Friction pair wear state observation method and system based on online abrasive particle detection

Abstract

The invention belongs to the field of fault diagnosis, and particularly discloses a friction pair wear state observation method and system based on online abrasive particle detection. According to the method, the dynamic modeling method on the macroscale and the model parameter estimation method combining the abrasive particle information and the machine learning thought are used, so that the initial state and the model parameters of the model can be further obtained on the basis of the theoretical model; according to the wear state observation method, the limitation that a microscopic model is prone to falling into actual wear complexity can be overcome, and a macrodynamic model capable of representing the actual wear state can be obtained through data-driven parameter estimation, so that the wear process of a mechanical system is diagnosed in real time by utilizing online abrasive particle detection information; and the purpose of tracking and predicting the wear state is achieved.

Description

Technical Field

[0001] The present application belongs to the field of fault diagnosis, and more specifically, relates to a method and system for observing the wear state of a friction pair based on online abrasive particle detection, which is suitable for scenarios where online abrasive particle detection information is used to diagnose the wear process of a mechanical system. Background Art

[0002] Faced with increasingly complex large-scale equipment, effective fault diagnosis technology is currently lacking. This has two profound consequences: First, faults cannot be detected promptly, leading to safety hazards during operation; second, the lack of monitoring data during the fault generation process makes it difficult to trace the fault formation mechanism, hindering effective material improvement, design optimization, and process enhancement to improve reliability and lifespan. In mechanical systems, friction pair wear is one of the main causes of mechanical failure. Abrasive particles are a key product of friction pair wear and are closely related to the condition of the friction pair. With the advancement of wear particle detection methods, wear particle characteristics in mechanical systems have become an important basis for fault diagnosis. In particular, large-scale equipment is often complex mechatronics products, with fluid lubrication and hydraulic transmission systems similar to the human cardiovascular system. Abrasive particles, like metabolic waste in the blood, are closely related to the condition of the friction pair. Therefore, utilizing wear particle characteristics to perform fault diagnosis and lifespan prediction for mechanical equipment is particularly important for the rise of advanced manufacturing and the development of high-tech industries.

[0003] Wear particle detection can be divided into three stages based on the detection method: simple offline detection, offline detection based on instrumental analysis, and real-time online detection. Simple offline detection collects wear particles through a filter device. Further, by periodically weighing the accumulated mass of the wear particles or measuring their size using a microscope, it can indirectly determine whether the mechanical system is functioning properly and thus assess its health. However, since weighing and size measurement are relatively crude and cannot determine the composition and quantity of the wear particles, this method is not very effective for assessment and prediction. To address these issues, instrumental detection methods such as spectrometers and ferrography have emerged. Spectroscopic technology uses the principle of interference to identify the composition and quantity of wear particles in oil, thereby determining the wear condition of the mechanical system. Ferrography utilizes a strong magnetic field with a high gradient to deposit wear particles in the oil in an orderly manner according to particle size. Other analytical methods are then used to determine the number, size, morphology, and composition of the wear particles, thereby understanding the wear condition of the system. Although currently advanced spectrometers and ferrography offer advantages such as fast analysis speed and high accuracy, and can identify a variety of wear particle materials, both methods are offline wear particle analysis methods and cannot meet the needs of real-time monitoring. Real-time detection of wear particle generation can effectively reveal the occurrence and development of faults. Therefore, online fault diagnosis technology has become an important branch of the fault diagnosis field. At present, online wear particle detection can provide information on wear particles throughout the life cycle of mechanical systems, reflect the development process of wear, and provide guarantees for reliable operation and on-demand maintenance. However, due to the lack of support from theoretical models, the setting of judgment criteria needs to rely on a lot of engineering experience. At the same time, due to the lack of in-depth research on the laws of wear particle generation, it is still difficult to effectively reveal the wear process of mechanical systems based on online wear particle detection information. The problem of mutual interference between wear particles from different friction pairs further exacerbates the difficulty of evaluating the wear process. How to reveal the evolution of wear from online wear particle information remains an unresolved problem. Summary of the Invention

[0004] In response to the defects of the existing technology, the purpose of this application is to provide a method and system for observing the wear state of friction pairs based on online abrasive particle detection, aiming to solve the problem that the existing technology lacks a model to describe the laws of macroscopic abrasive generation in mechanical systems and a method for observing the wear state using online abrasive particle information.

