CN108227676A - The online fault detect of valve-controlled cylinder electrohydraulic servo system, estimation and localization method - Google Patents

Abstract

The invention discloses a kind of online fault detect of valve-controlled cylinder electrohydraulic servo system, estimation and localization methods, include the following steps：1) mission nonlinear models；2) system normal parameter recognizes；3) data during system normal operation that will be obtained in step 2) bring the nonlinear state equation of system in step 1) into, obtain the expression formula of normal system parameter matrix and the nonlinear state equation of system；4) the correlated inputs output parameter of real-time acquisition system；5) fault detection and identification under varying load；6) robust Fault decision；If 7) fault-free in step 6), return to step 4)；If faulty, starting step 5) fault vectors estimation module in fault-detecting-observer, 8) Fault Isolation positioning；The beneficial effects of the invention are as follows：System running state can in real time be monitored, reduce economic loss in valve-controlled cylinder system there are being loaded outside unknown time-varying, in the case of non-linear and uncertain parameter.

Description

Online Fault Detection, Estimation and Location Method of Electro-hydraulic Servo System of Valve-controlled Cylinder

technical field

The invention relates to the field of fault diagnosis of an electro-hydraulic servo system, in particular to an online fault detection, estimation and positioning method of a valve-controlled cylinder electro-hydraulic servo system.

Background technique

The electro-hydraulic servo system is a feedback control system composed of an electric signal processing device and a hydraulic power mechanism. It is widely used in industrial Manufacturing, transportation, engineering machinery and other fields. Among them, the valve-controlled cylinder electro-hydraulic servo system is a common and widely used system. Its working principle is mainly to control the oil in and out of the hydraulic cylinder through a servo valve or a proportional valve, thereby controlling the position and speed of the cylinder piston, etc., such as The valve-controlled cylinder electro-hydraulic servo system is used for the control of aircraft and ship steering gear, the position control of the mechanical arm, and the thickness control of the strip mill. However, due to many unavoidable factors such as changes in the working environment, component wear and aging, etc., any system will inevitably fail. The electro-hydraulic servo system itself is a complex mechanical, electrical, and hydraulic integrated system, and the faults in it are often diverse and hidden. , coupling, and complex causality. With the development of modern manufacturing technology, control technology and automation technology, the electro-hydraulic servo system is developing in the direction of light weight, small size, high pressure and variable pressure. While the system scale, function, complexity and automation level are improving, the The status monitoring and fault diagnosis of the system also put forward higher requirements, and people are eager to improve the reliability and safety of the system.

At present, the fault diagnosis methods of electro-hydraulic servo system mainly include methods based on analytical model, methods based on knowledge and methods based on signal analysis. The method of signal analysis is mainly used for fault analysis and diagnosis of single hydraulic components, including hydraulic pumps, valves and oil cylinders; Knowledge-based methods mainly include neural networks, expert systems, etc. This method does not need to establish a system model and belongs to the field of intelligent diagnosis, but it needs to use typical data to train neural networks in advance or use expert experience to build knowledge bases; methods based on analytical models do not require Obtain a large amount of historical data in advance, but it is necessary to establish a system model, and there are certain requirements for the accuracy of the model. The above methods have their own advantages and disadvantages. In recent years, many fault methods based on model and data-driven fusion have emerged, but they are mainly aimed at the situation where the external load is constant and the load is known or measurable. Based on the complexity of the actual working conditions, the operating parameters of the system and the external load force are often not constant but time-varying, and it is difficult to obtain accurately. Under this unknown time-varying load, even under normal conditions, the output pressure of the system, The flow rate is not a fixed value, but fluctuates up and down, sometimes with a large fluctuation range, so the fault form of the system is easily confused with the normal form, and it is difficult to distinguish. In addition, there are many inherent nonlinear factors in the valve-controlled cylinder electro-hydraulic servo system, such as the nonlinear relationship between pressure and flow. challenge.

Contents of the invention

Aiming at the problems existing in the prior art, the invention provides an online fault detection, estimation and positioning method of a valve-controlled cylinder electro-hydraulic servo system with strong robustness and accurate detection.

Technical scheme of the present invention is as follows:

An online fault detection, estimation and location method for a valve-controlled cylinder electro-hydraulic servo system. The inventive method is based on a valve-controlled cylinder electro-hydraulic servo system. The valve-controlled cylinder electro-hydraulic servo system includes a hydraulic cylinder, a servo valve/proportional valve connected to the hydraulic cylinder, and The hydraulic pump connected to the servo valve/proportional valve, the piston of the hydraulic cylinder is equipped with a sensor, the sensor is connected to a controller through a circuit, and the controller is connected to the servo valve/proportional valve circuit;

It is characterized in that, comprising the steps of:

1) System nonlinear modeling: establish the mathematical model of valve-controlled cylinder electro-hydraulic servo system; select the system state variables; establish the nonlinear state equation of the system;

2) Normal parameter identification of the system: Obtain historical input and output data under different working conditions and environmental conditions during normal operation of the system, input the identification model, and obtain the parameters and their variation range during normal operation of the system from the identification model;

3) Bring the data obtained in step 2) into the nonlinear state equation of the system in step 1) to obtain the normal system parameter matrix and the expression of the nonlinear state equation of the system;

4) Real-time acquisition of relevant input and output parameters of the system;

5) Fault detection and estimation under variable load: establish a fault detection observer, which includes an external load force decoupling module, a fault dominant module, a nonlinear module, a stability module and a fault vector estimation module, and The collected signals of each module are input to the observer to obtain the estimated output of the observer, and the output residual is obtained by making a difference with the actual system output;

6) Robust fault decision-making: input the output residual error obtained in step 3) into the calculation error estimation function, and combine the adaptive threshold calculated in real time to diagnose whether there is a fault in the system;

7) If there is no fault in step 6), return to step 4); if there is a fault, start the fault vector estimation module in step 5) of the fault detection observer, and obtain the output residual value from the observer and the actual system for network weighting The real-time adjustment of the value coefficient finally obtains the estimated value of the fault vector;

8) Fault isolation and location: Combine the estimated value of the fault vector in step 7) with the movement direction of the hydraulic cylinder piston to judge the type and location of the possible fault, and output the diagnosis result.

