Method and apparatus for forging an automotive fuel injector forging

Abstract

This application relates to the field of automotive fuel injector forging technology, and particularly to a forging method and equipment for automotive fuel injector forgings. The method includes: acquiring metal flow resistance, metal flow velocity, forging vibration data, and lateral force; determining a target flow velocity based on the metal flow resistance, metal flow velocity, and forging vibration data; determining axial strain coefficient and radial strain coefficient based on the metal flow resistance, metal flow velocity, forging vibration data, and lateral force; generating a slider control command based on the axial strain coefficient and the target flow velocity, and generating a push rod control command based on the radial strain coefficient; and controlling a forging device and a hydraulic device to forge the automotive fuel injector forging according to the slider control command and the push rod control command, respectively. This method can significantly improve the forging accuracy, production efficiency, and the level of intelligence and automation in the forging process.

Description

Technical Field

[0001] This application belongs to the field of automotive fuel injector forging technology, and particularly relates to forging methods and equipment for automotive fuel injector forgings. Background Technology

[0002] Automotive fuel injectors are responsible for precisely injecting fuel into the engine's combustion chamber to ensure complete combustion and efficient engine operation. The quality of fuel injectors directly affects engine performance, emission standards, and fuel economy. As the automotive industry's requirements for environmental protection and performance increase, fuel injector design and manufacturing technologies are constantly advancing. Forging technology, due to its ability to provide higher material strength, good fatigue resistance, and superior wear resistance, has become one of the key technologies for manufacturing high-precision fuel injectors.

[0003] In the existing technology, in the traditional forging process of automotive fuel injector forgings, the process path is usually preset with fixed parameters in the process design stage, and is repeatedly run with the same loading path and motion trajectory in actual production. This cannot adapt to the real-time fluctuations in material properties and is prone to a series of quality defects, including folding, flow marks, and insufficient filling. Summary of the Invention

[0004] This application provides a forging method and equipment for automotive fuel injector forgings, which can solve the problem of insufficient forging accuracy caused by the reliance on fixed paths in traditional forging, which cannot adapt to material fluctuations.

[0005] In a first aspect, embodiments of this application provide a forging method for automotive fuel injector forgings, including:

[0006] Acquire metal flow resistance, metal flow velocity, forging vibration data, and lateral force; wherein, the forging vibration data includes die vibration amplitude and die vibration frequency, and the lateral force is used to characterize the intensity of radial metal flow, and the lateral force includes expansion thrust along the x-axis and expansion thrust along the y-axis;

[0007] The target flow velocity is determined based on the metal flow resistance, the metal flow velocity, and the forging vibration data.

[0008] Based on the metal flow resistance, the metal flow velocity, the forging vibration data, and the lateral force, the axial strain coefficient and the radial strain coefficient are determined.

[0009] Based on the axial strain coefficient and the target flow velocity, a slider control command is generated, and based on the radial strain coefficient, a push rod control command is generated; wherein, the slider control command is used to control the slider to apply pressure to the metal along the axial direction of the automotive fuel injector forging for forming, and the push rod control command is used to adjust the pressure of the four push rods to guide the flow of the metal in the radial direction, wherein the four push rods are all located in the horizontal direction and are respectively distributed in the positive and negative axis directions of the x-axis and the positive and negative axis directions of the y-axis;

[0010] According to the slider control command and the push rod control command, the forging device and the hydraulic device are respectively controlled to forge the automotive fuel injector forging.

[0011] The technical solutions described in this application embodiment have at least the following technical effects:

[0012] The forging method for automotive fuel injector forgings provided in this application first obtains metal flow resistance, metal flow velocity, forging vibration data (die vibration amplitude and die vibration frequency), and lateral force (characterizing the intensity of radial metal flow, including expansion thrust along the x-axis and expansion thrust along the y-axis). Then, based on the metal flow resistance, metal flow velocity, and forging vibration data, a target flow velocity is determined. Next, based on the metal flow resistance, metal flow velocity, forging vibration data, and lateral force, the axial strain coefficient and radial strain coefficient are determined. Then, based on the axial strain coefficient and target flow velocity, a slider control command is generated (used to control the slider to apply pressure to the metal along the axial direction of the automotive fuel injector forging for forming). Based on the radial strain coefficient, an ejector control command is generated (used to adjust the pressure of four ejector rods to guide the radial flow of metal; the four ejector rods are all located in the horizontal direction and are distributed in the positive and negative axes of the x-axis and y-axis, respectively). Finally, according to the slider control command and ejector control command, the forging device and hydraulic device are controlled to forge the automotive fuel injector forging. This method achieves precise control over metal flow behavior by comprehensively considering metal flow resistance, flow velocity, vibration parameters, and lateral forces, thereby significantly improving the uniformity of the internal structure and forming accuracy of forgings. By adjusting the movement of the slider and ejector pins through axial and radial strain coefficients, this method effectively controls the strain path and deformation coordination, reducing the probability of forming defects. Through multi-parameter dynamic optimization control of the forging process, this method can reduce unnecessary energy consumption and debugging time, improve production efficiency and system stability, and enhance the intelligence and automation level of the forging process. By employing a horizontally arranged four ejector pin structure combined with radial strain coefficient control, the uniform filling of metal in multiple radial directions can be effectively controlled, improving the coaxiality and structural symmetry of the forgings.

[0013] Secondly, embodiments of this application provide a forging apparatus for automotive fuel injector forgings, comprising:

[0014] The acquisition unit is used to acquire metal flow resistance, metal flow velocity, forging vibration data, and lateral force; wherein, the forging vibration data includes die vibration amplitude and die vibration frequency, and the lateral force is used to characterize the intensity of metal radial flow, and the lateral force includes expansion thrust along the x-axis direction and expansion thrust along the y-axis direction;

[0015] The target flow velocity determination unit is used to determine the target flow velocity based on the metal flow resistance, the metal flow velocity, and the forging vibration data;

[0016] The strain coefficient determination unit is used to determine the axial strain coefficient and the radial strain coefficient based on the metal flow resistance, the metal flow velocity, the forging vibration data and the lateral force.

[0017] A control command generation unit is used to generate slider control commands based on the axial strain coefficient and the target flow velocity, and to generate push rod control commands based on the radial strain coefficient; wherein, the slider control commands are used to control the slider to apply pressure to the metal along the axial direction of the automotive fuel injector forging for forming, and the push rod control commands are used to adjust the pressure of the four push rods to guide the radial flow of the metal, wherein the four push rods are all located in the horizontal direction and are respectively distributed in the positive and negative axis directions of the x-axis and the positive and negative axis directions of the y-axis;

[0018] The forging unit is used to control the forging device and the hydraulic device to forge the automotive fuel injector forging according to the slider control command and the push rod control command, respectively.

[0019] Thirdly, embodiments of this application provide a forging apparatus for automotive fuel injector forgings, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method described in any of the embodiments of the first aspect.

[0020] It is understood that the beneficial effects of the second and third aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic flowchart of a forging method for automotive fuel injector forgings according to an embodiment of this application;

[0023] Figure 2 This is a schematic diagram of the structure of the automotive fuel injector forging equipment provided in the embodiments of this application. Detailed Implementation

[0024] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0025] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0026] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0027] In related technologies, the plasticity, flow resistance, and strain hardening behavior of metallic materials often vary across different batches, heating states, or mold environments. Traditional forging processes generally employ fixed loading paths and preset process parameters, and repeat the same loading path and motion trajectory in actual production. This can lead to excessive or insufficient loading, resulting in inadequate or excessive metal flow; uneven deformation in localized areas of the part, especially in complex transition regions such as the needle valve seat, inclined surfaces, and arc parts of fuel injectors; and difficulty in ensuring overall forming accuracy, affecting the consistency of the parts and their subsequent machinability.

[0028] Fuel injector forgings typically have complex small cross-sections and precise contour structures, such as nozzle guides and tapered connecting sections. These parts require extremely high precision in metal filling behavior. However, in traditional forging processes, because the forming process relies solely on unidirectional loading control, defects such as insufficient filling and folding of the material in the mold cavity are prone to occur. Furthermore, under a fixed loading path, if the metal flow behavior deviates from the preset path, surface and internal defects are easily formed at mold corners or flow diversion zones.

[0029] Due to uncontrollable material fluctuations and flow paths, the pressure distribution and metal contact behavior of the same die vary greatly in different forging cycles. This can easily cause premature wear or even cracks in certain areas of the die due to repeated impacts; product consistency deteriorates, requiring additional subsequent inspection and repair processes; overall production efficiency decreases, and the pass rate fluctuates greatly, making it difficult to meet the requirements of large-scale high-quality forging.

[0030] To address the aforementioned problems, this application provides a forging method and equipment for automotive fuel injector forgings. The method first acquires metal flow resistance, metal flow velocity, forging vibration data (die vibration amplitude and die vibration frequency), and lateral force (characterizing the intensity of radial metal flow, including expansion thrust along the x-axis and expansion thrust along the y-axis). Then, based on the metal flow resistance, metal flow velocity, and forging vibration data, a target flow velocity is determined. Next, based on the metal flow resistance, metal flow velocity, forging vibration data, and lateral force, the axial strain coefficient and radial strain coefficient are determined. Then, based on the axial strain coefficient and target flow velocity, a slider control command is generated (used to control the slider to apply pressure to the metal along the axial direction of the automotive fuel injector forging for forming). Based on the radial strain coefficient, an ejector control command is generated (used to adjust the pressure of four ejector rods to guide the radial flow of the metal; the four ejector rods are all located horizontally and distributed along the positive and negative axes of the x-axis and y-axis, respectively). Finally, according to the slider control command and ejector control command, the forging device and hydraulic device are controlled to forge the automotive fuel injector forging. This method achieves precise control over metal flow behavior by comprehensively considering metal flow resistance, flow velocity, vibration parameters, and lateral forces, thereby significantly improving the uniformity of the internal structure and forming accuracy of forgings. By adjusting the movement of the slider and ejector pins through axial and radial strain coefficients, this method effectively controls the strain path and deformation coordination, reducing the probability of forming defects. Through multi-parameter dynamic optimization control of the forging process, this method can reduce unnecessary energy consumption and debugging time, improve production efficiency and system stability, and enhance the intelligence and automation level of the forging process. By employing a horizontally arranged four ejector pin structure combined with radial strain coefficient control, the uniform filling of metal in multiple radial directions can be effectively controlled, improving the coaxiality and structural symmetry of the forgings.

