A hydraulic cushion pre-acceleration control method

Abstract

The present invention provides a hydraulic pad pre-acceleration control method, which relates to the field of hydraulic pad control technology. The method includes: for each working condition of the hydraulic pad, optimizing the pre-acceleration parameters under the working condition based on the hydraulic pad simulation model to obtain the optimal pre-acceleration parameters under the working condition; using the working frequency and drawing stroke corresponding to each working condition as input, and the optimal pre-acceleration parameters under each working condition as output, training a neural network model to obtain an offline pre-acceleration control model of the hydraulic pad; inputting the working frequency and drawing stroke corresponding to the current working condition into the offline pre-acceleration control model to obtain the initial pre-acceleration parameters under the current working condition, and performing online optimization control on the hydraulic pad based on the initial pre-acceleration parameters and trust region reflection. The present invention does not require the design of the pre-acceleration trajectory, can optimize the effect of pre-acceleration during operation, reduce impact, and thus overcome the problems of difficult pre-acceleration trajectory design, difficulty in adapting to different working conditions, and poor control effect.

Description

Technical Field

[0001] The present invention relates to the technical field of hydraulic pad control, and in particular to a hydraulic pad pre-acceleration control method. Background Art

[0002] Hydraulic cushion technology is primarily used in CNC stamping equipment. Its core function is to cushion the contact between the die and the impact plate through pre-acceleration control, thereby reducing impact pressure, improving workpiece surface quality, and improving die life. Current technology approaches primarily include servo-valve-controlled hydraulic systems and pump-controlled hydraulic systems. Servo-valve-controlled systems utilize high-frequency servo valves to regulate oil flow for precise speed control, while pump-controlled systems utilize direct drive from an electro-hydraulic servo pump, offering higher energy efficiency.

[0003] During the operation of the hydraulic pad, it is first necessary to accurately support the mold and maintain its position while waiting for the press slide to descend at high speed. When the press is working, the hydraulic pad and the slide move in coordination, and the progressive compaction of the sheet material is achieved through closed-loop control to ensure that the sheet material is precisely formed on the mold. At the moment of contact, the relative speed between the two can reach 0.5-1.2m / s. The resulting hydraulic shock may cause the peak pressure of the system to exceed the rated value by more than 30%, causing cumulative damage to key components such as equipment guides and seals. As an effective means of reducing impact, the core of pre-acceleration technology is to accelerate the hydraulic pad in advance before impact through S-curve trajectory planning, thereby reducing the relative collision speed by 40%-60%. This technology can simultaneously improve the surface quality of the workpiece and extend the life of the mold by more than 30%.

[0004] However, current pre-acceleration typically designs a uniform acceleration trajectory based on the states of the pre-acceleration starting and ending points, and employs bidirectional control to achieve a smooth transition from pre-acceleration position control to blank holder force control. This design-then-control approach makes it difficult to quickly track the designed pre-acceleration trajectory if the mold allows for a small pre-acceleration stroke, or under high-frequency operating conditions, making it impossible to achieve the desired relative velocity, ultimately affecting the impact point's cushioning effect. Summary of the Invention

[0005] The embodiment of the present invention provides a hydraulic cushion pre-acceleration control method to solve the problems of difficult pre-acceleration trajectory design, difficulty in adapting to different working conditions, and poor control effect during the hydraulic cushion buffering control process.

[0006] In a first aspect, an embodiment of the present invention provides a hydraulic pad pre-acceleration control method, comprising:

[0007] For each working condition of the hydraulic cushion, the pre-acceleration parameters under that condition are optimized based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under that condition. One working condition corresponds to one working frequency and one drawing stroke. The pre-acceleration parameters include the height of the press slide from the hydraulic cushion at the start of pre-acceleration, the first pulse amplitude of the control instruction for the two-position, two-way proportional valve connected to the upper chamber of the hydraulic cushion, the second pulse amplitude of the control instruction for the two-position, three-way proportional valve connected to the lower chamber of the hydraulic cushion, and the pre-acceleration duration.

[0008] Taking the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameters under each working condition as output, the neural network model is trained to obtain the offline pre-acceleration control model of the hydraulic cushion.

[0009] The working frequency and drawing stroke corresponding to the current working condition are input into the offline pre-acceleration control model to obtain the initial pre-acceleration parameters under the current working condition, and the hydraulic cushion is optimized and controlled online based on the initial pre-acceleration parameters and trust region reflection.

[0010] In a possible implementation, for each working condition of the hydraulic cushion, optimizing the pre-acceleration parameters under the working condition based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under the working condition includes:

[0011] For each working condition of the hydraulic cushion, a set of pre-acceleration parameters under the working condition is randomly generated and recorded as the pre-acceleration parameters to be optimized;

[0012] Controlling the hydraulic pad simulation model to run a working cycle according to the pre-acceleration parameters to be optimized, and obtaining a cycle control result of the hydraulic pad simulation model, wherein the cycle control result includes the hydraulic pad speed, impact force, and pre-acceleration distance when the press slide collides with the hydraulic pad;

[0013] The pre-acceleration parameters to be optimized are optimized with the goal of achieving a comprehensive optimum among the press slide speed, the impact force, and the pre-acceleration distance at which the hydraulic pad speed reaches a preset pre-acceleration ratio, and the optimized pre-acceleration parameters to be optimized are used as new pre-acceleration parameters to be optimized, and the step of "controlling the hydraulic pad simulation model to run a working cycle according to the pre-acceleration parameters to be optimized" and subsequent steps are re-executed;

[0014] Until the hydraulic pad speed reaches the press slide speed of the preset pre-acceleration ratio and the impact force and the pre-acceleration distance are optimal, the current pre-acceleration parameter to be optimized is used as the optimal pre-acceleration parameter under this working condition.