The online fault detection, estimation and location method of the valve-controlled cylinder electro-hydraulic servo system is characterized in that, in the step 1), the mathematical model of the valve-controlled cylinder electro-hydraulic servo system is established: first, the valve-controlled cylinder electro-hydraulic servo system Each link is equivalent, and the flow equation of the servo valve/proportional valve, the dynamic equation of the servo valve/proportional valve, the flow continuity equation of the hydraulic cylinder, and the kinetic force balance equation of the hydraulic cylinder piston are established, and the overall system is expressed by the above four equations The mathematical model is as follows:

Flow equation for servo valve/proportional valve:

In the formula, q _± , p _± are the flow rate and pressure of the two chambers of the hydraulic cylinder respectively, p _s+ , p _s- are the oil supply and oil return pressure respectively, x _v and α are the deviation of the servo valve/proportional valve spool from the neutral position respectively Displacement and spool positive coverage, _kc is the flow coefficient;

Servo valve/proportional valve dynamic equation:

In the formula, k _v and τ are the gain and time coefficients describing the dynamic characteristics of the servo valve/proportional valve, and u is the input voltage;

Flow continuity equation of hydraulic cylinder:

a _± and V _± are the area and effective volume of the two chambers of the hydraulic cylinder respectively, c _i and c _e are the internal and external leakage coefficients of the hydraulic cylinder respectively, x _p is the piston displacement of the hydraulic cylinder, β _e is the effective bulk modulus, p ₊ and p _- is the pressure of the left and right chambers of the hydraulic cylinder;

The kinetic force balance equation of hydraulic cylinder piston:

In the formula, m is the total mass converted to the load, b _p is the viscous damping coefficient, f is the external load force acting on the piston, d is the variable load disturbance force, a ₊ and a _- are the areas of the two chambers of the hydraulic cylinder;

Select the system state variable: select the hydraulic cylinder piston movement speed The pressures p ₊ and p _- of the left and right chambers of the hydraulic cylinder and the displacement x _v of the spool of the servo valve/proportional valve are the state variables of the system, and the system state variables are defined as

Establish the nonlinear state equation of the system: regard the voltage input to the servo valve/proportional valve as the input u, and define the piston movement speed of the hydraulic cylinder The pressure p ₊ and p _- of the left and right chambers of the hydraulic cylinder are the output vectors The flow equation of the servo valve/proportional valve, the dynamic equation of the servo valve/proportional valve, the flow continuity equation of the hydraulic cylinder, and the kinetic force balance equation of the hydraulic cylinder piston are transformed into four equations, and the state variables x and non The linear term g(x) is converted into the following state equation form, and the matrices A, B, C, and D are the parameter matrices of the system;

The online fault detection, estimation and positioning method of the valve-controlled cylinder electro-hydraulic servo system is characterized in that, in the step 2), the input and output historical data under different working conditions and environmental conditions are obtained during the normal operation of the system: given continuous The variable servo valve/proportional valve voltage input signal collects the displacement value x _p of the hydraulic cylinder piston, and the pressure values p ₊ and p _- of the two chambers of the hydraulic cylinder;

Identification model: From the nonlinear state equation of the system, a linear equation system containing identification parameters is obtained:

Φθ=Γ,

θ=[β _e ^-1 c _i c _e k _q ] ^T is the parameter to be identified, k _q =k _c k _v ,

is a parameter matrix containing system state quantities, where k=1,2,3...;

Substitute the collected data into the above equations, and use the least squares method to identify the unknown parameter θ in the above equations; bring the identified system parameters into the nonlinear state equation of the system, and compare the obtained output with the actual system output After comparison and correction, the parameter values and variation ranges of the system in normal operation are finally obtained.

The online fault detection, estimation and positioning method of a valve-controlled cylinder electro-hydraulic servo system under variable load is characterized in that the step 5) decoupling module of the external load force: mainly realizes that the change of the external interference force will not affect the user The function that affects the residual error of fault judgment, the realization method: set the parameter matrix T of the observer =-D(CD) ⁺ +Y[I-(CD)(CD) ⁺ ], where (CD) ⁺ =( (CD) ^T (CD)) ^-1 (CD) ^T , Y is the parameter matrix to be set, obtained in the stability module, C, D are the system parameter matrix obtained in step 3), and I is the identity matrix;

Fault dominant module: mainly realizes the function that once a fault occurs in the system, it will affect the residual used for fault judgment. The realization method: set the parameter matrix M=I+TC, and MF _f ≠ 0, where F _f is the fault Matrix, expressed as F _f =[e ₁ e ₂ e ₃ ],

Nonlinear module: obtain the nonlinear expression g(x) of the original system to observe the state quantity Replace the original state quantity x, without linear simplification, directly used for stability module processing;

Stability module: mainly realizes the rapid and effective convergence of the observer, and the realization method is as follows:

a: Obtain a positive definite symmetric matrix P>0 by solving the following linear matrix inequality, and two matrices where P ₁ =P(I+T _a C), T _a ＝-D(CD)+, T _b ＝I-(CD)(CD) ⁺ , γ is a constant:

b: calculate the matrix Y and K, and

c: Get the rest of the observer parameter matrix, G=MB, N=MA-KC, L=K-NT;

Residual error generation module: the output signal of the actual system (including the hydraulic cylinder piston speed signal value And the output vector composed of the pressure signal values p ₊ and p _- of the two chambers of the hydraulic cylinder and the observer output vector Subtract to get the output residual vector

Fault vector estimation module: The fault vector estimation module is composed of a neural network, which does not work when the residual is less than or equal to the threshold, and is activated only when the residual is greater than the threshold to perform fault estimation.

