Hydraulic cylinder fault prediction and diagnosis method based on short memory grey model

Abstract

The invention provides a hydraulic cylinder fault prediction and diagnosis method based on a short memory grey model, and belongs to the technical field of fault prediction and diagnosis. According to the technical scheme, the method comprises the following steps that 1, a hydraulic cylinder fault prediction and diagnosis short memory grey model is constructed; 2, converting the hydraulic cylinder short memory grey model into an algebraic model; 3, collecting input and output data of the hydraulic cylinder in real time, and identifying model parameters in real time in combination with a deep learning algorithm; and 4, establishing a model parameter fault membership function, and performing fuzzy prediction and diagnosis on the hydraulic cylinder fault according to the parameter membership. The method has the advantages that the system response model is established from the input and output response characteristics of the hydraulic cylinder, the response model is cut off, the short memory gray model is obtained, gray model parameters are identified in real time through input and output data, and faults of the hydraulic cylinder are predicted and diagnosed in real time according to the gray model parameters.

Description

Technical Field

[0001] The invention relates to the technical field of fault prediction and diagnosis, and in particular to a hydraulic cylinder fault prediction and diagnosis method based on a short memory grey model. Background Art

[0002] The hydraulic cylinder has a simple structure and reliable operation. When realizing reciprocating motion, the deceleration device can be eliminated, and there is no transmission gap, and the movement is smooth. The hydraulic cylinder is basically composed of a cylinder barrel and a cylinder head, a piston and a piston rod, a sealing device, a buffer device and an exhaust device. The hydraulic cylinder has a large load and is widely used in the field of heavy machinery. However, as the load changes, it is often accompanied by problems such as inadequate or inaccurate movement, oil leakage, damage, insufficient thrust or unstable movement, creeping problems, leakage problems, entry of foreign objects, impact and buffering, abnormal noise and vibration.

[0003] However, common faults of hydraulic cylinder products, such as inaccurate action, insufficient thrust or unstable action, creeping problems, leakage problems, and foreign matter entry, often occur under working conditions, and it is difficult to reproduce the actual working condition of hydraulic cylinders under offline conditions. This requires online prediction and diagnosis of hydraulic cylinder faults, and emergency treatment when the fault has not occurred or is about to occur, to avoid the occurrence of faults and accidents.

[0004] Usually, a mathematical model is established to use input and output data to predict and diagnose system faults. However, the structure of the hydraulic cylinder is complex, and it is difficult to establish a model from the physical mechanism. Any system input and its output response have a process, and the system output is affected not only by the current input, but also by the historical input. However, it is difficult to study the output based on all historical inputs in engineering, and use input and output to predict and diagnose faults. The impulse response of a physical system often decays over time, and the impact of historical input on output decays over time. The impact of input on output that exceeds a certain historical time can be ignored. Summary of the invention

[0005] The purpose of the present invention is to provide a hydraulic cylinder fault prediction and diagnosis method based on a short memory gray model, which is aimed at the key safety issues of hydraulic cylinders. Starting from the input-output response characteristics of the hydraulic cylinder, the present invention establishes a system response model, truncates the response model to obtain a short memory gray model, uses input and output data to identify the gray model parameters in real time, and predicts and diagnoses hydraulic cylinder faults in real time according to the gray model parameters. Starting from the system input-output characteristics, the present invention does not rely on the physical characteristics of the hydraulic cylinder, establishes a hydraulic cylinder input-output gray model, and according to the attenuation characteristics, cuts a certain length to obtain a short memory gray model, and predicts and diagnoses hydraulic cylinder faults according to the model.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is specifically: a hydraulic cylinder fault prediction and diagnosis method based on short memory grey model, comprising the following steps:

[0007] Step 1: Construct a short memory grey model for hydraulic cylinder fault prediction and diagnosis;

[0008] Step 2: Convert the short memory grey model of the hydraulic cylinder into an algebraic model;

[0009] Step 3: Collect the input and output data of the hydraulic cylinder in real time, and identify the model parameters in real time by combining the deep learning algorithm;

[0010] Step 4: Establish the model parameter fault membership function and perform fuzzy prediction and diagnosis on the hydraulic cylinder fault according to the parameter membership.