[0015] In one possible implementation, the neural network model is trained using the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameter under each working condition as output to obtain an offline pre-acceleration control model for the hydraulic cushion, including:

[0016] The working frequency and drawing stroke corresponding to each working condition are used as input, and the optimal pre-acceleration parameters under each working condition are used as output to train the neural network model and obtain the fitting degree of the trained neural network model;

[0017] Determining whether the degree of fit meets the criteria;

[0018] If the degree of fit does not meet the standard, the hyperparameters of the trained neural network model are adjusted, and the neural network model after the hyperparameter adjustment is retrained;

[0019] If the degree of fit meets the criteria, the trained neural network model is used as the offline pre-acceleration control model of the hydraulic cushion.

[0020] In a possible implementation, the online optimization control of the hydraulic cushion based on the initial pre-acceleration parameter and the trust region reflection includes:

[0021] Controlling the hydraulic cushion to run a working cycle based on the initial pre-acceleration parameters, and obtaining the current impact force and current pre-acceleration distance corresponding to the hydraulic cushion running for one working cycle;

[0022] According to the current impact force and the current pre-acceleration distance, the initial pre-acceleration parameters are optimized and improved online based on trust region reflection to obtain improved pre-acceleration parameters;

[0023] The operation of the hydraulic cushion is controlled based on the improved pre-acceleration parameter.

[0024] In a possible implementation, performing online optimization and improvement on the initial pre-acceleration parameters based on trust region reflection according to the current impact force and the current pre-acceleration distance to obtain improved pre-acceleration parameters includes:

[0025] An objective function is constructed based on the optimal combination of impact force and pre-acceleration distance, and a finite difference method is used to calculate the gradient of the objective function according to the current impact force and the current pre-acceleration distance;

[0026] Determining whether the norm of the gradient is less than a first threshold;

[0027] If the norm of the gradient is greater than or equal to the first threshold, constructing a scaling matrix according to the upper and lower limits of the pre-acceleration parameter, calculating a Hessian matrix, constructing a trust subdomain problem based on the Hessian matrix and solving it to obtain an optimal solution to the subproblem;

[0028] The initial pre-acceleration parameters are optimized and improved online according to the optimal solution of the sub-problem to obtain improved pre-acceleration parameters.

[0029] In a possible implementation, after obtaining the improved pre-acceleration parameter, the method further includes:

[0030] Determining whether the number of the improved pre-acceleration parameters reaches a preset number;

[0031] If the number of the improved pre-acceleration parameters reaches the preset number, the offline pre-acceleration control model is retrained based on the improved pre-acceleration parameters and the working frequency and drawing stroke corresponding to the improved pre-acceleration parameters to obtain an optimized offline pre-acceleration control model.

[0032] In a second aspect, an embodiment of the present invention provides a hydraulic pad pre-acceleration control device, comprising:

[0033] An optimal parameter acquisition module is used to optimize the pre-acceleration parameters under each working condition of the hydraulic cushion based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under that working condition; wherein one working condition corresponds to one working frequency and one drawing stroke, and the pre-acceleration parameters include the height of the press slide from the hydraulic cushion at the start of pre-acceleration, the first pulse amplitude of the control instruction of the two-position two-way proportional valve connected to the upper chamber of the hydraulic cushion, the second pulse amplitude of the control instruction of the two-position three-way proportional valve connected to the lower chamber of the hydraulic cushion, and the pre-acceleration duration;

[0034] The model training module is used to train the neural network model using the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameters under each working condition as output to obtain the offline pre-acceleration control model of the hydraulic cushion;

[0035] The online optimization module is used to input the working frequency and drawing stroke corresponding to the current working condition into the offline pre-acceleration control model, obtain the initial pre-acceleration parameters under the current working condition, and perform online optimization control on the hydraulic pad based on the initial pre-acceleration parameters and trust region reflection.

[0036] In a third aspect, an embodiment of the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method in the first aspect or any possible implementation of the first aspect is implemented.

[0037] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method in the first aspect or any possible implementation of the first aspect.

[0038] In a fifth aspect, an embodiment of the present invention provides a computer program product, including a computer program, which, when executed by a processor, implements the method in the first aspect or any possible implementation of the first aspect.