[0031] The forging method for automotive fuel injector forgings provided in this application embodiment can be applied to automotive fuel injector forging equipment. In this case, the automotive fuel injector forging equipment is the main body for executing the automotive fuel injector forging method provided in this application embodiment. This application embodiment does not impose any restrictions on the specific type of automotive fuel injector forging equipment.

[0032] For example, such as Figure 2As shown in the figure, the four ejector pins act on the mold in the horizontal direction, while the slider acts on the mold in the vertical direction. The mold in the figure is the same automotive fuel injector mold. The dashed box in the figure indicates that the mold is not part of the hydraulic system. Forging equipment for automotive fuel injector forgings can include a forging press, a hydraulic system, and a control device that communicates with the forging press and the hydraulic system. The forging press is a device that applies pressure to the metal in the axial (vertical) direction via a slider, causing the metal to be integrally formed in the cavity of the automotive fuel injector mold. It can be a mechanical press, a servo press, or a hydraulic forging press, etc. The hydraulic system is a device that guides the metal to flow and form in the radial (horizontal) direction. It can include a liquid storage container (such as an oil tank), a power component (such as a hydraulic pump), a main control component (such as an overflow valve or a safety valve), four branch control components (such as pressure valves, flow valves, and directional valves), and four actuators (such as four hydraulic cylinders (where the ejector pins are located)). The liquid storage container is connected to the power element via an oil suction pipe. The power element is connected to the main control element via a pressure oil pipe. The oil outlet of the main control element is connected to four branch control elements via distribution lines. Each branch control element is connected to a corresponding actuator via an independent pressure oil pipe. Each actuator is connected to the liquid storage container via a separate or shared return oil line. The four actuators are evenly distributed around the mold in the horizontal direction, installed in the positive and negative axes of the horizontal x-axis (+X, -X) and the positive and negative axes of the horizontal y-axis (+Y, -Y). The power element provides pressure oil flow to the main control element, which distributes pressure oil flow to the four branch control elements. The four branch control elements independently control the opening, closing, and flow rate of their respective oil circuits. Each branch control element adjusts the pressure value of its corresponding actuator to control the extension or retraction of the actuator's push rod.

[0033] The overall shape of the automotive fuel injector mold is basically consistent with the shape of the fuel injector to be formed, including a main cavity and possibly multiple sub-cavities, channels, or internal holes. The mold surface has four holes with the same shape as the ejector pins (the shape of the contact surface between the ejector pin and the mold). An ejector pin is placed in each hole, and each ejector pin enters the mold through a channel and contacts the metal inside the mold. The ejector pins are tightly connected to the mold surface through the channels, but do not directly jam the mold; instead, they push the metal along the inner cavity of the mold through pressure. During the main forging process (the forging device applies pressure to the metal in a vertical direction), the ejector pins remain in place and are held in position on the inner surface of the mold to prevent metal from overflowing from the mold holes.

[0034] The control device is a device capable of controlling the forging and hydraulic devices and performing data processing. It can be a tablet computer, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), desktop computer, computing device or other processing device connected to a wireless modem, computer, laptop computer, customer premises equipment (CPE) and / or other devices for communication over wireless systems, as well as next-generation communication systems, such as mobile terminals in 5G networks or mobile terminals in future evolved Public Land Mobile Networks (PLMNs).

[0035] To better understand the forging method for automotive fuel injector forgings provided in the embodiments of this application, the specific implementation process of the forging method for automotive fuel injector forgings provided in the embodiments of this application will be described by way of example below.

[0036] Figure 1 A schematic flowchart of a forging method for automotive fuel injector forgings according to an embodiment of this application is shown. The forging method for automotive fuel injector forgings includes:

[0037] S100 acquires metal flow resistance, metal flow velocity, forging vibration data, and lateral force. The forging vibration data includes the die vibration amplitude and frequency. The lateral force characterizes the intensity of the radial flow of metal and includes the expansion thrust along the x-axis and the expansion thrust along the y-axis.

[0038] As can be understood, metal flow resistance refers to the internal resistance to the flow of metal material during the forging process when it undergoes plastic deformation under external force; it manifests as stress per unit area. The greater the metal flow resistance, the more difficult the metal is to deform.

[0039] For example, strain gauges or force sensors can be installed on the contact surface between the mold and the metal to monitor in real time the resistance per unit area (i.e., flow stress) experienced by the metal during deformation. The data can be transmitted to the control system through a data acquisition system, and the flow resistance of the metal at a specific time and location can be calculated.

[0040] It can be understood that metal flow velocity refers to the speed at which metal material moves plastically under the action of forging force. It directly reflects the rate at which the metal deforms and fills in the mold cavity. It is expressed as the distance the metal flows per unit time, and the unit can be mm / s or m / s.

[0041] For example, a three-dimensional laser Doppler velocimeter (3D LDV) or a three-dimensional laser displacement sensor array can be deployed on the mold surface or in key areas of the metal flow path to collect the flow velocity of the metal in the x, y, and z directions. Based on the flow velocity in the x, y, and z directions, the modulus of the total velocity vector is calculated, which is the metal flow velocity. The calculation formula is as follows: ,in, Indicates the metal flow rate. Represents the flow velocity along the x-axis. This represents the flow velocity along the y-axis. This represents the flow velocity along the z-axis.

[0042] It is understandable that forging vibration data describes the periodic vibration characteristics of the die during the forging process. It can include the die vibration amplitude and the die vibration frequency. The die vibration amplitude represents the maximum displacement generated by the die during the vibration process, and the unit is millimeters (mm). The die vibration frequency represents the number of vibrations per unit time, and the unit is Hertz (Hz).

[0043] For example, a high-precision acceleration sensor (such as a piezoelectric accelerometer) or a laser vibration sensor can be installed on the mold. The sensor signal is sent to the signal processing module through a data acquisition card (DAQ) to record the time-domain signal (acceleration-time curve) of the mold's acceleration during the forging process. The collected acceleration curve is numerically integrated to obtain the velocity curve, and then integrated again to obtain the displacement curve. The maximum absolute value of the displacement curve is the mold vibration amplitude.

[0044] The Fast Fourier Transform (FFT) algorithm can be used to convert the time-domain signal into a frequency-domain signal, and the frequency component with the largest amplitude in the frequency-domain signal is the main vibration frequency (mold vibration frequency).

[0045] Lateral force can be understood as the force exerted on the die wall along the x-axis and y-axis directions when metal expands radially after being subjected to axial pressure during forging. It is used to quantify the flow intensity and tendency of metal in the radial direction. The expansion thrust in the x-axis direction represents the radial pushing force generated by the metal along the x-axis during forging; the expansion thrust in the y-axis direction represents the radial pushing force generated by the metal along the y-axis during forging.

[0046] For example, four high-sensitivity pressure sensors are installed in both the x-axis and y-axis directions of the mold. The pressure sensor in the x-axis direction measures the expansion thrust along the x-axis, and the pressure sensor in the y-axis direction measures the expansion thrust along the y-axis. During forging, the metal expands radially under pressure, pushing the mold wall to generate a reaction force. The pressure sensors convert the expansion thrust in the four directions into electrical signals for recording. The measured values ​​from the pressure sensors in the x-axis and y-axis directions are averaged or weighted, respectively, to finally calculate the expansion thrust along the x-axis and the expansion thrust along the y-axis.

[0047] By simultaneously acquiring and processing the above four data items, comprehensive and accurate input parameters can be provided for subsequent forging process control, laying the foundation for intelligent forging control.

[0048] S200 determines the target flow velocity based on metal flow resistance, metal flow velocity, and forging vibration data.

[0049] For example, the flow velocity of metal is influenced by multiple factors. The greater the flow resistance, the more difficult it is for the metal to flow, and the lower the target velocity can be accordingly. The greater the vibration amplitude and frequency, the more significant the softening and lubrication effect on the metal, and the higher the target velocity can be appropriately increased. The current metal flow velocity provides the basis for the flow trend, and the target velocity can be fine-tuned based on the current metal flow velocity. A weighted model or a data-driven model can be used to construct the flow control model, such as... ,in, Indicates the target flow rate. Indicates the current metal flow rate. The adjustment coefficient can be obtained through experience or training with data. Indicates the amplitude of mold vibration. Indicates the vibration frequency of the mold. This indicates the resistance to metal flow.

[0050] The metal flow resistance, metal flow velocity, and forging vibration data obtained in step S100 can be substituted into the flow control model to calculate the target flow velocity. The target flow velocity can be used as a key input parameter for subsequent control strategies, such as slider speed control and deformation rate matching, which helps the metal flow at the optimal speed and fill the mold.