[0011] Furthermore, the step 1: constructing a short memory grey model for hydraulic cylinder fault prediction and diagnosis specifically includes the following steps:

[0012] Step 1.1 The sampling interval is T, and the pressure and displacement of the hydraulic cylinder are collected, which are recorded as: p and s respectively. p(0), p(1), p(2), ..., p(n) and s(0), s(1), s(2), ..., s(n) represent the pressure and displacement of the hydraulic cylinder at the initial time, 1T, 2T, ..., nT respectively. The gray model of hydraulic cylinder input and output calibration is established:

[0013]

[0014] Among them, (nT) α Indicates t α The sampling at time nT, k 1,n , k 2,n and α n is the time-varying parameter to be identified, the symbol * represents the convolution operation, It is the discretization processing of the hydraulic cylinder impulse response model.

[0015] Step 1.2 converts the convolutional model into an accumulation expression:

[0016]

[0017] The above reflects that the output of the hydraulic cylinder is the combined effect of all pressure signal inputs before the current moment, not just the current pressure signal.

[0018] Step 1.3: According to the memory decay characteristics, the memory length L is set to obtain the short memory grey model for hydraulic cylinder fault prediction and diagnosis.

[0019] For any hydraulic cylinder system, when the input changes within a bound, its output must converge quickly. That is, when the hydraulic cylinder pressure changes instantaneously, it will cause the hydraulic cylinder displacement to change, but the displacement will converge over time, and the displacement change beyond a certain time will be very small and can be ignored. Based on this, the historical information with a fixed length of L is intercepted for research, and the short memory grey model of the hydraulic cylinder is obtained:

[0020]

[0021] Where Sm represents the starting point of memory. When nL≥0, Sm=nL; when nL<0, Sm=0.

[0022] Furthermore, the step 2: converting the short memory grey model of the hydraulic cylinder into an algebraic model specifically comprises the following steps:

[0023] Step 2.1: Represent the model in formula (2) using a matrix.

[0024]

[0025] Step 2.2: Simplify the model further:

[0026]

[0027] where vector S = [s(0) s(1) … s(n)] T , K n =[k 1,n k 2,n ] T ,

[0028]

[0029] The symbol T represents the transpose of a vector.

[0030] Step 2.3: Use the least squares method to get the current parameter α n K under condition n , using formula (4) we can get k 1,n , k 2,n Identification algorithm:

[0031]

[0032] in Representation Matrix The inverse matrix of .

[0033] As a hydraulic cylinder fault prediction and diagnosis method based on a short memory grey model provided by the present invention, the step 3: real-time acquisition of hydraulic cylinder input and output data, and real-time identification of model parameters in combination with a deep learning algorithm, specifically includes the following steps:

[0034] Step 3.1: Construct the cost function:

[0035]

[0036] in is the parameter k 1,n , k 2,n and α n Under these conditions, we can obtain the value by solving the model in step 1.3.

[0037] It is the i-th data of the hydraulic cylinder collected in the experiment.

[0038] Step 3.2: Identify α using the improved quadratic learning particle swarm search algorithm n , specifically including the following steps:

[0039] B1: Initialize particle swarm parameters

[0040] B2: Get the output corresponding to the particle position

[0041] B3: Calculate the cost function

[0042] B4: Local Optimal Update

[0043] B5: Global Optimal Update

[0044] B6: Update particle speed

[0045] B7: Update particle position

[0046] B8: Check whether the end condition is met.

[0047] Furthermore, in B1, by first defining the number of particles as M, the particle speed and position limits as v max , v min , α max , α min , using α n,l,j , v l,j denote the position and velocity of the lth particle in the jth generation, respectively. n,l,j The position represents the parameter α to be identified corresponding to the lth particle in the jth generation n ;

[0048] Furthermore, in B2, data is collected by experiment p(i) and formula (5) in step 2.3, we get the current particle position α l,j and parameter K n (k 1,n , k 2,n ).

[0049] According to the current particle position and parameters (k1,n , k 2,n ) Using formula (2) to solve s(i) (i = S m ,S m +1,…,n).