Above-mentioned matrix T, M, G, N and L are the system parameter matrix to be designed, and matrix A, B, C and D are the system parameter matrix obtained in step 3);

The online fault detection, estimation and positioning method of the valve-controlled cylinder electro-hydraulic servo system is characterized in that the calculation error estimation function in the step 6) is: J(t)= ^rT (t)Hr(t), where H is a weighted diagonal function;

Fault detection is performed according to the following rules:

In the formula, λ(t) is the adaptive threshold, which is composed of two parts, one is the steady state threshold, and the other is the transient threshold, which is switched by the acceleration value of the hydraulic cylinder piston. When the threshold is set, the system Taking into account the normal fluctuation of parameters, the threshold value is calculated online based on statistical methods. In addition, the amount of data calculation and storage space is reduced by segmenting and weighting the residual sequence, while ensuring a certain degree of robustness. The specific implementation process:

a: Substitute the parameter variation range obtained by the identification model in step 2) into the fault detection observer to obtain the fluctuation range of the steady-state threshold, and take the upper and lower limits of the fluctuation as the steady-state threshold λ _o- , λ _o+ ;

b: Differentiate the obtained piston speed value or position value twice to obtain the piston motion acceleration value a _P (t);

c: Obtain a sequence pair of residual r _i (k) and piston acceleration a _i (k) with sample length S respectively, where i=(k-1)S/l+1,...,(k -1) S/l+S, calculate the piston mean value a _p (k) of this S data acceleration sequence;

d: Divide the residual sequence r _i (k) of S data length into l parts, and calculate its mean value μ _p (k) for each part of the residual data whose length is S/l, and then calculate the weighted value according to the following formula mean μ _r (k) and variance σ _r ² (k);

where w _p is the assigned weight coefficient, and

e: The adaptive threshold λ(k) is determined by the following formula;

Where ε is the bandwidth coefficient, c _a is the critical value of acceleration;

f: Repeat the above steps to calculate the threshold λ in real time.

The online fault detection, estimation and positioning method of the valve-controlled cylinder electro-hydraulic servo system is characterized in that the implementation process of the fault vector estimation module in the step 7) is specifically:

Set up a reasonable neural network structure, the number of network input nodes, hidden layer nodes and output nodes are respectively n+m, s, n, where n, m are the state quantities and input quantities of the fault diagnosis observer, and the number of hidden layers is s Determined offline, the input of the neural network is the estimated value of the system state quantity and the input voltage of the servo valve/proportional valve, and the neural network is established In the formula, W is the output weight coefficient matrix of the neural network to be designed, is the neural network basis function;

Obtaining State Estimators of Fault Diagnosis Observers in Real Time And the output residual vector r _n (t), the real-time adjustment of the neural network output weight coefficient is carried out according to the following formula In the formula, the parameter matrix η is a normal number, P is a positive definite matrix, and u is the input voltage of the servo valve/proportional valve;

Record the output value of the neural network at this time, as the estimated value of the current fault vector f(t)=f _nn (t);

Get the observer estimated output vector after starting the neural network fault vector estimation module and get the output residuals Where y(t) is the actual output vector;

Detect output residual r _n (t), if r _n (t)>λ _n (t) return to step 6), if r _n (t)≤λ _n (t) enter the next step, where λ _n (t) Decision Adaptive Thresholds for Fault Estimation;

The output fault vector estimate is f(t)=[f ₁ f ₂ f ₃ ] ^T .

The online fault detection, estimation and location method of the valve-controlled cylinder electro-hydraulic servo system is characterized in that the fault type and location in the step 8) are judged as follows:

The fault location method based on the fault estimation vector f(t)=[f ₁ f ₂ f ₃ ] ^T is as follows:

, if |f| ₁ ＞δ ₁ , |f ₂ |<δ ₂ , |f ₃ |<δ ₃ or, When |f ₁ |<δ ₁ , |f ₂ |>δ ₂ , |f ₃ |<δ ₃ , it is judged that the oil supply pressure of the system is abnormal, check the hydraulic pump and pump outlet relief valve;

, if |f ₁ |<δ ₁ , |f ₂ |>δ ₂ , |f ₃ |<δ ₃ or, When |f ₁ |>δ ₁ , |f ₂ |<δ ₂ , |f ₃ |<δ ₃ , it is judged that the oil return pressure of the system is abnormal, check whether there is blockage in the circuit pipeline;

If |f ₁ |＞δ ₁ , |f ₂ |＞δ ₂ , |f ₃ |<δ ₃ , it is judged as hydraulic cylinder leakage failure, check the hydraulic cylinder;

If |f ₁ |<δ ₁ , |f ₂ |<δ ₂ , |f ₃ |>δ ₃ , it is judged that the servo valve/proportional valve is faulty, check the valve;

Among them, each fault judgment value δ ₁ , δ ₂ , δ ₃ is a constant.