[0005] In a first aspect, the present application provides a method for observing the wear state of a friction pair based on online abrasive particle detection, comprising: The initial state and parameters of the kinetic model of wear particle generation are dynamically fitted through the real-time online detected wear particle number sequence; A plurality of samples are randomly selected in the actual state space and the actual parameter space, each sample is composed of a state variable and a parameter, and the wear state evolution pattern existing in the actual parameter space is determined; Calculate the fitness of each sample for the wear particle number sequence detected online, where the fitness reflects the degree of consistency between the wear particle generation rates of the two; Find the optimal sample that satisfies the highest degree of compliance; Based on the parameters of the optimal sample, critical parameter perturbation is used to analyze the influence of each parameter on the wear evolution pattern.

[0006] Preferably, the kinetic model of abrasive particle generation is constructed in the following manner: (1) Based on the conservation of energy and mass on a macroscopic time scale, the state variables of the wear process are abstracted from three aspects: power loss, friction pair characteristics, and wear debris characteristics. The state variables include the mean value of the friction pair roughness, the standard deviation of the friction pair roughness, the mass of the abrasive particles, and the power loss of each type of wear behavior; (2) Abstracting multiple factors affecting wear into model parameters, including the mean effect coefficient, standard deviation effect coefficient, mass effect coefficient, and loss power correlation coefficient of each type of wear behavior; (3) Based on the linear damage mechanism and using differential equations to describe the dynamic relationship between state variables and model parameters, a dynamic model of wear particle generation is established.

[0007] Preferably, when only rough wear, fatigue wear and adhesive wear are considered, the state space is specifically as follows:

[0008] in, is the adhesion loss power, is the fatigue loss power, is the roughness loss power, and the three are state variables of wear. is the parameter matrix of the state equation, , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, , , are the roughness mean effect coefficients of rough wear, fatigue wear and adhesive wear, , , They represent the standard deviation effect coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the fatigue loss power correlation coefficient, is the roughness loss power correlation coefficient.

[0009] Preferably, a plurality of samples are randomly sampled by a Monte Carlo method.

[0010] Preferably, the fitness function is constructed based on the least squares method, as follows:

[0011] in, The abrasive particle generation rate is obtained based on the abrasive particle number sequence and is directly calculated from the online detected data. is the wear particle generation rate calculated based on the dynamic model, , is the adhesion loss power, is the fatigue loss power, is the roughness loss power, and the three together constitute the state variable of wear , , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the roughness loss power correlation coefficient, is the morphology factor of the abrasive particle, which is the ratio of the bottom area to the height of the abrasive particle. is the time length of the wear particle number sequence.

[0012] Preferably, the influence of each parameter on the wear evolution pattern is analyzed by using critical parameter perturbation, specifically as follows: Put the system in a critical state and perturb the model parameters in sequence to analyze the effect of single parameter change on power loss, friction pair roughness characteristics, and wear particle characteristics; Based on the perturbation results, the effects of the mean action coefficient of rough wear, the mean action coefficient of fatigue wear, the mean action coefficient of adhesive wear, the standard deviation action coefficient of rough wear, the standard deviation action coefficient of adhesive wear, the fatigue wear power coefficient, the mass action coefficient, the standard deviation action coefficient of fatigue wear, the rough wear power factor coefficient and the adhesive wear power factor coefficient on the development direction and development rate of the loss power and the roughness characteristics of the friction pair are analyzed.

[0013] Preferably, the method further includes determining the stability of the mechanical system, specifically as follows: (1) Obtain the dynamic model parameters of abrasive particle generation and generate the closed-loop transfer function characteristic equation of the system; (2) Construct the corresponding Routh array according to the characteristic equation; (3) Determine the system stability based on Routh criterion.

