A neural network adaptive super-spiral sliding mode control method for a hydraulic press leveling system

Abstract

The present application relates to a kind of hydraulic machine leveling system neural network adaptive super-helix sliding mode control method, comprising the following steps: the mathematical model of hydraulic machine passive four-corner leveling system considering movable beam model is established;Based on the principle of RBF neural network, the unknown part in the system model is approximated;According to the mathematical model established, the above-mentioned RBF neural network is combined with sliding mode control, and a super-helix sliding mode position closed loop is designed, with the highest cylinder displacement as virtual axis, and four leveling cylinders respectively track virtual axis, to meet the four-cylinder synchronous control requirement;Based on the designed neural network super-helix sliding mode controller, the controller gain is online set;For the designed neural network adaptive super-helix sliding mode controller, based on Lyapunov method, the stability and convergence of the entire closed-loop system are guaranteed.The present application can obtain higher synchronous control precision and good robustness under the condition of large unknown disturbance and model uncertainty during the pressing load process of hydraulic machine.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hydraulic control systems, and particularly relates to a neural network adaptive super-spiral sliding mode control method for a hydraulic press leveling system. BACKGROUND

[0002] Composite parts have outstanding advantages such as light weight, high strength, corrosion resistance and fatigue resistance, and are widely used in key fields such as automobiles, aerospace, high-speed rail and ships. During the molding process of a composite press, the movable beam will be subjected to a sudden unknown disturbance force, and the generated eccentric load moment will cause the movable beam plane to tilt, affecting the pressing precision. The hydraulic passive four-corner leveling system is an important method to solve this problem. The passive four-corner leveling system is a typical electro-hydraulic servo system with strong nonlinearity and many model uncertainties, including parameter uncertainties and uncertain nonlinearities, making it difficult to establish an accurate mathematical model. In addition, the leveling time of the hydraulic press is short, and a faster response speed is required. During the leveling process, the four leveling cylinders need to follow the highest cylinder at all times to meet the synchronous control requirements. Therefore, a new control method is needed to achieve accurate and effective leveling of the leveling system under external disturbance and nonlinear characteristics. Referring to patent CN201710521828.5, a passive torque leveling control method is proposed, which designs a sliding mode control structure, replaces the traditional sign function with a hyperbolic tangent function, and improves the reaching law to obtain the optimal target output leveling force and improve the leveling system precision. For example, referring to patent CN202010448451.7, a PID hydraulic leveling system control method based on an improved sine-cosine algorithm is proposed, which obtains the mathematical model of the system through parameter identification, combines the sine-cosine algorithm with PID to optimize the parameters, and further improves the control performance. Although the above control methods can achieve anti-interference characteristics of the four-corner leveling system and obtain high synchronous control precision, there are still some deficiencies and limitations that need to be further improved.

[0003] The four-corner leveling system of the hydraulic press has parameter uncertainties and uncertain nonlinearities. Parameter uncertainties include changes in the bulk modulus of oil and oil density, and uncertain nonlinearities mainly include unknown external disturbances, unmodeled hydraulic cylinder and valve leakage, and nonlinear friction. These nonlinear characteristics make it difficult to establish an accurate mathematical model that conforms to reality. In addition, the four cylinders are mutually coupled and belong to a multiple-input multiple-output system. Traditional electro-hydraulic servo system control usually uses an approximate linearization method to handle nonlinear characteristics for controller design. As the response speed, control precision and robustness requirements of the leveling system continue to improve, this control strategy has gradually failed to meet the high performance requirements of the system.

[0004] The sliding mode control has good robustness to parameter variation and external disturbance, but the discontinuous switching characteristic of the traditional sliding mode variable structure control can cause chattering of the system, and affect the accuracy of the control. For the leveling system, the synchronization control of the four cylinders makes the chattering of the sliding mode control more severe, and even causes instability of the system. On the other hand, due to the real-time variation of the external load, the fixed control gain cannot meet the synchronization accuracy in the whole leveling stage, and the expected performance requirement cannot be achieved. Therefore, how to weaken the chattering of the sliding mode variable structure and adaptively adjust the gain according to the actual external load is the key to improve the control performance. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a hydraulic press leveling system neural network adaptive super-hyper-spiral sliding mode control method, which can obtain high synchronization control accuracy and good robustness under the condition of large disturbance and model uncertainty in the hydraulic press pressing load process.