[0039] In an embodiment of the present invention, the pre-acceleration parameters for each working condition of the hydraulic cushion are first optimized based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters for the working condition. Then, the working frequency and drawing stroke corresponding to each working condition are used as input and the optimal pre-acceleration parameters for each working condition are used as output to train a neural network model to obtain an offline pre-acceleration control model for the hydraulic cushion. The working frequency and drawing stroke corresponding to the current working condition are then input into the offline pre-acceleration control model to obtain the initial pre-acceleration parameters for the current working condition. Based on the initial pre-acceleration parameters and trust region reflection, the hydraulic cushion is optimized and controlled online, thereby transforming the trajectory planning and closed-loop control problems into optimization problems. Therefore, there is no need to design a pre-acceleration trajectory. Instead, only instructions for the two valves (i.e., pre-acceleration parameters) need to be given according to fixed parameters (i.e., the working frequency and drawing stroke corresponding to the working condition). Furthermore, the system has the ability of online learning and can adapt to different molds and different working conditions. The pre-acceleration effect can be optimized during operation and the impact can be reduced, thereby overcoming the problems of difficult pre-acceleration trajectory design, difficulty in adapting to different working conditions, and poor control effect in the hydraulic cushion buffering control process. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a diagram of an application scenario of the hydraulic pad pre-acceleration control method provided by an embodiment of the present invention;

[0041] Figure 2 This is a flow chart of an implementation of a hydraulic pad pre-acceleration control method provided by an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the pre-acceleration process control provided by an embodiment of the present invention;

[0043] Figure 4 is a schematic diagram of the impact force during the pre-acceleration phase provided by an embodiment of the present invention;

[0044] Figure 5 is a schematic diagram of the motion process in the pre-acceleration stage provided by an embodiment of the present invention;

[0045] Figure 6 1 is a flow chart of obtaining the optimal pre-acceleration parameters under each working condition provided by an embodiment of the present invention;

[0046] Figure 7 is a schematic diagram of a process for training a neural network model provided by an embodiment of the present invention;

[0047] Figure 81 is a flow chart of online optimization control of a hydraulic cushion provided by an embodiment of the present invention;

[0048] Figure 9 2 is a schematic structural diagram of a hydraulic cushion pre-acceleration control device provided by an embodiment of the present invention;

[0049] Figure 10 is a schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] Figure 1 This is an application scenario diagram of the hydraulic pad pre-acceleration control method provided by the embodiment of the present invention. Figure 1 As shown, a significant challenge in controlling the hydraulic pad is the impact contact between the press slide and the module supported by the pad. The pad must withstand the immense impact force and switch from position control to force control within tens of milliseconds, achieving control of the blank holder force. Switching control modes and buffering impact forces during high-inertia, high-rigidity collisions significantly increase the impact velocity as processing frequency (and therefore operating frequency), significantly increasing the difficulty of state transition and buffering.

[0052] The current solution is to use pre-acceleration control to accelerate the hydraulic pad downward in advance so that when the hydraulic pad collides with the slider, it has the same movement direction as the slider and the relative speed of the collision is low. Currently, in the pre-acceleration stage, a pre-acceleration trajectory curve is generated, and then the position control mode is performed according to the position control mode. It is necessary to control the two-position two-way proportional valve connected to the upper chamber of the hydraulic pad and the two-position three-way proportional valve connected to the lower chamber separately. In this process, two quantities need to be controlled simultaneously, and the hydraulic pad has many working conditions. Each working condition will require a different pre-acceleration curve. The mold quality and oil temperature are constantly changing, which will also cause the control parameters to be unable to adapt to the changing working conditions. Therefore, there are problems such as difficulty in designing the pre-acceleration trajectory and poor adaptability to working conditions, resulting in poor pre-acceleration control effect under different working conditions, and the pre-acceleration stroke is still too large. Therefore, a buffering control method that is easier to implement and has stronger adaptability is needed.

[0053] See also Figure 2 , which shows a flow chart of the implementation of the hydraulic pad pre-acceleration control method provided by an embodiment of the present invention, and is described in detail as follows:

[0054] Step 201 : For each working condition of the hydraulic cushion, the pre-acceleration parameters under the working condition are optimized based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under the working condition.

[0055] Among them, one working condition corresponds to a working frequency and a drawing stroke, and the pre-acceleration parameters include the height of the press slider from the hydraulic pad at the start of pre-acceleration, the first pulse amplitude of the control instruction of the two-position two-way proportional valve connected to the upper chamber of the hydraulic pad, the second pulse amplitude of the control instruction of the two-position three-way proportional valve connected to the lower chamber of the hydraulic pad, and the pre-acceleration duration.

[0056] Optionally, for each working condition of the hydraulic cushion, optimizing the pre-acceleration parameters under the working condition based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under the working condition may include:

[0057] For each working condition of the hydraulic cushion, a set of pre-acceleration parameters under the working condition is randomly generated and recorded as the pre-acceleration parameters to be optimized.

[0058] The hydraulic pad simulation model is controlled to run a working cycle according to the pre-acceleration parameters to be optimized, and the cycle control results of the hydraulic pad simulation model are obtained, wherein the cycle control results include the hydraulic pad speed, impact force and pre-acceleration distance when the press slider collides with the hydraulic pad.

[0059] The pre-acceleration parameters to be optimized are optimized with the goal of achieving the optimal comprehensive result of the press slider speed, impact force and pre-acceleration distance when the hydraulic pad speed reaches the preset pre-acceleration ratio. The optimized pre-acceleration parameters to be optimized are used as new pre-acceleration parameters to be optimized, and the steps of "controlling the hydraulic pad simulation model to run a working cycle according to the pre-acceleration parameters to be optimized" and subsequent steps are re-executed.

