Accurate adjusting method of engine fuel injection system

Abstract

The invention relates to the technical field of engine fuel injection control, in particular to a precise adjustment method for an engine fuel injection system, which comprises the following steps: S1, multi-physics coupling modeling; s2, a real-time state sensing system; s3, a deep reinforcement learning decision framework; s4, performing hierarchical prediction control; s5, carrying out closed-loop verification and sustainable evolution; the oil injection error in the full working condition range is reduced to be within + / -1% from + / -3%, and particularly, the precision is improved more remarkably during small-oil-quantity injection. And more accurate air-fuel ratio control can be achieved through higher injection precision, the combustion efficiency is improved, and oil consumption and emission are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of engine fuel injection control, and in particular to a precise adjustment method for an engine fuel injection system. Background Art

[0002] Traditional engine fuel injection systems face the following technical bottlenecks:

[0003] Static calibration limitations: MAP calibration methods based on steady-state bench tests are difficult to cover all actual operating conditions, especially in transient conditions (such as rapid acceleration and cold starts), where significant injection deviations occur. Traditional calibration methods are typically optimized based on a limited number of test points and are unable to adapt to the complex operating conditions experienced in actual driving. This results in reduced fuel injection control accuracy and combustion efficiency under non-calibrated operating conditions.

[0004] Strong coupling of multiple parameters: Parameters such as injection volume, injection pressure, and injection timing influence each other, making it difficult for traditional PID control to achieve global optimization. For example, changes in injection pressure can affect the atomization quality and injection rate of the fuel, which in turn affects the combustion process. However, traditional PID controllers cannot effectively handle the complex coupling relationships between these parameters, which can easily lead to oscillation and overshoot during the control process.

[0005] Complex physical field interactions: The fuel injection process involves complex physical phenomena such as gas-liquid two-phase flow, turbulent mixing, and high-pressure phase transitions, which are difficult for existing models to accurately describe. Under high-pressure injection conditions, the flow characteristics and atomization process of the fuel are affected by multiple factors, such as cavitation and turbulent breakup. Traditional models are unable to accurately capture these physical phenomena, resulting in inaccurate predictions of the fuel injection process.

[0006] Time-varying system characteristics: Factors such as injector wear, fuel quality fluctuations, and ambient temperature changes cause system characteristics to drift, making traditional control strategies less adaptable. As the engine runs longer, injectors wear, causing injection characteristics to change. Fuel quality also varies between batches, all of which affect the performance of the fuel injection system. However, traditional control strategies are unable to adapt to these changes in real time.

[0007] Emissions regulations are tightening: Increasingly stringent emissions regulations are placing higher demands on fuel injection precision, pushing traditional technologies close to their physical limits. To meet these stringent regulations, fuel injection accuracy and control performance must be further improved, but traditional technologies struggle to meet these challenges. Therefore, a method for precisely adjusting engine fuel injection systems is proposed. Summary of the Invention

[0008] In view of this, the present invention provides a precise adjustment method for an engine fuel injection system to solve or alleviate the technical problems existing in the prior art and at least provide a beneficial option.

[0009] The technical solution of the present invention is implemented as follows: A method for accurately adjusting an engine fuel injection system comprises the following steps:

[0010] S1, multi-physics coupling modeling;

[0011] S2, real-time status perception system;

[0012] S3, deep reinforcement learning decision framework;

[0013] S4, hierarchical predictive control execution;

[0014] S5. Closed-loop verification and continuous evolution.

