A method for monitoring wear evolution and leakage of a hydraulic cylinder and a digital twin system thereof

Abstract

The present application belongs to the field of hydraulic cylinder monitoring device, and particularly relates to a method for monitoring wear evolution and leakage of a hydraulic cylinder and a digital twin system thereof. The method comprises the following steps: 1. constructing a leakage model representing the mapping relationship between the leakage amount and the contact pressure of the hydraulic cylinder; 2. performing CFD simulation analysis on the three-dimensional model of the hydraulic cylinder to obtain a pressure distribution model under different wear depths; 3. performing wavelet packet analysis on the pressure signal to determine the wear characteristic value, and retaining the corresponding characteristic value extraction model; 4. collecting real data and creating a data-driven model from the characteristic value to the wear depth; 5. fusing the base model to obtain a wear evolution monitoring model and a leakage amount monitoring model; and 6. using the wear evolution monitoring model and the leakage amount monitoring model to predict the state of the collected pressure and displacement signals. The present application solves the problems of difficult detection, high detection cost, insufficient accuracy and real-time performance of internal wear and leakage of the hydraulic cylinder.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of hydraulic cylinder monitoring devices, and particularly relates to a method for monitoring the wear evolution and leakage of a hydraulic cylinder and a digital twin system thereof. BACKGROUND

[0002] A hydraulic cylinder is a mechanical actuator widely used in different industries such as construction, manufacturing, aerospace and offshore oil and gas, and can realize linear motion, jacking operation, etc. The safe and stable operation of the hydraulic cylinder cannot be separated from a reliable sealing system. However, due to factors such as oil pollution, the sealing element is severely worn, which eventually leads to leakage and affects production efficiency. In addition, due to the existence of the "back-pumping effect", the wear of the sealing element does not necessarily lead to the occurrence of leakage. In order to obtain the wear evolution law and leakage state of the sealing element, the two states often need to be monitored separately.

[0003] In the prior art, the wear or leakage of the sealing element is mainly monitored through various sensor signal features, among which vibration, pressure and acoustic emission signals are considered to be able to effectively reflect the wear state of the sealing element and to analyze the signal features when the leakage fault occurs, so as to judge the running state of the hydraulic cylinder. However, there is a certain lag in the analysis through the signal features, and in addition, the monitoring model obtained through the data-driven method depends on the quantity and quality of the data, and has certain limitations. In recent years, with the development of new sensors, the stress distribution of the sealing element contact surface is obtained through the embedded FBG sensor, and the leakage amount is obtained through the mechanism model, thereby improving the accuracy and real-time performance of the monitoring. However, due to the limitation of the internal size of the hydraulic cylinder, the embedded FBG sensor has high cost and is difficult to maintain, and the sensing data of the local position may have fault misjudgment.

[0004] How to monitor the leakage while capturing the wear evolution state of the sealing element of the hydraulic cylinder through easy-to-measure signals has become a big problem, and the digital twin technology provides a new real-time monitoring function. Through high-fidelity simulation, the internal running state of the moving part can be obtained, thereby saving the design and maintenance cost while realizing stable real-time monitoring, so the application of the digital twin technology in the monitoring field is helpful to solve the difficulties in monitoring the wear and leakage of the hydraulic cylinder. SUMMARY

[0005] In order to solve the problems that the internal wear and leakage of the hydraulic cylinder are difficult to detect, the detection cost is high, and the accuracy and real-time performance are insufficient, the application provides a method for monitoring the wear evolution and leakage of a hydraulic cylinder and a digital twin system thereof.

[0006] The application adopts the following technical scheme:

[0007] A hydraulic cylinder wear evolution and leakage monitoring method for generating monitoring results of wear depth V and leakage amount q based on signals of real-time collected rodless cavity pressure p1, rod cavity pressure p2 and displacement x of the piston rod of the hydraulic cylinder, the hydraulic cylinder wear evolution and leakage monitoring method comprising the following steps: L w L

[0008] I. Based on the theoretical model of the current type of hydraulic cylinder, a leakage model q(p) is constructed to represent the mapping relationship between the internal leakage amount q and the external leakage amount q and the contact pressure p. L1 L2 f L f L L1 L2

[0009] II. A three-dimensional model of the current type of hydraulic cylinder is created, and CFD simulation analysis is performed on the hydraulic cylinder using the three-dimensional model, and the wear depth V is calculated at each state based on the Archard equation during the simulation process; thereby obtaining a pressure distribution model p(x, V) representing the global contact pressure distribution of the sealing surface of each network node x at different wear depths V, and reducing the one-dimensional model. w w f w

[0010] III. The two collected pressure signals are subjected to wavelet packet analysis, and the wavelet packet energy entropy H, wavelet packet variance V and wavelet packet energy potential P are extracted as candidate characteristic values. Then the candidate characteristic values with the strongest correlation with the wear state of the hydraulic cylinder evaluated by the Sobol sensitivity analysis method are used as the wear characteristic value WT, and the corresponding characteristic value extraction model is retained. wavelet wavelet j,k

[0011] IV. A large amount of real pressure test data of the current type of hydraulic cylinder in different wear states is collected, and the wear characteristic value samples are generated according to the pressure test data. Then a data-driven model V(WT) is trained using a machine learning algorithm, with the wear characteristic value WT as the input and the wear depth V as the output. w w

[0012] ​​​​​​​​​​​​​​​​​​​​Fifth, a wear evolution monitoring model is obtained by fusing the feature value extraction model and the data-driven model. The leakage model, pressure distribution model, feature value extraction model, and data-driven model are fused to obtain a leakage monitoring model. Then, the discrete parts of the fused model are interpolated and fitted using a joint training strategy in machine learning algorithms.