[0060] Until the hydraulic pad speed reaches the press slide speed of the preset pre-acceleration ratio, and the impact force and pre-acceleration distance are optimal, the current pre-acceleration parameters to be optimized are used as the optimal pre-acceleration parameters under this working condition.

[0061] The method proposed in this embodiment is that during the pre-acceleration process, a pre-acceleration position curve is no longer designed for tracking, but the control instructions of the two-position three-way proportional valve connected to the lower chamber are set. For a fixed opening, set the control command of the two-position two-way proportional valve connected to the upper chamber For a pulse that rises first and then falls (for example Figure 3 The triangular pulse shown, or half-cycle, quarter-cycle sine pulse, or bell-shaped pulse, Gaussian pulse, etc.), as well as the height of the press slide from the hydraulic pad at the time of starting pre-acceleration and pre-acceleration duration ,like Figure 3 As shown, there are 4 variables ( ) needs to be determined to achieve downward acceleration to achieve the purpose of pre-acceleration. In practical applications, different operating conditions will require different parameters. Therefore, the plan is to transform pre-acceleration trajectory tracking control into an optimization problem through a combination of offline pre-optimization and online optimization.

[0062] To determine the four variables, an optimization algorithm was first used based on the hydraulic pad simulation model to generate a list of operating conditions with varying operating frequencies and drawing strokes. For each condition in this list, the four variables were then optimized offline, with the goal of achieving the optimal combination of hydraulic pad speed, press slide speed, and impact force and pre-acceleration distance at the preset pre-acceleration ratio. This yielded a parameter list tailored to the specific condition.

[0063] For example, a hydraulic cushion simulation model can be established based on simulation software such as AMESim or Matlab. The preset pre-acceleration ratio can be 80%, and the objective function corresponding to the optimal comprehensive impact force and pre-acceleration distance can be: , and is the weight coefficient, For impact force, To expect impact, is the pre-acceleration distance, is the expected pre-acceleration distance.

[0064] Among them, the impact force in the pre-acceleration stage can be determined as the peak error of the first overshoot obtained by the force instruction in the simulation results of the hydraulic cushion simulation model, such as Figure 4 The pre-acceleration distance is the distance from the position where the hydraulic pad starts to accelerate to the impact point when the hydraulic pad collides with the press slide, as shown in Figure 5 As shown. For example, combined Figure 6 During offline optimization, the pre-acceleration parameter table can be segmented according to the operating frequency and drawing stroke to form a list of working conditions. Each condition in this list can then be optimized. For example, a working condition can be selected, the hydraulic cushion simulation model can be invoked, and a cycle can be run to obtain the results of the cycle control. The hydraulic cushion speed, the pressure impact (impact force) after pre-acceleration, and the travel distance at the impact point (pre-acceleration distance) can then be comprehensively optimized to achieve an optimal state. This set of parameters can then be recorded and the next working condition can be moved on to the next working condition until all working conditions are cycled and optimized.

[0065] For example, each working condition may be optimized based on a particle swarm algorithm or the like.

[0066] For example, after obtaining the results of cycle control, the hydraulic pad speed and impact force are obtained. and pre-acceleration distance Then, based on the particle swarm algorithm, the control parameters are modified in one direction. After modification, the hydraulic pad speed and impact force are obtained. and pre-acceleration distance ,if If it increases, it means that the optimization direction is wrong and needs to be reversed. If it decreases, it means the direction is correct. Is a constraint that needs to be met , if the pre-acceleration distance Greater than This means that the optimization direction is wrong. The ultimate goal is Cannot be reduced further.

[0067] In step 202 , the working frequency and drawing stroke corresponding to each working condition are used as inputs, and the optimal pre-acceleration parameters under each working condition are used as outputs to train the neural network model and obtain an offline pre-acceleration control model of the hydraulic cushion.

[0068] Optional, such as Figure 7 As shown, the working frequency and drawing stroke corresponding to each working condition are used as input, and the optimal pre-acceleration parameters under each working condition are used as output. The neural network model is trained to obtain the offline pre-acceleration control model of the hydraulic cushion, which may include:

[0069] The working frequency and drawing stroke corresponding to each working condition are used as input, and the optimal pre-acceleration parameters under each working condition are used as output. The neural network model is trained and the fitting degree of the trained neural network model is obtained.

[0070] Determine whether the fit meets the standard.

[0071] If the degree of fit does not meet the standards, the hyperparameters of the trained neural network model are adjusted, and the neural network model with the adjusted hyperparameters is retrained.

[0072] If the fitting degree meets the requirements, the trained neural network model will be used as the offline pre-acceleration control model of the hydraulic cushion.

[0073] In this embodiment, a neural network model is used to fit the parameter list. For example, the input layer of the neural network can be set to have two nodes, corresponding to the working frequency and the drawing stroke, and then 2-3 hidden layers and 4 nodes in the output layer are set, corresponding to the 4 parameters of pre-acceleration. The data fitting flow chart is as follows: Figure 7 shown.

[0074] This embodiment fits the parameter list through a neural network model to generate a fitted neural network, which can make the parameters continuous.

