optimizing the algorithm, so as to ensure that balanced optimization solution sets are obtained under the conflict condition of performance indicators; The final Pareto optimal solution set is output to form a trade-off solution between different optimization objectives for subsequent fuzzy comprehensive evaluation method to select the optimal solution.

[0011] Preferably, the performance indicators include: Engine output power; Nitrogen oxide emissions; Air-fuel ratio; Combustion chamber peak pressure.

[0012] Preferably, the step of determining the optimal prediction solution under a specific working condition in the Pareto solution set by a comprehensive evaluation method to realize the prediction of the engine degradation state comprises: Based on the multiple performance indicators output by the digital twin model, the performance indicator values corresponding to all solutions in the Pareto optimal solution set are obtained; Determine the weight distribution of each performance indicator in combination with the operating target requirements under a specific working condition, for example, power optimization priority or emission control priority; Comprehensive evaluation is performed on each solution in the Pareto optimal solution set, and the performance indicators are weighted and calculated according to the weight distribution; Select the solution with the highest comprehensive evaluation score as the optimal prediction solution under the current working condition; Based on the optimal prediction solution, calculate the key offset parameters corresponding to the current degradation state of the engine, including fuel injection timing offset, fuel supply variation and intake valve opening time offset, to form a degradation characteristic description; Output the prediction result to evaluate the degradation state of the engine.

[0013] Preferably, the comprehensive evaluation method is a fuzzy comprehensive evaluation method, comprising the following steps: Determine the evaluation indicators of fuzzy comprehensive evaluation, select the performance indicators corresponding to the solutions in the Pareto optimal solution set as the evaluation indicators, including power, nitrogen oxide emissions, air-fuel ratio and combustion chamber peak pressure; According to the specific operating condition requirements, set the weight of each performance indicator, and the weight is adjusted dynamically according to the current working condition of the engine; Construct a fuzzy evaluation matrix for each solution in the Pareto solution set, combine the normalized performance indicator values of the solution with the weight distribution, and calculate the comprehensive evaluation score; Use a fuzzy membership function to fuzz the comprehensive score; Sort the fuzzy score results, and select the solution with the highest score as the optimal prediction solution under the current working condition; According to the results of the optimal prediction solution, the key degradation parameters and offset characteristics of the engine are identified, and the degradation state prediction is realized.

[0014] The application also provides a fault diagnosis model of a marine engine, comprising: A data acquisition module is configured to acquire operating parameters of the marine engine. A digital twin module is configured to construct a baseline health state model of the marine engine based on the operating parameters. A degradation evaluation module is configured to identify an actuator degradation offset by comparing the difference between the output of the digital twin module and the actual operating parameters. A multi-objective optimization module is configured to generate a Pareto optimal solution set. A decision-making module is configured to select an optimal prediction solution under a target working condition by using a fuzzy comprehensive evaluation method.

[0015] Preferably, the data acquisition module communicates with the engine through an electronic control unit and acquires operating parameters in real time.

[0016] The application provides a self-adaptive aging prediction method and a fault diagnosis model for a marine engine. 1. The application constructs a digital twin model based on a high-precision simulation model and experimental data, dynamically updates the baseline health state of the engine, and realizes high-precision prediction of the degradation state. Through difference analysis and time series fitting, the application can accurately identify the degradation offset and trend, significantly improve the accuracy of degradation evaluation, and avoid the limitations of traditional methods that rely on expensive experimental equipment or artificial experience.

[0017] 2. Through real-time monitoring of the degradation state and multi-objective optimization, the application can accurately predict the development state of the degradation and provide a scientific maintenance time window. Compared with the traditional passive maintenance method, the application significantly reduces unplanned downtime and unnecessary maintenance costs, improves the economy and reliability of ship operation.

[0018] 3. The application uses the NSGA-III multi-objective optimization algorithm to generate a Pareto optimal solution set, comprehensively balances the conflicts between multiple performance indicators such as fuel consumption rate, power, emissions, and combustion chamber peak pressure, and selects the optimal solution by using a fuzzy comprehensive evaluation method. The application provides a dynamic adjustment global optimization solution for the operation of the marine engine under complex working conditions.

[0019] 4. The application acquires operating parameters in real time and dynamically adjusts the control strategy to realize optimal adjustment of key input parameters (such as injection timing, fuel supply, and intake valve opening time). Through degradation compensation and performance recovery, the application effectively improves the operating efficiency of the marine engine, prolongs the service life of key components, and reduces the risk of failure.

[0020] 5、The application combines digital twin technology and data-driven algorithms, can dynamically adapt to complex working conditions such as high load, medium load and low load, and generate optimized control strategies matching the working condition requirements. Whether it is single parameter deviation or multiple parameter composite deviation, the application can realize accurate prediction and dynamic optimization of the degradation state of the marine engine, and shows strong adaptability and robustness. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is a schematic diagram of the method of the application; Figure 2 is a flow chart of the overall framework of the method of the application; Figure 3 is Figure 1 a solid engine parameter diagram; Figure 4 is a cylinder pressure curve verification result diagram of the application; Figure 5 is a second-order response model calculation flow chart of the application; Figure 6 is a simulation working condition point distribution situation diagram of the application generated by Sobol sequence and V-optimal design method; Figure 7 is an adaptive prediction simulation optimization framework diagram of the application; Figure 8 is a simulation result diagram of a typical deviation condition of the application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the application will be described in detail below with reference to the drawings of the application specification. Obviously, the described embodiments are only a part of the embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the application.