[0050] Further, in B3, the cost function is calculated according to step 3.1 to obtain the cost function corresponding to the current particle position

[0051] Furthermore, the local optimal update method is as follows: update the individual optimal fitness one by one, compare the individual cost function with the best cost function corresponding to the individual's historical best position, and retain the best cost function and the corresponding position α n,l,best ;

[0052] The global optimal update method is as follows: compare the cost function corresponding to each individual position with the cost function corresponding to the global optimal position, and retain the global optimal position α n,best .

[0053] Furthermore, the particle velocity in B6 is expressed as:

[0054] v l,j+1 =v l,j +c1r1(α n,l,j -α n,best )+c2r2(α n,l,j -α n,l,best ) (7)

[0055] In the above formula, c1 and c2 are normal numbers for adjusting the learning speed, r1 and r2 are random numbers in (0,1).

[0056] The particle position of B7 is expressed as:

[0057] α n,l,j+1 =α n,l,j +v n,l,j+1 (8)

[0058] Furthermore, the specific step of B8 is: determine whether the maximum number of iterations has been reached, if so, α n,best is the global optimal solution, and the optimal identification parameter α is obtained n , otherwise jump to B2 calculation.

[0059] Combined with the results in B2, we can get the grey model parameter k 1,n , k 2,n and α n .

[0060] Furthermore, the step 4: establishing a model parameter fault membership function, and performing fuzzy prediction and diagnosis on the hydraulic cylinder fault according to the parameter membership, specifically includes the following steps:

[0061] Step 4.1, determine the working range of the hydraulic cylinder parameters and establish the fault membership function of each parameter:

[0062] Determine the normal working range of hydraulic cylinder parameters, which are: [k 1,min ,k 1,max ]、[k 2,min ,k 2,max ] and [α min ,α max ].

[0063] Establish parameters k respectively 1,n , k 2,n and α n The corresponding fault membership function and

[0064]

[0065] w1, w2 and w α The membership weight factors corresponding to k1, k2 and α respectively adjust the parameter fault membership.

[0066] According to the above definition of membership, and The value range is (0,1). The membership degree represents the degree of hydraulic cylinder failure. The larger the membership degree, the more serious the failure.

[0067] Step 4.2, establish the comprehensive evaluation function Ev(n) of hydraulic cylinder fault:

[0068]

[0069] in represents the multi-parameter abnormal evaluation function of the hydraulic cylinder, 0<Ce1<1 and 0<Ce2<1 are ’s weight, and satisfies 0<Ce1+Ce2<1.

[0070] The comprehensive evaluation function in formula (12) indicates that any abnormality in any parameter or the comprehensive evaluation abnormality of multiple parameters indicates that the hydraulic cylinder is abnormal.

[0071] Step 4.3, set the hydraulic cylinder fault threshold Th to predict and diagnose the hydraulic cylinder fault.

[0072] When Ev(n) is greater than Th, it is determined that the hydraulic cylinder is faulty. The larger it is, the more serious the fault is. When Ev(n) is less than Th, the closer it is to Th, the closer the hydraulic cylinder is to fault.

[0073] Step 4.4, n plus 1, go to step 3 to achieve real-time prediction and diagnosis of hydraulic cylinder faults.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] 1. The present invention proposes a hydraulic cylinder fault prediction and diagnosis method based on a short memory grey model. The method builds a model based on the dynamic response characteristics of the hydraulic cylinder, identifies model parameters based on input and output data, and predicts and diagnoses the hydraulic cylinder operating conditions based on the parameter range. The diagnostic method is more in line with the hydraulic cylinder fault mechanism.

[0076] 2. The present invention uses input-output response characteristics to establish a hydraulic cylinder memory model instead of a traditional differential model, and combines the model attenuation characteristics to obtain a short memory model, which reduces the complexity of the algorithm and improves real-time performance. It also takes into account the real-time performance and accuracy of hydraulic cylinder fault diagnosis.

[0077] 3. The fault prediction and diagnosis method proposed by the present invention based on the system response model parameters is based on the hydraulic cylinder, but is not limited to the hydraulic cylinder fault prediction and diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0079] Figure 1 The present invention is a flow chart of the hydraulic cylinder fault prediction and diagnosis method based on the short memory grey model.

[0080] Figure 2 A short memory grey model flow chart for hydraulic cylinder fault prediction and diagnosis is constructed for the present invention.