The beneficial effects of the present invention are:

1) This method can monitor the operating status of the system in real time when there are unknown time-varying external loads, nonlinear and uncertain parameters in the valve-controlled cylinder system, and can effectively judge whether the system is faulty or not, and can further analyze Estimate the location and size of the fault, which is more conducive to the acquisition of fault information, rapid analysis and processing of faults, and reduces economic losses;

2) It avoids the space and cost problems caused by the need to add additional sensors to measure the external load force in the previous method, and also avoids the inaccurate measurement and estimation of the external load force, which affects the accuracy of fault diagnosis, and also realizes system failure. The integrated process of online detection, positioning and estimation, and the establishment of an online fault diagnosis system for valve-controlled cylinder electro-hydraulic servo systems are more conducive to early diagnosis of system faults;

3) This method does not need to install additional sensors to measure the magnitude of the time-varying external load force, nor does it need to be estimated in advance, but uses mathematical analysis to decouple the time-varying external load, saving costs and avoiding installation Trouble, improve the accuracy of diagnosis.

4) This method can reduce the amount of sample training, and does not need to build multiple observers for fault detection and location estimation. Only one fault diagnosis observer can realize fault detection, occurrence location, category judgment and size estimation , and is not affected by time-varying external loads, has strong robustness, small amount of online calculation, and is very convenient to use.

Description of drawings

Fig. 1 is the structural representation of valve-controlled cylinder electro-hydraulic servo system of the present invention;

Fig. 2 is a fault detection observer work flow chart of the present invention;

Fig. 3 is the overall flowchart of the present invention;

Fig. 4 is a construction diagram of the fault detection observer of the present invention;

In the figure: 1-hydraulic pump, 2-servo valve/proportional valve, 3-hydraulic cylinder, 4-piston, 5-sensor, 6-controller.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings of the description.

As shown in Figure 1-4, the valve-controlled cylinder electro-hydraulic servo system includes a hydraulic pump 1, a servo valve/proportional valve 2, a hydraulic cylinder 3 (also a double-rod cylinder), a piston 4, a sensor 5 and a controller 6. The servo valve/proportional valve 2 is connected to the hydraulic cylinder 3, and the hydraulic pump 1 is connected to the servo valve/proportional valve 2. The piston 4 of the hydraulic cylinder 3 is provided with a sensor 5, and the sensor 5 is connected to a controller 6 through a circuit. Servo valve/proportional valve 2 circuits are connected; the external load f of the system changes with time and is unknown. Input the command displacement or speed, the controller 6 outputs an adjustment voltage u to the servo valve/proportional valve 2 according to the current displacement or speed collected by the sensor, and the servo valve/proportional valve 2 generates the spool displacement x _v , when the oil supply pressure of the system is Under p _s+ , the flow q ₊ passes through the servo valve/proportional valve 2 and then enters the left chamber of the hydraulic cylinder 3 to generate pressure p ₊ , which pushes the load (synthetic effective mass m) to the right through the piston 4 (effective area a ₊ ). Correspondingly, the flow q _- is pushed out of the right chamber of the hydraulic cylinder 3 to generate the pressure p _- , and then returns to the oil tank through the oil return line (pressure p _s- ), and the operation in the opposite direction is similar, so that the closed-loop control can finally achieve the required The piston 4 displacement or velocity.

The on-line fault detection, estimation and location method of valve-controlled cylinder electro-hydraulic servo system includes the following steps:

1) System nonlinear modeling: establish the mathematical model of valve-controlled cylinder electro-hydraulic servo system; select the system state variables; establish the nonlinear state equation of the system;

First, the valve-controlled cylinder electro-hydraulic servo system is equivalent to each link, and the flow equation of the servo valve/proportional valve, the dynamic equation of the servo valve/proportional valve, the flow continuity equation of the hydraulic cylinder and the kinetic force balance equation of the piston of the hydraulic cylinder are established. The mathematical model of the overall system is expressed by the above four equations, as follows:

Flow equation for servo valve/proportional valve:

In the formula, q _± , p _± are the flow rate and pressure of the two chambers of the hydraulic cylinder respectively, p _s+ , p _s- are the oil supply and oil return pressure respectively, x _v and α are the deviation of the servo valve/proportional valve spool from the neutral position respectively Displacement and spool positive coverage, _kc is the flow coefficient;

Servo valve/proportional valve dynamic equation:

In the formula, k _v and τ are the gain and time coefficients describing the dynamic characteristics of the servo valve/proportional valve, and u is the input voltage;

Flow continuity equation of hydraulic cylinder:

a _± and V _± are the area and effective volume of the two chambers of the hydraulic cylinder respectively, c _i and c _e are the internal and external leakage coefficients of the hydraulic cylinder respectively, x _p is the piston displacement of the hydraulic cylinder, β _e is the effective bulk modulus, p ₊ and p _- is the pressure of the left and right chambers of the hydraulic cylinder;

The kinetic force balance equation of hydraulic cylinder piston:

In the formula, m is the total mass converted to the load, b _p is the viscous damping coefficient, f is the external load force acting on the piston, d is the variable load disturbance force, a ₊ and a _- are the areas of the two chambers of the hydraulic cylinder;

Select the system state variable: select the hydraulic cylinder piston movement speed The pressures p ₊ and p _- of the left and right chambers of the hydraulic cylinder and the displacement x _v of the spool of the servo valve/proportional valve are the state variables of the system, and the system state variables are defined as

Establish the nonlinear state equation of the system: regard the voltage input to the servo valve/proportional valve as the input u, and define the piston movement speed of the hydraulic cylinder The pressure p ₊ and p _- of the left and right chambers of the hydraulic cylinder are the output vectors The flow equation of the servo valve/proportional valve, the dynamic equation of the servo valve/proportional valve, the flow continuity equation of the hydraulic cylinder, and the kinetic force balance equation of the hydraulic cylinder piston are transformed into four equations, and the state variables x and non The linear term g(x) is converted into the following state equation form, and the matrices A, B, C, and D are the parameter matrices of the system;