[0075] In step 203 , the working frequency and drawing stroke corresponding to the current working condition are input into the offline pre-acceleration control model to obtain the initial pre-acceleration parameters under the current working condition, and the hydraulic cushion is optimized online based on the initial pre-acceleration parameters and trust region reflection.

[0076] Based on the above steps, this embodiment uses a deployed neural network model (i.e., an offline pre-acceleration control model) to input the working frequency and drawing stroke corresponding to the current working conditions, and outputs four parameters of pre-acceleration. Based on this set of parameters, pre-acceleration control can be achieved.

[0077] Optionally, online optimization control of the hydraulic cushion based on initial pre-acceleration parameters and trust region reflection may include:

[0078] The hydraulic cushion is controlled to run a working cycle based on the initial pre-acceleration parameters, and the current impact force and the current pre-acceleration distance corresponding to the hydraulic cushion running for one working cycle are obtained.

[0079] According to the current impact force and the current pre-acceleration distance, the initial pre-acceleration parameters are optimized and improved online based on trust region reflection to obtain the improved pre-acceleration parameters.

[0080] The hydraulic pad operation is controlled based on the improved pre-acceleration parameters.

[0081] Optionally, based on the current impact force and the current pre-acceleration distance, the initial pre-acceleration parameters are optimized online based on trust region reflection to obtain improved pre-acceleration parameters, which may include:

[0082] The objective function is constructed by comprehensively optimizing the impact force and the pre-acceleration distance, and the gradient of the objective function is calculated using the finite difference method according to the current impact force and the current pre-acceleration distance.

[0083] Determine whether the norm of the gradient is less than a first threshold.

[0084] If the norm of the gradient is greater than or equal to the first threshold, a scaling matrix is ​​constructed according to the upper and lower limits of the pre-acceleration parameter, and the Hessian matrix is ​​calculated. A trust subdomain problem is constructed based on the Hessian matrix and solved to obtain the optimal solution to the subproblem.

[0085] The initial pre-acceleration parameters are optimized and improved online according to the optimal solution of the sub-problem to obtain the improved pre-acceleration parameters.

[0086] In this embodiment, considering that there are certain differences between the actual system and the simulation model, the parameters of offline training may not be optimal. Therefore, during the online operation, the parameters need to be calibrated. The online calibration process can be performed using the Trust Region Reflective (TRR) method. The specific process is as follows: Figure 8As shown in the figure, after a working condition begins, the offline optimized parameters are read from the neural network. After running for one cycle, the optimization process of the TRR method begins. Gradient values ​​are calculated, a scaling matrix is ​​constructed to handle boundary reflections, the trust subdomain problem is solved, and finally, the improved parameters are calculated. If the optimal termination condition has been reached, the current working condition and the optimal value are recorded.

[0087] The following is a detailed description of the key steps in the TRR method online correction process:

[0088] 1. Gradient value calculation

[0089] The finite difference method is used to calculate the gradient of the objective function. (That is, the objective function of the optimal combination of impact force and pre-acceleration distance), gradient for:

[0090] ;

[0091] in, is the small perturbation step size, is a unit vector. If the parameter Close to the boundary, such as 4 variables ( ) feasible scope, adjust Ensure that the perturbation remains within the feasible region.

[0092] 2. Construct a scaling matrix to handle boundary reflections

[0093] Convert the bounded constraints into an unconstrained problem.

[0094] Define the scaling matrix : diagonal matrix, elements ,in For parameters lower and upper limits.

[0095] Reflection transformation: When the parameter approaches the boundary, , ( is the midpoint of the boundary) to transform the problem into the reflection space to avoid crossing the boundary.

[0096] 3. Solve the trust subdomain problem

[0097] Sub-problem form:

[0098] ;

[0099] in, is the trust region radius, is the Hessian matrix, which is the second-order partial derivative of the objective function with respect to each variable, helping to determine the optimization direction and the optimal step size within the trust region boundary.

[0100] 3. Calculate the improved parameters:

[0101] Parameter update:

[0102] ;

[0103] in, To improve the pre-acceleration parameters, is the optimal solution to the subproblem.

[0104] Boundary reflection back mapping: If Beyond the boundary, it is projected back into the feasible region by the following reflection operation:

[0105] ;

[0106] 4. Termination optimization conditions

[0107] The optimization is terminated when one of the following three conditions is met:

[0108] (1) Gradient norm .

[0109] (2) Parameter change .

[0110] (3) The maximum number of iterations is reached.

[0111] Optionally, after obtaining the improved pre-acceleration parameters, the following steps may also be included:

[0112] Determine whether the number of improved pre-acceleration parameters reaches a preset number.

[0113] If the number of improved pre-acceleration parameters reaches a preset number, the offline pre-acceleration control model is retrained based on the improved pre-acceleration parameters and the working frequency and drawing stroke corresponding to the improved pre-acceleration parameters to obtain an optimized offline pre-acceleration control model.