[0014] Preferably, the characteristic equation is expressed as follows:

[0015] in,

[0016]

[0017]

[0018] in, , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, , , are the roughness mean effect coefficients of rough wear, fatigue wear and adhesive wear, , , They represent the standard deviation effect coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the fatigue loss power correlation coefficient, is the roughness loss power correlation coefficient, is the Laplace operator, is the coefficient of the characteristic equation.

[0019] Preferably, the conditions that the state parameters need to meet when the system is stable are:

[0020] When the system is unstable, the parameters must meet any of the following eleven conditions:

[0021]

[0022]

[0023]

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030]

[0031] The system is in a critically stable state. There are two possible combinations of model parameters: when hour:

[0032] when and hour:

[0033] in, Represents a small positive number, used to replace the value 0.

[0034] In the second aspect, the present application provides a friction pair wear state observation system based on online abrasive particle detection, comprising: at least one memory for storing programs; at least one processor for entering the program stored in the memory, and when the program stored in the memory is entered, the processor is used to enter the state observation method as described in the first aspect.

[0035] It can be understood that the beneficial effects of the second aspect mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0036] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the existing technologies: (1) This application proposes a method for observing the wear state of friction pairs based on online abrasive particle detection, focusing on the conservation of energy and mass at a macroscopic time scale. By abstracting the state variables of wear loss power, friction pair characteristics, and abrasive particle characteristics, the abstract core state variables can avoid being trapped in the complexity of real wear. Then, differential equations are used to characterize the dynamic relationship between the core state variables, and a wear dynamics model is established, which ultimately solves the problem of describing the wear evolution behavior at a macroscopic time scale.

[0037] (2) This application proposes a method for observing the wear state of a friction pair based on online abrasive particle detection. By designing a model parameter estimation method for the abrasive particle time series, the initial state and model parameters are identified online through wear data. Further drawing on the current popular machine learning ideas, the error between the model output and the actual detection data is used as the evaluation function. Through advanced machine learning algorithms, the model parameters and initial state are continuously iteratively searched in the possible solution space in reality, and the optimal model parameters are finally obtained, thereby realizing the diagnosis of the wear process of a specific object. This innovation overcomes the limitations of traditional methods and provides a more accurate dynamic evaluation method that is suitable for practical application scenarios.

[0038] (3) This application proposes a method for observing the wear state of a friction pair based on online abrasive particle detection. By using a dynamic modeling method at a macroscopic scale and a model parameter estimation method combining abrasive particle information and machine learning ideas, the initial state and model parameters of the model can be further obtained on the basis of the theoretical model. This wear state observation method can overcome the limitation that the microscopic model is prone to fall into the complexity of actual wear. Through data-driven parameter estimation, a macroscopic dynamic model that can characterize the actual wear state is obtained, thereby realizing the real-time use of online abrasive particle detection information to diagnose the wear process of the mechanical system, and achieving the purpose of tracking and predicting the wear state.

[0039] (4) This application proposes a method for observing the wear state of friction pairs based on online abrasive particle detection. The wear evolution process is simulated by Monte Carlo simulation within a reasonable model parameter range, and the influence of changes in various factors on the wear evolution is analyzed by perturbation near the critical parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a flow chart of a method for observing the wear state of a friction pair based on online abrasive particle detection provided in an embodiment of the present application.

[0041] Figure 2 This is a diagram showing the evolution of the wear system provided in an embodiment of the present application.

[0042] Figure 3 is a result diagram of the wear evolution mode ratio provided by the embodiment of the present application; Figure 4 is a result diagram of the critical parameter perturbation provided by the embodiment of the present application; Figure 5 It is a schematic diagram of the optimal parameter estimation provided by the embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0044] The term "and / or" in this application describes an association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, or B exists alone. The symbol " / " in this application indicates that the associated objects are in an "or" relationship, for example, A / B means A or B.

[0045] In this application, the terms "first" and "second" are used to distinguish different objects, rather than to describe a specific order of objects. For example, the terms "first response message" and "second response message" are used to distinguish different response messages, rather than to describe a specific order of response messages.

[0046] The term "electrical connection" in this application can be a direct circuit connection or signal transmission through a communication protocol.