[0023] Please refer to the drawings of the application Figure 1 -attached Figure 8 , the application provides a marine engine adaptive aging prediction method, based on digital twin technology and multi-objective optimization algorithm, an intelligent system capable of predicting engine aging state in real time, optimizing running performance, and providing fault diagnosis basis is constructed. The steps of the method of the application will be described in detail below with reference to the drawings.

[0024] As Figure 1 shown, the marine engine adaptive aging prediction method can include the following steps: S1, a digital twin model of the marine engine is constructed, and calibration and verification are performed based on experimental data and high-precision simulation; S2, collect the operating parameters of the marine engine and input them into the digital twin model to calculate the baseline health state; S3, compare the digital twin model with the performance output of the physical engine to identify the degradation offset of the key parameters; S4, based on the degradation offset and the performance indicators of the digital twin model, use a multi-objective optimization algorithm to generate a Pareto optimal solution set; S5, filter the Pareto optimal solution by a comprehensive evaluation method, and output the current degradation state of the engine and the optimization control strategy.

[0025] For step S1, in this embodiment, a digital twin model is constructed based on the digital twin diesel engine degradation prediction system of the marine engine, which is used to simulate the baseline health state of the marine engine, so as to realize dynamic evaluation and degradation prediction of the performance state of the engine.

[0026] As an option, the construction of the digital twin model is based on experimental data under typical operating conditions combined with high-precision simulation models, and the engine system is modeled by a data-driven algorithm. Specifically, the experimental data covers various typical operating conditions of the engine, such as high load, medium load, low load, etc., and also considers key operating parameters such as fuel injection quantity, fuel injection timing, intake valve relative opening time, fuel pressure, speed, and throttle opening degree.

[0027] It should be noted that in the modeling process, a one-dimensional diesel engine simulation tool (such as GT-Suite) is used to construct an offline simulation model of the engine. For example, in the constructed one-dimensional simulation model, cylinder pressure and heat release process are used as core indicators for performance verification to ensure that the model accurately reflects the physical characteristics and operating rules of the engine under various operating conditions.

[0028] Specifically, in combination with the attached Figure 2 Through the baseline health state modeling method, first, the one-dimensional simulation model is preliminarily calibrated based on typical operating condition data, and compared with experimental data. For example, in the simulation verification of cylinder pressure, the simulation results obtained under the condition that the sub-model parameters remain unchanged have good consistency with the experimental data, and the relative error of the cylinder pressure is controlled within 8% under various test conditions, proving the high precision and reliability of the model.

[0029] In one possible implementation, the calibration step of the baseline model includes the following steps: First, by inputting the experimental data of the marine engine under the baseline operating condition, the fuel injection timing, fuel pressure and other parameters in the simulation model are preliminarily adjusted; Then, based on the output results of the one-dimensional simulation model, the error ranges of various performance indicators (such as power, fuel consumption rate, nitrogen oxide emissions, combustion chamber peak pressure, and air-fuel ratio) are calculated; Finally, through iterative optimization, the simulation errors of all performance indicators are controlled within a preset range (for example, 5% to 8%).

[0030] It should be noted that after the simulation verification is completed, the digital twin model in this embodiment is further extended based on typical working conditions. As an extension, as shown in Figure 6 The simulation working condition point distribution generated by the Sobol sequence and the V-optimal design method is shown. The offline simulation model uses the Sobol sequence design method and the V-optimal design method to determine experimental points, generating 70 global input points and 26 global input points, respectively, thereby forming an input data set covering a wide range of operating states. Subsequently, a multiple-point scan (such as 11 points) of the injection timing is performed on each experimental point using an arithmetic sequence, resulting in a total of 1056 sample data points, further enriching the input space of the simulation model.

[0031] In the process of establishing the baseline health state, the present application innovatively combines the radial basis function (RBF) second-order response surface modeling technique to improve the accuracy of the model in complex parameter space. Specifically, the following formula is used in the model to fit the nonlinear relationship between the input variables and the output performance indicators:

[0032] wherein, represents the radial basis function, , , is the regression coefficient to be fitted, is the input variable (such as injection timing, cycle fuel supply amount, intake valve opening time, etc.), is the output performance indicator of the model (such as power or fuel consumption rate).

[0033] As shown in Figure 5 , a flowchart of the second-order response model is shown. This baseline health state model not only evaluates the current engine performance, but also provides an important reference benchmark for subsequent degradation assessment and control optimization. Through the above method, the digital twin model in this embodiment can be constructed with high accuracy and efficiency, and provides a solid foundation for subsequent degradation offset identification and optimization.