[0114] In this embodiment, based on the above steps, if the optimal termination condition has been reached, the current operating conditions and optimal values ​​are recorded. After accumulating a certain amount of data, these recorded operating conditions and optimal values ​​can be used to further adjust the neural network model (i.e., the offline pre-acceleration control model) to better align it with the actual system. For example, the neural network model's network weights can be updated through backpropagation to minimize the mean squared error between the actual optimal parameters corresponding to the operating conditions and the network-predicted optimal parameters of the neural network model, thereby obtaining an optimized offline pre-acceleration control model.

[0115] Optionally, during the online optimization control of the hydraulic cushion, a safety protection link can be added to avoid abnormal working conditions such as excessive movement distance and excessive speed of the hydraulic cushion caused by inappropriate instructions corresponding to the pre-acceleration parameters.

[0116] The shock suppression control strategy with offline optimization and online correction proposed in this embodiment transforms trajectory planning and closed-loop control into an optimization problem. This strategy achieves excellent shock suppression, simplifies system parameter adjustment, and enables self-learning adaptation to different frequencies and operating conditions. This intelligent and practical method enables online learning optimization for complex operating conditions, eliminating the need for manual adjustment of trajectory parameters for each condition. Furthermore, this shock suppression control strategy fully utilizes the throttling effect of the three-way valve, resulting in rapid acceleration and significantly shortening the stroke (i.e., pre-acceleration distance) compared to traditional pre-acceleration methods, thus expanding the range of molds for which pre-acceleration can be applied.

[0117] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean 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 the present invention.

[0118] The following are device embodiments of the present invention. For details not fully described therein, reference may be made to the corresponding method embodiments described above.

[0119] Figure 9 The structure diagram of the hydraulic pad pre-acceleration control device provided by an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, which are described in detail as follows:

[0120] like Figure 9 As shown, the hydraulic pad pre-acceleration control device includes: an optimal parameter acquisition module 91, a model training module 92 and an online optimization module 93.

[0121] An optimal parameter acquisition module 91 is configured to optimize the pre-acceleration parameters for each working condition of the hydraulic cushion based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters for that working condition. Each working condition corresponds to a working frequency and a drawing stroke, and the pre-acceleration parameters include the height of the press slide from the hydraulic cushion at the start of pre-acceleration, the first pulse amplitude of the control instruction for the two-position, two-way proportional valve connected to the upper chamber of the hydraulic cushion, the second pulse amplitude of the control instruction for the two-position, three-way proportional valve connected to the lower chamber of the hydraulic cushion, and the pre-acceleration duration.

[0122] The model training module 92 is used to train the neural network model using the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameters under each working condition as output to obtain an offline pre-acceleration control model for the hydraulic cushion;

[0123] The online optimization module 93 is used to input the working frequency and drawing stroke corresponding to the current working condition into the offline pre-acceleration control model, obtain the initial pre-acceleration parameters under the current working condition, and perform online optimization control on the hydraulic pad based on the initial pre-acceleration parameters and trust region reflection.

[0124] In one possible implementation, the optimal parameter acquisition module 91 can be used to randomly generate a set of pre-acceleration parameters under each working condition of the hydraulic pad, recorded as pre-acceleration parameters to be optimized; control the hydraulic pad simulation model to run a working cycle according to the pre-acceleration parameters to be optimized, and obtain the periodic control results of the hydraulic pad simulation model, wherein the periodic control results include the hydraulic pad speed, impact force and pre-acceleration distance when the press slider collides with the hydraulic pad; optimize the pre-acceleration parameters to be optimized with the goal of achieving the optimal combination of the press slider speed, the impact force and the pre-acceleration distance when the hydraulic pad speed reaches a preset pre-acceleration ratio, and use the optimized pre-acceleration parameters to be optimized as new pre-acceleration parameters to be optimized, and re-execute the step of "controlling the hydraulic pad simulation model to run a working cycle according to the pre-acceleration parameters to be optimized" and subsequent steps; until the hydraulic pad speed reaches the press slider speed of the preset pre-acceleration ratio, and the impact force and the pre-acceleration distance are optimal, use the current pre-acceleration parameters to be optimized as the optimal pre-acceleration parameters under this working condition.

[0125] In one possible implementation, the model training module 92 can be used to train the neural network model with the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameters under each working condition as output, and obtain the fit of the trained neural network model; determine whether the fit meets the standard; if the fit does not meet the standard, adjust the hyperparameters of the trained neural network model, and re-train the neural network model after the hyperparameter adjustment; if the fit meets the standard, use the trained neural network model as the offline pre-acceleration control model of the hydraulic pad.

[0126] In one possible implementation, the online optimization module 93 can be used to control the hydraulic pad to run a working cycle based on the initial pre-acceleration parameters, and obtain the current impact force and current pre-acceleration distance corresponding to the hydraulic pad after running a working cycle; based on the current impact force and the current pre-acceleration distance, the initial pre-acceleration parameters are online optimized and improved based on trust region reflection to obtain improved pre-acceleration parameters; and the operation of the hydraulic pad is controlled based on the improved pre-acceleration parameters.