2) Normal parameter identification of the system: Obtain historical input and output data under different working conditions and environmental conditions during normal operation of the system, input the identification model, and obtain the parameters and their variation range during normal operation of the system from the identification model;

Given a continuously changing servo valve/proportional valve voltage input signal, collect the piston displacement value x _p of the hydraulic cylinder, and the pressure values p ₊ and p _- of the two chambers of the hydraulic cylinder;

From the nonlinear state equation of the system, a system of linear equations with identification parameters is obtained:

Φθ=Γ,

θ=[β _e ^-1 c _i c _e k _q ] ^T is the parameter to be identified, k _q =k _c k _v ,

is a parameter matrix containing system state quantities, where k=1,2,3...;

Substitute the collected data into the above equations, and use the least squares method to identify the unknown parameter θ in the above equations; bring the identified system parameters into the nonlinear state equation of the system, and compare the obtained output with the actual system output After comparison and correction, the parameter values and variation ranges of the system in normal operation are finally obtained.

3) Bring the data obtained in step 2) into the nonlinear state equation of the system in step 1) to obtain the normal system parameter matrix and the expression of the nonlinear state equation of the system;

4) Real-time acquisition of relevant input and output parameters of the system;

5) Fault detection and estimation under variable load: establish a fault detection observer, which includes an external load force decoupling module, a fault dominant module, a nonlinear module, a stability module and a fault vector estimation module, and The collected signals of each module are input to the observer to obtain the estimated output of the observer, and the output residual is obtained by making a difference with the actual system output;

External load force decoupling module: it mainly realizes the function that the change of external disturbance force will not affect the residual error used for fault judgment. The realization method is to set the parameter matrix T of the observer T=-D(CD) ⁺ +Y[I -(CD)(CD) ⁺ ], where (CD) ⁺ ＝((CD) ^T (CD)) ^-1 (CD) ^T , Y is the parameter matrix to be set, obtained in the stability module, C, D are In the system parameter matrix obtained in step 3), I is an identity matrix;

Nonlinear module: obtain the nonlinear expression g(x) of the original system to observe the state quantity Replace the original state quantity x, without linear simplification, directly used for stability module processing;

Stability module: mainly realizes the rapid and effective convergence of the observer, and the realization method is as follows:

a: Obtain a positive definite symmetric matrix P>0 by solving the following linear matrix inequality, and two matrices where P ₁ =P(I+T _a C), T _a ＝-D(CD) ⁺ , T _b ＝I-(CD)(CD) ⁺ , γ is a constant:

b: calculate the matrix Y and K, and

c: Get the rest of the observer parameter matrix, G=MB, N=MA-KC, L=K-NT;

Residual error generation module: the output signal of the actual system (including the hydraulic cylinder piston speed signal value And the output vector composed of the pressure signal values p ₊ and p _- of the two chambers of the hydraulic cylinder and the observer output vector Subtract to get the output residual vector

Fault vector estimation module: The fault vector estimation module is composed of a neural network, which does not work when the residual is less than or equal to the threshold, and is activated only when the residual is greater than the threshold to perform fault estimation.

Above-mentioned matrix T, M, G, N and L are the system parameter matrix to be designed, and matrix A, B, C and D are the system parameter matrix obtained in step 3);

6) Robust fault decision-making: input the output residual obtained in step 3) into the error estimation function, and combine with the adaptive threshold calculated in real time to diagnose whether there is a fault in the system;

The calculation error estimation function is: J(t)=r ^T (t)Hr(t), where H is a weighted diagonal function; fault detection is performed according to the following rules:

In the formula, λ(t) is the adaptive threshold, which is composed of two parts, one is the steady state threshold, and the other is the transient threshold, which is switched by the acceleration value of the hydraulic cylinder piston. When the threshold is set, the system Taking into account the normal fluctuation of parameters, the threshold value is calculated online based on statistical methods. In addition, the amount of data calculation and storage space is reduced by segmenting and weighting the residual sequence, while ensuring a certain degree of robustness. The specific implementation process:

a: Substitute the parameter variation range obtained by the identification model in step 2) into the fault detection and estimation observer to obtain the fluctuation range of the steady-state threshold, and take the upper and lower limits of the fluctuation as the steady-state threshold λ _o- , λ _o+ ;

b: Differentiate the obtained piston speed value or position value twice to obtain the piston motion acceleration value a _P (t);

c: Obtain a sequence pair of residual r _i (k) and piston acceleration a _i (k) with sample length S respectively, where i=(k-1)S/l+1,...,(k -1) S/l+S, calculate the piston mean value a _p (k) of this S data acceleration sequence;

d: Divide the residual sequence r _i (k) of S data length into l parts, and calculate its mean value μ _p (k) for each part of the residual data whose length is S/l, and then calculate the weighted value according to the following formula mean μ _r (k) and variance σ _r ² (k);

where w _p is the assigned weight coefficient, and

e: The adaptive threshold λ(k) is determined by the following formula;

Where ε is the bandwidth coefficient, c _a is the critical value of acceleration;

f: Repeat the above steps to calculate the threshold λ in real time.