[0034] In some embodiments, the model can be further extended to dynamic simulation under specific working conditions. For example, the timing offset of the fuel injection caused by the nozzle hole diameter wear or the performance degradation of the fuel pump in actual operation is simulated, and the performance difference before and after the offset is evaluated in combination with the benchmark state output by the digital twin model, to verify the robustness and adaptability of the benchmark model.

[0035] For step S2, in the present embodiment, the operating parameters of the marine engine are collected based on the digital twin diesel engine degradation prediction system, and are input into the digital twin model to calculate the benchmark health state. This process aims to generate a health state benchmark covering the current working condition of the engine through the association of the real-time operating data of the physical engine with the digital twin model.

[0036] As shown in FIG. 2, the physical engine parameter diagram is shown as an option. As an option, the collected operating parameters include but are not limited to key variables such as fuel injection quantity, fuel injection timing, intake valve relative opening time, fuel pressure, speed, and throttle opening degree. These parameters are obtained through communication with the engine electronic control unit (ECU), ensuring real-time and high accuracy. Figure 3 Figure 2 Specifically, when collecting operating parameters, the sensor network monitored by the ECU is used to dynamically capture the core operating state of the engine. For example, the fuel injection timing parameter is recorded in real time by a high-precision position sensor, and the fuel pressure is directly measured by a pressure sensor installed in the fuel supply pipeline. The collection frequency and accuracy of these parameters can be adjusted according to the engine operating requirements to meet the precise analysis requirements under high load or dynamic changing working conditions.

[0037] In a possible implementation, the collected operating parameters are input into the digital twin model after preprocessing. The preprocessing includes denoising and outlier filtering of the collected data to ensure the accuracy of the input data. For example, the sliding window mean method is used to smooth the mutation of the fuel injection quantity and fuel pressure signals, eliminating possible transient interference signals. It should be noted that the data format after preprocessing needs to match the input interface of the digital twin model to ensure efficient transmission of data flow.

[0038] It can be understood that the digital twin model calculates the benchmark health state of the engine using the above-mentioned collected operating parameters. The benchmark health state is represented by multiple key performance indicators generated by the model, including but not limited to power, fuel consumption rate, nitrogen oxide emission, air-fuel ratio, and combustion chamber peak pressure. As a reference, the calculation formulas of these performance indicators are consistent with the formulas described in step S1, and are realized by a second-order response surface model based on radial basis functions.

[0039] It can be understood that the digital twin model calculates the benchmark health state of the engine using the above-mentioned collected operating parameters. The benchmark health state is represented by multiple key performance indicators generated by the model, including but not limited to power, fuel consumption rate, nitrogen oxide emission, air-fuel ratio, and combustion chamber peak pressure. As a reference, the calculation formulas of these performance indicators are consistent with the formulas described in step S1, and are realized by a second-order response surface model based on radial basis functions.

[0040] ​It should be noted that the output performance indicators of the model not only reflect the current running state of the engine, but also provide benchmark data for subsequent degradation offset identification and multi-objective optimization.

[0041] As an extended application, in multi-condition testing, the digital twin model can adapt to complex conditions such as high load, medium load and low load. For example, in high load conditions (such as engine speed of 750 rpm and fuel injection of 2943 mg), the model dynamically updates the baseline health state output by combining real-time data of fuel injection timing and fuel pressure, ensuring the accuracy and robustness of the model.

[0042] It should be noted that the baseline health state output by the model is also consistent with the cylinder pressure curve verification results in the Figure 4 In multiple test conditions, the simulation error of cylinder pressure and heat release process is controlled within 8%, which further verifies the reliability of the digital twin model in calculating the baseline health state.

[0043] In some embodiments, to further expand the applicability of the model, additional operating parameters can also be introduced as inputs. For example, by combining the dynamic change trend of exhaust gas temperature and intake air flow, the health state simulation capability of the digital twin model for complex operating conditions can be optimized. The introduction of these extended input parameters provides more comprehensive reference for subsequent offset analysis.

[0044] Through the above process, the embodiment realizes the accurate collection of operating parameters and the calculation of baseline health state, providing basic data support for the identification of degradation offset in the subsequent steps, while ensuring the adaptability and high precision of the digital twin model in dynamic operation.

[0045] For step S3, in the embodiment, by comparing the baseline health state output of the digital twin model with the actual operating performance parameters of the physical engine, the degradation offset of the actuator of the marine engine is identified and the degradation characteristics are quantified.

[0046] As an option, the baseline health state output of the digital twin model is based on the key performance indicators calculated in the aforementioned step S2, including power, fuel consumption rate, nitrogen oxide emission, air-fuel ratio and combustion chamber peak pressure; the actual operating performance parameters of the physical engine are collected through the ECU and corresponding sensors, and are compared with the above-mentioned baseline health state to identify the degradation offset.