[0127] In one possible implementation, the online optimization module 93 can be used to optimally construct an objective function based on the impact force and pre-acceleration distance, and calculate the gradient of the objective function using the finite difference method according to the current impact force and the current pre-acceleration distance; determine whether the norm of the gradient is less than a first threshold; if the norm of the gradient is greater than or equal to the first threshold, construct a scaling matrix according to the upper and lower limits of the pre-acceleration parameters, and calculate the Hessian matrix, construct a trust subdomain problem based on the Hessian matrix and solve it to obtain the optimal solution to the sub-problem; and perform online optimization and improvement on the initial pre-acceleration parameters according to the optimal solution to the sub-problem to obtain improved pre-acceleration parameters.

[0128] In one possible implementation, the model training module 92 can also be used to determine whether the number of the improved pre-acceleration parameters reaches a preset number; if the number of the improved pre-acceleration parameters reaches the preset number, the offline pre-acceleration control model is retrained based on the improved pre-acceleration parameters and the working frequency and drawing stroke corresponding to the improved pre-acceleration parameters to obtain an optimized offline pre-acceleration control model.

[0129] Figure 10 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 10 As shown, the electronic device 10 of this embodiment includes: a processor 100 and a memory 101. The memory 101 stores a computer program 102. When the processor 100 executes the computer program 102, the steps of the above-mentioned method embodiments are implemented. Alternatively, when the processor 100 executes the computer program 102, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0130] For example, the computer program 102 may be divided into one or more modules / units, which are stored in the memory 101 and executed by the processor 100 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program 102 in the electronic device 10.

[0131] The electronic device 10 may include, but is not limited to, a processor 100 and a memory 101. Those skilled in the art will appreciate that Figure 10 This is merely an example of the electronic device 10 and does not constitute a limitation of the electronic device 10. The electronic device 10 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 10 may also include input and output devices, network access devices, buses, etc.

[0132] The processor 100 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0133] The memory 101 may be an internal storage unit of the electronic device 10, such as the hard disk or memory of the electronic device 10. The memory 101 may also be an external storage device of the electronic device 10, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the electronic device 10. Furthermore, the memory 101 may include both an internal storage unit of the electronic device 10 and an external storage device. The memory 101 is used to store the computer program 102 and other programs and data required by the electronic device 10. The memory 101 may also be used to temporarily store data that has been output or is about to be output.

[0134] For the sake of convenience and brevity, the division of the above functional modules / units is only used as an example. In actual applications, the above functions can be assigned to different functional modules / units as needed. The above modules / units can be implemented in the form of hardware, software, or a combination of hardware and software.

[0135] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods in the above-mentioned method embodiments.

[0136] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the methods in the above-mentioned method embodiments.

[0137] The term "computer program" includes computer program code, which may be in source code form, object code form, executable file, or some intermediate form. Computer-readable media may include any entity or device capable of carrying computer program code, recording media, USB flash drives, removable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunications signals, and software distribution media.

[0138] In the above embodiments, the descriptions of each embodiment have their own focus. For parts not described or recorded in detail in one embodiment, please refer to the relevant descriptions of other embodiments. Unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other. The technical features of different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0139] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A hydraulic cushion pre-acceleration control method, characterized in that: include: For each working condition of the hydraulic cushion, the pre-acceleration parameters under that condition are optimized based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under that condition. One working condition corresponds to one working frequency and one drawing stroke. The pre-acceleration parameters include the height of the press slide from the hydraulic cushion at the start of pre-acceleration, the first pulse amplitude of the control instruction for the two-position, two-way proportional valve connected to the upper chamber of the hydraulic cushion, the second pulse amplitude of the control instruction for the two-position, three-way proportional valve connected to the lower chamber of the hydraulic cushion, and the pre-acceleration duration. Taking the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameters under each working condition as output, the neural network model is trained to obtain the offline pre-acceleration control model of the hydraulic cushion. Inputting the working frequency and drawing stroke corresponding to the current working condition into the offline pre-acceleration control model to obtain the initial pre-acceleration parameters under the current working condition, and performing online optimization control on the hydraulic cushion based on the initial pre-acceleration parameters and trust region reflection; For each working condition of the hydraulic cushion, the pre-acceleration parameters under the working condition are optimized based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under the working condition, including: For each working condition of the hydraulic cushion, a set of pre-acceleration parameters under the working condition is randomly generated and recorded as the pre-acceleration parameters to be optimized; Controlling the hydraulic pad simulation model to run a working cycle according to the pre-acceleration parameters to be optimized, and obtaining a cycle control result of the hydraulic pad simulation model, wherein the cycle control result includes the hydraulic pad speed, impact force, and pre-acceleration distance when the press slide collides with the hydraulic pad; The pre-acceleration parameters to be optimized are optimized with the goal of achieving a comprehensive optimum among the press slide speed, the impact force, and the pre-acceleration distance so that the hydraulic cushion speed reaches a preset pre-acceleration ratio, and the optimized pre-acceleration parameters to be optimized are used as new pre-acceleration parameters to be optimized, and the step of "controlling the hydraulic cushion simulation model to run a working cycle according to the pre-acceleration parameters to be optimized" and subsequent steps are re-executed; Until the hydraulic pad speed reaches the press slide speed of the preset pre-acceleration ratio and the impact force and the pre-acceleration distance are optimal, the current pre-acceleration parameter to be optimized is used as the optimal pre-acceleration parameter under this working condition.