[0013] 6. The real-time collected hydraulic cylinder pressures p1 (rodless chamber), p2 (rod chamber), and piston rod displacement x are then processed. L The signal is input into a surrogate model that includes a wear evolution monitoring model and a leakage monitoring model, and the corresponding wear depth V is output. w and leakage amount q L .

[0014] As a further improvement to this invention, the construction process of the leakage model in step one is as follows:

[0015] (1) Obtain the structural parameters of the hydraulic cylinder and substitute them into the operating parameters of the hydraulic cylinder, and then calculate the internal leakage q using the following leakage equation. L1 and external leakage q L2 :

[0016]

[0017] In the above formula, l y The radial contact length between the seal and the hydraulic cylinder is approximately equal to the circumference of the hydraulic cylinder's inner diameter; x h is the axial contact length between the seal and the hydraulic cylinder. δ Oil film thickness; μ c is the channel correction factor; μ is the kinematic viscosity; S is the piston rod stroke length; d is the piston rod diameter; w A w represents the maximum pressure gradient when the piston rod extends. E x represents the maximum pressure gradient during the piston rod's return stroke. L q represents the piston rod displacement; o q represents the leakage rate during the outer stroke of the hydraulic cylinder. i x is the return flow rate during the hydraulic cylinder's internal stroke. o x represents the external stroke displacement; i For internal stroke displacement;

[0018] (2) The relationship between oil film thickness and contact pressure gradient is obtained through the one-dimensional Reynolds fluid equation:

[0019]

[0020] In the above formula, Indicates the pressure gradient; The film thickness at the point of maximum pressure is calculated using the following formula:

[0021]

[0022] (3) using the cubic equation root formula to solve the equation of the above step, obtaining an expression of the oil film thickness h(p f ) related to the contact pressure;

[0023] (4) substituting the expression of the oil film thickness into the leakage equation, obtaining the required leakage model q L (p f ).

[0024] As a further improvement of the application, in the CFD simulation process of step two, the contact of the sealing element and the inner wall of the cylinder and the contact of the sealing element and the piston rod are simulated under different wear states by modifying the grid node parameters; and then the state parameters under different conditions are obtained.

[0025] As a further improvement of the application, in step two, the Archard equation is used to calculate the friction depth V w , and the formula is as follows:

[0026]

[0027] In the above formula, K w represents the wear coefficient, F l represents the contact pressure of the simulation network node, x w represents the slip distance of the grid node, and H m represents the material hardness set in the simulation process.

[0028] As a further improvement of the application, in step three, first, a wavelet packet tree is constructed by using the obtained pressure signal to select a wavelet packet basis function, and then features are extracted from the layer signals of the hydraulic cylinder inner and outer cavity pressure signals by wavelet packet decomposition, obtaining three types of characteristic values;

[0029] The calculation formula of the small packet energy potential P j,k is as follows:

[0030]

[0031] In the above formula, E j,k represents the energy of each wavelet packet sub-signal; and E total represents the energy of the entire signal.

[0032] The calculation formula of the small packet energy entropy H wavelet is as follows:

[0033]

[0034] The calculation formula of the small packet energy variance V wavelet is as follows:

[0035]

[0036] In the above formula, N represents the number of wavelet packet sub-signals decomposed; M wavelet represents the average value of the wavelet packet energy.

[0037] As a further improvement of the application, in step three, Sobol sensitivity analysis method is adopted, the main effect variance and first-order interaction effect variance of the input variables are calculated through variance decomposition method, and the Sobol index is calculated using the ratio of them to the total variance, so as to evaluate the signal features with the strongest correlation with the wear state.

[0038] As a further improvement of the application, in step four, the machine learning algorithms adopted in the process of generating the required data-driven model based on the wear eigenvalues and the corresponding friction depth data include: support vector machine SVM, extreme gradient boosting algorithm XGBoost, K nearest neighbor method KNN, and long short-term memory neural network LSTM.

[0039] The application also includes a digital twin system for hydraulic cylinder wear evolution and leakage monitoring, which is used to monitor the wear evolution and leakage state of the hydraulic cylinder by adopting the method for hydraulic cylinder wear evolution and leakage monitoring as described above, combined with digital twin technology, and visualize the monitoring results. The digital twin system includes a mechanism model creation module, a CFD simulation module, a feature extraction module, a proxy model construction module, and a virtual-real interaction module.

[0040] The mechanism model creation module is used to create a leakage model for representing the mapping relationship between the internal and external leakage of the hydraulic cylinder and the contact pressure according to the geometric model, the dynamics model, the one-dimensional Reynolds flow equation, and the leakage beam calculation equation of the hydraulic cylinder.