7) Start the fault vector estimation module in step 5) of the fault detection observer, and carry out real-time adjustment of the network weight coefficients by the residual value obtained by the observer and the actual system output, and finally obtain the fault vector estimation value;

The implementation process of the fault vector estimation module is as follows: set a reasonable neural network structure, and the number of network input nodes, hidden layer nodes and output nodes are respectively n+m, s, n, where n, m are the state quantities and input of the fault diagnosis observer The number of measurements and the number of hidden layers s are determined offline, the input of the neural network is the estimated value of the system state quantity and the input voltage of the servo valve/proportional valve, and the neural network is established In the formula, W is the output weight coefficient matrix of the neural network to be designed, is the neural network basis function;

Obtaining State Estimators of Fault Diagnosis Observers in Real Time And the output residual vector r _n (t), the real-time adjustment of the neural network output weight coefficient is carried out according to the following formula In the formula, the parameter matrix η is a normal number, P is a positive definite matrix, and u is the input voltage of the servo valve/proportional valve;

Record the output value of the neural network at this time, as the estimated value of the current fault vector f(t)=f _nn (t);

Get the observer estimated output vector after starting the neural network fault vector estimation module and get the output residuals Where y(t) is the actual output vector;

Detect output residual r _n (t), if r _n (t)>λ _n (t) return to step 6), if r _n (t)≤λ _n (t) enter the next step, where λ _n (t) Decision Adaptive Thresholds for Fault Estimation;

The output fault vector estimate is f(t)=[f ₁ f ₂ f ₃ ] ^T .

8) Fault isolation and location: Combining the estimated value of the fault vector and the movement direction of the hydraulic cylinder piston to judge the type and location of the possible fault, and output the diagnosis result.

The fault location method based on the fault estimation vector f(t)=[f ₁ f ₂ f ₃ ] ^T is as follows:

, if |f| ₁ ＞δ ₁ , |f ₂ |<δ ₂ , |f ₃ |<δ ₃ or, When |f ₁ |<δ ₁ , |f ₂ |>δ ₂ , |f ₃ |<δ ₃ , it is judged that the oil supply pressure of the system is abnormal, check the hydraulic pump and pump outlet relief valve;

, if |f ₁ |<δ ₁ , |f ₂ |>δ ₂ , f ₃ |<δ ₃ or, When |f ₁ |>δ ₁ , |f ₂ |<δ ₂ , |f ₃ |<δ ₃ , it is judged that the oil return pressure of the system is abnormal, check whether there is blockage in the circuit pipeline;

If |f ₁ |＞δ ₁ , |f ₂ |＞δ ₂ , |f ₃ |<δ ₃ , it is judged as hydraulic cylinder leakage failure, check the hydraulic cylinder;

If |f ₁ |<δ ₁ , |f ₂ |<δ ₂ , |f ₃ |>δ ₃ , it is judged that the servo valve/proportional valve is faulty, check the valve;

Among them, each fault judgment value δ ₁ , δ ₂ , δ ₃ is a constant.

Through the above method, without any measurement or estimation of the external load force, the operating state of the system can be effectively judged, the existing fault can be detected, and the size of the fault can be estimated, and the location and type of the fault can be estimated. It is very economical, convenient and reliable for practical application to judge.

The construction of the fault detection observer in the patent of the present invention is the key point. The observer includes an external load force decoupling module, a fault dominant module, a nonlinear module, a stability module, a residual generation module and a fault vector estimation module. The external load force decoupling module is used for the effective decoupling of the unknown time-varying external load, so that it will not affect the residual used for fault judgment; the fault explicit module is used to reflect the system fault information in the residual, so that Fault judgment; the nonlinear module mainly deals with the nonlinear relationship of the system; the stability module is used to realize the rapid and effective convergence of the fault detection observer; the residual error generation module is used to generate the residual error for fault judgment; the fault vector estimation module consists of The neural network structure does not work when the residual is less than or equal to the threshold, and is only activated when the residual is greater than the threshold to perform fault estimation. Through the cooperation of the above modules, effective processing of time-varying external loads and system nonlinearity can be realized, and the accuracy of fault detection and estimation can be improved.

The advantage of the present invention is that there is no need to install additional sensors to measure the size of the time-varying external load force, and it does not need to be estimated in advance, but the time-varying external load is decoupled by a mathematical analysis method, saving costs and avoiding Installation is troublesome, and the accuracy of diagnosis is improved. At the same time, the present invention can reduce the amount of sample training, and does not need to build multiple observers for fault detection and location estimation. Only one fault diagnosis observer can realize fault detection, occurrence position, category judgment and size estimation. And it is not affected by time-varying external loads, has strong robustness, has a small amount of online calculation, and is very convenient to use.

Claims (7)