[0015] Further preferably, in said S1, based on computational fluid dynamics (CFD), taking into account physical phenomena such as cavitation effect, turbulent breakup, and evaporation phase change, the VOF (Volume of Fluid) method is used to track the gas-liquid interface. This model can accurately simulate the atomization process of the fuel after it is sprayed from the injector, including phenomena such as the formation, breakup, and evaporation of oil droplets, providing an accurate physical basis for subsequent fuel injection control. It considers pressure fluctuations in the high-pressure oil pipe, fuel compressibility, and viscosity effects, and establishes a one-dimensional unsteady flow equation. This model can accurately describe the flow characteristics of the fuel in the high-pressure oil pipe, predict the impact of pressure fluctuations on fuel injection, provide theoretical support for optimizing the fuel injection system, and model the dynamic response characteristics of the injector solenoid valve, taking into account the coil inductance change, hysteresis effect, and mechanical motion characteristics. This model can accurately predict the opening and closing time of the solenoid valve and the motion characteristics of the needle valve, providing key parameters for achieving precise fuel injection control. It uses Proper Orthogonal Decomposition (POD) combined with the Galerkin projection method to reduce the high-dimensional CFD model to a low-order state-space model, improving computational efficiency by more than 90%. Through order reduction processing, the calculation speed of the model is greatly improved while ensuring model accuracy, enabling it to meet the needs of real-time control.

[0016] Further preferably, in said S2, the multi-sensor fusion layout is:

[0017] Piezoelectric pressure sensor: measures common rail pressure (accuracy ±0.5MPa) and fuel line pressure fluctuations (frequency response ≥10kHz), and can accurately obtain fuel system pressure information in real time, providing an important basis for fuel injection control;

[0018] High-speed camera: Captures spray morphology (frame rate ≥ 10,000 fps) for model verification and correction. Spray images captured by high-speed cameras can be used to visually observe the atomization effect of the fuel, evaluate the accuracy of the model, and correct and optimize the model.

[0019] Ion current sensor: Detects the ion current in the combustion chamber, reflecting the combustion quality. The ion current sensor can monitor the changes in ion current during the combustion process in real time, thereby determining the stability and sufficiency of combustion and providing feedback information for adjusting fuel injection parameters;

[0020] Vibration sensor: monitors the impact of the injector needle valve seating and determines the actual opening time. The vibration sensor can detect the vibration signal generated when the injector needle valve seats, accurately determine the actual opening time of the needle valve, and improve the accuracy of fuel injection control;

[0021] Data preprocessing algorithm:

[0022] Wavelet transform denoising: eliminates high-frequency electromagnetic interference and improves the quality and reliability of sensor data.

[0023] Kalman filter fusion: It processes the time delay and inconsistency of multi-sensor data and fuses the data from different sensors to obtain more accurate and reliable state estimation;

[0024] Outlier detection: Identify sensor faults based on the isolation forest algorithm, detect sensor anomalies in a timely manner, and ensure the normal operation of the system.

[0025] Further preferably, S3 includes an 18-dimensional state vector, covering engine speed, load, coolant temperature, common rail pressure, fuel line pressure fluctuation characteristics, etc. This fully and accurately describes the operating state of the engine, providing rich information for the deep reinforcement learning algorithm. The continuous control space, including injection pulse width (0.5-20ms, resolution 0.01ms), injection pressure (50-250MPa, resolution 0.1MPa), and multiple injection intervals (0.1-5ms, resolution 0.05ms), provides a sophisticated control method to achieve precise adjustment of the fuel injection process.

[0026] Reward function design: Consider the following factors:

[0027] Deviation between actual injection amount and target value (weight 0.4): Ensuring the accuracy of fuel injection amount is the basis for achieving precise control;

[0028] Combustion stability index (COVofIMEP) (weight 0.3): ensures the stability of the combustion process and improves engine performance and reliability;

[0029] Fuel consumption rate (weight 0.2): optimize fuel economy and reduce fuel consumption;

[0030] Emission index (NOx+PM) (weight 0.1): reduce pollutant emissions and meet environmental protection requirements;

[0031] Algorithm Architecture: This algorithm uses the TwinDelayedDDPG (TD3) algorithm, introducing delayed policy updates and target network smoothing mechanisms to improve training stability. The TD3 algorithm has advantages in handling continuous control problems, effectively avoiding overestimation and improving algorithm training efficiency and stability.