[0041] The CFD simulation module is used to continuously test the global contact pressure of the sealing contact surface under different network nodes and wear depth conditions through three-dimensional modeling and CFD simulation analysis, and then generate a pressure distribution model; and reduce the three-dimensional pressure distribution model to a one-dimensional model.

[0042] The feature extraction module is used to perform wavelet packet analysis on the collected hydraulic cylinder rodless cavity and rod cavity pressures, extract eigenvalues using the feature extraction model, and determine wear eigenvalues through sensitivity analysis; and then train a required data-driven model using a large number of experimentally obtained wear eigenvalues and their corresponding friction depth values.

[0043] The agent model construction module is used for fusing the obtained leakage model, pressure distribution model, characteristic value extraction model and data driven model, and performing data interpolation fitting through a machine learning training method.

[0044] The virtual-real interaction module includes a physical space and a twin space; the physical space includes a pressure sensor, a displacement sensor, an oil conveying pipe, a data transmission line, a signal acquisition card, an industrial computer and a display module.

[0045] In the digital twin system, the sensors of the physical space detect the pressure signal and the displacement signal of the hydraulic cylinder in real time, the pressure signal and the displacement signal are input into the industrial computer in real time through serial communication and are imported into the database software in the twin space, data connection is realized through the interface with the digital twin software, data synchronous updating and interaction mapping between the physical space and the signal space are realized.

[0046] As a further improvement of the present application, the preset wear updating time t r to realize the timing updating of the wear depth, and by setting the leakage safety threshold q r to realize the leakage abnormality alarm.

[0047] As a further improvement of the present application, the database software in the twin space adopts MySQL, and the visualization model of the hydraulic cylinder is created through Unity3D.

[0048] The technical scheme provided by the present application has the following beneficial effects:

[0049] 1. The present application obtains the contact pressure distribution of the sealing contact surface through high-fidelity simulation, and obtains the internal and external leakage of the hydraulic cylinder through a mechanism model; in addition, the signal characteristics and mechanism model input are obtained through easily installed pressure and displacement sensors, avoiding the problems of installation limitation and maintenance difficulty caused by embedded sensors.

[0050] 2、The present application carries out sensitive analysis on the signal characteristics of the decomposed signals of the import and export oil cavity pressure and displacement signals by adopting the wavelet packet analysis method, establishes a data-driven model between the signal characteristics and the wear depth value, thereby obtaining the real-time wear state, and updates in the virtual-real interaction module, improves the monitoring accuracy of the wear evolution state, and avoids the uncertainty of wear and leakage caused by the "back-off effect".

[0051] 3、The present application provides a kind of leakage fault monitoring method for wear evolution of hydraulic cylinder by digital twin technology, wear state is improved by real-time solution leakage and interval update, the perception ability of hydraulic cylinder to leakage fault and wear evolution is improved, so that the operating state of hydraulic cylinder can be identified more quickly and accurately. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 It is the step flow chart of the method for monitoring wear evolution and leakage of hydraulic cylinder provided in embodiment 1 of the present application.

[0053] Figure 2 It is the signal flow diagram of wear evolution monitoring model and leakage monitoring model constructed in embodiment 1 of the present application.

[0054] Figure 3 It is the module architecture diagram of the digital twin system for monitoring wear evolution and leakage of hydraulic cylinder provided in embodiment 2 of the present application.

[0055] Figure 4 It is the working principle diagram of the process of signal acquisition, data analysis and visualization of monitoring results realized in physical space and twin space in virtual-real interaction module in embodiment 2 of the present application. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0057] Embodiment 1

[0058] The present embodiment provides a method for monitoring wear evolution and leakage of hydraulic cylinder, which is used to generate monitoring results of wear depth V L and leakage q L from signals of real-time collected hydraulic cylinder rodless cavity pressure p1, rod cavity pressure p2 and displacement x w of piston rod, as shown in Figure 1 , the method for monitoring wear evolution and leakage of hydraulic cylinder comprises the following steps:

[0059] I. Based on the theoretical model of the current type of hydraulic cylinder, a model is constructed to characterize the leakage qL and contact pressure p of the hydraulic cylinder. f Leakage model of inter-mapping relationship q L (p f In this embodiment, the leakage amount q L Divided into internal leakage amount q L1 and external leakage q L2 Two parts, namely: q L ={q L1 ,q L2}

[0060] In this embodiment, the process of constructing the leakage model specifically includes the following steps:

[0061] (1) Obtain the structural parameters of the hydraulic cylinder, such as the actual dimensions of the seals and piston chamber, and substitute them into the operating parameters of the hydraulic cylinder. Then, calculate the internal leakage q using the following leakage equation. L1 and external leakage q L2 :

[0062]

[0063] In the above formula, l y The radial contact length between the seal and the hydraulic cylinder is approximately equal to the circumference of the hydraulic cylinder's inner diameter; x h is the axial contact length between the seal and the hydraulic cylinder. δ Oil film thickness; μ c is the channel correction factor; μ is the kinematic viscosity; S is the piston rod stroke length; d is the piston rod diameter; w A w represents the maximum pressure gradient when the piston rod extends. E x represents the maximum pressure gradient during the piston rod's return stroke. L q represents the piston rod displacement; o q represents the leakage rate during the outer stroke of the hydraulic cylinder. i x is the return flow rate during the hydraulic cylinder's internal stroke. o x represents the external stroke displacement; i This refers to the internal stroke displacement.