[0047] Specifically, the operating performance parameter collection process of the physical engine includes: Fuel injection timing is recorded by a high-precision crankshaft position sensor in terms of crank angle (CA); Fuel pressure is collected by a pressure sensor installed on the high-pressure fuel pump; The relative opening time of the intake valve is monitored by the camshaft sensor, and is output in the form of time or angle in combination with the engine operating speed; Other parameters (such as throttle opening and speed) are sampled and recorded in real time by the ECU.

[0048] In one possible implementation, the collected physical engine performance parameters are compared item by item with the output of the digital twin model. For example, the difference value of fuel consumption rate can be calculated by the following formula:

[0049] wherein, represents the difference value of fuel consumption rate, is the fuel consumption rate of the physical engine, is the fuel consumption rate of the digital twin model.

[0050] It should be noted that the calculation of the difference value is not limited to fuel consumption rate, and other performance parameters (such as nitrogen oxide emissions and power) can be processed in a similar manner.

[0051] It can be understood that based on the above difference value, the degradation characteristics of the engine actuator are analyzed, including the offset amount of key variables such as injection timing, fuel supply amount, and intake valve opening time. For example, when the injection timing offset exceeds its calibrated range (for example, -12°CA to +8°CA), it may indicate degradation or needle valve wear of the fuel injection system; when the fuel supply amount decreases by more than 5%, it may indicate blockage of the high-pressure fuel pump or nozzle hole.

[0052] In some embodiments, the identification of the degradation characteristics uses regression analysis or statistical methods to fit the difference value, thereby generating a trend curve of the degradation offset. For example, a multiple linear regression method is used to fit the degradation trend of the injection timing:

[0053] wherein, is the time of the injection timing offset, is the injection timing of the initial state, and is the regression coefficient, is the random error.

[0054] As an extended application, the present application also comprehensively analyzes the degradation characteristics under different working conditions. For example, under high load conditions (speed 750 rpm, fuel injection amount 2943 mg), the combined offset of injection timing and intake valve opening time can significantly affect combustion efficiency; under low load conditions (speed 350 rpm, fuel injection amount 1500 mg), the effect of fuel supply amount reduction can be more significant.

[0055] It should be noted that the accompanying drawings Figure 8 Simulation results of various typical offset cases are shown, including the following single parameter or combination of multiple parameters: Injection timing offset; Reduced fuel supply per cycle; Intake valve relative opening time lag; Compound offset of injection timing offset and reduced fuel supply; Compound offset of injection timing offset and intake valve relative opening time lag; Compound offset of reduced fuel supply and intake valve relative opening time lag.

[0056] These simulation results verify the accuracy and applicability of the degradation offset identification method.

[0057] Through the above method, the embodiment realizes the accurate identification of the degradation offset of the marine engine actuator, and provides necessary parameter support for the multi-objective optimization and degradation trend prediction in the subsequent steps. At the same time, the reliability and practicability of the digital twin model in dynamic degradation state evaluation are further verified.

[0058] For step S4, in the embodiment, based on the digital twin diesel engine degradation prediction system, the output performance indicators and related operating parameters of the digital twin model are used to construct a multi-objective optimization problem, and a Pareto optimal solution set is generated through a multi-objective optimization algorithm, providing guidance for the degradation state optimization of the marine engine.

[0059] As an option, the objective function of the multi-objective optimization problem includes the output performance indicators of the digital twin model, such as engine power (Power), fuel consumption rate (BSFC), nitrogen oxide emission (NOx), combustion chamber peak pressure (P_max) and air-fuel ratio (AFR). These performance indicators have been calculated by the digital twin model in the aforementioned steps S2 and S3, and combined with the operating parameter input as the input variables of the optimization problem.

[0060] Specifically, the construction process of the multi-objective optimization problem is as follows: First, a mapping relationship is established between the input variables of the digital twin model (such as injection timing, fuel supply, intake valve opening time) and the corresponding performance indicators, and the optimization objectives are to minimize BSFC, NOx emission and P_max, while maximizing AFR and power; Second, the optimization objective function is described by formula. For example, for the target optimization of fuel consumption rate, it can be described as:

[0061] Wherein, The digital twin model represents the input variables. Calculated fuel consumption rate.

[0062] It should be noted that the optimization algorithm adopts the NSGA-III multi-objective optimization algorithm, which generates the Pareto optimal solution set based on the principle of non-dominated sorting and ensures the uniformity of solution set distribution through a diversity preservation mechanism.

[0063] In one possible implementation, the solution process for multi-objective optimization includes the following steps: Generate an initial solution set using a Sobol sequence or random initialization method; For each set of solutions, the objective function value is calculated using a digital twin model; The solution set is classified according to the non-dominated sorting principle, retaining the solutions on the Pareto front and discarding the rest; The next generation of solution sets is generated through crossover and mutation operations, gradually approaching the Pareto front.

[0064] For example, in the appendix Figure 7 In the optimization framework shown, the optimization input space is constructed using performance indicators and operating parameters output by the digital twin model. In some embodiments, the optimization input variables can be expanded to more parameters, such as fuel injection pressure and exhaust temperature, to adapt to more complex operating conditions.