[0047] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0048] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more, for example, multiple processing units means two or more processing units, etc.; multiple elements means two or more elements, etc.

[0049] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0050] like Figure 1 As shown, the present application provides a method for observing the wear state of a friction pair based on online abrasive particle detection, comprising: The initial state and parameters of the kinetic model of wear particle generation are dynamically fitted through the real-time online detected wear particle number sequence; A plurality of samples are randomly selected in the actual state space and the actual parameter space, each sample is composed of a state variable and a parameter, and the wear state evolution pattern existing in the actual parameter space is determined; Calculate the fitness of each sample for the wear particle number sequence detected online, where the fitness reflects the degree of consistency between the wear particle generation rates of the two; Find the optimal sample that satisfies the highest degree of compliance; Based on the parameters of the optimal sample, critical parameter perturbation is used to analyze the influence of each parameter on the wear evolution pattern.

[0051] Preferably, the kinetic model of abrasive particle generation is constructed in the following manner: (1) Based on the conservation of energy and mass on a macroscopic time scale, the state variables of the wear process are abstracted from three aspects: power loss, friction pair characteristics, and wear debris characteristics. The state variables include the mean value of the friction pair roughness, the standard deviation of the friction pair roughness, the mass of the abrasive particles, and the power loss of each type of wear behavior.

[0052] Wear manifests itself at the microscopic level as surface fracture or detachment caused by the forces acting on the friction pair, while at the macroscopic level, it manifests as structural damage to the friction pair due to loss of transmission power. Although the wear evolution process is random and discrete in reality, it is governed by macroscopic dynamic mechanisms.

[0053] To describe the wear evolution mechanism, we first analyze the macroscopic wear behavior and relate its influencing factors at the dynamic level. From the perspective of abrasive particle generation, wear can be divided into three types: asperity wear, fatigue wear, and adhesive wear.

[0054] Rough wear usually occurs between a hard spike and a softer friction surface. When the spike lacks sufficient strength, it is peeled off and produces fragments; conversely, a strong spike will scratch the other surface, similar to cutting.

[0055] Fatigue wear is common on cyclically contacting friction surfaces, such as bearings and gears. Cyclic stress causes fatigue in the friction surface, resulting in surface damage characterized by pits and grooves, and the generation of a large number of irregular pieces. Although this type of wear rate is moderate, the formation of grooves will quickly increase vibration and may lead to local mechanical system failure.

[0056] Adhesive wear, in which friction is intensified due to the interaction of the spikes, resulting in flakes of material being detached from the surfaces of the friction pair; this type of wear rapidly increases the wear rate, ultimately leading to increased wear.

[0057] Rough wear, fatigue wear and adhesive wear have different degrees of influence on mechanical systems, but they all manifest as power loss on a macro scale. Therefore, according to the wear mechanism, the power loss can be divided into roughness loss power, fatigue loss power and adhesive loss power. According to the law of conservation of energy and mass, the wear evolution mechanism is determined by the combined effect of power loss, friction pair characteristics and abrasive particle characteristics. The mean and standard deviation of the normal distribution can be used to describe the roughness characteristics of the friction pair. The characteristics of the abrasive particles are mainly size and number, which are closely related to the roughness characteristics of the friction pair. Therefore, this application abstracts the wear process into 6 state quantities: Adhesion loss power , fatigue loss power , roughness loss power , average roughness of friction pair , Standard deviation of friction pair roughness and the quality of wear Based on the actual test data, a test quantity is abstracted - the number of abrasive particles. .

[0058] If other types of wear need to be considered, the power loss of this type of wear should also be included in the state variables. At the same time, the subsequent steps need to consider the relationship between each power loss and the last three quantities to establish the corresponding state space.

[0059] (2) The multiple factors affecting wear are abstracted into model parameters, which include the mean effect coefficient, standard deviation effect coefficient, mass effect coefficient and loss power correlation coefficient of each type of wear behavior.

[0060] (3) Based on the linear damage mechanism and using differential equations to describe the dynamic relationship between state variables and model parameters, a dynamic model of wear particle generation is established.