[0006] To achieve the above purpose, the present application adopts the following technical scheme: a hydraulic press leveling system neural network adaptive super-hyper-spiral sliding mode control method, comprising the following steps:

[0007] Step S1: under the condition of parameter uncertainty and unknown disturbance, a dynamic equation of the passive four-corner leveling system of the hydraulic press considering the movable beam model is established;

[0008] Step S2: based on the principle of RBF neural network, the approximation of the unknown part in the system model is realized;

[0009] Step S3: according to the established mathematical model, the RBF neural network is combined with the sliding mode control, a super-hyper-spiral sliding mode position closed loop is designed, the highest cylinder displacement is taken as a virtual axis, and the four leveling cylinders respectively track the virtual axis to meet the synchronization control requirement of the four cylinders;

[0010] Step S4: based on the designed neural network super-hyper-spiral sliding mode controller, an adaptive law is designed to online set the controller gain;

[0011] Step S5: for the designed neural network adaptive super-hyper-spiral sliding mode controller, based on the Lyapunov stability theory, the stability and convergence of the whole closed loop system are ensured.

[0012] In a preferred embodiment, step S1 specifically comprises:

[0013] A mathematical model of the movable beam coupled with the four cylinders is established; the hydraulic press is specifically a three-beam four-column structure, the three beams are divided into an upper beam, a lower beam and a movable beam, the upper and lower beams are connected with four supporting columns to form the overall frame of the press; the passive four-corner leveling dynamically levels the four corners of the movable beam;

[0014] There is a movement x along the Z axis z , θ rotation around the X axis x and the rotation θ around the Y axis y ;

[0015] Define the upward movement direction along the Z axis as positive, and the counterclockwise rotation direction around the X axis and the Y axis as positive. According to Newton's second law and the law of rotation of a rigid body about a fixed axis, the dynamic model of the movable beam is as follows:

[0016]

[0017] Where m z is the total mass of the movable beam and the upper mold; x z is the displacement of the movable beam; is the acceleration of the movable beam; F z is the leveling fluid pressure of the driving cylinder on the movable beam; f j F is the leveling force of leveling cylinder No. j; L is the external load on the movable beam; J x 、J y are the moments of inertia around the X-axis and Y-axis respectively; θ x ,θ y are the deflection angles around the X-axis and Y-axis respectively; are the deflection angular accelerations around the X-axis and Y-axis respectively; l x 、l y They are half of the distance between the two leveling cylinders of X axis and Y axis respectively; M x 、M y are the eccentric load moments in the X-axis and Y-axis directions received by the movable beam respectively;

[0018] The relationship between the acceleration of the movable beam and the angular acceleration:

[0019]

[0020] In the four-corner leveling system, the rod chamber pressure is set to a constant value through a pressure reducing valve, and the leveling force output of each cylinder is controlled by adjusting the rodless chamber pressure to achieve the control target;

[0021] A hydraulic model of the four-corner leveling system is established, and the asymmetric cylinder is controlled by a servo proportional valve. The hydraulic model of one leveling cylinder is expressed as:

[0022]

[0023] Where q Aj is the flow rate of the rodless chamber of the leveling cylinder No. j; x pj is the displacement of leveling cylinder No. j; is the speed of leveling cylinder No. j; is the acceleration of the jth leveling cylinder; A and a are the areas of the rodless chamber and the rod chamber of the leveling cylinder, respectively; C ip , C ep are the internal and external leakage coefficients, respectively; p Aj , p aj are the pressures in the rodless chamber and the rod chamber of the jth leveling cylinder, respectively; V0 is the initial volume of the rodless chamber; β e is the effective bulk modulus of the oil; C d is the flow coefficient of the servo proportional valve; W is the area gradient; x vj is the displacement of the valve core; p is the density of the oil; p s is the pump pressure; m p is the mass of the piston rod; b p is the viscous damping coefficient; k p is the load spring stiffness; f Lj is the external load force acting on the jth leveling cylinder;

[0024] The dynamic equations of each leveling cylinder can be obtained from equations (1)-(3):

[0025]

[0026] According to modern control theory, the state variable is selected as the valve opening x v , and the output is the hydraulic cylinder speed y; the 3rd order nonlinear state equation of the leveling system based on control orientation is written as follows:

[0027]

[0028] In the equation,

[0029] f j (·) and g j (·) are nonlinear functions; d j (t) is the disturbance.