[0051] This step, through a comprehensive analysis of the current state of the metal (metal flow resistance, metal flow velocity) and external excitation (forging vibration data), dynamically calculates a target flow velocity that makes the forming process more stable, uniform, and defect-free, providing a quantitative basis for subsequent pressure regulation and path control, and is a key link in realizing intelligent forging.

[0052] In one possible implementation, S200, based on metal flow resistance, metal flow velocity, and forging vibration data, determines the target flow velocity, including:

[0053] S210, based on data of metal flow resistance, metal flow velocity and forging vibration, determines the rheological resistance control coefficient.

[0054] For example, during hot forging, the plastic deformation behavior of metals is influenced by both the material's inherent rheological properties and external loading conditions. To achieve accurate prediction and control of metal deformation behavior, a rheological resistance control coefficient can be dynamically calculated as an important parameter for constructing the forming force model and control strategy. A calculation model for the rheological resistance control coefficient can be constructed by introducing metal flow resistance, metal flow velocity, and forging vibration data. ,in, This represents the rheological resistance control coefficient. Indicates the basic rheological regulation coefficient. Indicates the resistance to metal flow. Indicates the metal flow rate. Indicates the amplitude of mold vibration. Indicates the vibration frequency of the mold. and This represents an empirical adjustment coefficient, which is set based on the material type and process conditions.

[0055] By conducting hot compression or tensile tests on the target material under high-temperature conditions, its stress-strain-strain rate curve can be obtained. From this curve, the constitutive model parameters of the material can be fitted, and the basic rheological coefficients can be extracted. , representing the plastic resistance response of a material under vibration-free conditions.

[0056] Based on the obtained data on metal flow resistance, metal flow velocity, and forging vibration, as well as the calculation model of the rheological resistance control coefficient, the rheological resistance control coefficient can be calculated, which reflects the actual response of the metal to deformation under the current vibration, flow velocity, and resistance.

[0057] This step, by introducing metal flow resistance, metal flow velocity, and forging vibration data to jointly regulate rheological parameters, achieves a more accurate modeling of metal plastic behavior, providing a dynamic response basis for subsequent intelligent forging control strategies.

[0058] S220, based on metal flow resistance and target forming force, determines the forming driving force. The forming driving force characterizes the net force actually applied to metal deformation.

[0059] It can be understood that during the forging process, the total force output by the equipment (i.e. the target forming force) can be decomposed into metal flow resistance (resistance to overcome internal material friction, mold friction and interface adhesion, etc., which hinder flow) and forming driving force (representing the net effective force actually used by the equipment to drive the material to undergo plastic deformation).

[0060] For example, the total forming load (target forming force) required at the current stage can be determined based on process design, empirical values, or finite element simulation results. The force remaining after subtracting the metal flow resistance from the target forming force is the forming driving force, i.e. ,in, Indicates the molding driving force. This represents the target forming force.

[0061] S230 determines the target flow velocity based on the rheological resistance control coefficient and the forming driving force.

[0062] For example, the target flow rate achieved during the current forming process can be determined by inversely solving the relationship between the rheological resistance control coefficient, the forming driving force, and the target flow rate. The relationship between the rheological resistance control coefficient, the forming driving force, and the target flow rate is as follows: The target flow rate can be used to generate subsequent control commands to ensure uniform plastic deformation while taking into account both cavity filling quality and energy efficiency.

[0063] Through these steps, at each forging time point, the equipment output force is clearly decomposed into the resistance-overcoming part and the effective deformation part, the material properties are organically coupled with the forming process, and the target flow rate is calculated, thereby achieving precise control and intelligent forging.

[0064] S300 determines the axial strain coefficient and radial strain coefficient based on metal flow resistance, metal flow velocity, forging vibration data, and lateral force.

[0065] For example, metal flow resistance, metal flow velocity, forging vibration data, and lateral force can be transformed into the [0,1] interval using a normalization function to construct a standard input vector, thereby eliminating the influence of dimensions and facilitating subsequent modeling, such as X=[ , , , , , ],in, Indicates the resistance to metal flow. Indicates the metal flow rate. Indicates the amplitude of mold vibration. Indicates the vibration frequency of the mold. This represents the expansion thrust along the x-axis. This represents the expansion thrust along the y-axis.

[0066] Data-driven modeling methods such as multiple regression, neural networks, or support vector machines (SVM) can be used to establish a nonlinear mapping relationship between input parameters and strain coefficients. The axial strain coefficient calculation model is as follows: ,in, The axial strain coefficient represents the unit strain of the metal in the axial direction of the forging (i.e., the direction of the sliding block), reflecting the degree of axial deformation. The radial strain coefficient calculation model is as follows: ,in, The radial strain coefficient reflects the flow and expansion capacity of metal in the radial direction (i.e., perpendicular to the axial direction) of the mold, determining the magnitude of the supplementary effect of the ejector pin. The computational model can be trained based on actual production data or finite element simulation data. The model parameters are optimized using error backpropagation algorithms or the least squares method, ensuring that the error between the predicted and actual values ​​remains within an acceptable range.

[0067] Since the various influencing factors in the forging process have different degrees of dominance, weighting factors can be introduced for coupling and adjustment. For example, for metals with high flow resistance, the weight of vibration frequency in the model can be increased to promote forming; when lateral force... , When the value exceeds the set threshold, the model can strengthen the radial strain weight to avoid metal overflow or insufficient cavity filling.

[0068] During the forging process, the metal flow resistance and vibration state can be monitored in real time. The data is fed back to the control system via industrial sensors. The system then updates the input vector X cyclically based on the latest data and updates the strain coefficient in real time. and This enables dynamic and adaptive deformation control.

[0069] This step provides core parameter support for the generation of subsequent control commands, which helps the forging achieve the best forming state in both the axial and radial directions.

[0070] In one possible implementation, S300, based on metal flow resistance, metal flow velocity, forging vibration data, and lateral force, determines the axial strain coefficient and radial strain coefficient, including:

[0071] S310: Based on the current control cycle, acquire the time window data prior to the current control cycle, and predict the defect probability of the automotive fuel injector forging based on the time window data. The time window data includes the time series of metal flow resistance, metal flow velocity, and forging vibration data.

[0072] For example, a fixed-length time window [t0, t] can be set to capture the process dynamics before the current control cycle t. The window length Δt(t-t0) can cover the characteristic change cycle within a typical forging stage (such as pre-pressing, main pressing, or vibration-assisted stage), and can be adjusted according to the forging cycle, such as being set to 0.5~2 seconds.

[0073] The system can read metal flow resistance data, metal flow velocity data, and forging vibration data from the time window in real time from the data storage module, and synchronize and align them with the timestamps to form time series of metal flow resistance, metal flow velocity, and forging vibration data. The three types of time series are then uniformly encapsulated into a time series input matrix. ,in, This represents the time series of metal flow resistance. Represents the time series of metal flow rate. This represents the time series of mold vibration amplitude. This represents the time series of mold vibration frequencies.

[0074] Preprocessing and feature extraction of time window data allows for the calculation of the mean, standard deviation, rate of change, and extreme value locations. Fast Fourier Transform (FFT) can be used to identify periodic anomalies, and time-delay embedding or sliding window interpolation can enhance short-term dynamic representation. Temporal modeling methods can be employed to learn and predict based on the extracted features. Optional models include LSTM (Long Short-Term Memory) networks, suitable for learning continuous process data with strong temporal dependencies; GRU, 1D-CNN, or Transformer, balancing real-time performance with prediction accuracy; and Random Forest / Support Vector Machine (SVM), a lightweight solution for rapid deployment (statistical features must be extracted first). Predictive model training can be based on historical labeled datasets (whether the forging is defective), using defect labels as supervisory signals to build classification or regression models.

[0075] After the model is trained, the time series input matrix X can be input into the prediction model. The prediction model can predict the probability of defects. If the model has a multi-class structure, it can also output the defect type (such as incomplete filling, tissue stratification, eccentricity, etc.).

[0076] By constructing a defect prediction model based on time window data, potential defect risks can be identified in advance before the forming process is completed, and intelligent forging quality control can be achieved through dynamic adjustment. This mechanism improves the consistency and yield of fuel injector forgings and is an important component of building an intelligent manufacturing closed-loop quality management system.

[0077] Optionally, in step S310, based on time window data, the defect probability of the automotive fuel injector forging is predicted, including:

[0078] S311, the total deformation energy is obtained by integrating the time series of metal flow resistance and metal flow velocity.

[0079] For example, the work (i.e., energy) per unit time can be expressed as By continuously sampling throughout the entire time window and integrating the power, the total deformation energy is obtained, i.e. ,in, This represents the total deformation energy.

[0080] Discrete numerical integration methods can be used to calculate the cumulative energy of time series data. For example, the trapezoidal integral method has the following formula: The left rectangle method (when the data accuracy is sufficient) has the following calculation formula: ,in, Indicates the sampling time interval. This represents the total number of data points (sampling points) in the time series.

[0081] By performing point-by-point product integration of the time series of metal flow resistance and flow velocity, the actual deformation energy within the time window can be accurately calculated. This not only quantifies the energy consumption of the process but also provides key basic data for forging quality control and energy efficiency optimization.

[0082] S312, based on the time series of metal flow resistance, calculates the mean resistance and standard deviation of resistance, and calculates the ratio between the standard deviation of resistance and the mean resistance to obtain the force fluctuation coefficient.

[0083] For example, the mean resistance represents the average resistance to deformation of a metal under controlled loading within a time window. The mean resistance can be calculated as the average of all metal flow resistances in a time series. ,in, This represents the average resistance.