[0081] Figure 3 The present invention converts the short memory grey model into an algebraic model flow chart.

[0082] Figure 4 This is a flow chart of the hydraulic cylinder model parameter identification algorithm of the present invention.

[0083] Figure 5 This is a flow chart of real-time fault prediction and diagnosis of hydraulic cylinders according to the present invention.

[0084] Figure 6 This is a curve chart of fault prediction and diagnosis results in Example 1 of the present invention.

[0085] Figure 7 This is a curve chart of fault prediction and diagnosis results in Example 2 of the present invention. DETAILED DESCRIPTION

[0086] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Of course, the specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0087] Example 1

[0088] See also Figures 1 to 6 The technical solution provided by this embodiment is a hydraulic cylinder fault prediction and diagnosis method based on a short memory grey model, comprising the following steps:

[0089] Step 1: Construct a short memory grey model for hydraulic cylinder fault prediction and diagnosis;

[0090] Step 2: Convert the short memory grey model of the hydraulic cylinder into an algebraic model;

[0091] Step 3: Collect the input and output data of the hydraulic cylinder in real time, and identify the model parameters in real time by combining the deep learning algorithm;

[0092] Step 4: Establish the model parameter fault membership function and perform fuzzy prediction and diagnosis on the hydraulic cylinder fault according to the parameter membership.

[0093] Specifically, step 1: constructing a short memory grey model for hydraulic cylinder fault prediction and diagnosis, which specifically includes the following steps:

[0094] Step 1.1 The sampling interval is T, and the pressure and displacement of the hydraulic cylinder are collected, which are recorded as: p and s respectively. p(0), p(1), p(2), ..., p(n) and s(0), s(1), s(2), ..., s(n) represent the pressure and displacement of the hydraulic cylinder at the initial time, 1T, 2T, ..., nT respectively. The gray model of hydraulic cylinder input and output calibration is established:

[0095]

[0096] Among them, (nT) α Indicates t α The sampling at time nT, k 1,n , k 2,n and α n is the time-varying parameter to be identified, the symbol * represents the convolution operation, It is the discretization processing of the hydraulic cylinder impulse response model.

[0097] Step 1.2 converts the convolutional model into an accumulation expression:

[0098]

[0099] The above reflects that the output of the hydraulic cylinder is the combined effect of all pressure signal inputs before the current moment, not just the current pressure signal.

[0100] Step 1.3: According to the memory decay characteristics, the memory length L is set to obtain the short memory grey model for hydraulic cylinder fault prediction and diagnosis.

[0101] For any hydraulic cylinder system, when the input changes within a bound, its output must converge quickly. That is, when the hydraulic cylinder pressure changes instantaneously, it will cause the hydraulic cylinder displacement to change, but the displacement will converge over time, and the displacement change beyond a certain time will be very small and can be ignored. Based on this, the historical information with a fixed length of L is intercepted for research, and the short memory grey model of the hydraulic cylinder is obtained:

[0102]

[0103] Where Sm represents the starting point of memory. When nL≥0, Sm=nL; when nL<0, Sm=0.

[0104] Specifically, step 2: converting the short memory grey model of the hydraulic cylinder into an algebraic model, specifically includes the following steps:

[0105] Step 2.1: Represent the model in formula (2) using a matrix.

[0106]

[0107] Step 2.2: Simplify the model further:

[0108]

[0109] where vector S = [s(0) s(1) … s(n)] T , K n =[k 1,n k 2,n ] T ,

[0110]

[0111] The symbol T represents the transpose of a vector.

[0112] Step 2.3: Use the least squares method to get the current parameter α n K under condition n , using formula (4) we can get k 1,n , k 2,n Identification algorithm:

[0113]

[0114] in Representation Matrix The inverse matrix of .

[0115] Specifically, step 3: real-time collection of hydraulic cylinder input and output data, combined with deep learning algorithms to identify model parameters in real time, specifically includes the following steps:

[0116] Step 3.1: Construct the cost function:

[0117]

[0118] in is the parameter k 1,n , k 2,n and α n Under these conditions, we can obtain the value by solving the model in step 1.3. It is the i-th data of the hydraulic cylinder collected in the experiment.