[0032] Further preferably, in S4, based on model predictive control (MPC), the injection parameter sequence is optimized in a rolling manner within the 100ms prediction time domain, taking into account system constraints such as the minimum injection pulse width limit and the upper and lower limits of the common rail pressure. This layer can optimize the fuel injection parameters from a global perspective to achieve optimal overall performance. The compensation adjustment layer: Based on the online identification of system characteristics, it compensates for deviations caused by factors such as injector wear and fuel temperature changes in real time. The recursive least squares (RLS) method is used for system parameter identification. This layer can adaptively adjust the control strategy to compensate for changes in system characteristics and ensure the stability of control accuracy. The injector drive circuit is designed based on sliding mode control theory to achieve an injection timing control accuracy of ±0.05ms. An adaptive feedforward compensation strategy is used to eliminate the impact of solenoid valve response delay. This layer can accurately execute control instructions to ensure the accuracy and consistency of fuel injection.

[0033] Further preferably, in the S5, a HIL platform including real injectors and sensors is constructed to verify the real-time performance of the control algorithm. Through HIL testing, the actual operation of the engine can be simulated in a laboratory environment to verify the real-time performance and effectiveness of the control algorithm. Gaussian process regression (GPR) is used to perform online correction on the reduced-order model to compensate for modeling errors. As the engine running time increases, the model may deviate. Through online model correction, the model parameters can be adjusted in time to improve the accuracy of the model. Through vehicle-cloud collaboration, operating data of different vehicles are collected, and a federated learning algorithm is used to continuously optimize the control strategy. The powerful computing power and big data analysis technology of the cloud are used to continuously optimize the control strategy to improve the performance and adaptability of the system.

[0034] The embodiment of the present invention adopts the above technical solution, which has the following advantages:

[0035] The present invention reduces injection error from ±3% to within ±1% across the full operating range, with particularly significant improvements when injecting small amounts of fuel. This higher injection accuracy enables more precise air-fuel ratio control, improving combustion efficiency and reducing fuel consumption and emissions.

[0036] Second, the present invention triples the response speed of injection parameter adjustments under transient operating conditions, effectively reducing emissions under transient conditions. Faster dynamic response enables the system to rapidly adjust fuel injection parameters as operating conditions change, maintaining combustion stability and reducing pollutant emissions.

[0037] Third, the present invention can automatically adapt to factors such as injector wear and changes in fuel quality, maintaining an injection accuracy of ±1.5% after 1,000 hours of durability testing. This enhanced adaptability ensures the stability and reliability of the system during long-term operation and reduces maintenance costs.

[0038] Fourth, while meeting the same emission regulations, this invention improves fuel economy by 5-8% and reduces power output fluctuation by 40%. This comprehensive performance optimization can enhance the overall performance and competitiveness of the engine, meeting user demands for efficient and reliable engine operation.

[0039] 5. This invention uses virtual simulation and cloud-based collaborative optimization to shorten development time by more than 50% compared to traditional calibration methods. This shortened development cycle can speed up the launch of new products and reduce R&D costs.

[0040] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0042] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0043] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.

[0044] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, an embodiment of the present invention provides a method for accurately adjusting an engine fuel injection system, comprising the following steps:

[0046] S1, multi-physics coupling modeling;

[0047] S2, real-time status perception system;

[0048] S3, deep reinforcement learning decision framework;

[0049] S4, hierarchical predictive control execution;

[0050] S5. Closed-loop verification and continuous evolution.