[0061] Specifically, the dynamic relationship of state parameters is constructed, and the model parameters are associated according to the relationship between loss power, friction pair characteristics and abrasive characteristics at the dynamic level. Based on the first principles of conservation of energy and mass, a macroscopic dynamic model of wear is constructed.

[0062] The power loss from rough wear, fatigue wear, and adhesive wear will lead to an increase in the mass of the abrasive particles. Therefore, the increment of the abrasive particle mass can be expressed as:

[0063] in, , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, respectively.

[0064] Power loss also changes the standard deviation and mean of the roughness of the friction pair. From the perspective of the three wear mechanisms, the peeling or cutting of spikes during rough wear and the pitting or groove formation during fatigue wear cause the originally smooth surface to become rough and uneven. Adhesive wear generally occurs when there are many spikes. The peeling of spikes after engagement can make the surface relatively smooth, improving flatness. Therefore, the changes in the roughness mean and standard deviation can be expressed as:

[0065]

[0066] in, , , are the roughness mean effect coefficients of rough wear, fatigue wear and adhesive wear, respectively. , , They represent the standard deviation effect coefficients of rough wear, fatigue wear and adhesive wear respectively.

[0067] The standard deviation of the surface roughness of the friction pair is positively correlated with the adhesion loss power. Therefore, the adhesion loss power can be expressed as:

[0068] As the mass of the abrasive increases, the profile of the friction pair becomes more uneven, resulting in an increase in fatigue loss power. Therefore, the fatigue loss power can be expressed as:

[0069] The roughness mean of the friction pair also determines the roughness loss power:

[0070] in, is the adhesion loss power correlation coefficient, is the fatigue loss power correlation coefficient, is the roughness loss power correlation coefficient.

[0071] The formulas of the above six state variables together constitute the dynamic model of wear particle generation.

[0072] Preferably, when only rough wear, fatigue wear, and adhesive wear are considered, the state space obtained by combining the above six formulas is as follows:

[0073] in, is the adhesion loss power, is the fatigue loss power, is the roughness loss power, and the three are state variables of wear. is the parameter matrix of the state equation.

[0074] Preferably, a plurality of samples are randomly sampled by a Monte Carlo method.

[0075] Preferably, the fitness function is constructed based on the least squares method, as follows:

[0076] in, The abrasive particle generation rate is obtained based on the abrasive particle number sequence and is directly calculated from the online detected data. is the wear particle generation rate calculated based on the dynamic model, , is the adhesion loss power, is the fatigue loss power, is the roughness loss power, the state variable of the three wear together , , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the roughness loss power correlation coefficient, is the morphology factor of the abrasive particle, which is the ratio of the bottom area to the height of the abrasive particle. is the time length of the wear particle number sequence. As the wear particle time series is updated, the optimal estimate can also be updated accordingly. , so the wear state of the friction pair can be observed.

[0077] Using machine learning ideas, the error between the model output and the actual detection data is used as the evaluation function. Through advanced machine learning algorithms, the model parameters and initial states are continuously iteratively searched in the possible solution space in reality, and finally the optimal model parameters are obtained.

[0078] Since the wear evolution results are related to both the model form and the model parameters, and the values ​​of each state parameter are not fixed in reality but have their own value range, we determined the reasonable range of the model parameters involved by researching relevant literature, and then conducted numerical simulations within the realistic model parameter range.

[0079] Preferably, the influence of each parameter on the wear evolution pattern is analyzed by using critical parameter perturbation, specifically as follows: Put the system in a critical state and perturb the model parameters in sequence to analyze the effect of single parameter change on power loss, friction pair roughness characteristics, and wear particle characteristics; Based on the perturbation results, the effects of the mean action coefficient of rough wear, the mean action coefficient of fatigue wear, the mean action coefficient of adhesive wear, the standard deviation action coefficient of rough wear, the standard deviation action coefficient of adhesive wear, the fatigue wear power coefficient, the mass action coefficient, the standard deviation action coefficient of fatigue wear, the rough wear power factor coefficient and the adhesive wear power factor coefficient on the development direction and development rate of the loss power and the roughness characteristics of the friction pair are analyzed.