[0119] Step 3.2: Identify α using the improved quadratic learning particle swarm search algorithm n , specifically including the following steps:

[0120] B1: Initialize particle swarm parameters

[0121] B2: Get the output corresponding to the particle position

[0122] B3: Calculate the cost function

[0123] B4: Local Optimal Update

[0124] B5: Global Optimal Update

[0125] B6: Update particle speed

[0126] B7: Update particle position

[0127] B8: Check whether the end condition is met.

[0128] Furthermore, in B1, by first defining the number of particles as M, the particle speed and position limits as v max , v min , α max , α min , using α n,l,j , v l,j denote the position and velocity of the lth particle in the jth generation, α n,l,j The position represents the parameter α to be identified corresponding to the lth particle in the jth generation n ;

[0129] Furthermore, in B2, data is collected using experiments p(i) and formula (5) in step 2.3, we get the current particle position α l,j and parameter K n(k 1,n , k 2,n ).

[0130] According to the current particle position and parameters (k 1,n , k 2,n ) Using formula (2) to solve s(i) (i = S m ,S m +1,…,n).

[0131] Furthermore, in B3, the cost function is calculated according to step 3.1 to obtain the cost function corresponding to the current particle position

[0132] Furthermore, the local optimal update method is as follows: update the individual optimal fitness one by one, compare the individual cost function with the best cost function corresponding to the individual's historical best position, and retain the best cost function and the corresponding position α n,l,best ;

[0133] The global optimal update method is as follows: compare the cost function corresponding to each individual position with the cost function corresponding to the global optimal position, and retain the global optimal position α n,best .

[0134] Furthermore, the particle velocity in B6 is expressed as:

[0135] v l,j+1 =v l,j +c1r1(α n,l,j -α n,best )+c2r2(α n,l,j -α n,l,best ) (7)

[0136] In the above formula, c1 and c2 are normal numbers for adjusting the learning speed, r1 and r2 are random numbers in (0,1).

[0137] The particle position of B7 is expressed as:

[0138] α n,l,j+1 =α n,l,j +v n,l,j+1 (8)

[0139] Furthermore, the specific steps of B8 are: determine whether the maximum number of iterations has been reached, if so, α n,best is the global optimal solution, and the optimal identification parameter α is obtained n , otherwise jump to B2 calculation.

[0140] Combined with the results in B2, we can get the grey model parameter k 1,n , k 2,n and α n .

[0141] Specifically, step 4: establish a model parameter fault membership function, and perform fuzzy prediction and diagnosis on the hydraulic cylinder fault according to the parameter membership, which specifically includes the following steps:

[0142] Step 4.1, determine the working range of the hydraulic cylinder parameters and establish the fault membership function of each parameter:

[0143] Determine the normal working range of hydraulic cylinder parameters, which are: [k 1,min ,k 1,max ]、[k 2,min ,k 2,max ] and [α min ,α max ].

[0144] Establish parameters k respectively 1,n , k 2,n and α n The corresponding fault membership function and

[0145]

[0146] w1, w2 and w α The membership weight factors corresponding to k1, k2 and α respectively adjust the parameter fault membership.

[0147] According to the above definition of membership, and The value range is (0,1). The membership degree represents the degree of hydraulic cylinder failure. The larger the membership degree, the more serious the failure.

[0148] Step 4.2, establish the comprehensive evaluation function Ev(n) of hydraulic cylinder fault:

[0149]

[0150] in represents the multi-parameter abnormal evaluation function of the hydraulic cylinder, 0<Ce1<1 and 0<Ce2<1 are ’s weight, and satisfies 0<Ce1+Ce2<1.

[0151] The comprehensive evaluation function in formula (12) indicates that any abnormality in any parameter or the comprehensive evaluation abnormality of multiple parameters indicates that the hydraulic cylinder is abnormal.

[0152] Step 4.3, set the hydraulic cylinder fault threshold Th to predict and diagnose the hydraulic cylinder fault.

[0153] When Ev(n) is greater than Th, it is determined that the hydraulic cylinder is faulty. The larger it is, the more serious the fault is. When Ev(n) is less than Th, the closer it is to Th, the closer the hydraulic cylinder is to fault.

[0154] Step 4.4, n plus 1, go to step 3 to achieve real-time prediction and diagnosis of hydraulic cylinder faults.