[0030] In a preferred embodiment, step S2 specifically comprises:

[0031] The uncertain terms f(·) and g(·) are approximated based on the RBF neural network, and the input and output algorithm of the RBF network is:

[0032]

[0033] wherein x is the network input; i is the i-th network input of the hidden layer of the network; h = [h i ] T is the output of the Gaussian basis function; W * and V * are the approximations of f j(·) and g j (·) of the ideal network weight, ε f and ε g is the network approximation error, |ε f |≤ε Mf , |ε g |≤ε Mg ; take x = [x1 x2 x3] T , then the RBF output is:

[0034]

[0035] where h f (x) and h g (x) are the Gaussian basis functions of the RBF network.

[0036] In a preferred embodiment, step S3 specifically comprises:

[0037] According to the mathematical model of the leveling system, output y j = x 1j , the highest cylinder virtual axis is x max , and the error is defined as Then the sliding mode variable surface is defined as:

[0038]

[0039] where λ is a normal number, and n is the order of the system. Since the valve-controlled cylinder system is 3-order, the sliding mode surface is expressed as:

[0040]

[0041] The sliding mode surface gain is defined as c 1j = 2λ j , c 2j = λ j 2 Then the sliding mode of one leveling cylinder of the leveling system is:

[0042]

[0043] Taking the first-order derivative of the above formula, we get:

[0044]

[0045] The control input of the system is composed of the nominal control law u eqcj (t) and the switching control law u astcj (t), which can be designed as:

[0046] u j (t) = u eqcj (t) - u astcj (t) (12)

[0047] wherein:

[0048]

[0049]

[0050] In a preferred embodiment, step S4 specifically comprises:

[0051] The external disturbance d(t) of the system satisfies the following form:

[0052] |d j (t)|≤δs 1 / 2 (15)

[0053] δ is an unknown positive number;

[0054] α 1j , α 2j can be updated by the following adaptive law:

[0055]

[0056] wherein ω1, κ1, μ1, ε are all arbitrary normal numbers.

[0057] In a preferred embodiment, step S5 specifically comprises

[0058] The Lyapunov function is designed as:

[0059]

[0060] The parameters of the sliding mode surface (10) are selected, the adaptive law of the adaptive super-hyperbolic sliding mode control is (16), and the parameters are designed so that the following formula is established:

[0061]

[0062] Then the designed adaptive super-hyperbolic sliding mode controller can make the closed-loop system asymptotically stable in a finite time, and the convergence time is:

[0063]

[0064] Compared with the prior art, the present application has the following beneficial effects:

[0065] The present application proposes a neural network adaptive super-hyperbolic sliding mode control method for the hydraulic press leveling system, which can realize high-precision leveling when the system has strong nonlinearity, coupling characteristics and is subjected to external unknown disturbance. The significant advantages include:

[0066] Aiming at the strong nonlinearity and four-cylinder coupling characteristics in the leveling system, a radial basis function neural network (RBF neural network) algorithm was designed to effectively solve the nonlinearity in the leveling system, realize the approximation of nonlinear functions in the model, reduce the influence of unmodeled errors and coupling characteristics on synchronization accuracy, and improve the control performance.

[0067] To address the high-frequency chattering problem of traditional sliding mode control, an adaptive super-helical sliding mode controller was designed. This controller uses a high-order sliding mode and applies discontinuous sign terms to the high-order derivatives of the sliding mode, effectively reducing chattering. By using an adaptive law to adjust the control gain online, the control accuracy of the leveling system is improved while maintaining robustness. This provides a new solution for improving the control performance of hydraulic press leveling systems in complex external environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 is a flowchart of a method implementation of an embodiment of the present invention;

[0069] Figure 2 is a system principle diagram of an embodiment of the present invention;

[0070] Figure 3 is a hydraulic principle diagram of an embodiment of the present invention;

[0071] Figure 4 2 is a schematic diagram of a method for a neural network adaptive super-helical sliding mode controller according to an embodiment of the present invention;

[0072] Figure 5 Schematic diagram of synchronization error of the leveling system in an embodiment of the present invention. DETAILED DESCRIPTION

[0073] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0074] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present application belongs.