[0084] The standard deviation of resistance indicates the degree to which the resistance to metal flow fluctuates around its mean; its calculation formula is: ,in, This represents the mean and standard deviation.

[0085] The force fluctuation coefficient can be obtained by calculating the ratio between the standard deviation of resistance and the mean resistance. ,in, The force fluctuation coefficient is a dimensionless index that can be used to measure the smoothness of the forging process.

[0086] S313, based on the metal flow velocity time series, determines whether the difference between adjacent flow velocities is greater than a threshold, counts the number of times the difference between adjacent flow velocities is greater than the threshold, and obtains the number of flow velocity abrupt changes.

[0087] For example, a flow velocity change threshold (e.g., 0.5) can be set to identify abnormal fluctuations. The flow velocity at adjacent time points is differentially calculated to obtain the difference between adjacent flow velocities. It is then determined whether the difference exceeds the flow velocity change threshold. If it does, it is identified as a sudden change event. The cumulative number of times the judgment condition is met is the number of flow velocity sudden changes. The number of flow velocity mutations can be used as an auxiliary indicator for anomaly prediction, vibration control judgment, and mold damage risk assessment.

[0088] S314, calculate vibration energy based on the time series of forging vibration data.

[0089] For example, under the ideal elastic vibration model, the instantaneous energy density can be expressed as: The vibrational energy can be obtained by integrating the instantaneous vibrational energy density within a time window, i.e. ,in, Represents vibrational energy. The equivalent vibrational mass (which can be considered a constant or obtained through modeling) can be set. =1 is used as a normalization reference to obtain the relative vibration energy index, which is used for comparison under different working conditions.

[0090] By statistically analyzing the number of abrupt changes in the metal flow velocity time series and quantitatively calculating the vibration energy, we have achieved quantitative analysis of two important dynamic factors in the forging process, providing a directly usable data foundation for process control, defect prediction, and intelligent parameter adjustment.

[0091] S315, construct the defect vector based on the total deformation energy, force fluctuation coefficient, flow velocity mutation number, and vibration energy.

[0092] For example, the total deformation energy, force fluctuation coefficient, number of velocity abrupt changes, and vibration energy can be normalized (e.g., min-max normalization or Z-score standardization), and the normalized total deformation energy, force fluctuation coefficient, number of velocity abrupt changes, and vibration energy can be concatenated into a defect vector. .

[0093] S316, based on the defect vector, uses a convolutional neural network to predict the defect probability of automotive fuel injector forgings, and obtains the defect probability.

[0094] For example, a defect vector can be input into a trained convolutional neural network, which can predict the current defect probability in real time.

[0095] The training process of convolutional neural networks: Process data can be collected from historical forging batches, and four key indicators, namely total deformation energy, force fluctuation coefficient, flow velocity mutation number, and vibration energy, can be extracted to construct defect vectors. Each defect vector corresponds to a label (sample pair). If the forging has defects (such as insufficient filling, cracks, etc.), the label is 1; if the forging has no defects, the label is 0.

[0096] One-dimensional convolutional neural networks (1D-CNNs) can be used to model the feature relationships in defect vectors. The structure can include an input layer, where the input can be a four-dimensional feature vector; a convolutional layer (Conv1D), which can use multiple 1D convolutional kernels to extract local feature combinations (such as 16 kernels with a kernel size of 2); the ReLU activation function, which can improve non-linear modeling capabilities; a pooling layer (MaxPooling1D), which can reduce dimensionality and enhance model stability; a fully connected layer (Dense), which can combine and classify the extracted features; and an output layer, which is a single neuron and can use the Sigmoid activation function to output the defect probability.

[0097] All sample pairs can be divided into a training set (approximately 70%), a validation set (approximately 15%), and a test set (approximately 15%). Binary cross-entropy can be used as the loss function, and the Adam optimizer can be used for backpropagation and weight updates. The learning rate can be set to 0.001 and automatically adjusted to accelerate convergence. The network's hyperparameters, such as batch size, can be set to 32; epochs can be set to 50-100, with early stopping determined based on the validation set loss curve; a dropout layer can be added before the fully connected layers to prevent overfitting; and L2 regularization can be applied to the convolutional kernels to enhance generalization ability.

[0098] The system can plot loss function curves and accuracy curves for the training and validation sets to determine if overfitting has occurred (the validation set loss increases while the training set loss decreases), and save the model parameters that perform best on the validation set (model checkpoint). The trained model can be evaluated on the test set using metrics such as accuracy, precision, recall, and F1 score. If the performance is satisfactory, the convolutional neural network can be deployed to the forging system for real-time prediction.

[0099] By constructing a defect vector reflecting abnormal energy behavior and process disturbances, and combining it with a lightweight convolutional neural network structure, a rapid and accurate prediction of defect probabilities in automotive fuel injector forgings was achieved. This not only possesses high real-time performance and scalability, but can also be integrated as a key quality monitoring module in intelligent forging control systems.

[0100] S320 calculates the real-time filling rate of the mold cavity based on the metal flow rate and the mold cavity pressure.

[0101] For example, in the forging process of automotive fuel injector forgings, the filling rate of the mold cavity is an important parameter for measuring whether the metal fully occupies the mold space and whether there are quality problems such as insufficient filling or dead corner defects. The real-time filling rate represents the ratio of the volume of metal filled inside the mold at the current moment to the total effective volume of the mold. The total effective volume of the mold is a fixed value determined by the forging structure.

[0102] The volume of metal entering the mold cavity per unit time can be calculated based on the metal flow velocity. Let the initial time be t0, then the cumulative inflow volume within the time interval [t0, t] is... ,in, This indicates the volume of metal flowing into the mold cavity. This represents the instantaneous effective flow cross-sectional area of ​​the metal entering the mold cavity (which can be approximated as the cross-sectional area of ​​the mold cavity inlet or dynamically estimated by pressure feedback). This represents the flow velocity of metal at the mold inlet (measured or calculated by a displacement sensor). In numerical implementation, volume calculation can be performed using discrete integration within a sliding time window (such as the trapezoidal integral method).

[0103] Piezoelectric sensors can be placed in key areas of the mold cavity (e.g., nozzle seats, flange transition areas) to measure the mold cavity pressure. This pressure reflects the degree of obstruction in metal filling. By combining the trends in flow rate and pressure, it can be determined whether the metal is encountering variable cross-section stagnation or dead zone accumulation. If the metal flow rate is high while the mold cavity pressure remains low, it indicates that the mold cavity is still in the rapid filling stage. If the metal flow rate decreases while the mold cavity pressure increases significantly, it indicates that the metal has begun to squeeze out the remaining unfilled areas, and filling is nearing completion. Therefore, a dynamic correction factor can be set based on the mold cavity pressure to adjust the filling process. Perform real-time compensation, that is ,in, This represents the compensated volume of metal flowing into the mold cavity. Indicates the dynamic correction factor. The specific functional form of ∈ (0.95, 1.05) can be obtained by experimental calibration or machine learning fitting.

[0104] The ratio between the compensated volume of metal flowing into the mold cavity and the total effective volume of the mold can be calculated to obtain the real-time filling rate of the mold cavity. ,in, Indicates the real-time fill rate. This indicates the total effective volume of the mold.

[0105] By integrating metal flow rate with mold pressure information, a real-time calculation mechanism for mold filling rate was established, which can dynamically and accurately assess the mold cavity filling progress, providing key technical support for intelligent forging control and defect early warning of fuel injector forgings.

[0106] Optionally, S320 calculates the real-time filling rate of the mold cavity based on the metal flow rate and the mold cavity pressure, including:

[0107] S321, Determine the pressure filling rate based on the mold cavity pressure.

[0108] For example, n pressure sensors can be deployed in multiple key areas (such as nozzle seats, flange transition areas, etc.) within the mold cavity, with each pressure sensor responsible for collecting metal pressure data of its area in real time.

[0109] For each pressure sensor area, a pressure reference value (standard saturation pressure) can be preset, which is the pressure reached in the area under ideal filling conditions. The pressure reference value can be determined by historical test data, finite element simulation or empirical formula. The pressure reference value of different areas can be different, reflecting their geometric characteristics and forming difficulty.

[0110] To reflect the contribution of different regions to the overall filling state, weighting factors can be introduced. For example, the nozzle region is difficult to form due to restricted flow, so a weight of 1.2 can be set; the flange region is relatively easy to form, so a weight of 0.8 can be set; and other regions can be set to 1.0.

[0111] The formula for calculating the pressure filling rate is as follows: ,in, Indicates pressure filling rate, Indicates the first The actual pressure value is collected in real time by a pressure sensor. Indicates the first Standard saturation pressure in the area where each pressure sensor is located; Indicates the first The weights corresponding to the areas where each pressure sensor is located. The pressure filling rate can reflect whether the mold cavity is close to saturation and compaction during the current forming process.

[0112] By integrating local pressure sensing, regional importance weighting, and normalization, an effective indicator reflecting the overall filling quality of the mold cavity, namely the pressure filling rate, is constructed. This indicator has advantages such as strong real-time performance, clear physical meaning, and easy integration into control systems, and can be widely used in fields such as intelligent forging, process diagnosis, and quality prediction.

[0113] S322, based on the metal flow rate and the total length of the flow path, determines the filling progress.

[0114] For example, the mold cavity can be analyzed in advance using a three-dimensional CAD model to determine the longest path segment from the initial contact point to the final filling of the mold cavity. The length of this longest path segment is the theoretical maximum flow distance required for metal filling, i.e., the total length of the flow path.