[0051] In one embodiment, in S1, the VOF (Volume of Fluid) method is used to track the gas-liquid interface based on computational fluid dynamics (CFD), taking into account physical phenomena such as cavitation, turbulent breakup, and evaporation phase transition. This model accurately simulates the atomization process of fuel after it is sprayed from the injector, including phenomena such as droplet formation, breakup, and evaporation, providing an accurate physical basis for subsequent fuel injection control. It considers pressure fluctuations within the high-pressure fuel line, fuel compressibility, and viscosity effects, and establishes a one-dimensional unsteady flow equation. This model accurately describes the flow characteristics of fuel within the high-pressure fuel line, predicts the impact of pressure fluctuations on fuel injection, provides theoretical support for optimizing the fuel injection system, and models the dynamic response characteristics of the injector solenoid valve, taking into account coil inductance changes, hysteresis effects, and mechanical motion characteristics. This model can accurately predict the opening and closing time of the solenoid valve and the motion characteristics of the needle valve, providing key parameters for achieving precise fuel injection control. It uses Proper Orthogonal Decomposition (POD) combined with the Galerkin projection method to reduce the high-dimensional CFD model to a low-order state-space model, improving computational efficiency by more than 90%. Through order reduction processing, the calculation speed of the model is greatly improved while ensuring model accuracy, enabling it to meet the needs of real-time control.

[0052] In one embodiment, in S2, the multi-sensor fusion layout:

[0053] Piezoelectric pressure sensor: measures common rail pressure (accuracy ±0.5MPa) and fuel line pressure fluctuations (frequency response ≥10kHz), and can accurately obtain fuel system pressure information in real time, providing an important basis for fuel injection control;

[0054] High-speed camera: Captures spray morphology (frame rate ≥ 10,000 fps) for model verification and correction. Spray images captured by high-speed cameras can be used to visually observe the atomization effect of the fuel, evaluate the accuracy of the model, and correct and optimize the model.

[0055] Ion current sensor: Detects the ion current in the combustion chamber, reflecting the combustion quality. The ion current sensor can monitor the changes in ion current during the combustion process in real time, thereby determining the stability and sufficiency of combustion and providing feedback information for adjusting fuel injection parameters;

[0056] Vibration sensor: monitors the impact of the injector needle valve seating and determines the actual opening time. The vibration sensor can detect the vibration signal generated when the injector needle valve seats, accurately determine the actual opening time of the needle valve, and improve the accuracy of fuel injection control;

[0057] Data preprocessing algorithm:

[0058] Wavelet transform denoising: eliminates high-frequency electromagnetic interference and improves the quality and reliability of sensor data.

[0059] Kalman filter fusion: It processes the time delay and inconsistency of multi-sensor data and fuses the data from different sensors to obtain more accurate and reliable state estimation;

[0060] Outlier detection: Identify sensor faults based on the isolation forest algorithm, detect sensor anomalies in a timely manner, and ensure the normal operation of the system.

[0061] In one embodiment, S3 contains an 18-dimensional state vector covering engine speed, load, coolant temperature, common rail pressure, fuel line pressure fluctuation characteristics, etc. This fully and accurately describes the engine's operating state, providing rich information for deep reinforcement learning algorithms. The continuous control space, including injection pulse width (0.5-20ms, resolution 0.01ms), injection pressure (50-250MPa, resolution 0.1MPa), and multiple injection intervals (0.1-5ms, resolution 0.05ms), provides a sophisticated control method to achieve precise adjustment of the fuel injection process.

[0062] Reward function design: Consider the following factors:

[0063] Deviation between actual injection amount and target value (weight 0.4): Ensuring the accuracy of fuel injection amount is the basis for achieving precise control;

[0064] Combustion stability index (COVofIMEP) (weight 0.3): ensures the stability of the combustion process and improves engine performance and reliability;

[0065] Fuel consumption rate (weight 0.2): optimize fuel economy and reduce fuel consumption;

[0066] Emission index (NOx+PM) (weight 0.1): reduce pollutant emissions and meet environmental protection requirements;

[0067] Algorithm Architecture: This algorithm uses the TwinDelayedDDPG (TD3) algorithm, introducing delayed policy updates and target network smoothing mechanisms to improve training stability. The TD3 algorithm has advantages in handling continuous control problems, effectively avoiding overestimation and improving algorithm training efficiency and stability.