[0064] (2) The relationship between oil film thickness and contact pressure gradient is obtained through the one-dimensional Reynolds fluid equation. This relationship can be transformed from the relationship between internal leakage and oil film thickness to the relationship with contact pressure gradient:

[0065]

[0066] In the above formula, Indicates the pressure gradient; The film thickness at the point of maximum pressure is calculated using the following formula:

[0067]

[0068] (3) Using the cubic equation root formula to solve the equation of the above step, obtaining the expression of the oil film thickness h(p f ) related to the contact pressure;

[0069] (4) Substituting the expression of the oil film thickness into the leakage equation, obtaining the required leakage model q L (p f ). The leakage model can be used to represent the mapping relationship between the internal leakage q L1 or the external leakage q L2 and the contact pressure p f .

[0070] II. Create a three-dimensional model of the current type of hydraulic cylinder, use the three-dimensional model to perform CFD simulation analysis on the hydraulic cylinder, and calculate the wear depth V w of each state based on the Archard equation during the simulation process; thereby obtaining a pressure distribution model p f (x, V w ) representing the global contact pressure distribution of the sealing surface of each network node x under different wear depths V w , and reducing it to a one-dimensional model.

[0071] In the embodiment, the three-dimensional model of the hydraulic cylinder can be created using existing various types of three-dimensional modeling software, and then imported into the CFD simulation software. During the CFD simulation process, the technician can continuously modify the grid node parameters of the sealing element and the inner wall of the cylinder, and the grid node parameters of the sealing element and the piston rod, to simulate the contact of the sealing element under different wear states; and then obtain the state parameters under different conditions.

[0072] During the CFD simulation process, the Archard equation is used to calculate the friction depth V w corresponding to each group of state parameters, and the calculation formula is as follows:

[0073]

[0074] In the above formula, K w represents the wear coefficient, F l represents the contact pressure of the simulation network node, x w represents the slip distance of the grid node, and H m represents the material hardness set in the simulation process.

[0075] Combined with the simulation data and the calculated friction depth V w , a pressure distribution model p f (x, V w ) representing the global contact pressure distribution of the sealing surface of each network node x under different wear depths V wThe pressure distribution model of the global contact pressure distribution of the seal contact surface under the condition is a three-dimensional parameter model because the pressure distribution model in the embodiment is obtained by using three-dimensional model simulation data. In actual application, the three-dimensional parameter model is reduced to a one-dimensional parameter model for storage.

[0076] III. The two collected pressure signals are subjected to wavelet packet analysis, and the wavelet packet energy entropy H wavelet , the wavelet packet variance V wavelet and the wavelet packet energy potential P j,k are extracted as candidate characteristic values. Then, the candidate characteristic value with the strongest correlation with the hydraulic cylinder wear state evaluated by the Sobol sensitivity analysis method is taken as the wear characteristic value WT; and the corresponding characteristic value extraction model is retained.

[0077] In the embodiment, considering that it is difficult to directly realize the prediction of the wear depth by directly using the pressure signal, the pressure signal containing other signal characteristics is extracted and analyzed by wavelet packet analysis in the embodiment. Specifically, the pressure signals of the hydraulic cylinder of the different seals in their respective wear states are collected in advance, a wavelet packet tree is constructed by selecting a wavelet packet basis function, and signal extraction is performed on the signal decomposition layers of the inner and outer cavity pressure signals of the hydraulic cylinder by wavelet packet decomposition. Specifically, the wavelet packet energy entropy H wavelet , the wavelet packet variance V wavelet and the wavelet packet energy potential P j,k of the three characteristic parameters are extracted as candidate characteristic values; the calculation formulas of the three are as follows:

[0078] The calculation formula of the wavelet packet energy potential P j,k is as follows:

[0079]

[0080] In the above formula, E j,k represents the energy of each wavelet packet sub-signal; E total represents the energy of the entire signal.

[0081] The calculation formula of the wavelet packet energy entropy H wavelet is as follows:

[0082]

[0083] The calculation formula of the wavelet packet energy variance V wavelet is as follows:

[0084]

[0085] In the above formula, N represents the number of wavelet packet sub-signals decomposed; M wavelet represents the average value of the wavelet packet energy.

[0086] Based on the extracted three candidate feature values, the embodiment adopts Sobol sensitivity analysis method to analyze the correlation between each feature value and the friction depth index. Specifically, first, the main effect variance and the first-order interaction effect variance of the input variable are calculated by the variance decomposition method, and the Sobol index is calculated using the ratio of the total variance, so as to evaluate the signal feature with the strongest correlation with the wear state.

[0087] After evaluating the signal feature with the strongest correlation, the feature value is directly used as the wear feature value WT in the embodiment, and in subsequent work, the mapping relationship between the wear feature value and the wear depth is constructed, so as to realize the prediction of the wear depth according to the pressure signal or the wear feature value. It needs to be particularly emphasized that after selecting the wear feature value, the signal extraction tool for extracting the corresponding feature value is also saved as the required feature value extraction model. The feature value extraction model in the embodiment is essentially a signal processing tool that can realize wavelet packet decomposition.