[0115] To quantify the progress of the metal filling process, the current distance the metal travels in the mold cavity can be calculated based on the metal flow rate. This distance is then compared to the total length of the flow path to obtain the percentage of filling progress. ,in, Indicates the filling progress. This represents the total distance the metal front has advanced from the start of forming to the current moment. This indicates the total length of the flow path.

[0116] By combining metal flow velocity and CAD path information, an intuitive and physically meaningful flow front tracking system was constructed. This system can not only quantify the forging filling progress in real time, but also provide key decision-making basis for forging process control, helping to achieve intelligent and high-quality metal forming.

[0117] S323: Obtain the current average temperature, and compensate the filling progress based on the current average temperature and the reference temperature to obtain temperature compensation data.

[0118] It is understandable that during the hot forging process of metal, temperature changes can cause thermal expansion and contraction of the metal material, which can directly affect the accuracy of the filling path length and the estimation of the forming volume. In order to more realistically reflect the actual filling degree of the metal in the mold cavity, temperature compensation correction can be applied to the original filling progress.

[0119] For example, a temperature sensor array can be arranged in the metal flow path or a critical area of ​​the mold, and temperature data from multiple points can be collected through the temperature sensor array. The average value of the temperature data from multiple points can be calculated to obtain the current average temperature.

[0120] A reference temperature (e.g., 1150°C) can be preset. The reference temperature can be derived from the ideal molding temperature recommended by the material. If the current average temperature is lower than the reference temperature, the actual volume may be smaller than expected (shrinkage). If the current average temperature is higher than the reference temperature, it may be due to an overestimation of the filling progress caused by expansion.

[0121] A temperature compensation coefficient can be determined based on the current average temperature and the reference temperature, and the filling progress can be compensated according to the temperature compensation coefficient. ,in, This indicates the filling progress after temperature compensation (temperature compensation data). This indicates the current average temperature. Indicates reference temperature. Indicates the temperature compensation coefficient. This represents the coefficient of thermal expansion (unit: 1 / °C), which can be determined based on the material's thermophysical properties (e.g., steel is approximately 1.2 × 10⁻⁶). ~2.0× ).

[0122] This step significantly improves the accuracy of characterizing the actual forming state, providing a more reliable basis for decision support in the intelligent forging process.

[0123] S324 weights and fuses the pressure filling rate, filling progress, and temperature compensation data to obtain the real-time filling rate.

[0124] For example, to adapt to the sensing characteristics of different stages of the forging process, the weight allocation can be dynamically adjusted according to the current filling level, forming a stage-based weighted strategy. See the table below:

[0125]

[0126] The current working stage can be determined based on the filling progress, and the corresponding weight can be selected from the table above based on the current working stage. , , For weights and pressure fill rate, weight and filling progress, The real-time fill rate is obtained by weighted summation of the temperature compensation data.

[0127] By integrating pressure field distribution, path tracking progress, and thermal field compensation effects, a dynamic weight-driven real-time filling rate estimation system was constructed, which improved the perception accuracy and stability of the forging process and provided solid data support for building an intelligent forging closed-loop control system.

[0128] S330 constructs a state vector based on defect probability, real-time fill rate, metal flow resistance, metal flow velocity, forging vibration data, and lateral force.

[0129] For example, to unify parameters of different dimensions and units, each data point can be standardized (e.g., Z-score normalization or min-max scaling), and the standardized data can be concatenated into an 8-dimensional state vector. .

[0130] S340, based on state vectors, uses a policy network to calculate the axial strain coefficient and radial strain coefficient.

[0131] It can be understood that the strategy network is a deep neural network structure used to directly output control decisions based on the state vector. Its function is to predict the deformation strategy executed by the slider and the ejector based on the current state of the forging system. That is, the axial strain coefficient determines the deformation adjustment intensity of the slider in the Z-axis direction; the radial strain coefficient determines the adjustment intensity of the ejector in inducing metal flow in the x and y directions.

[0132] For example, the policy network can adopt a multilayer perceptron (MLP) architecture. For instance, the input layer is an 8-dimensional state vector, the hidden layers are 2-3 layers with 32-128 neurons per layer, the activation function can be ReLU, and the output layer has two consecutive output values, representing the axial strain coefficient and the radial strain coefficient, which can be normalized to [0, 1] or [-1, 1] using Sigmoid or Tanh. The output format can be... ,in, Represents the policy function. This represents the trainable parameters of the policy network. The policy network can be trained using supervised learning, which uses historical process data (including optimal strain output) to train the policy network; or using reinforcement learning, which treats the forging process as the environment, constructs a reward function (such as reducing defect rate or improving filling uniformity), and learns online or offline through policy gradients (such as PPO, DDPG).

[0133] In each control cycle, the input state vector The strategy network can output axial strain coefficients in real time. and radial strain coefficient .

[0134] This step enables data-driven decision control for complex forging processes, allowing for real-time response to changes in process status. It can adapt to different molds, materials, and working conditions, and can access more sensor data to expand the state dimensions.

[0135] Optionally, step S340, based on the state vector, calculates the axial strain coefficient and radial strain coefficient using a policy network, including:

[0136] S341, based on the state vector, uses a policy network to forward propagate and calculate the mean axial strain and mean radial strain.

[0137] For example, the state vector can be normalized to a uniform interval of [0, 1], which is beneficial for effective training and inference of the neural network. The normalized state vector can be input into the policy network. The policy network uses linear mapping and ReLU activation function to process the normalized state vector to obtain a 64-dimensional feature representation. Normalizing the 64-dimensional feature representation improves training stability and generalization ability. Tanh activation is used to process the normalized 64-dimensional feature representation to obtain the final feature representation (32-dimensional feature representation). The final feature representation is mapped to three strain mean values ​​through linear transformation: the axial strain mean value and the radial strain mean value (x-axis strain mean value and y-axis strain mean value).

[0138] This step uses a multi-layer feedforward neural network to map multi-dimensional state perception information into continuous control quantities (average axial and radial strain values), realizing intelligent, multi-axis, and multi-factor control and adjustment capabilities for the forging process.

[0139] S342 performs motion sampling on the mean axial strain and mean radial strain to obtain the sampled mean axial strain and mean radial strain.

[0140] For example, theoretically, the mean strain output by the strategy network is only an expected value, while the actual strain applied to the forging can fluctuate around the expected value. Therefore, a normal distribution model can be used to model the output of the strategy network for sampling. Specifically, a two-dimensional Gaussian distribution with the mean of the strategy network output and a covariance of a diagonal matrix can be constructed to generate sampled values, i.e. ,in, This represents the mean axial strain after sampling. This represents the average x-axis strain after sampling. This represents the average y-axis strain after sampling. This represents the mean axial strain. This represents the mean strain along the x-axis. This represents the mean strain along the y-axis. Let represent the covariance matrix.

[0141] To control the intensity of sampling perturbation, a diagonal covariance matrix can be defined. Each direction corresponds to a fixed standard deviation. The standard deviation indicates the permissible level of fluctuation. A larger standard deviation means the sampled values ​​deviate more from the mean, indicating a more exploratory system; conversely, a smaller standard deviation indicates greater stability. A value of 0.1 can be selected, indicating a smaller disturbance, which is suitable for fine-tuning in a more stable phase.

[0142] Samples can be taken once in both the axial and radial directions. For the axial strain control value, in Nearby sampling yielded For radial strain control values, in and Samples were taken from nearby locations to obtain... , The sampled values ​​are the axial and radial deformation strain values ​​that the system expects to apply to the metal under the current state, and can be used as direct inputs for the generation of subsequent control commands.

[0143] This step enables the system to withstand certain random disturbances, preventing the strategy from getting trapped in local optima. During the strategy training phase, the sampling mechanism supports diverse exploration, which can improve the comprehensiveness of strategy convergence. The sampling mechanism can simulate the perturbations caused by uncontrollable factors in actual equipment, enhancing the control system's adaptability to real-world operating conditions.

[0144] S343, constrains are applied to the sampled mean axial strain and mean radial strain to obtain the axial strain coefficient and radial strain coefficient. The constraint processing includes range constraints, rate of change constraints, and safety constraints. The radial strain coefficient includes x-axis strain components and y-axis strain components.

[0145] For example, the numerical range of strain coefficients can be limited to a physically permissible and equipment-controllable range to avoid illegal or unexecutable control values ​​due to sampling offsets or strategy errors. For instance, axial strain control values ​​can be limited to the range [0.5, 2.0]; radial strain control values ​​can be limited to the range [0.4, 1.8]. Range constraints ensure that strain does not exceed the design limits of the mold, material, or hydraulic system, preventing equipment damage or abnormal flow. For example, If it is 1.30, then =clip(1.30, 0.5, 2.0)=1.30.

[0146] To prevent abrupt changes in the strain coefficient within adjacent control cycles, which could lead to drastic system fluctuations or abnormal metal flow, the variation range of axial and radial strain values ​​can be limited. For example, a maximum allowable variation can be set. (e.g., 0.3), if the difference between the new sampled value and the previous period value exceeds... This limits the range of change of the new sampled value to Internally, the rate of change constraint can effectively smooth the control behavior, improve the continuity and stability of the forming process, and reduce frequent fine-tuning impacts.

[0147] To ensure the proper flow direction and distribution characteristics of metal during critical stages, specific logical safety rules can be introduced based on forming process requirements. For example, when the real-time fill rate... When the fill rate is >15% and the radial strain coefficient is below 0.6, to ensure the lateral distribution of the metal, the radial strain coefficient can be maintained at least as high as the previous cycle or increased to the minimum requirement; if the real-time fill rate The axial strain has reached 80%, but it is still too low (e.g., <1.2). The axial strain coefficient can be appropriately increased to promote further longitudinal forming of the metal and prevent insufficient intermediate compaction. Safety constraints can be based on experience or experimental rules, ensuring the effectiveness and controllability of the forging process under complex boundary conditions.