[0075] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form, and it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0076] like Figure 1-5As shown, this embodiment provides a neural network adaptive super-helical sliding mode control method for a hydraulic press leveling system, comprising the following steps:

[0077] Step S1: In the presence of parameter uncertainty and unknown disturbances, the dynamic equations of the hydraulic press passive four-corner leveling system considering the movable beam model are established;

[0078] Step S2: Based on the principle of RBF neural network, the unknown part of the system model is approximated;

[0079] Step S3: Based on the established mathematical model, the above-mentioned RBF neural network is combined with the sliding mode control to design a super-helical sliding mode position closed loop. The highest cylinder displacement is used as the virtual axis, and the four leveling cylinders track the virtual axis respectively to achieve the four-cylinder synchronous control requirement;

[0080] Step S4: designing an adaptive law based on the designed neural network super-helical sliding mode controller and adjusting the controller gain online;

[0081] Step S5: For the designed neural network adaptive super-helical sliding mode controller, based on Lyapunov stability theory, the stability and convergence of the entire closed-loop system are guaranteed.

[0082] In this embodiment, the 1500t press leveling system is taken as an example, and the parameters are as follows:

[0083]

[0084]

[0085] The following is a further detailed explanation of the relevant contents involved in this method.

[0086] like Figure 1 As shown, in step S1, considering the uncertainty parameters of the leveling system and the influence of complex unknown external disturbances, a hydraulic model of the coupling between the movable beam and the four leveling cylinders is established, and the dynamic model of the whole machine is obtained through the relationship between various variables.

[0087] Figure 2 、 Figure 3 This is the principle diagram of the leveling system in this embodiment. To simplify the dynamic model of the movable beam and ignore factors with minor influence, the following assumptions are made:

[0088] 1) Assuming that the movable beam is a rigid body, its elastic deformation can be ignored;

[0089] 2) Assume that the only freedom of motion of the movable beam is movement along the Z axis. z , θ rotation around the X axis x and the rotation θ around the Y axis y .

[0090] The upward movement along the Z axis is defined as positive, and the rotation around the X axis and the Y axis is defined as counterclockwise. According to Newton's second law and the rigid body rotation law, the dynamic model of the movable beam is as follows:

[0091]

[0092] In the formula, m z is the total mass of the movable beam and the upper mold; x z is the displacement of the movable beam; F z is the leveling hydraulic pressure of the driving cylinder on the movable beam; f j is the leveling force of the jth leveling cylinder; F L is the external load force received by the movable beam; J x , J y are the moments of inertia around the X axis and the Y axis, respectively; θ x , θ y are the deflection angles around the X axis and the Y axis, respectively; l x , l y are half of the distance between the X axis and the Y axis; M x , M y are the deflection moments received by the movable beam in the X axis and Y axis directions, respectively.

[0093] The relationship between the acceleration of the movable beam and the angular acceleration can be approximately considered as:

[0094]

[0095] In the four-corner leveling system, the rod cavity pressure is set to a constant value through a pressure reducing valve, and the leveling force output of each cylinder is controlled by adjusting the non-rod cavity pressure to achieve the control target.

[0096] The hydraulic model of the four-corner leveling system is established, and the asymmetric cylinder is controlled by a servo proportional valve. The hydraulic model of one of the leveling cylinders can be represented as:

[0097]

[0098] In the formula, q Aj is the non-rod cavity flow of the jth leveling cylinder; x pj is the displacement of the jth leveling cylinder; A and a are the areas of the non-rod cavity and the rod cavity of the leveling cylinder, respectively; C ip , C ep are the internal and external leakage coefficients, respectively; p Aj , p aj are the non-rod cavity and rod cavity pressures of the jth leveling cylinder, respectively; V0 is the initial volume of the non-rod cavity; β e is the effective bulk modulus of the oil; C d is the flow coefficient of the servo proportional valve; W is the area gradient; xvj is the displacement of the spool; p is the oil density; p s is the pump pressure; m p is the mass of the piston rod; b p is the viscous damping coefficient; k p is the load spring stiffness; f Lj is the external load force on the jth leveling cylinder.