[0155] Furthermore, when the hydraulic cylinder input is 0.5t, the hydraulic cylinder pressure output is measured and the identification is performed using the method of this embodiment. The identification result is as follows: Figure 6 As shown, Figure 6 The parameter identification results show that the parameters are abnormal near t = 0.5, which indicates that the hydraulic cylinder has a fault warning.

[0156] Example 2

[0157] Based on Example 1, see Figure 7 , the hydraulic cylinder input is 1+0.05t, the hydraulic cylinder pressure output is measured, and the above method is used for identification. The identification results are as follows Figure 7 As shown, Figure 7 The parameter identification results show that the parameters are abnormal near t = 0.4, which indicates that the hydraulic cylinder has a fault warning.

[0158] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A hydraulic cylinder fault prediction and diagnosis method based on short memory grey model, characterized in that: The following steps are involved: Step 1: Construct a short memory grey model for hydraulic cylinder fault prediction and diagnosis; Step 2: Convert the short memory grey model of the hydraulic cylinder into an algebraic model; Step 3: Collect the input and output data of the hydraulic cylinder in real time, and identify the model parameters in real time by combining the deep learning algorithm; Step 4: Establish the model parameter fault membership function and perform fuzzy prediction and diagnosis on the hydraulic cylinder fault according to the parameter membership.

2. The hydraulic cylinder fault prediction and diagnosis method based on short memory grey model according to claim 1 is characterized in that: In step 1, a short memory grey model for hydraulic cylinder fault prediction and diagnosis is constructed, including the following steps: Step 1.1: The sampling interval is T, and the pressure and displacement of the hydraulic cylinder are collected, which are recorded as p and s respectively. p(0), p(1), p(2), ..., p(n) and s(0), s(1), s(2), ..., s(n) represent the pressure and displacement of the hydraulic cylinder at the initial time, 1T, 2T, ..., nT respectively. The grey model of hydraulic cylinder input and output calibration is established: Among them, (nT) α Indicates t α The sampling at time nT, k 1,n , k 2,n and αn are time-varying parameters to be identified, the symbol * represents the convolution operation, It is the discretization of the hydraulic cylinder impulse response model; Step 1.2: Convert the convolutional model into an accumulation expression: The above reflects that the output of the hydraulic cylinder is the combined effect of all pressure signal inputs before the current moment, not just the current pressure signal; Step 1.3: According to the memory decay characteristics, the memory length L is set to obtain the short memory grey model for hydraulic cylinder fault prediction and diagnosis; For any hydraulic cylinder system, when the input changes within a bound, its output must converge quickly, that is, when the hydraulic cylinder pressure changes instantaneously, it will cause the hydraulic cylinder displacement to change, but the displacement will converge over time, and the displacement change beyond a certain time will be very small and negligible. Based on this, the historical information with a fixed length of L is intercepted for research, and the short memory grey model of the hydraulic cylinder is obtained. Where Sm represents the starting point of memory. When nL≥0, Sm=nL; when nL<0, Sm=0.

3. The hydraulic cylinder fault prediction and diagnosis method based on short memory grey model according to claim 1 is characterized in that: In the step 2, the short memory grey model of the hydraulic cylinder is converted into an algebraic model, which includes the following steps: Step 2.1: Represent the model in formula (2) using a matrix; Step 2.2: Simplify the model further: Among them, vector S = [s(0) s(1)…s(n)] T , K n =[k 1,n k 2,n ] T , The symbol T represents the transpose of a vector; Step 2.3: Use the least squares method to get the current parameter α n K under condition n , using formula (4) to obtain k 1,n , k 2,n Identification algorithm: in Representation Matrix The inverse matrix of .

4. The hydraulic cylinder fault prediction and diagnosis method based on short memory grey model according to claim 1 is characterized in that: In step 3, the input and output data of the hydraulic cylinder are collected in real time, and the model parameters are identified in real time by combining the deep learning algorithm, including the following steps: Step 3.1: Construct the cost function: in is the parameter k 1,n , k 2,n and α n Under these conditions, we can obtain the value by solving the model in step 1.