[0148] The axial strain coefficient and radial strain coefficient will be used to generate control commands for the slider and push rod, enabling fine adjustment and control of the metal flow behavior.

[0149] S400 generates slider control commands based on the axial strain coefficient and the target flow velocity, and generates ejector control commands based on the radial strain coefficient. The slider control commands control the slider to apply pressure to the metal along the axis of the automotive fuel injector forging for forming. The ejector control commands adjust the pressure of the four ejector rods to guide the radial flow of the metal. All four ejector rods are horizontally located and distributed along the positive and negative x-axis and y-axis directions, respectively.

[0150] It is understandable that in the process of forging automotive fuel injector forgings, the slider is the actuator that applies the main pressure along the axis of the forging. Its function is to make the metal flow along a predetermined path in a confined space, fill the mold cavity, and complete the final forming.

[0151] Definitions of axial and radial directions: Assuming the working surface of the automotive fuel injector mold faces upwards, the X-axis extends along the left and right sides of the mold, parallel to the working surface; the Y-axis extends along the front and back of the mold, also parallel to the working surface; the Z-axis is perpendicular to the horizontal plane formed by the X and Y axes, and perpendicular to the working surface. Radial direction is the direction parallel to the horizontal plane formed by the X and Y axes, and axial direction is the direction parallel to the Z-axis.

[0152] For example, the control objectives of the slider may include achieving stable deformation of the metal in the axial direction at a target flow rate; ensuring that the axial strain is controlled within the process setting range, and avoiding over-compaction or metal backflow. Based on the control objectives, the ideal motion parameters of the slider (slider pressing speed, slider displacement, and slider pressure) can be calculated in reverse. According to the target flow rate and metal deformation characteristics, the actual pressing speed of the slider satisfies the principle of material volume conservation and considers the axial strain requirements of the metal. The formula for calculating the slider pressing speed is as follows: ,in, Indicates the downward speed of the slider. This indicates the target flow velocity of the metal in the axial direction. It can be based on the flow control model and the flow velocity in the z-axis direction. Calculate ,Right now .when When the value is positive, it indicates that the material is compressed and deformed, and the slider can provide additional displacement; when When the pressure is low, the slider speed can be reduced relatively to prevent overpressure.

[0153] The slider displacement can be calculated based on the slider pressing speed and the set deformation time (which can be given by the process cycle). The formula for calculating the slider displacement is as follows: ,in, Indicates the displacement of the slider. This indicates the deformation time. The upper limit of the slide stroke and the safe stop range can be set according to the mold cavity geometry and forging structural characteristics to prevent over-stroke of the metal extrusion die or slide.

[0154] The slider can provide sufficient pressure to overcome the flow resistance of the metal and factors such as friction and vibration. The formula for calculating the slider pressure is as follows: ,in Indicates the slider pressure. The safety factor, representing the safety margin, can be set based on factors such as mold friction and heat loss. The control system can then adjust accordingly. Set the output pressure of the hydraulic device and monitor it in real time to avoid overload.

[0155] The above calculation results can be summarized to form a set of standardized slider control instructions. For example, {"component": "slider","target_velocity": "target_displacement": "target_pressure": `,"execution_mode": "adaptive_real_time","feedback_enabled": true}.` Slider control commands can be sent to the servo control system or hydraulic control module to guide the slider in executing specific motion trajectories and pressure control, ensuring that the metal completes the required plastic flow and cavity filling along the axial direction.

[0156] During the slider's movement, the actual speed can be continuously monitored and... The deviation, actual pressure and The degree of matching is determined. If a significant deviation is detected, the slider control parameters can be adjusted to dynamically correct the command, thereby achieving closed-loop control.

[0157] By combining the axial strain coefficient with the target flow rate, the coordinated scheduling of pressure and speed, the unification of forming rhythm and target flow rate, and the precision control and dynamic compensation of mold cavity filling are achieved, providing a fundamental guarantee for the high-precision and high-consistency forming of complex small forgings such as automotive fuel injectors.

[0158] In traditional forging processes, both the slide block and ejector pin move along the Z-axis (vertical direction). The slide block handles the main deformation, while the ejector pin ejects the finished product or, when necessary, adjusts the metal flow at the bottom of the die cavity. However, this structure has significant limitations: it cannot effectively induce directional metal flow in the horizontal direction (X, Y axes), especially in complex or irregularly shaped forgings, easily leading to uneven metal filling, dead-angle voids, or microstructure segregation. Therefore, this solution transforms the traditional single ejector pin structure into a horizontal ejector pin array and uses differential control to achieve localized, directional metal flow induction.

[0159] By differentially loading four independent ejector pins distributed in the four quadrants of the mold (the positive and negative axes of the horizontal X-axis and the horizontal Y-axis), radial flow of metal along the horizontal X and Y axes can be effectively induced. The three-dimensional coupling control strategy significantly improves the ability to control complex flow paths, which is beneficial for achieving full filling of the mold cavity edges and corners.

[0160] By adjusting the relative positions and forces of the four ejector pins, the flow of metal at the bottom of the mold can be actively guided, thereby alleviating dead zones, reducing cold shuts and voids, and improving filling integrity. This is particularly effective for complex forgings with asymmetrical shapes or prone to flow deviation, helping to improve dimensional accuracy and microstructure uniformity.

[0161] This solution does not alter the slider structure or affect the Z-axis main pressure system; it only modifies the original single push rod on the Z-axis into a horizontally controllable push rod array (four push rods). This modification is a partial upgrade with high structural compatibility, requiring no major modifications to the main frame and main drive system, and possesses extremely high engineering feasibility and cost-effectiveness.

[0162] For example, the four push rods consist of four independent hydraulic plungers distributed along the positive and negative axes of the horizontal X-axis and the horizontal Y-axis (+X, -X, +Y, –Y). To achieve local differential adjustment, the target pressure coefficient (partial pressure factor model) of each push rod can be calculated based on the radial strain coefficient and lateral force. The partial pressure factor model is as follows: , , , ,in, , , , These represent the target pressure coefficients of the push rod in the +x, -x, +y, and -y directions, respectively. To adjust the gain factor, To prevent the division by zero of tiny positive numbers, the target pressure of each of the four push rods can be calculated based on the set maximum system output pressure and the target pressure coefficients in the four directions. ,in, This indicates the target pressure on the push rod. This indicates the maximum output pressure. The target pressures of the four ejector rods can be encapsulated as standard control commands, for example: {"component":"ejector_array","control_mode":"independent_pressure","target_pressures":{"X_pos": "X_neg": "Y_pos": "Y_neg": },"feedback_enabled":true,"response_profile":"adaptive_real_time"}.

[0163] The ejector control commands can be sent to the hydraulic servo system to control the movement of the four ejector rods, enabling them to dynamically adjust the direction of the metal within the die cavity. During the forging process, pressure and displacement sensors continuously acquire the actual ejector rod response and compare it with the set target. If the metal flow in a certain direction is insufficient, the feedback system can increase the ejector rod pressure in that direction. If the metal flow rate exceeds expectations or causes local bulging, the corresponding ejector rod can be appropriately retracted to prevent overpressure in the die cavity or uncontrolled deformation of the forging.

[0164] By converting the radial strain coefficient into independent pressure control signals for four ejector pins, precise induction and zoned regulation of the radial flow of metal are achieved. Compared with the traditional single ejector pin structure, this strategy not only improves the cavity filling efficiency but also significantly enhances the system's adaptability to complex flow fields.

[0165] In one possible implementation, in step S400, based on the axial strain coefficient and the target flow velocity, a slider control command is generated, including:

[0166] S410, calculate the slider speed based on the axial strain coefficient and the target flow velocity.

[0167] For example, the product of the axial strain coefficient and the target flow velocity can be calculated, which is the slider velocity. The calculation formula is as follows: ,in, Indicates the slider speed. This represents the axial strain coefficient.

[0168] This step enables dynamic and adaptive slider speed control, avoiding uneven flow or defect accumulation caused by simple uniform pressing.

[0169] S420 determines the slider displacement based on the slider speed and generates slider control commands based on the slider displacement.

[0170] For example, the slider displacement can be calculated by multiplying the slider speed with the time step of the current control cycle (the time interval between each round of control command updates, such as 10ms). If the slider displacement is a cumulative value, the slider speed can be updated step by step by integrating.

[0171] The target position of the slider can be determined by adding the calculated slider displacement to the current initial slider position, or by continuously updating the current slider position for closed-loop control in continuous control. If position control mode is used, the target position can be directly converted into a standard control command format (slider control instruction). The control system can compare the actual position with the target position and output adjustment signals (voltage, current, or hydraulic pressure) to drive the slider to the desired position. If speed and displacement combined control is used, the slider speed and target position can be converted into slider control instructions. The generated slider control instructions can be transmitted to the actuator through specific interfaces, such as the PLC control bus, servo controller API, or hydraulic proportional valve analog signals (voltage or current).

[0172] This step transforms the theoretically derived slider displacement into actual physical control commands, completing the closed-loop connection between the strategy network control and the forging equipment. By accurately generating slider control commands, high-quality forming of the fuel injector forging in the axial direction is ensured, providing a stable foundation for subsequent radial flow control and defect suppression.

[0173] In another possible implementation, in step S400, based on the radial strain coefficient, a push rod control command is generated, including:

[0174] S401, calculate the target pressure based on the x-axis strain components, y-axis strain components, and reference pressure. The target pressure includes the target pressure on the four push rods along the positive and negative axes of the horizontal x-axis and y-axis.