[0068] In one embodiment, in S4, based on model predictive control (MPC), the injection parameter sequence is optimized in a rolling manner within a 100ms prediction time domain, taking into account system constraints such as the minimum injection pulse width limit and the upper and lower limits of the common rail pressure. This layer can optimize fuel injection parameters from a global perspective to achieve optimal overall performance. The compensation and adjustment layer: Based on the online identification of system characteristics, it compensates for deviations caused by factors such as injector wear and fuel temperature changes in real time. The recursive least squares (RLS) method is used for system parameter identification. This layer can adaptively adjust the control strategy to compensate for changes in system characteristics and ensure the stability of control accuracy. The injector drive circuit is designed based on sliding mode control theory to achieve an injection timing control accuracy of ±0.05ms. An adaptive feedforward compensation strategy is used to eliminate the impact of solenoid valve response delay. This layer can accurately execute control instructions to ensure the accuracy and consistency of fuel injection.

[0069] In one embodiment, in S5, a HIL platform including real injectors and sensors is constructed to verify the real-time performance of the control algorithm. Through HIL testing, the actual operation of the engine can be simulated in a laboratory environment to verify the real-time performance and effectiveness of the control algorithm. Gaussian process regression (GPR) is used to perform online correction on the reduced-order model to compensate for modeling errors. As the engine operating time increases, the model may deviate. Through online model correction, the model parameters can be adjusted in time to improve the accuracy of the model. Through vehicle-cloud collaboration, operating data of different vehicles are collected, and a federated learning algorithm is used to continuously optimize the control strategy. The powerful computing power and big data analysis technology of the cloud are used to continuously optimize the control strategy to improve the performance and adaptability of the system.

[0070] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various modifications and substitutions within the technical scope disclosed in the present invention, and such modifications and substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A method for accurately adjusting an engine fuel injection system, characterized in that: The following steps are involved: S1, multi-physics coupling modeling; S2, real-time status perception system; S3, deep reinforcement learning decision framework; S4, hierarchical predictive control execution; S5. Closed-loop verification and continuous evolution.

2. The precise adjustment method of an engine fuel injection system according to claim 1, characterized in that: In S1, based on computational fluid dynamics (CFD), the VOF (Volume of Fluid) method is used to track the gas-liquid interface, taking into account physical phenomena such as cavitation, turbulent breakup, and evaporation phase transition. This model accurately simulates the atomization process of fuel after it is ejected from the injector, including phenomena such as droplet formation, breakup, and evaporation, providing an accurate physical basis for subsequent fuel injection control. It considers pressure fluctuations within the high-pressure fuel line, fuel compressibility, and viscosity effects, and establishes a one-dimensional unsteady flow equation. This model accurately describes the flow characteristics of fuel within the high-pressure fuel line, predicts the impact of pressure fluctuations on fuel injection, and provides theoretical support for optimizing the fuel injection system. It also models the dynamic response characteristics of the injector solenoid valve, taking into account coil inductance changes, hysteresis effects, and mechanical motion characteristics. This model can accurately predict the opening and closing time of the solenoid valve and the motion characteristics of the needle valve, providing key parameters for achieving precise fuel injection control. It uses Proper Orthogonal Decomposition (POD) combined with the Galerkin projection method to reduce the high-dimensional CFD model to a low-order state-space model, improving computational efficiency by more than 90%. Through order reduction processing, the calculation speed of the model is greatly improved while ensuring model accuracy, enabling it to meet the needs of real-time control.