[0080] Specifically, the macroscopic wear behavior of the simulation analysis model is analyzed, and the actual solution domain of the state parameters is determined according to the actual working conditions. Within a reasonable parameter range, the Monte Carlo method is used to simulate the evolution trend of various wear systems, and the critical parameter perturbation is used to analyze the influence of each parameter on wear evolution.

[0081] Furthermore, the Monte Carlo method is used to randomly generate a large number of parameters as different wear samples in a reasonable parameter space, and their evolution process under the same initial state is simulated to summarize the evolution pattern existing in the actual parameter space. Then, the wear evolution behavior near the critical parameter is studied, and the system is placed in a critical state and the parameters are perturbed in sequence. , , , , , , , , , , as well as , thereby analyzing the impact of single parameter changes on power loss, friction pair roughness characteristics, and abrasive particle characteristics. Based on the perturbation results, the effects of the rough wear mean effect coefficient, fatigue wear mean effect coefficient, adhesive wear mean effect coefficient, rough wear standard deviation effect coefficient, adhesive wear standard deviation effect coefficient, fatigue wear power coefficient, mass effect coefficient, fatigue wear standard deviation effect coefficient, rough wear power factor coefficient, and adhesive wear power factor coefficient on the development direction and rate of power loss and friction pair roughness characteristics are analyzed. Based on these findings, the influence of multiple real-world factors on the evolution of the wear system is revealed at a macro level, thereby achieving fault diagnosis and life optimization of equipment.

[0082] Preferably, the method further includes determining the stability of the mechanical system, specifically as follows: (1) Obtain the dynamic model parameters of abrasive particle generation and generate the closed-loop transfer function characteristic equation of the system.

[0083] The stability condition of the system is determined by the parameters and symbols of the characteristic equation of the closed-loop transfer function of the system, according to the parameter matrix of the wear state equation. Preferably, the characteristic equation is expressed as follows:

[0084] in,

[0085]

[0086]

[0087] in, , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, , , are the roughness mean effect coefficients of rough wear, fatigue wear and adhesive wear, , , They represent the standard deviation effect coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the fatigue loss power correlation coefficient, is the roughness loss power correlation coefficient, is the Laplace operator, is the coefficient of the characteristic equation.

[0088] (2) Construct the corresponding Routh array according to the characteristic equation.

[0089] (3) Determine the system stability based on Routh criterion.

[0090] Specifically, the relationship between model eigenvalues ​​and parameters was analyzed, and the Routh criterion was used to derive a mathematical criterion for wear evolution trends. This mathematically determined the parameter relationships corresponding to different system evolution trends. Using the Routh stability criterion to describe the wear process, three scenarios can be preliminarily categorized: stable, unstable, and critically stable. Each of these three scenarios was analyzed sequentially to ensure compliance with the Routh criterion. The scenarios in which all state parameters must meet the requirements were listed, and some scenarios where these requirements do not exist were proposed based on actual operating conditions.

[0091] Preferably, the conditions that the state parameters need to meet when the system is stable are:

[0092] When the system is unstable, the parameters must meet any of the following eleven conditions:

[0093]

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103] The system is in a critically stable state. There are two possible combinations of model parameters: when hour:

[0104] when and hour:

[0105] in, Represents a small positive number, used to replace the value 0.

[0106] Example In this example, a Monte Carlo method was used to randomly generate 10,000 parameter sets within the actual parameter space as different wear samples. Their evolutionary processes under the same initial conditions were simulated to summarize the evolutionary patterns present in the actual parameter space. The parameter sample size used here provides a simulation example based on a large number of samples and is not limited to this value. Since the transmission power of a mechanical system typically ranges from 1 kW to 100 kW, with a transmission efficiency between 80% and 100%, and considering only friction pair losses, the power loss is set within 10%. This means that for a 100 kW load, the sum of the three power losses should be within 10 kW. Furthermore, considering the contribution of different wear types to the mechanical wear process under actual operating conditions, rough wear typically accounts for 50%, fatigue wear for 20%, and adhesive wear for 30%. Therefore, the simulation is terminated when the roughness power loss exceeds 5 kW, the fatigue power loss exceeds 2 kW, the adhesive power loss exceeds 3 kW, or when the power loss falls below 0 W. In addition, the life of mechanical systems is generally between several thousand and ten thousand hours, so the time of evolution simulation is set to 5000 hours. In order to simulate the evolution from the state when the roughness peak appears, it is assumed that the total power loss at the beginning is 500W, and the initial states of the three types of power loss are , , .