[0175] It can be understood that the x-axis strain component characterizes the deformation demand of the metal along the positive and negative axes of the horizontal x-axis (defined as east-west), reflecting the strength adjustment of the pressure applied to the push rod in the east and west directions.

[0176] The y-axis strain component characterizes the deformation demand of the metal along the positive and negative axes of the horizontal y-axis, which is horizontal and perpendicular to the x-axis (defined as the north-south direction), and determines the pressure distribution of the push rod in the north-south direction.

[0177] The reference pressure is a preset reference pressure, determined by material properties, mold geometry, and process objectives, and serves as the baseline for adjusting the ejector pin pressure in all directions.

[0178] For example, the theoretical reference pressure can be calculated based on the material constitutive model, i.e. ,in, Indicates the theoretical reference pressure. Indicates the mold constraint coefficient. Indicates the yield strength of the material. This indicates the compression ratio of the forging thickness. The yield strength of a material can be determined based on its yield properties at high temperatures; for example, the yield strength of 42CrMo steel at 1150℃ is approximately 80 MPa. The initial thickness of the forging can be calculated. With final thickness The ratio reflects the degree of plastic deformation; for example, the typical compression ratio of an automotive fuel injector is about 2.5. Based on the constraint effect of the die structure, the die constraint coefficient in closed-die forging is in the range of 1.2 to 1.5, used to correct for theoretical pressure and reflect factors such as friction and die rigidity.

[0179] The theoretical reference pressure can be constrained by the ratio of the maximum thrust of the ejector rod to the area of ​​a single ejector rod. That is, the theoretical reference pressure must be less than or equal to the ratio of the maximum thrust of the ejector rod to the area of ​​a single ejector rod, thus preventing mechanical damage caused by exceeding the maximum thrust of the equipment. To ensure uniform metal flow in the radial direction, the pressure difference between the ejector rods in the east-west and north-south directions can be limited to a certain proportion of the material's yield strength, such as a pressure difference of less than or equal to 0.3. The reference pressure can be obtained by constraining the maximum thrust of the push rod, the area of ​​a single push rod, and limiting the pressure difference, based on the theoretical reference pressure.

[0180] The reference pressure obtained through the above method can meet both the plastic deformation requirements of the ejector pin and the requirements of equipment safety and forging quality. The reference pressure serves as the basic parameter for subsequent pressure regulation and is crucial for achieving high-quality injector forging control.

[0181] A linear interpolation and symmetric mapping method can be used to calculate the target pressure in four directions based on the reference pressure and the x-axis and y-axis strain components. A linear interpolation structure can be used, and the target pressure in the four directions satisfies... This equation satisfies the condition that the total thrust remains constant, and the deformation direction can be controlled by adjusting only the distribution ratio. This indicates the target pressure on the push rod in the east direction. This indicates the target pressure on the push rod in the west direction. The target pressure of the south-facing push rod. The target pressure on the north-facing push rod. Taking the east-west direction as an example, , ,in, Represents the x-axis strain component, when When = 0, symmetrical loading occurs. =1.5 , =0.5 ;when When =1, balanced loading is achieved. = = ;when When =2, antisymmetric loading is used. =0.5 , =1.5 The calculation method for target pressure in the east-west direction represents a symmetrical-offset-anti-symmetrical adjustment method, which can achieve continuous and controllable flow direction. Similarly, the calculation method for the north-south direction is... , ,in, This represents the y-axis strain component.

[0182] By dynamically adjusting the ejector pin pressure using a linear function, precise control of radial flow can be achieved in the intelligent forging system, effectively improving forging quality and the stability of the forming process.

[0183] S402, obtain the actual pressure in four directions, and calculate the error value between the actual pressure and the target pressure in each direction.

[0184] For example, pressure sensors located at four key positions along the positive and negative axes of the horizontal X-axis and the horizontal Y-axis can be used to collect the actual pressure values ​​in the four directions at the current moment. The pressure error (error value) between the target pressure and its corresponding actual pressure in each of the four directions is then calculated.

[0185] S403 generates push rod control commands for each direction based on the error value in each direction.

[0186] For example, if the error is positive, it means that the actual pressure is lower than the target pressure, indicating that the metal is not supporting enough in that direction. The push rod can be pushed forward to increase the forming force in that area. If the error is negative, it means that the actual pressure is higher than the target pressure, indicating that the area is under too much force. The push rod can be decelerated, stopped, or retracted to avoid excessive local compaction or folding. If the error is zero or within the tolerance range, it means that the actual pressure is close to the target pressure and the current state can be maintained without adjustment.

[0187] The output thrust of the push rod is controlled by hydraulic pressure. The error value can be mapped to the adjustment value of the hydraulic valve (such as pressurization, pressure holding, and pressure reduction). A linear proportional gain can be set so that the push rod thrust changes linearly with the error. The error value can also be mapped to a reference value for the propulsion speed to control the propulsion rhythm of the push rod. A larger error corresponds to a higher propulsion speed, and the speed is reduced when approaching the target to prevent overshoot.

[0188] For each push rod direction, the final generated push rod control command can include the target hydraulic pressure or thrust value, push rod advance speed, control duration, or threshold judgment condition (such as stopping when the target pressure is reached). The push rod control command can be sent in real time to the corresponding push rod control valve group through the control system (such as PLC or hydraulic servo system) to achieve precise pressure regulation and advance control.

[0189] These steps can adjust the direction of metal flow, compensate for differences in pressure distribution within the mold cavity, and achieve a more uniform forging process. They also have feedback closed-loop control capabilities, making them a key component of intelligent control in precision forging.

[0190] The S500, based on slider control commands and push rod control commands, controls the forging device and hydraulic device respectively to forge automotive fuel injector forgings.

[0191] For example, the slider control command can be transmitted to the forging device control unit. The forging device control unit can set the initial position of the slider in the forging device according to the slider control command to ensure alignment with the metal billet. The slider pressing action is started according to the slider pressing speed in the slider control command, and the slider automatically decelerates when the displacement reaches the specified value (slider displacement). The pressing pressure can be dynamically adjusted to match the slider pressure in the slider control command. The hydraulic valve group in the forging device can be adjusted through the feedback of the pressure sensor. The slider load-stroke curve can be monitored to realize closed-loop control of the entire forming process.

[0192] The ejector control command can be transmitted to the hydraulic device control unit. The hydraulic device control unit can set the initial extension position according to the ejector control command to ensure that the four ejector rods contact the metal blank synchronously. According to the target pressure of the four ejector rods, the four hydraulic cylinders in the hydraulic device are driven to output different pressures. During the forming process, the position and force of the ejector rods can be adjusted in real time according to the metal flow state to induce the metal to expand in an orderly manner in the radial direction in the mold cavity. Through the feedback of displacement and load sensors, differential coordination between the ejector rods can be achieved to avoid metal accumulation or flow deviation.

[0193] The control of the slider and ejector pin is not an isolated operation, but rather a coordinated linkage mechanism. The control system continuously monitors the metal flow rate, mold stress state, and displacement progress. If local slow flow, insufficient filling, or off-center loading is detected, the ejector pin pressure distribution can be adjusted, or the slider stroke can be fine-tuned. The control strategy can be switched at different stages (such as pre-pressure, main pressure, and holding pressure) to adapt to the metal strain behavior and flow characteristics. Through this real-time collaborative control of the slider and ejector pin, highly adaptive three-dimensional spatial forming pressure field scheduling can be achieved.

[0194] After forming is completed, the slider stops applying pressure and quickly returns to its original position, while the ejector pin maintains a slight pressure state to keep the forging stably positioned.

[0195] This step achieves efficient collaboration between the forging and hydraulic systems by using slider control commands for main forming control and ejector control commands for flow guidance and compensation. This not only significantly improves the cavity filling quality and forming accuracy of complex forgings but also maintains a simple equipment structure, possessing good industrial application value and adaptability to intelligent manufacturing.

[0196] In one possible implementation, the forging method for automotive fuel injector forgings also includes:

[0197] S10: Construct a reward function based on defect probability, real-time filling rate, and forging vibration data.

[0198] For example, the reward function can consist of positive rewards and negative penalties, and may include infill integrity and forming efficiency (positive rewards), defect risk suppression (negative penalties), and vibration stability (negative penalties). The expression for the reward function is as follows: ,in Represents the reward function, , , The weight coefficients can be adjusted empirically or dynamically adapted during training. The weights can be set according to the task preferences at different stages, such as emphasizing filling efficiency in the early stage and focusing more on defects and stability in the middle and later stages.

[0199] This term can be used to encourage the control strategy to achieve cavity filling as quickly and fully as possible. When the real-time fill rate approaches 1 (i.e., filling is nearly complete), this term tends to its maximum value. A positive result indicates a good process; conversely, if the filling rate is severely insufficient or excessive, the reward will decay rapidly, encouraging the system to quickly return to the target filling rate.

[0200] This factor can indirectly optimize metal flow paths and load configurations by suppressing the predicted defect probability. As the defect probability increases, this factor decreases rapidly, effectively penalizing potential irrational operations or strain distributions.

[0201] This can be used to suppress severe mechanical vibrations or system instability during the forging process. This represents the vibration amplitude or frequency energy within a time window. It is expressed in square form, strongly penalizing violent vibrations.