3. The i-th data of the hydraulic cylinder is obtained by experimental collection; Step 3.2: Identify α using the improved quadratic learning particle swarm search algorithm n , including the following steps: Step 3.2.1: Initialize particle swarm parameters; Step 3.2.2: Get the output corresponding to the particle position; Step 3.2.3: Calculate the cost function Step 3.2.4: Local Optimum Update Step 3.2.5: Global Optimum Update Step 3.2.6: Update particle velocity Step 3.2.7: Update particle positions Step 3.2.8: Check whether the end condition is met.

5. The hydraulic cylinder fault prediction and diagnosis method based on short memory grey model according to claim 1 is characterized in that: In step 3.2.1, the number of particles is defined as M, and the particle speed and position limits are defined as v. max , v min , α max , α min , using α n,l,j , v l,j denote the position and velocity of the lth particle in the jth generation, respectively. n,l,j The position represents the parameter α to be identified corresponding to the lth particle in the jth generation n ; In step 3.2.2, use the experimental data collection p(i) and formula (5) in step 2.3, we get the current particle position α l,j and parameter K n (k 1,n , k 2,n ); According to the current particle position and parameters (k 1,n , k 2,n ) Using formula (2) to solve s(i) (i = S m ,S m +1,…,n); In step 3.2.3, the cost function is calculated according to step 3.1 to obtain the cost function corresponding to the current particle position The local optimal update method is as follows: update the individual optimal fitness one by one, compare the individual cost function with the best cost function corresponding to the individual's historical best position, and retain the best cost function and the corresponding position α n,l,best ; The global optimal update method is as follows: compare the cost function corresponding to each individual position with the cost function corresponding to the global optimal position, and retain the global optimal position α n,best ; The particle velocity in step 3.2.6 is expressed as: v l,j+1 =v l,j +c1r1(a n,l,j -a n,best )+c2r2(a n,l,j -a n,l,best ) (7) In the above formula, c1 and c2 are normal numbers for adjusting the learning speed, r1 and r2 are random numbers in (0,1). The particle position of step 3.2.7 is expressed as: α n,l,j+1 ＝α n,l,j +v n,l,j+1 (8) Step 3.2.8 The specific steps are: determine whether the maximum number of iterations has been reached, if so, α n,best is the global optimal solution, and the optimal identification parameter α is obtained n , otherwise jump to B2 calculation; Combined with the results in step 3.2.2, the grey model parameter k is obtained 1,n , k 2,n and α n .

6. The hydraulic cylinder fault prediction and diagnosis method based on short memory grey model according to claim 1 is characterized in that: The step 4: establishing a model parameter fault membership function, and performing fuzzy prediction and diagnosis on the hydraulic cylinder fault according to the parameter membership, comprises the following steps: Step 4.1 Determine the working range of the hydraulic cylinder parameters and establish the fault membership function of each parameter: Determine the normal working range of hydraulic cylinder parameters, which are: [k 1,min ,k 1,max ]、[k 2,min ,k 2,max ] and [α min ,α max ]; Establish parameters k respectively 1,n , k 2,n and α n The corresponding fault membership function and w1, w2 and w α The membership weight factors corresponding to k1, k2 and α respectively adjust the parameter fault membership; According to the above definition of membership, and The value range is (0,1). The membership degree represents the degree of hydraulic cylinder failure. The larger the membership degree, the more serious the failure. Step 4.2 Establish the comprehensive evaluation function Ev(n) of hydraulic cylinder fault: in represents the multi-parameter abnormal evaluation function of the hydraulic cylinder, 0<Ce1<1 and 0<Ce2<1 are The weight of , and satisfy 0<Ce1+Ce2<1; The comprehensive evaluation function in formula (12) indicates that any abnormal parameter or the comprehensive evaluation abnormality of multiple parameters indicates that the hydraulic cylinder is abnormal; Step 4.3 sets the hydraulic cylinder fault threshold Th to predict and diagnose the hydraulic cylinder fault; When Ev(n) is greater than Th, it is determined that the hydraulic cylinder is faulty. The larger it is, the more serious the fault is. When Ev(n) is less than Th, the closer it is to Th, the closer the hydraulic cylinder is to fault. Step 4.4n plus 1, go to step 3, and realize real-time prediction and diagnosis of hydraulic cylinder faults.

Images