[0088] Four, a large number of real pressure test data of the current type of hydraulic cylinder in different wear states are collected, and wear feature value samples are generated according to the pressure test data; then a data-driven model V w is trained by using a machine learning algorithm, which takes the wear feature value WT as input and takes the wear depth V w (WT) as output.

[0089] In step three, the wear feature value with the strongest correlation with the wear depth has been determined, so in the embodiment, in order to create the required data-driven model, a large amount of sample data of wear feature values is needed, and the pressure signal of the hydraulic cylinder in the non-wear state is obtained by actual test in the embodiment, and the wear depth corresponding to each pressure signal is recorded, and the wear depth is taken as the label of the pressure signal. Then the feature value of the pressure signal is extracted to obtain the corresponding wear feature value. At this time, a large number of wear feature values with wear depth labels are obtained in the embodiment.

[0090] Next, by using various existing machine learning algorithms, a data-driven model representing the mapping relationship between the wear feature value and the wear depth can be generated. By using the obtained data-driven model, a wear feature value is input, and a corresponding wear depth prediction result can be obtained. In the embodiment, the machine learning algorithms used in the process of generating the required data-driven model based on the wear feature value and the corresponding wear depth data include support vector machine SVM, extreme gradient boosting algorithm XGBoost, K nearest neighbor method KNN, and long short-term memory neural network LSTM, etc.

[0091] Fifth, the eigenvalue extraction model and the data-driven model are fused to obtain a wear evolution monitoring model; and the leakage model, pressure distribution model, eigenvalue extraction model and data-driven model are fused together, and machine learning algorithms are used to perform data interpolation and fitting on the discrete parts, thereby obtaining a leakage monitoring model.

[0092] In the four steps described above in this embodiment, four basic models for data analysis and processing were obtained: a leakage model, a pressure distribution model, a feature extraction model, and a data-driven model. The leakage model, in particular, is a parameter q representing the leakage amount of the hydraulic cylinder, derived from the physical and dynamic models of the hydraulic cylinder. L With contact pressure p f A mathematical model of the mapping relationship between the piston rod and the piston. This model can be based on the input piston rod displacement x. L and contact pressure p f This outputs a corresponding leakage amount. In fact, the leakage amount q output by the leakage model... L Divided into internal leakage amount q L1 and external leakage q L2 Two parts.

[0093] The pressure distribution model is another parametric model constructed in this embodiment using state parameters obtained through 3D modeling and CFD simulation. The input to this parametric model is the wear depth V of the hydraulic cylinder. w The output is V for different wear depths. w The global contact pressure p of the sealing contact surface f .

[0094] The eigenvalue extraction model is a signal processing tool that performs wavelet packet analysis and decomposition on pressure signals. This tool can convert pressure signals into corresponding wear characteristic values, which are the wavelet packet energy entropy H extracted from the pressure signal. wavelet Wavelet packet variance V wavelet and wavelet packet energy potential P j,k The eigenvalue is the one most strongly correlated with the friction depth. That is, the input to the eigenvalue extraction model is the pressure signal, including the rodless chamber pressure p1 and the rod chamber pressure p2 of the hydraulic cylinder; the output is the corresponding wear eigenvalue.

[0095] The data-driven model, constructed in this embodiment, is a parametric model that can output the corresponding wear depth result based on the input wear feature values. The data-driven model constructed in this embodiment uses a large amount of validated, measured data, thus achieving high prediction accuracy. The wear feature values ​​input to the data-driven model are those extracted from the pressure signal using a feature value extraction model.

[0096] Therefore, as shown in Figure 2 the embodiment, the characteristic value extraction model and the data-driven model are cascaded, the data processing channel of "pressure signal → wear characteristic value → wear depth" is opened, and a required wear evolution monitoring model is obtained.

[0097] The characteristic value extraction model, the data-driven model, the pressure distribution model, and the leakage model can also be cascaded in sequence to open the data processing channel of "pressure signal → wear characteristic value → wear depth → contact pressure + displacement signal → leakage amount", and a required leakage amount monitoring model is further obtained.

[0098] It should be particularly emphasized that in the fusion process of the wear evolution monitoring model and the leakage amount monitoring model, the creation conditions of each basic model are different, and the parameters of each basic model also differ, so they cannot be directly associated and applied. There may be data abnormalities or conflicts, resulting in the problem of being unable to output correct results. In order to solve this problem, after the various basic models are associated with each other, the training strategy in the machine learning algorithm is used to jointly train the various basic models, and then the method of combining neural networks and linear interpolation is used to perform data interpolation and fitting on the missing parts of the fused parameter model. Further, a required wear evolution monitoring model and a required leakage amount monitoring model with complete functions are obtained.

[0099] Six, input the signals of the real-time collected pressure p1 of the rodless cavity of the hydraulic cylinder, the pressure p2 of the rod cavity, and the displacement x of the piston rod L to the surrogate model containing the wear evolution monitoring model and the leakage amount monitoring model, and output the corresponding wear depth V w and the leakage amount Q.

[0100] Embodiment 2

[0101] The embodiment provides a digital twin system for monitoring the wear evolution and leakage of a hydraulic cylinder, which is used to monitor the wear evolution and leakage state of the hydraulic cylinder by adopting the method for monitoring the wear evolution and leakage of the hydraulic cylinder as in embodiment 1 and combining digital twin technology, and visualize the monitoring results.