[0099] The dynamic equations of each leveling cylinder can be obtained from equations (1)-(3):

[0100]

[0101] According to modern control theory, the state variable is selected as the valve opening x v , and the output is the hydraulic cylinder speed y. The 3rd order nonlinear state equation of the leveling system based on control orientation can be written as follows:

[0102]

[0103] where, f j (·) and g j (·) are nonlinear functions; d j (t) is the disturbance.

[0104] In step S2, the unknown part of the system model is approximated based on the RBF neural network.

[0105] The leveling system is a typical electro-hydraulic servo system, which has parameter uncertainty and uncertain nonlinearity, such as pressure-flow nonlinearity of the servo valve, friction nonlinearity, and unknown leakage parameters, so the above model has uncertain nonlinear functions. In order to ensure the control effect, real-time dynamic approximation is needed, so the RBF neural network is used to approximate the uncertain terms f(·) and g(·). The input and output algorithm of the RBF network is:

[0106]

[0107] where x is the network input; i is the i th network input of the network hidden layer; h = [h i ] T is the output of the Gaussian basis function; W * and V * are the ideal network weights for approximating f j (·) and g j (·), respectively, and ε f and ε g are the network approximation errors, |ε f |≤εMf ,|ε g |≤ε Mg .

[0108] Take x = [x1 x2 x3] T , then the RBF output is:

[0109]

[0110] Among them, h f (x) and h g (x) is the Gaussian basis function of the RBF network.

[0111] like Figure 4 As shown, in step S3, according to the established mathematical model, the above-mentioned RBF neural network is combined with the sliding mode control to design a super-helical sliding mode position closed loop. The highest cylinder displacement is used as the virtual axis, and the four leveling cylinders track the virtual axis respectively to achieve the four-cylinder synchronous control requirements.

[0112] The super-helical sliding mode control obtains continuous system control input through integration, thus avoiding the chattering problem of the traditional sliding mode controller. At the same time, it can cope with the influence of system uncertainty and unknown disturbances and has good robustness.

[0113] According to the mathematical model of the leveling system, the output y j =x 1j , the highest cylinder virtual axis is x max , define the error as Then the sliding mode variable surface is defined as:

[0114]

[0115] Where λ is a positive constant and n is the system order. Since the valve-controlled cylinder system is of order 3, the sliding surface can be expressed as:

[0116]

[0117] Define the sliding surface gain as c 1j =2λ j , c 2j =λ j 2 , then the sliding mode of a leveling cylinder in the leveling system is:

[0118]

[0119] Taking the first-order derivative of the above formula, we can get:

[0120]

[0121] The control input of the system is given by the nominal control law u eqcj(t) and switching control law u astcj (t) can be designed as:

[0122] u j (t) = u eqcj (t) - u astcj (t) (12)

[0123] wherein:

[0124]

[0125]

[0126] In step S4, the adaptive law is designed based on the designed neural network hyper-spiral sliding mode controller to continuously adjust the controller gain.

[0127] Suppose that the external disturbance d(t) of the system satisfies the following form:

[0128] |d j (t)|≤δs 1 / 2 (15)

[0129] δ is an unknown positive number.

[0130] α 1j 、α 2j can be updated by the following adaptive law:

[0131]

[0132] wherein ω1, κ1, μ1, ε are all arbitrary normal numbers.

[0133] Step S5: For the designed neural network adaptive hyper-spiral sliding mode controller, the stability and convergence of the entire closed-loop system are ensured based on the Lyapunov method.