3. The precise adjustment method of an engine fuel injection system according to claim 1, characterized in that: In the S2, the multi-sensor fusion layout: Piezoelectric pressure sensor: measures common rail pressure (accuracy ±0.5MPa) and fuel line pressure fluctuations (frequency response ≥10kHz), and can accurately obtain fuel system pressure information in real time, providing an important basis for fuel injection control; High-speed camera: Captures spray morphology (frame rate ≥ 10,000 fps) for model verification and correction. Spray images captured by high-speed cameras can be used to visually observe the atomization effect of the fuel, evaluate the accuracy of the model, and correct and optimize the model. Ion current sensor: Detects the ion current in the combustion chamber, reflecting the combustion quality. The ion current sensor can monitor the changes in ion current during the combustion process in real time, thereby determining the stability and sufficiency of combustion and providing feedback information for adjusting fuel injection parameters; Vibration sensor: monitors the impact of the injector needle valve seating and determines the actual opening time. The vibration sensor can detect the vibration signal generated when the injector needle valve seats, accurately determine the actual opening time of the needle valve, and improve the accuracy of fuel injection control; Data preprocessing algorithm: Wavelet transform denoising: eliminates high-frequency electromagnetic interference and improves the quality and reliability of sensor data. Kalman filter fusion: It processes the time delay and inconsistency of multi-sensor data and fuses the data from different sensors to obtain more accurate and reliable state estimation; Outlier detection: Identify sensor faults based on the isolation forest algorithm, detect sensor anomalies in a timely manner, and ensure the normal operation of the system.

4. The precise adjustment method of an engine fuel injection system according to claim 1, characterized in that: The S3 contains an 18-dimensional state vector covering engine speed, load, coolant temperature, common rail pressure, fuel line pressure fluctuation characteristics, etc. This fully and accurately describes the engine's operating state, providing rich information for deep reinforcement learning algorithms. The continuous control space, including injection pulse width (0.5-20ms, resolution 0.01ms), injection pressure (50-250MPa, resolution 0.1MPa), and multiple injection intervals (0.1-5ms, resolution 0.05ms), provides sophisticated control means to achieve precise adjustment of the fuel injection process. Reward function design: Consider the following factors: Deviation between actual injection amount and target value (weight 0.4): Ensuring the accuracy of fuel injection amount is the basis for achieving precise control; Combustion stability index (COVofIMEP) (weight 0.3): ensures the stability of the combustion process and improves engine performance and reliability; Fuel consumption rate (weight 0.2): optimize fuel economy and reduce fuel consumption; Emission index (NOx+PM) (weight 0.1): reduce pollutant emissions and meet environmental protection requirements; Algorithm Architecture: This algorithm uses the TwinDelayedDDPG (TD3) algorithm, introducing delayed policy updates and target network smoothing mechanisms to improve training stability. The TD3 algorithm has advantages in handling continuous control problems, effectively avoiding overestimation and improving algorithm training efficiency and stability.

5. The precise adjustment method of an engine fuel injection system according to claim 1, characterized in that: In S4, based on model predictive control (MPC), the injection parameter sequence is continuously optimized within a 100ms prediction window, taking into account system constraints such as the minimum injection pulse width and the upper and lower limits of common rail pressure. This layer optimizes fuel injection parameters from a global perspective to achieve optimal overall performance. The compensation and adjustment layer, based on online identification of system characteristics, compensates for deviations caused by factors such as injector wear and fuel temperature fluctuations in real time. Recursive least squares (RLS) is used for system parameter identification. This layer adaptively adjusts the control strategy to compensate for changes in system characteristics and ensure stable control accuracy. The injector drive circuit is designed based on sliding mode control theory to achieve an injection timing control accuracy of ±0.05ms. An adaptive feedforward compensation strategy is used to eliminate the effects of solenoid valve response delay. This layer precisely executes control commands to ensure accurate and consistent fuel injection.

6. The precise adjustment method of an engine fuel injection system according to claim 1, characterized in that: In the S5, a HIL platform including real injectors and sensors is constructed to verify the real-time performance of the control algorithm. Through HIL testing, the actual operation of the engine can be simulated in a laboratory environment to verify the real-time performance and effectiveness of the control algorithm. Gaussian process regression (GPR) is used to perform online correction of the reduced-order model to compensate for modeling errors. As the engine operating time increases, the model may deviate. Through online model correction, the model parameters can be adjusted in time to improve the accuracy of the model. Through vehicle-cloud collaboration, operating data of different vehicles are collected, and a federated learning algorithm is used to continuously optimize the control strategy. The powerful computing power and big data analysis technology of the cloud are used to continuously optimize the control strategy to improve the performance and adaptability of the system.

Images