[0202] This reward function comprehensively considers real-time filling rate (positive feedback), defect probability (negative feedback), and vibration stability (negative feedback), embodying the multi-objective optimization concept in forging control. By continuously adjusting the control strategy to maximize this reward function, the reinforcement learning algorithm can gradually learn the optimal forming path and force control strategy, thereby improving the quality and consistency of the fuel injector forgings.

[0203] S20, determine the temporal difference residual based on the reward function and the state vector.

[0204] For example, after a control action is executed, the state vector after the action is executed can be collected. State vector The current state vector The current state vector and the state vector after the action are executed can be input into the Critic network, which can then evaluate the value of the current and post-action states. , The Critic network is a value function estimator used to determine the long-term performance value of a system under a given state.

[0205] Temporal difference residuals (TD error) are used to measure the difference between the current policy behavior and the expected optimal behavior. The formula for calculating this residual is: ,in, This represents the discount factor (e.g., set to 0.95), used to control the importance of future returns. The time-series difference residuals can be calculated using the TD error calculation formula.

[0206] S30, based on the time-series differential residuals, update the strategy network parameters. The updated strategy network parameters are used to calculate the axial strain coefficient and radial strain coefficient for the next control cycle.

[0207] It can be understood that a policy network takes the current state vector as input and outputs the probability distribution of an action, i.e. ,in, This represents the parameters of the current policy network. This indicates the axial strain coefficient and radial strain coefficient output during the current control cycle.

[0208] For example, it can be based on time-series differential residuals and policy network Calculate the policy gradient, which guides the direction of parameter adjustment. If the action receives high positive feedback (TD error is positive), the probability of that action can be increased; if the feedback is poor (TD error is negative), the probability of that action can be decreased.

[0209] The policy network parameters can be updated based on the policy gradient and a pre-set learning rate (e.g., 0.0003). ,Right now ,in, This represents the learning rate of the policy network, used to control the update magnitude.

[0210] New policy network parameters It will be used in the next control cycle to propagate the current state forward based on the updated strategy network, outputting better axial strain coefficients and radial strain coefficients to achieve smarter and more adaptive control of the forging process.

[0211] Through a strategy network update mechanism based on reward feedback and state evaluation, the system can continuously learn from operational results, iteratively improve the intelligence and forming stability of the control strategy, thereby achieving high-quality, low-defect intelligent forging forming.

[0212] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0213] Corresponding to the forging method for automotive fuel injector forgings described in the above embodiments, this application also provides an automotive fuel injector forging apparatus, the various units of which can realize the various steps of the automotive fuel injector forging method.

[0214] The device includes:

[0215] The acquisition unit is used to acquire metal flow resistance, metal flow velocity, forging vibration data, and lateral force. The forging vibration data includes the die vibration amplitude and frequency. The lateral force characterizes the intensity of the radial flow of metal and includes the expansion thrust along the x-axis and the expansion thrust along the y-axis.

[0216] The target flow velocity determination unit is used to determine the target flow velocity based on metal flow resistance, metal flow velocity, and forging vibration data.

[0217] The strain coefficient determination unit is used to determine the axial strain coefficient and radial strain coefficient based on metal flow resistance, metal flow velocity, forging vibration data, and lateral force.

[0218] The control command generation unit generates slider control commands based on the axial strain coefficient and the target flow velocity, and generates ejector control commands based on the radial strain coefficient. The slider control commands control the slider to apply pressure to the metal along the axis of the automotive fuel injector forging for forming. The ejector control commands adjust the pressure of the four ejector rods to guide the radial flow of the metal. All four ejector rods are horizontally located and distributed along the positive and negative x-axis and y-axis directions, respectively.

[0219] The forging unit is used to control the forging device and the hydraulic device to forge automotive fuel injector forgings according to the slider control command and the push rod control command, respectively.

[0220] It should be noted that the information interaction and execution process between the above-mentioned units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0221] This application also provides a forging equipment for automotive fuel injector forgings. Figure 2 This is a schematic diagram of a forging equipment for automotive fuel injectors, provided as an embodiment of this application. The forging equipment includes a forging press, a hydraulic system, and a control device communicatively connected to the forging press and the hydraulic system. Figure 2 As shown, the control device 6 of the automotive fuel injector forging equipment in this embodiment includes: at least one processor 60 ( Figure 2 Only one is shown in the image), at least one memory 61 ( Figure 2 (Only one is shown in the image) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60, wherein when the processor 60 executes the computer program 62, it causes the automotive fuel injector forging equipment to perform the steps in any of the above-described automotive fuel injector forging method embodiments, or causes the automotive fuel injector forging equipment to perform the functions of each unit in the above-described device embodiments.

[0222] Exemplarily, the computer program 62 may be divided into one or more modules / units, which are stored in the memory 61 and executed by the processor 60 to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 62 in the control device 6 of the automotive fuel injector forging equipment.

[0223] The control device 6 of the automotive fuel injector forging equipment can be a desktop computer, laptop, handheld computer, or cloud server, etc. The automotive fuel injector forging equipment may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will understand that... Figure 2 This is merely an example of forging equipment for automotive fuel injectors and does not constitute a limitation on such equipment. It may include more or fewer parts than shown in the illustration, or a combination of certain parts, or different parts, such as input / output devices, network access devices, buses, etc.

[0224] The processor 60 can be a Central Processing Unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0225] In some embodiments, the memory 61 may be an internal storage unit of the control device 6 of the automotive fuel injector forging equipment, such as the hard drive or memory of the automotive fuel injector forging equipment. In other embodiments, the memory 61 may be an external storage device of the automotive fuel injector forging equipment, such as a plug-in hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the automotive fuel injector forging equipment. Furthermore, the memory 61 may include both internal and external storage units of the automotive fuel injector forging equipment. The memory 61 is used to store operating systems, applications, bootloaders, data, and other programs, such as the program code of computer programs. The memory 61 can also be used to temporarily store data that has been output or will be output.

[0226] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0227] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method of forging an automotive fuel injector forging, comprising: The method comprises: acquiring metal flow resistance, metal flow rate, forging vibration data and lateral force; wherein the forging vibration data comprises die vibration amplitude and die vibration frequency, and the lateral force is used to represent the strength of the radial flow of the metal, and the lateral force comprises expansion thrust in the x-axis direction and expansion thrust in the y-axis direction; determining a target flow rate based on the metal flow resistance, the metal flow rate and the forging vibration data; determining axial strain coefficients and radial strain coefficients based on the metal flow resistance, the metal flow rate, the forging vibration data and the lateral force; wherein the radial strain coefficients comprise x-axis strain components and y-axis strain components; generating slider control instructions based on the axial strain coefficients and the target flow rate, and generating top rod control instructions based on the radial strain coefficients; wherein the slider control instructions are used to control the slider to apply pressure to the metal along the axis direction of the automobile fuel nozzle forging for shaping, and the top rod control instructions are used to adjust the pressure of the four top rods to guide the flow of the metal in the radial direction, and the four top rods are located in the horizontal direction and are distributed in the positive and negative axis directions of the x-axis and the positive and negative axis directions of the y-axis; controlling the forging device and the hydraulic device to forge the automobile fuel nozzle forging according to the slider control instructions and the top rod control instructions; wherein the determination of the target flow rate based on the metal flow resistance, the metal flow rate and the forging vibration data comprises: determining a flow resistance regulation coefficient based on the metal flow resistance, the metal flow rate and the forging vibration data; determining a forming driving force based on the metal flow resistance and the target forming force; wherein the forming driving force is used to represent the net acting force actually used for metal deformation; determining the target flow rate based on the flow resistance regulation coefficient and the forming driving force; the determination of the axial strain coefficients and the radial strain coefficients based on the metal flow resistance, the metal flow rate, the forging vibration data and the lateral force comprises: acquiring time window data before the current control period according to the current control period, and predicting the defect probability of the automobile fuel nozzle forging based on the time window data; wherein the time window data comprises a metal flow resistance time series, a metal flow rate time series and a forging vibration data time series; calculating the real-time filling rate of the die cavity according to the metal flow rate and the die cavity pressure; constructing a state vector according to the defect probability, the real-time filling rate, the metal flow resistance, the metal flow rate, the forging vibration data and the lateral force; calculating the axial strain coefficients and the radial strain coefficients by using a strategy network based on the state vector; the generation of the slider control instructions based on the axial strain coefficients and the target flow rate comprises: calculating the slider speed according to the axial strain coefficients and the target flow rate; determining the slider displacement according to the slider speed, and generating the slider control instructions according to the slider displacement; the generation of the top rod control instructions based on the radial strain coefficients comprises: According to the x-axis strain component, the y-axis strain component and the reference pressure, a target pressure is calculated; wherein the target pressure includes target pressures of four top rods in positive and negative axial directions of horizontal x-axis and positive and negative axial directions of horizontal y-axis; Actual pressures in four directions are obtained, and error values between actual pressures and target pressures in each direction are calculated respectively; Based on the error values in each direction, top rod control instructions in each direction are generated.

2. The method of claim 1 wherein, The method further comprises: The method further comprises: The method further comprises: The method further comprises: The method further comprises: The method further comprises: The method further comprises:

3. The method of claim 1 wherein the fuel nozzle forging is an automotive fuel nozzle forging. The method further comprises: The processor implements the method according to any one of claims 1 to 5 when executing the computer program. The processor implements the method according to any one of claims 1 to 5 when executing the computer program. ​ ​ 4. The method of claim 1 wherein the fuel nozzle forging is an automotive fuel nozzle forging. ​ ​ ​ ​ 5. The method of claim 1 wherein the step of forging the automotive fuel nozzle forging comprises the step of: ​ ​ ​ ​ 6. An automotive fuel nozzle forging apparatus comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein, ​

Images