[0065] Understandably, the optimization process also needs to ensure the diversity and uniformity of the Pareto optimal solution set. In this embodiment, congestion distance and uniformity indices are introduced as auxiliary constraints to ensure a good balance solution can be obtained when performance indices conflict. For example, when there is a conflict between power and NOx emissions, the Pareto solution set should contain multiple different trade-off schemes.

[0066] It should be noted that the final Pareto optimal solution set is a set of equilibrium solutions, each solution corresponding to a specific set of input variable values ​​(such as injection timing, fuel supply quantity, and intake valve opening time) and their output performance index values. These solution sets are not only used to evaluate the current degradation state of the engine, but also provide a basis for comprehensive evaluation and optimization of control strategies in subsequent steps.

[0067] As an extension, this embodiment also verifies the applicability of the Pareto solution set under different operating conditions (such as high load, medium load, and low load). For example, under high load conditions, the optimization objective in some solution sets tends to reduce NOx emissions; while under low load conditions, the optimization objective focuses more on minimizing fuel consumption rate. These verification results are attached. Figure 8 As shown, this demonstrates the applicability and robustness of the optimization algorithm.

[0068] Through the above process, the embodiment realizes the application of the multi-objective optimization algorithm in the digital twin model, and generates a Pareto optimal solution set covering different optimization objectives, thereby providing strong support for the prediction and control optimization of the engine degradation state.

[0069] For step S5, in the embodiment, the optimal solution under a specific working condition is screened out through comprehensive evaluation of the Pareto optimal solution set, so as to determine the current degradation state of the marine engine and the optimal control strategy.

[0070] As an option, the comprehensive evaluation method adopts a fuzzy comprehensive evaluation method, which realizes dynamic screening and sorting of the Pareto optimal solution by introducing multiple performance indicators (such as power, fuel consumption rate, nitrogen oxide emission, air-fuel ratio, and combustion chamber peak pressure) and combining weight distribution.

[0071] Specifically, the implementation of the fuzzy comprehensive evaluation includes the following steps: Selection of evaluation indexes: As described in the foregoing step S4, each solution in the Pareto optimal solution set corresponds to a group of performance indicator values. As an example, power (Power), fuel consumption rate (BSFC), nitrogen oxide emission (NOx), combustion chamber peak pressure (P_max), and air-fuel ratio (AFR) are selected as the core indicators for comprehensive evaluation.

[0072] Weight distribution: The weights of the performance indicators are dynamically adjusted according to the optimization requirements of the current working condition. For example, in a high-load working condition, power and combustion chamber peak pressure may have higher weights; while in a low-load working condition, fuel consumption rate and nitrogen oxide emission are preferentially optimized.

[0073] In a possible implementation manner, the fuzzy comprehensive evaluation method first normalizes the performance indicators of each solution in the Pareto optimal solution set, so that indicators of different dimensions have comparability.

[0074] After the normalization process is completed, the performance indicators are weighted and calculated using weight distribution. The specific weight distribution method is determined according to the working condition requirements, and the total weight is 1.

[0075] It should be noted that the fuzzy comprehensive evaluation also combines a fuzzy membership function to fuzz the comprehensive score. The membership function usually adopts the form of a triangular or trapezoidal function, and the specific form can be designed according to actual optimization requirements. For example, when the score of a certain indicator is higher than a certain threshold, its membership degree can be set to 1, and vice versa.

[0076] As an extended application, the fuzzy comprehensive evaluation method also supports dynamic decision-making under multiple working conditions. In some embodiments, the Pareto optimal solution set under different working conditions can be input into the fuzzy evaluation model respectively to generate the optimal solution corresponding to the working condition. For example, under high load working condition, the optimal solution screened out may correspond to higher power output and moderate fuel consumption rate; while under low load working condition, the optimal solution may pay more attention to the optimization of emission indicators.

[0077] Exemplary, the Pareto optimal solution set and the corresponding fuzzy comprehensive evaluation results under different offset states are shown in the following table. Under the compound offset state of fuel injection timing lag and fuel supply reduction, the optimal solution shows the relative optimization of fuel consumption rate; while under the single offset state of intake valve opening lag, the optimal solution tends to reduce nitrogen oxide emissions. Figure 8

[0078] It can be understood that the screened optimal solution not only clearly shows the best optimization scheme under the current working condition, but also outputs the optimization control strategy by offset adjustment of key input parameters (such as fuel injection timing, fuel supply, and intake valve opening time).

[0079] Through the above steps, the embodiment realizes efficient screening and dynamic decision-making of the Pareto optimal solution under complex working conditions, and provides clear guidance for the degradation state optimization and performance recovery of the marine engine. At the same time, through the introduction of the fuzzy comprehensive evaluation method, the applicability and engineering value of the optimization result are further improved.

[0080] In summary, the present application proposes a self-adaptive aging prediction method for marine engines based on digital twinning. Through the construction and optimization of the digital twinning model, combined with the collection of operating parameters, the identification of degradation offset, multi-objective optimization and comprehensive evaluation, high-precision prediction and optimization control of the degradation state of the marine engine are realized. The specific process is as follows: 1. Construction of digital twinning model The present application first constructs a digital twinning model of the marine engine based on experimental data and high-precision simulation models (such as GT-Suite, etc.).