[0102] As shown in Figure 3 the embodiment, the digital twin system provided by the embodiment includes a mechanism model creation module, a CFD simulation module, a feature extraction module, a surrogate model construction module, and a virtual-real interaction module.

[0103] The mechanism model creation module is used to create a leakage model for representing the mapping relationship between the internal and external leakage amounts and the contact pressure of the hydraulic cylinder according to the geometric model, the dynamics model, the one-dimensional Reynolds flow equation, and the leakage amount calculation equation of the hydraulic cylinder.

[0104] That is, the mechanism model creation module is used to implement the relevant tasks in step one of the method for monitoring the wear evolution and leakage of hydraulic cylinders in Example 1.

[0105] The CFD simulation module is used to continuously test the global contact pressure of the sealing contact surface under different network nodes and wear depths through 3D modeling and CFD simulation analysis, thereby generating a pressure distribution model; and reducing the 3D pressure distribution model to a 1D model.

[0106] That is, the CFD simulation module is used to implement the relevant tasks in step two of the method for monitoring the wear evolution and leakage of hydraulic cylinders in Example 1.

[0107] The feature extraction module is used to perform wavelet packet analysis on the collected pressure of the rodless and rod chambers of the hydraulic cylinder, extract feature values ​​using the feature extraction model, and perform sensitivity analysis on the features to determine wear feature values; then, a required data-driven model is trained using the wear feature values ​​and their corresponding friction depth values ​​obtained from a large number of experiments.

[0108] That is, the feature extraction module is used to implement the relevant tasks in steps three and four of the method for monitoring the wear evolution and leakage of hydraulic cylinders in Example 1.

[0109] The proxy model building module is used to fuse the acquired leakage model, pressure distribution model, feature extraction model, and data-driven model, and to perform data interpolation and fitting through machine learning training methods. This results in a wear evolution monitoring model that takes pressure signal as input and wear depth as output; and a leakage monitoring model that takes pressure signal and displacement signal as input and leakage amount as output.

[0110] That is, the proxy model construction module is used to implement the relevant tasks in step five of the method for monitoring the wear evolution and leakage of hydraulic cylinders in Example 1.

[0111] The virtual-real interaction module is the key to realizing the digital twin system in this embodiment, such as... Figure 4 As shown, the virtual-real interaction module includes a physical space and a twin space. The physical space includes pressure sensors, displacement sensors, oil pipes, data transmission lines, signal acquisition cards, industrial control computers, and display modules. The hydraulic cylinder is the monitoring object in this embodiment. There are two pressure sensors: pressure sensor 1 and pressure sensor 2. Pressure sensor 1 is installed on the oil pipe on the rodless side of the hydraulic cylinder to measure the pressure p1 in the rodless side. Pressure sensor 2 is installed on the oil pipe on the rod side of the hydraulic cylinder to measure the pressure p2 in the rod side. The displacement sensor is installed on the piston rod of the hydraulic cylinder to measure the displacement x of the piston rod. L .

[0112] The pressure sensor and the displacement sensor are connected to the data acquisition card through the data transmission line, the data acquisition card collects signals according to a preset sampling frequency, and then the collected pressure signals and displacement signals are sent to the twin space for data analysis.

[0113] The twin space includes database software, a leakage monitoring model and a wear evolution monitoring model imported from the agent model construction module, and a three-dimensional hydraulic cylinder visualization model in the digital twin software. In the digital twin system, the sensors in the physical space detect the pressure signals and displacement signals of the hydraulic cylinder in real time, the pressure signals and displacement signals are input into the industrial computer in real time through serial communication and are imported into the database software in the twin space, data connection is realized through the interface with the digital twin software, and data synchronous updating and interactive mapping of the physical space and the signal space are realized. The database software in the embodiment adopts MySQL; the digital twin software adopts Unity3D. Data connection with the digital twin software Unity3D is realized through MySQL Connector / NET.

[0114] The database software inputs the acquired pressure signals and displacement signals into the wear evolution monitoring model and the leakage monitoring model, and the wear evolution monitoring model and the leakage monitoring model respectively predict the wear depth and the leakage of the hydraulic cylinder according to the input data. The visualization model is used for visualizing the running state of the hydraulic cylinder, the predicted wear depth and the predicted leakage, and outputting to the display module in the physical space for display.

[0115] In addition, in the digital twin system provided in the embodiment, a wear update time t r is preset to realize the timing update of the wear depth. In the digital twin system, a leakage safety threshold q r can also be set to realize leakage abnormality alarm, and when the digital twin system detects that the leakage exceeds the safety threshold, the corresponding alarm is issued through the display module in the physical space or other alarm.

[0116] At the same time, during each maintenance, the technician can obtain some data related to the equipment parameters and the running state of the hydraulic cylinder. Including the real leakage and the wear depth and the like. At this time, the technician also feeds back the measured state parameters to the agent model construction module to realize parameter correction of the wear evolution monitoring model and the leakage monitoring model fused by the agent model construction model. Further improve the prediction accuracy of the wear evolution monitoring model and the leakage monitoring model.