1. On-line fault detection, estimation and location method of the valve-controlled cylinder electro-hydraulic servo system. The invented method is based on the valve-controlled cylinder electro-hydraulic servo system. The valve-controlled cylinder electro-hydraulic servo system includes a hydraulic cylinder and a servo valve/proportion connected to the hydraulic cylinder The valve and the hydraulic pump connected to the servo valve/proportional valve, the piston of the hydraulic cylinder is equipped with a sensor, the sensor is connected to the controller through the circuit, and the controller is connected to the servo valve/proportional valve circuit; It is characterized in that, comprising the steps of: 1) System nonlinear modeling: establish the mathematical model of valve-controlled cylinder electro-hydraulic servo system; select the system state variables; establish the nonlinear state equation of the system; 2) Normal parameter identification of the system: Obtain the input and output historical data under different working conditions and environmental conditions during the normal operation of the system, input the identification model, and obtain the parameters and their variation range during the normal operation of the system from the identification model; 3) Bring the data obtained in step 2) into the nonlinear state equation of the system in step 1) to obtain the normal system parameter matrix and the expression of the nonlinear state equation of the system; 4) Real-time acquisition of relevant input and output parameters of the system; 5) Fault detection and estimation under variable load: establish a fault detection observer, which includes an external load force decoupling module, a fault dominant module, a nonlinear module, a stability module and a fault vector estimation module, and The collected signals of each module are input to the observer to obtain the estimated output of the observer, and the output residual is obtained by making a difference with the actual system output; 6) Robust fault decision-making: input the output residual error obtained in step 3) into the calculation error estimation function, and combine the adaptive threshold calculated in real time to diagnose whether there is a fault in the system; 7) If there is no fault in step 6), return to step 4); if there is a fault, start the fault vector estimation module of the fault detection observer in step 5), and obtain the output residual value from the observer and the actual system for network weighting The real-time adjustment of the value coefficient finally obtains the estimated value of the fault vector; 8) Fault isolation and location: Combine the estimated value of the fault vector in step 7) with the movement direction of the hydraulic cylinder piston to judge the type and location of the possible fault, and output the diagnosis result. 2. The on-line fault detection, estimation and location method of the valve-controlled cylinder electro-hydraulic servo system according to claim 1, characterized in that, in the step 1), the mathematical model of the valve-controlled cylinder electro-hydraulic servo system is established: first the valve Each link of the cylinder-controlled electro-hydraulic servo system is equivalent, and the flow equation of the servo valve/proportional valve, the dynamic equation of the servo valve/proportional valve, the flow continuity equation of the hydraulic cylinder, and the kinetic force balance equation of the piston of the hydraulic cylinder are established. The equation expresses the mathematical model of the overall system as follows: Flow equation for servo valve/proportional valve: In the formula, q _± , p _± are the flow rate and pressure of the two chambers of the hydraulic cylinder respectively, p _s+ , p _s- are the oil supply and oil return pressure respectively, x _v and α are the deviation of the servo valve/proportional valve spool from the neutral position respectively Displacement and spool positive coverage, _kc is the flow coefficient; Servo valve/proportional valve dynamic equation: In the formula, k _v and τ are the gain and time coefficients describing the dynamic characteristics of the servo valve/proportional valve, and u is the input voltage; Flow continuity equation of hydraulic cylinder: a _± and V _± are the area and effective volume of the two chambers of the hydraulic cylinder respectively, c _i and c _e are the internal and external leakage coefficients of the hydraulic cylinder respectively, x _p is the piston displacement of the hydraulic cylinder, β _e is the effective bulk modulus, p ₊ and p _- is the pressure of the left and right chambers of the hydraulic cylinder; The kinetic force balance equation of hydraulic cylinder piston: In the formula, m is the total mass converted to the load, b _p is the viscous damping coefficient, f is the external load force acting on the piston, d is the variable load disturbance force, a ₊ and a _- are the areas of the two chambers of the hydraulic cylinder; Select the system state variable: select the hydraulic cylinder piston movement speed The pressures p ₊ and p _- of the left and right chambers of the hydraulic cylinder and the displacement x _v of the spool of the servo valve/proportional valve are the state variables of the system, and the system state variables are defined as Establish the nonlinear state equation of the system: regard the voltage input to the servo valve/proportional valve as the input u, and define the piston movement speed of the hydraulic cylinder The pressure p ₊ and p _- of the left and right chambers of the hydraulic cylinder are the output vectors The flow equation of the servo valve/proportional valve, the dynamic equation of the servo valve/proportional valve, the flow continuity equation of the hydraulic cylinder, and the kinetic force balance equation of the hydraulic cylinder piston are transformed into four equations, and the state variables x and non The linear term g(x) is converted into the following state equation form, and the matrices A, B, C, and D are the parameter matrices of the system; 3. The valve-controlled cylinder electro-hydraulic servo system online fault detection, estimation and location method according to claim 1, characterized in that, in the step 2), the input and output under different working conditions and environmental conditions are obtained when the system is in normal operation Historical data: Given a continuously changing servo valve/proportional valve voltage input signal, collect the piston displacement value x _p of the hydraulic cylinder, and the pressure values p ₊ and p _- of the two chambers of the hydraulic cylinder; Identification model: From the nonlinear state equation of the system, a linear equation system containing identification parameters is obtained: Φθ=Γ, θ=[β _e ^-1 c _i c _e k _q ] ^T is the parameter to be identified, k _q =k _c k _v , is a parameter matrix containing system state quantities, where k=1,2,3...; Substitute the collected data into the above equations, and use the least squares method to identify the unknown parameter θ in the above equations; bring the identified system parameters into the nonlinear state equation of the system, and compare the obtained output with the actual system output After comparison and correction, the parameter values and variation ranges of the system in normal operation are finally obtained. 