[0081] The model is parameterized through experimental data under typical working conditions, so that the model can accurately reflect the baseline health state of the marine engine.

[0082] The model is calibrated and verified to ensure that the error of its output performance indicators (such as power, fuel consumption rate, nitrogen oxide emissions, etc.) and actual experimental results is controlled within 8%.

[0083] Finally, a digital twinning model covering multiple working conditions such as high load, medium load and low load is constructed, providing a basis for degradation prediction and optimization analysis.

[0084] ​2. Acquisition of operating parameters and calculation of baseline health status Real-time acquisition of operating parameters of marine engines through the engine control unit (ECU) and related sensors, including fuel injection quantity, injection timing, intake valve opening time, fuel pressure, speed, and throttle opening.

[0085] Input the operating parameters into the digital twin model to calculate the baseline health status performance indicators of the engine, which serve as the comparison baseline for subsequent degradation analysis.

[0086] 3. Identification of degradation offset By comparing the baseline health status output of the digital twin model with the actual performance parameters of the physical engine (such as power, fuel consumption rate, nitrogen oxide emissions, etc.), the difference values of key performance parameters are calculated.

[0087] Based on the performance difference values, identify the degradation offset of the engine actuator, including single or combined offset conditions such as injection timing offset, fuel supply reduction, and intake valve opening time lag.

[0088] Quantitative analysis of the degradation offset, combined with time series fitting method to generate degradation trend curve and predict the degradation development state.

[0089] 4. Multi-objective optimization and Pareto optimal solution set generation Based on the performance indicators output by the digital twin model, construct a multi-objective optimization problem and generate a Pareto optimal solution set using the NSGA-III algorithm.

[0090] The objective function of multi-objective optimization includes minimizing fuel consumption rate, nitrogen oxide emissions, and combustion chamber peak pressure, while maximizing power and air-fuel ratio.

[0091] Through the non-dominated sorting and diversity preservation mechanism, ensure that the Pareto optimal solution set is evenly distributed and covers the trade-off schemes between different performance indicators.

[0092] 5. Comprehensive evaluation and optimal solution selection Perform fuzzy comprehensive evaluation on each solution in the Pareto optimal solution set, dynamically allocate weights according to actual working condition requirements, and perform weighted calculation on performance indicators.

[0093] Through the membership function, the comprehensive score is fuzzified to select the optimal solution under specific working conditions.

[0094] Output the key input variable values (such as injection timing, fuel supply, intake valve opening time, etc.) and optimized performance indicator values corresponding to the optimal solution.

[0095] 6. Degradation state assessment and optimization control strategy generation Based on the optimal solution screened out, the current degradation state of the engine is evaluated, and the degradation characteristic parameters are quantitatively described.

[0096] According to the optimization result, a dynamic adjustment strategy is generated, for example, adjusting key parameters such as injection timing, fuel supply amount and intake valve opening time, to realize degradation compensation and performance optimization.

[0097] The running state is continuously monitored, and the digital twin model is dynamically updated to ensure the real-time and high precision of prediction and optimization.

[0098] In summary, the present application realizes the dynamic degradation prediction and optimization control of marine engine through the organic combination of digital twin model construction, degradation state identification, multi-objective optimization and comprehensive evaluation. This method has the characteristics of high precision, low cost and strong adaptability, which can significantly improve the reliability and economy of engine operation, and provides effective support for the intelligent maintenance of marine power system.

[0099] Correspondingly, the present application also provides a fault diagnosis model for a marine engine, which includes a data acquisition module, a digital twin module, a degradation evaluation module, a multi-objective optimization module and a decision module. Through the cooperative work of these modules, accurate diagnosis and optimization control of the marine engine fault are realized.

[0100] Data acquisition module The data acquisition module in this embodiment communicates with the electronic control unit (ECU) of the marine engine to acquire the key operating parameters of the engine in real time. The acquired operating parameters include but are not limited to injection amount, injection timing, intake valve relative opening time, fuel pressure, speed and throttle opening.

[0101] As an option, the data acquisition module can dynamically adjust the sampling frequency according to the engine operating conditions. For example, when operating at high load, the data acquisition module can increase the sampling frequency of injection amount and injection timing to more accurately capture the rapidly changing operating characteristics.

[0102] It should be noted that the collected parameters are transmitted to the digital twin module after preprocessing to ensure the accuracy and integrity of the input data. The preprocessing includes denoising, outlier filtering and time alignment of the sampled data.

[0103] Digital twin module The digital twin module in this embodiment constructs a baseline health state model of the marine engine based on the operating parameters. This module combines high-precision simulation models (such as GT-Suite) and experimental data under typical operating conditions to accurately reflect the operating state of the engine.