[0117] The mechanism model creation module, the CFD simulation module, the feature extraction module, the proxy model construction module, and the virtual space part of the virtual-real interaction module in the digital twin system for monitoring the wear evolution and leakage of the hydraulic cylinder provided in the embodiment are essentially a computer device for realizing data processing and instruction generation, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0118] The computer device provided in the embodiment can be an intelligent terminal, a tablet computer, a notebook computer, a desktop computer, a rack server, a blade server, a tower server, or a cabinet server (including a standalone server or a server cluster composed of multiple servers), etc. The computer device of the embodiment at least includes but is not limited to a memory and a processor which can be connected to each other for communication through a system bus.

[0119] In the embodiment, the memory (i.e., a readable storage medium) includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory can be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device.

[0120] In other embodiments, the memory can also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device. Of course, the memory can also include both the internal storage unit and the external storage device of the computer device. In the embodiment, the memory is generally used to store the operating system and various application software installed on the computer device, etc. In addition, the memory can also be used to temporarily store various data that have been output or will be output.

[0121] The processor in some embodiments can be a central processing unit (CPU), a graphics processing unit (GPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor is generally used to control the overall operation of the computer device. In the embodiment, the processor is used to run the program code or process the data stored in the memory.

[0122] The above merely describes preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement and improvement within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method of hydraulic cylinder wear evolution and leakage monitoring, characterized by, for generating a monitoring result of a wear depth V w and a leakage amount q L from signals of a real-time acquired rodless chamber pressure p 1, a rod chamber pressure p 2, and a displacement of a piston rod x L of a hydraulic cylinder; the method for monitoring the wear evolution and the leakage of the hydraulic cylinder comprising the following steps: I. Based on the current type of hydraulic cylinder theoretical model, a model is constructed for representing the internal leakage of hydraulic cylinder q L1 and external leakage q L2 and contact pressure p f mapping relationship between the leakage model q L ( p f ); q L ={ q L1 , q L2}; 2. Create a 3D model of the current type of hydraulic cylinder, perform CFD simulation analysis on the hydraulic cylinder using the 3D model, and calculate the wear depth for each state based on the Arcard equation during the simulation process. V w This allows us to obtain a representation of each network node. x At different wear depths V w Pressure distribution model of global contact pressure distribution on the sealing contact surface p f ( x , V w ), and reduce it to a one-dimensional model; III. Wavelet packet analysis is performed on the two collected pressure signals, and the wavelet packet energy entropy is extracted H wavelet Wavelet packet energy variance V wavelet Wavelet packet energy potential P j,k As a candidate feature value, and then the candidate feature value with the strongest correlation with the hydraulic cylinder wear state after evaluation by the Sobol sensitivity analysis method is used as the wear feature value WT ; and retain the corresponding feature value extraction model; Four, collect a large number of current type hydraulic cylinder in different wear state of the real pressure test data, and according to the pressure test data to generate wear characteristic value sample; then use machine learning algorithm to train a wear characteristic value WT As input, wear depth V w Data driven model V w ( WT ) as output Five, the feature value extraction model and the data driven model are fused to obtain a wear evolution monitoring model, the leakage model, the pressure distribution model, the feature value extraction model and the data driven model are fused to obtain a leakage quantity monitoring model; and the discrete parts in the fused model are data interpolated and fitted by using the joint training strategy in the machine learning algorithm; Six, the real-time collected hydraulic cylinder rodless cavity pressure p 1, the pressure of the rod cavity p 2 and the displacement of the piston rod x L The signal is input into the agent model containing the wear evolution monitoring model and the leakage monitoring model, and the corresponding wear depth V w And the leakage q L .

2. The method of hydraulic cylinder wear evolution and leakage monitoring of claim 1, wherein: In step one, the construction process of the leakage model is as follows: (1) Obtain the structure parameters of the hydraulic cylinder and substitute them into the operating parameters of the hydraulic cylinder, and then calculate the internal leakage quantity through the following leakage quantity equation q L1 and the external leakage quantity q L2 : In the above formula, l y is the radial contact length of the seal with the hydraulic cylinder, approximately equal to the circumference of the hydraulic cylinder inner diameter; l x is the axial contact length of the seal with the hydraulic cylinder; is the oil film thickness; is the passage correction factor; is the kinematic viscosity; S is the piston rod stroke length; d is the piston rod diameter; w A is the maximum pressure gradient when the piston rod is extended, w E is the maximum pressure gradient when the piston rod is retracted, x L is the piston rod displacement; q o is the hydraulic cylinder outer stroke leakage; q i is the hydraulic cylinder inner stroke return; x o is the outer stroke displacement; x i is the inner stroke displacement; (2) Obtain the corresponding relationship between the oil film thickness and the contact pressure gradient through the one-dimensional Reynolds fluid equation: ； In the above formula, represents the pressure gradient; is the film thickness at the maximum pressure point, and the calculation formula is as follows: ； (3) using cubic equation root formula to solve the equation of the above step, obtaining an expression of the oil film thickness related to the contact pressure ; (4) Substitute the expression of oil film thickness into the leakage equation to obtain the required leakage model q L ( p f ).