4. The online fault detection, estimation and location method of a valve-controlled cylinder electro-hydraulic servo system under variable load according to claim 1, characterized in that, in the step 5) decoupling module of external load force: mainly realizes external interference The function that the force change will not affect the residual error used for fault judgment, the realization method: set the parameter matrix T of the observer = -D(CD) ⁺ +Y[I-(CD)(CD) ⁺ ], where (CD) ⁺ ＝((CD) ^T (CD)) ^-1 (CD) ^T , Y is the parameter matrix to be set, obtained in the stability module, C and D are the system parameter matrices obtained in step 3), I is the identity matrix; Fault dominant module: mainly realizes the function that once a fault occurs in the system, it will affect the residual used for fault judgment. The realization method: set the parameter matrix M=I+TC, and MF _f ≠ 0, where F _f is the fault Matrix, expressed as F _f =[e ₁ e ₂ e ₃ ], e _i =[0 e _i1 e _i2 e _i3 ] ^T , Nonlinear module: obtain the nonlinear expression g(x) of the original system to observe the state quantity Replace the original state quantity x, without linear simplification, directly used for stability module processing; Stability module: mainly realizes the rapid and effective convergence of the observer, and the realization method is as follows: a: Obtain a positive definite symmetric matrix P>0 by solving the following linear matrix inequality, and two matrices where P ₁ =P(I+T _a C), T _a ＝-D(CD) ⁺ , T _b ＝I-(CD)(CD) ⁺ , γ is a constant: b: calculate the matrix Y and K, and c: Get the rest of the observer parameter matrix, G=MB, N=MA-KC, L=K-NT; Residual error generation module: the output signal of the actual system (including the hydraulic cylinder piston speed signal value And the output vector composed of the pressure signal values p ₊ and p _- of the two chambers of the hydraulic cylinder and the observer output vector Subtract to get the output residual vector Fault vector estimation module: The fault vector estimation module is composed of a neural network, which does not work when the residual is less than or equal to the threshold, and is activated only when the residual is greater than the threshold to perform fault estimation; The above matrices T, M, G, N and L are the system parameter matrices to be designed, and the matrices A, B, C and D are the system parameter matrices obtained in step 3). 5. The valve-controlled cylinder electro-hydraulic servo system online fault detection, estimation and location method according to claim 1, characterized in that the calculation error estimation function in the step 6) is: J (t)=r ^T (t )Hr(t), where H is a weighted diagonal function; Fault detection is performed according to the following rules: In the formula, λ(t) is the adaptive threshold, which is composed of two parts, one is the steady state threshold, and the other is the transient threshold, which is switched by the acceleration value of the hydraulic cylinder piston. When the threshold is set, the system Taking into account the normal fluctuation of parameters, the threshold value is calculated online based on statistical methods. In addition, the amount of data calculation and storage space is reduced by segmenting and weighting the residual sequence, while ensuring a certain degree of robustness. The specific implementation process: a: Substitute the parameter variation range obtained by the identification model in step 2) into the fault detection observer to obtain the fluctuation range of the steady-state threshold, and take the upper and lower limits of the fluctuation as the steady-state threshold λ _o- , λ _o+ ; b: Differentiate the obtained piston speed value or position value twice to obtain the piston motion acceleration value a _P (t); c: Obtain a sequence pair of residual r _i (k) and piston acceleration a _i (k) with sample length S respectively, where i=(k-1)S/l+1,...,(k -1) S/l+S, calculate the piston mean value a _p (k) of this S data acceleration sequence; d: Divide the residual sequence r _i (k) of S data length into l parts, and calculate its mean value μ _p (k) for each part of the residual data whose length is S/l, and then calculate the weighted value according to the following formula mean μ _r (k) and variance σ _r ² (k); where w _p is the assigned weight coefficient, and e: The adaptive threshold λ(k) is determined by the following formula; Where ε is the bandwidth coefficient, c _a is the critical value of acceleration; f: Repeat the above steps to calculate the threshold λ in real time. 6. The valve-controlled cylinder electro-hydraulic servo system online fault detection, estimation and location method according to claim 1, is characterized in that, the implementation process of the fault vector estimation module in the described step 7) is specifically: Set up a reasonable neural network structure, the number of network input nodes, hidden layer nodes and output nodes are respectively n+m, s, n, where n, m are the state quantities and input quantities of the fault diagnosis observer, and the number of hidden layers is s Determined offline, the input of the neural network is the estimated value of the system state quantity and the input voltage of the servo valve/proportional valve, and the neural network is established In the formula, W is the output weight coefficient matrix of the neural network to be designed, is the neural network basis function; Obtaining State Estimators of Fault Diagnosis Observers in Real Time And the output residual vector r _n (t), the real-time adjustment of the neural network output weight coefficient is carried out according to the following formula In the formula, the parameter matrix η is a normal number, P is a positive definite matrix, and u is the input voltage of the servo valve/proportional valve; Record the output value of the neural network at this time, as the estimated value of the current fault vector f(t)=f _nn (t); Get the observer estimated output vector after starting the neural network fault vector estimation module and get the output residuals Where y(t) is the actual output vector; Detect output residual r _n (t), if r _n (t)>λ _n (t) return to step 6), if r _n (t)≤λ _n (t) enter the next step, where λ _n (t) Decision Adaptive Thresholds for Fault Estimation; The output fault vector estimate is f(t)=[f ₁ f ₂ f ₃ ] ^T . 7. The on-line fault detection, estimation and location method of valve-controlled cylinder electro-hydraulic servo system according to claim 1, characterized in that, the fault type and location judgment in the step 8) are as follows: The fault location method based on the fault estimation vector f(t)=[f ₁ f ₂ f ₃ ] ^T is as follows: , if |f| ₁ ＞δ ₁ , |f ₂ |<δ ₂ , |f ₃ |<δ ₃ or, When |f ₁ |<δ ₁ , |f ₂ |>δ ₂ , |f ₃ |<δ ₃ , it is judged that the oil supply pressure of the system is abnormal, check the hydraulic pump and pump outlet relief valve; , if |f ₁ |<δ ₁ , |f ₂ |>δ ₂ , |f ₃ |<δ ₃ or, When |f ₁ |>δ ₁ , |f ₂ |<δ ₂ , |f ₃ |<δ ₃ , it is judged that the oil return pressure of the system is abnormal, check whether there is blockage in the circuit pipeline; If |f ₁ |＞δ ₁ , |f ₂ |＞δ ₂ , |f ₃ |<δ ₃ , it is judged as hydraulic cylinder leakage failure, check the hydraulic cylinder; If |f ₁ |<δ ₁ , |f ₂ |<δ ₂ , |f ₃ |>δ ₃ , it is judged that the servo valve/proportional valve is faulty, check the valve; where the fault judgment values δ ₁ , δ ₂ , δ ₃ are constants.