[0134] Based on the leveling system model established above, the neural network adaptive hyper-spiral sliding mode controller designed by formula (12) is used, and the adaptive law is updated by (16), then for any initial condition, there exists a finite time to make the sliding mode surface approach to any small range near zero. The detailed proof is given as follows:

[0135] Substitute (12) into (11) to obtain:

[0136]

[0137] Define a new state variable: j = [Γ 1j Γ 2j ] T = [|sj | 1 / 2 sgn(s j ) j ] T ,but:

[0138]

[0139] Because |d j (t)|≤δs 1 / 2 , so d(t) must have the following form:

[0140] d j (t) = γ | s j | 1 / 2 sgn(s j ) (19)

[0141] Combining (18) and (19) we can get:

[0142]

[0143] Define the Lyapunov function:

[0144]

[0145] Z is a symmetric positive definite matrix defined as λ, κ1, κ2, α 10 , α 20 Is a positive constant, the above conditions can ensure that V is positive. Find the derivative of V0 with respect to time t:

[0146]

[0147] Where:

[0148]

[0149] Among them, Q 11 =-2(γ-α1)(λ+4ε 2 )-4εα2,Q 12 =Q 21 =2ε(γ-α1)+α2-λ-4ε 2 , Q 22 =4ε. When α1 satisfies the following conditions, the minimum eigenvalue of the matrix Q is λ min (Q)≥2ε:

[0150]

[0151] For formula (22), the following inequality holds:

[0152]

[0153] By generalizing the Rayleigh-Ritz theorem, we know that:

[0154]

[0155] Combining (25)-(26) we can get:

[0156]

[0157] Find the time derivative of V in (21):

[0158]

[0159] According to the inequality (a 2 +b 2 +c 2 ) 1 / 2 ≤|a|+|b|+|c|, so formula (28) can be rewritten as:

[0160]

[0161] In the formula, ξ=min{ζ,ω1,ω2}, α1 and α2 are bounded, so there must be a constant α 10 , α 20 , so that α1-α 10 <0,α2-α 20 <0 holds true, so formula (29) is transformed into:

[0162]

[0163] When|s j When |>μ, make but Substituting into (30) we get:

[0164]

[0165] Formula (31) shows that when |s j When |>μ, the control gains α1 and α2 will gradually increase until they satisfy Equation (24), so that Q satisfies the positive definite condition, ensuring that the sliding surface can converge to the vicinity of zero in a finite time and achieve asymptotic stability of the system. j |≤μ, the control gains α1 and α2 decrease until |s is satisfied again. j |>μ. Integrating both sides of equation (31), we can get the convergence time as:

[0166]

[0167] Figure 5The figure is a four-cylinder synchronization error schematic diagram of a hydraulic machine leveling system in the embodiment. According to the neural network self-adaptive super-helix sliding mode control method, precise leveling can be realized under a complex external environment.