[0104] As a possible implementation, the digital twin module utilizes real-time data provided by the acquisition module, in combination with parameters such as fuel injection timing, fuel pressure, and rotational speed, to dynamically update the output performance indicators of the health state model. These performance indicators include power, specific fuel consumption, nitrogen oxide emissions, air-fuel ratio, and combustion chamber peak pressure, among others.

[0105] It can be understood that the digital twin module provides a standard benchmark for the subsequent degradation assessment and optimization modules, ensuring high accuracy of the diagnostic model.

[0106] Degradation assessment module The degradation assessment module in this embodiment identifies the degradation offset of the actuator by comparing the benchmark health state output by the digital twin module with the actual operating parameters of the physical engine.

[0107] As an option, the degradation assessment module can identify single-parameter offset (such as fuel injection timing lag, fuel supply reduction) or multi-parameter composite offset (such as fuel injection timing lag combined with intake valve opening time lag). In addition, the module quantifies the degradation trend through time series analysis methods and predicts the degradation development state of the actuator.

[0108] It should be noted that the output results of the degradation assessment module will serve as the input of the optimization module, providing a reference basis for degradation compensation and optimization control.

[0109] Multi-objective optimization module The multi-objective optimization module in this embodiment generates a Pareto optimal solution set by constructing a multi-objective optimization problem. The optimization objectives include, but are not limited to, minimizing specific fuel consumption, nitrogen oxide emissions, and combustion chamber peak pressure, while maximizing power and air-fuel ratio.

[0110] In a possible implementation, the module employs the NSGA-III optimization algorithm, which ensures uniform distribution of the generated Pareto solution set through non-dominated sorting and diversity preservation mechanisms, and can cover the trade-off schemes between different optimization objectives.

[0111] As an example, under high load conditions, the Pareto solution set generated by the module may be more inclined to optimize power output; while under low load conditions, it may prioritize optimization of specific fuel consumption and emission indicators.

[0112] Decision module The decision module in this embodiment selects the optimal predicted solution under the target operating condition from the Pareto optimal solution set through fuzzy comprehensive evaluation method. The module assigns weights to the optimization objectives according to the requirements of the actual operating condition, and conducts comprehensive evaluation and sorting of the Pareto solution set through membership functions.

[0113] It should be noted that the output of the decision module not only includes the input variable values (such as fuel injection timing, fuel supply amount and intake valve opening time) corresponding to the optimal prediction solution, but also includes the optimized performance index value. These results provide direct guidance for engine fault diagnosis and degradation compensation.

[0114] In some embodiments, the decision module also supports dynamic decision-making under different working conditions. For example, in high load working conditions, the decision module will preferentially select the optimal solution with higher power; while in low load working conditions, it will be more inclined to the solution with better emission index optimization.

[0115] Overall, through the synergistic effect of the above modules, the marine engine fault diagnosis model of the present embodiment realizes a complete closed loop from data acquisition to degradation identification, multi-objective optimization and optimal solution. The model can dynamically adjust the optimization target to adapt to the complex operation requirements of different working conditions, and provides efficient and reliable technical support for the health management and fault diagnosis of marine engines.

[0116] Although embodiments of the present application have been shown and described, it will be understood by those having ordinary skill in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method of adaptive aging prediction for a marine engine, characterized by, The method comprises the following steps: constructing a digital twin model of the marine engine, the model being constructed based on experimental data and high-precision simulation models under typical working conditions, and used for simulating a benchmark health state of the marine engine; comparing the output of the digital twin model with actual operation parameters of the physical engine to identify degradation offset of the actuator; based on multiple performance indicators output by the digital twin model, using a multi-objective optimization algorithm to perform optimization, and obtaining a multi-objective Pareto optimal solution set; in the Pareto solution set, determining an optimal prediction solution under a specific working condition by a comprehensive evaluation method, and realizing prediction of the degradation state of the engine.

2. A method of adaptive aging prediction of a marine engine according to claim 1, characterized in that, The step of constructing the digital twin model of the marine engine comprises: determining benchmark working condition parameters of the marine engine, the benchmark working condition parameters comprising fuel injection quantity, fuel injection timing, intake valve relative opening time, rotation speed, throttle opening degree, and fuel pressure; constructing the digital twin model based on a one-dimensional engine simulation tool, and integrating according to physical characteristics of the marine engine; performing preliminary calibration on the digital twin model, and adjusting key parameters of the model by using experimental data collected by the actual engine under the benchmark working condition; verifying the digital twin model under typical working conditions, so that simulation error of the performance indicators is controlled within a preset range; based on the verified digital twin model, performing dynamic operation simulation under high-load, medium-load, and low-load working conditions, and generating benchmark model output data covering multiple actual operation states; constructing a complete digital twin model, and used for comparison and evaluation with actual operation parameters.