2. The hydraulic cushion pre-acceleration control method according to claim 1, characterized in that: The method uses the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameters under each working condition as output to train the neural network model to obtain an offline pre-acceleration control model of the hydraulic cushion, including: The working frequency and drawing stroke corresponding to each working condition are used as input, and the optimal pre-acceleration parameters under each working condition are used as output to train the neural network model and obtain the fitting degree of the trained neural network model; Determining whether the degree of fit meets the criteria; If the degree of fit does not meet the standard, the hyperparameters of the trained neural network model are adjusted, and the neural network model after the hyperparameter adjustment is retrained; If the degree of fit meets the criteria, the trained neural network model is used as the offline pre-acceleration control model of the hydraulic cushion.

3. The hydraulic cushion pre-acceleration control method according to claim 1, characterized in that: The online optimization control of the hydraulic cushion based on the initial pre-acceleration parameters and the trust region reflection includes: Controlling the hydraulic cushion to run a working cycle based on the initial pre-acceleration parameters, and obtaining the current impact force and current pre-acceleration distance corresponding to the hydraulic cushion running for one working cycle; According to the current impact force and the current pre-acceleration distance, the initial pre-acceleration parameters are optimized and improved online based on trust region reflection to obtain improved pre-acceleration parameters; The operation of the hydraulic cushion is controlled based on the improved pre-acceleration parameter.

4. The hydraulic cushion pre-acceleration control method according to claim 3, characterized in that: The online optimization and improvement of the initial pre-acceleration parameters based on trust region reflection according to the current impact force and the current pre-acceleration distance to obtain the improved pre-acceleration parameters includes: An objective function is constructed based on the optimal combination of impact force and pre-acceleration distance, and a finite difference method is used to calculate the gradient of the objective function according to the current impact force and the current pre-acceleration distance; Determining whether the norm of the gradient is less than a first threshold; If the norm of the gradient is greater than or equal to the first threshold, constructing a scaling matrix according to the upper and lower limits of the pre-acceleration parameter, calculating a Hessian matrix, constructing a trust subdomain problem based on the Hessian matrix and solving it to obtain an optimal solution to the subproblem; The initial pre-acceleration parameters are optimized and improved online according to the optimal solution of the sub-problem to obtain improved pre-acceleration parameters.

5. The hydraulic cushion pre-acceleration control method according to claim 3, characterized in that: After obtaining the improved pre-acceleration parameters, it also includes: Determining whether the number of the improved pre-acceleration parameters reaches a preset number; If the number of the improved pre-acceleration parameters reaches the preset number, the offline pre-acceleration control model is retrained based on the improved pre-acceleration parameters and the working frequency and drawing stroke corresponding to the improved pre-acceleration parameters to obtain an optimized offline pre-acceleration control model.

6. A hydraulic cushion pre-acceleration control device, characterized in that: include: An optimal parameter acquisition module is used to optimize the pre-acceleration parameters under each working condition of the hydraulic cushion based on the hydraulic cushion simulation model to obtain the optimal pre-acceleration parameters under that working condition; wherein one working condition corresponds to one working frequency and one drawing stroke, and the pre-acceleration parameters include the height of the press slide from the hydraulic cushion at the start of pre-acceleration, the first pulse amplitude of the control instruction of the two-position two-way proportional valve connected to the upper chamber of the hydraulic cushion, the second pulse amplitude of the control instruction of the two-position three-way proportional valve connected to the lower chamber of the hydraulic cushion, and the pre-acceleration duration; The model training module is used to train the neural network model using the working frequency and drawing stroke corresponding to each working condition as input and the optimal pre-acceleration parameters under each working condition as output to obtain the offline pre-acceleration control model of the hydraulic cushion; an online optimization module, configured to input the working frequency and drawing stroke corresponding to the current working condition into the offline pre-acceleration control model, obtain initial pre-acceleration parameters under the current working condition, and perform online optimization control of the hydraulic cushion based on the initial pre-acceleration parameters and trust region reflection; The optimal parameter acquisition module is specifically used for: For each working condition of the hydraulic cushion, a set of pre-acceleration parameters under the working condition is randomly generated and recorded as the pre-acceleration parameters to be optimized; Controlling the hydraulic pad simulation model to run a working cycle according to the pre-acceleration parameters to be optimized, and obtaining a cycle control result of the hydraulic pad simulation model, wherein the cycle control result includes the hydraulic pad speed, impact force, and pre-acceleration distance when the press slide collides with the hydraulic pad; The pre-acceleration parameters to be optimized are optimized with the goal of achieving a comprehensive optimum among the press slide speed, the impact force, and the pre-acceleration distance so that the hydraulic cushion speed reaches a preset pre-acceleration ratio, and the optimized pre-acceleration parameters to be optimized are used as new pre-acceleration parameters to be optimized, and the step of "controlling the hydraulic cushion simulation model to run a working cycle according to the pre-acceleration parameters to be optimized" and subsequent steps are re-executed; Until the hydraulic pad speed reaches the press slide speed of the preset pre-acceleration ratio and the impact force and the pre-acceleration distance are optimal, the current pre-acceleration parameter to be optimized is used as the optimal pre-acceleration parameter under this working condition.

7. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

9. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 5 when being executed by a processor.

Images