3. The method of hydraulic cylinder wear evolution and leakage monitoring of claim 1, wherein: In the CFD simulation process of step two, the contact of the sealing element and the inner wall of the cylinder and the grid node parameters of the sealing element and the piston rod are modified to simulate the sealing element under different wear conditions; and then the state parameters under different conditions are obtained.

4. The method of hydraulic cylinder wear evolution and leakage monitoring of claim 1, wherein: The wear depth is calculated in step two using the Archard equation V w The formula is as follows: ； in the above formulae, K w denotes the wear coefficient, F l denotes the contact pressure of the simulation network node, x w denotes the slip distance of the grid node, H m denotes the material hardness set by the simulation process.

5. The method of hydraulic cylinder wear evolution and leakage monitoring of claim 1, wherein: In step three, first, the wavelet packet tree is constructed by using the obtained pressure signal to select the wavelet packet basis function, and then the features are extracted from the layer signals of the hydraulic cylinder inner and outer cavity pressure signals by wavelet packet decomposition to obtain three types of feature values; The wavelet packet energy potential is calculated by the following formula: P j,k The calculation formula is as follows: ； In the above formula, E j,k denotes the energy of each wavelet packet sub-signal; E total denotes the energy of the entire signal; The wavelet packet energy entropy H wavelet The calculation formula is as follows: ； The wavelet packet energy variance V wavelet The calculation formula is as follows: ； In the above formula, N represents the number of wavelet packet sub-signals decomposed; M wavelet represents the average value of the wavelet packet energy.

6. The method of hydraulic cylinder wear evolution and leakage monitoring of claim 1, wherein: In step three, the Sobol sensitivity analysis method is adopted, the main effect variance and the first-order interaction effect variance of the input variables are calculated by the variance decomposition method, and the Sobol index is calculated by using the ratio of them to the total variance, so as to evaluate the signal features with the strongest correlation with the wear state.

7. The method of hydraulic cylinder wear evolution and leakage monitoring of claim 1, wherein: In step four, the machine learning algorithms used in the process of generating the required data driven model based on the wear feature values and the corresponding wear depth data include: support vector machine SVM, extreme gradient boosting algorithm XGBoost, K nearest neighbor method KNN, and long short-term memory neural network LSTM.

8. A digital twin system for hydraulic cylinder wear evolution and leakage monitoring, characterized by: It is used for the hydraulic cylinder wear evolution and leakage monitoring method as claimed in any one of claims 1-7, which combines digital twin technology to realize monitoring of the wear evolution and leakage state of the hydraulic cylinder, and visualizes the monitoring results; the digital twin system comprises: A mechanism model creation module is used to create a leakage model for representing the mapping relationship between the inner and outer leakage quantities of the hydraulic cylinder and the contact pressure according to the geometric model, the dynamics model, the one-dimensional Reynolds flow equation, and the leakage beam calculation equation of the hydraulic cylinder; A CFD simulation module is used to continuously test the global contact pressure of the sealing contact surface under different network nodes and wear depth conditions by means of three-dimensional modeling and CFD simulation analysis, and then generate a pressure distribution model; and the three-dimensional pressure distribution model is reduced to a one-dimensional model; A feature extraction module is used to perform wavelet packet analysis on the collected hydraulic cylinder rodless cavity and rod cavity pressure, extract feature values by using a feature extraction model, and determine wear feature values by performing sensitivity analysis on the features; and then a required data driven model is trained by using a large number of experimentally obtained wear feature values and their corresponding wear depth values; The agent model construction module is used for fusing the obtained leakage model, pressure distribution model, eigenvalue extraction model and data-driven model, and performing data interpolation fitting through a machine learning training method; and then obtains a wear evolution monitoring model taking pressure signals as input and wear depth as output, and a leakage amount monitoring model taking pressure signals and displacement signals as input and leakage amount as output. The virtual-real interaction module includes a physical space and a twin space; the physical space includes a pressure sensor, a displacement sensor, an oil pipeline, a data transmission line, a signal acquisition card, an industrial computer, and a display module; and the twin space includes a database software, the imported leakage amount monitoring model and wear evolution monitoring model, and a three-dimensional hydraulic cylinder visualization model in the digital twin software. In the digital twin system, the sensors of the physical space detect the pressure signals and displacement signals of the hydraulic cylinder in real time, the pressure signals and displacement signals are input into the industrial computer in real time through serial communication and are imported into the database software in the twin space, data connection is performed through the interface with the digital twin software, data synchronous updating and interaction mapping between the physical space and the signal space are realized, the wear depth and leakage amount of the hydraulic cylinder are predicted through the wear evolution monitoring model and the leakage amount monitoring model in the twin space, and visual output is performed through the visualization model.

9. The digital twin system for hydraulic cylinder wear evolution and leakage monitoring of claim 8, wherein: The preset wear updating time in the digital twin system t r To realize the timing update of the wear depth, and realize the leakage abnormality alarm by setting the leakage amount safety threshold q r At the same time, during each maintenance, the measured parameters are also fed back to the agent model construction module to realize the parameter correction of the fused wear evolution monitoring model and the leakage amount monitoring model.

10. The digital twin system for hydraulic cylinder wear evolution and leakage monitoring of claim 8, wherein: The database software in the twin space adopts MySQL, and the visualization model of the hydraulic cylinder is created through Unity3D.

Images