3. A method of adaptive aging prediction for a marine engine according to claim 1, characterized in that, The step of comparing the output of the digital twin model with actual operation parameters of the physical engine to identify degradation offset of the actuator comprises: collecting real-time operation parameters of the marine engine, including fuel injection quantity, fuel injection timing, intake valve relative opening time, fuel pressure, rotation speed, and throttle opening degree; inputting the collected real-time operation parameters into the digital twin model, and calculating a benchmark health state output of the engine based on the digital twin model; simultaneously collecting actual output performance parameters of the physical engine under the same working condition, including power, fuel consumption rate, nitrogen oxide emission, air-fuel ratio, and combustion chamber peak pressure; comparing the output performance parameters of the digital twin model with the actual output performance parameters of the physical engine, and calculating difference values between key performance parameters; based on the difference values, analyzing degradation characteristics of the actuator, and identifying degradation features; quantitatively analyzing the degradation offset, combining with operation working condition parameters to generate a degradation trend curve, and predicting a degradation development state of the actuator.

4. A method of adaptive aging prediction of a marine engine according to claim 3, characterized in that, The degradation offset comprises the following single parameter or combination of multiple parameters: fuel injection timing offset; reduction of fuel supply per cycle; lag of intake valve relative opening time; compound offset of fuel injection timing offset and reduction of fuel supply; compound offset of fuel injection timing offset and lag of intake valve relative opening time; compound offset of reduction of fuel supply and lag of intake valve relative opening time.

5. A method of adaptive aging prediction of a marine engine according to claim 1, characterized in that, The step of based on multiple performance indicators output by the digital twin model, using a multi-objective optimization algorithm to perform optimization, and obtaining a multi-objective Pareto optimal solution set comprises: The performance indicators and related operating parameters output by the digital twin model are used as input data to construct the objective function of the multi-objective optimization problem; An initial solution set is generated based on a multi-objective optimization algorithm, and each solution is calculated according to the optimization objectives to form a preliminary solution set; In each iteration, the solution set is updated based on the principle of non-dominated sorting, and non-Pareto optimal solutions are removed while the Pareto optimal solution set is retained; A diversity preservation mechanism is introduced to control the uniformity of the Pareto solution set through the optimization algorithm, ensuring that a balanced optimization solution set is obtained under conflicting performance indicators; The final Pareto optimal solution set is output to form a trade-off solution between different optimization objectives for subsequent fuzzy comprehensive evaluation method to select the optimal solution.

6. A method of adaptive aging prediction of a marine engine according to claim 5, characterized in that, The performance indicators include: Engine output power; Nitrogen oxide emissions; Air-fuel ratio; Peak pressure in the combustion chamber.

7. A method of adaptive aging prediction of a marine engine according to claim 1, characterized in that, The step of determining the optimal prediction solution under a specific operating condition in the Pareto solution set through a comprehensive evaluation method to realize the prediction of the engine degradation state includes: Based on the multiple performance indicators output by the digital twin model, the performance indicator values corresponding to all solutions in the Pareto optimal solution set are obtained; The weight distribution of each performance indicator is determined based on the operating target requirements under a specific operating condition, such as power optimization priority or emission control priority; Each solution in the Pareto optimal solution set is comprehensively evaluated, and the performance indicators are weighted calculated according to the weight distribution; The solution with the highest comprehensive evaluation score is selected as the optimal prediction solution for the current operating condition; Based on the optimal prediction solution, the key offset parameters corresponding to the current degradation state of the engine are calculated, including fuel injection timing offset, fuel supply variation, and intake valve opening time offset, forming a degradation characteristic description; The prediction result is output to evaluate the degradation state of the engine.

8. A method of adaptive aging prediction of a marine engine according to claim 7, characterized in that, The comprehensive evaluation method is a fuzzy comprehensive evaluation method, including the following steps: Determine the evaluation indicators for fuzzy comprehensive evaluation, select the performance indicators corresponding to the solutions in the Pareto optimal solution set as the evaluation indicators, including power, nitrogen oxide emissions, air-fuel ratio, and peak pressure in the combustion chamber; According to the specific operating condition requirements, set the weight of each performance indicator, and the weight is adjusted dynamically according to the current operating condition of the engine; Construct a fuzzy evaluation matrix for each solution in the Pareto solution set, combine the normalized performance indicator values of the solution with the weight distribution, and calculate the comprehensive evaluation score; Use the fuzzy membership function to fuzz the comprehensive score; Sort the fuzzy score results, and select the solution with the highest score as the optimal prediction solution for the current operating condition; According to the results of the optimal prediction solution, identify the key degradation parameters and offset characteristics of the engine to realize the prediction of the degradation state.

9. A failure diagnosis model of a marine engine, characterized by, It includes: A data acquisition module for acquiring operating parameters of a marine engine; A digital twin module for constructing a baseline health state model of the marine engine based on the operating parameters; A degradation evaluation module for identifying actuator degradation offsets by comparing the differences between the digital twin module output and the actual operating parameters; A multi-objective optimization module for generating a Pareto optimal solution set; A decision module is configured to select the optimal prediction solution in the target working condition by using a fuzzy comprehensive evaluation method.

10. The marine engine fault diagnosis model according to claim 9, characterized in that, The data acquisition module communicates with the engine through an electronic control unit and acquires the operating parameters in real time.

Images