Claims

1. A neural network adaptive super-helical sliding mode control method for a hydraulic press leveling system, characterized in that: The following steps are involved: Step S1: In the presence of parameter uncertainty and unknown disturbances, the dynamic equations of the hydraulic press passive four-corner leveling system considering the movable beam model are established; Step S2: Based on the principle of RBF neural network, the unknown part of the system model is approximated; Step S3: Based on the established mathematical model, the above-mentioned RBF neural network is combined with the sliding mode control to design a super-helical sliding mode position closed loop. The highest cylinder displacement is used as the virtual axis, and the four leveling cylinders track the virtual axis respectively to achieve the four-cylinder synchronous control requirement; Step S4: designing an adaptive law based on the designed neural network super-helical sliding mode controller and adjusting the controller gain online; Step S5: for the designed neural network adaptive super-helical sliding mode controller, based on Lyapunov stability theory, ensure the stability and convergence of the entire closed-loop system; Step S1 specifically includes: A mathematical model for the coupling of a movable beam and four cylinders was established. The hydraulic press has a three-beam, four-column structure. The three beams are divided into an upper beam, a lower beam, and a movable beam. The upper and lower beams are connected to four supporting columns to form the overall frame of the press. Passive four-corner leveling dynamically adjusts the four corners of the movable beam. 1) There is movement along the Z axis , rotating around the X axis and rotation around the Y axis ; Define the upward movement direction along the Z axis as positive, and the counterclockwise rotation direction around the X axis and the Y axis as positive. According to Newton's second law and the law of rotation of a rigid body about a fixed axis, the dynamic model of the movable beam is as follows: , where is the total mass of the movable beam and the upper mold; is the displacement of the movable beam; is the acceleration of the movable beam; It is the leveling hydraulic pressure of the driving cylinder on the movable beam; for Leveling force of leveling cylinder; is the external load force on the movable beam; 、 are the moments of inertia around the X-axis and Y-axis respectively; 、 are the deflection angles around the X-axis and Y-axis respectively; 、 are the deflection angular accelerations around the X-axis and Y-axis respectively; 、 They are half of the distance between the two leveling cylinders of X-axis and Y-axis respectively; 、 are the eccentric load moments in the X-axis and Y-axis directions received by the movable beam respectively; The relationship between the acceleration of the movable beam and the angular acceleration: In the four-corner leveling system, the rod chamber pressure is set to a constant value through a pressure reducing valve, and the leveling force output of each cylinder is controlled by adjusting the rodless chamber pressure to achieve the control target; A hydraulic model of the four-corner leveling system is established, and the asymmetric cylinder is controlled by a servo proportional valve. The hydraulic model of one leveling cylinder is expressed as: , where is the flow rate of the rodless chamber of the leveling cylinder No. J; is the displacement of leveling cylinder No. j; is the speed of leveling cylinder No. j; is the acceleration of leveling cylinder No. j; A and a are the area of ​​the rodless cavity and the rod cavity of the leveling cylinder respectively; 、 are the internal and external leakage coefficients, 、 They are the pressures in the rodless chamber and the rod chamber of the leveling cylinder No. J respectively; is the initial volume of the rodless cavity; is the effective bulk elastic modulus of the oil; is the flow coefficient of the servo proportional valve; W is the area gradient; is the valve core displacement; is the oil density; is the pump source pressure; is the mass of the piston rod; is the viscous damping coefficient; is the load spring stiffness; is the external load force on the leveling cylinder No. J; By - The dynamic equations of each leveling cylinder can be obtained: , according to modern control theory, select the state variable , the input is the valve opening , the output is the hydraulic cylinder velocity y; the third-order control-oriented nonlinear state equation of the leveling system is written as follows: , where ; ; and They are all nonlinear functions; For interference.

2. The neural network adaptive super-helical sliding mode control method for a hydraulic press leveling system according to claim 1 is characterized in that: Step S2 specifically includes: Approximation of Uncertain Term Based on RBF Neural Network and , the RBF network input and output algorithm is: ,in, Input for the network; The hidden layer of the network Network input; is the output of the Gaussian basis function; and Approximation and The ideal network weights are and is the network approximation error, , ;Pick , then the RBF output is: ,in, and is the Gaussian basis function of the RBF network.

3. The neural network adaptive super-helical sliding mode control method for a hydraulic press leveling system according to claim 1 is characterized in that: Step S3 specifically includes: According to the mathematical model of the leveling system, the output , the highest cylinder virtual axis is , define the error as , then the sliding mode variable surface is defined as: ,in, is a positive constant, n is the system order, because the valve-controlled cylinder system is 3rd order, the sliding surface is expressed as: , define the sliding surface gain as , , then the sliding mode of a leveling cylinder in the leveling system is: , taking the first-order derivative of the above formula we can get: , the control input of the system is given by the nominal control law and switching control law Composition, can be designed as: ,in: ， 。 4. The neural network adaptive super-helical sliding mode control method for a hydraulic press leveling system according to claim 3 is characterized in that: Step S4 specifically includes: External interference to the system Satisfy the following form: , is an unknown positive number; 、 It can be updated by the following adaptive law: ,in, 、 、 、 、 are all arbitrary positive numbers.

5. The neural network adaptive super-helical sliding mode control method for a hydraulic press leveling system according to claim 4 is characterized in that: Step S5 specifically includes: Design the Lyapunov function as: , select the sliding surface Parameters, ensuring that the adaptive law of adaptive super-helical sliding mode control is , the design parameters make the following equation hold: , then the designed adaptive super-helical sliding mode controller can make the closed-loop system asymptotically stable within a finite time, and the convergence time is: 。

Images