[0107] To better illustrate the simulation results, the wear states are divided into four categories based on the change in power loss between the end and initial simulation times. Based on the wear macrodynamic model and combined with the simulation results, four wear evolution patterns are identified. Pattern 1 refers to an increase in both roughness and adhesion power loss, indicating an increase in the mean and standard deviation of the friction pair roughness. Pattern 2 refers to a decrease in roughness power loss but an increase in adhesion power loss at the end of the simulation, corresponding to a decrease in the mean roughness of the friction pair but an increase in the standard deviation of the friction pair roughness. Pattern 3 refers to an increase in roughness power loss and a decrease in adhesion power loss, corresponding to an increase in the mean roughness of the friction pair and a decrease in the standard deviation of the friction pair roughness. Pattern 4 refers to a decrease in both roughness power loss and adhesion power loss at the end of the simulation, corresponding to a decrease in the mean and standard deviation of the friction pair roughness. For all four wear patterns, fatigue power loss increases, indicating that as the abrasive mass increases, the friction pair profile changes and fatigue loss increases, which is consistent with reality.

[0108] For these four wear modes, fatigue loss power increases, which means that as the abrasive mass increases, the friction pair profile changes and fatigue loss increases, which is consistent with reality. The four modes are represented by black, cyan, orange and pink respectively. After entering the appropriate parameters, Monte Carlo simulation is performed. After the simulation is completed, the wear evolution is output as follows Figure 2 As shown. Figure 2 In the figure, the black curve represents the case where the system is completely divergent; the cyan curve represents the reduction of roughness loss power, that is, the case where the mean value of the friction pair roughness decreases during the wear process; the orange curve represents the reduction of adhesion loss power, that is, the case where the standard deviation of the friction pair roughness decreases during the wear process; the pink curve represents the case where both the roughness loss power and adhesion loss power decrease, and the corresponding mean and standard deviation of the friction pair roughness decrease.

[0109] Furthermore, the ratios of the four wear evolution modes in 10,000 wear samples are as follows: Figure 3 As shown in Figure 2, 50% of the wear samples evolved according to Mode 1, with the power loss completely diverging. That is, all three power loss factors increased continuously during the wear process, indicating that the mean and standard deviation of the friction pair roughness of the system continued to increase over the simulation period, causing rapid damage to the mechanical equipment. 28% of the wear samples evolved according to Mode 2, and after 5000 hours of evolution, the roughness power loss decreased. That is, the mean value of the friction pair roughness decreased, indicating that in this case, the friction pair surface was flat but a jagged wear profile appeared. 7% of the wear samples evolved according to Mode 3 after 5000 hours, and the adhesion power loss decreased. That is, the standard deviation of the friction pair roughness decreased. This indicates that in this case, the flatness of the friction pair surface gradually improved with the wear process, resulting in a more consistent wear texture. 15% of the wear samples evolved according to Mode 4, with both the roughness power loss and the adhesion power loss decreasing. That is, the friction pair surface characteristics continued to improve, and in fact, tended to be smoother with the wear process.

[0110] Critical parameter perturbations such as Figure 4 The friction pair state and abrasive particle characteristic phase trajectory are shown in the upper left, reflecting the overall influence trend of parameter changes on abrasive particle mass, roughness mean and standard deviation; the abrasive particle mass phase trajectory is shown in the lower left, reflecting the influence of parameter changes on the evolution of abrasive particle mass; the roughness mean phase trajectory is shown in the upper right, reflecting the influence of parameter changes on the evolution of roughness mean; the roughness standard deviation phase trajectory is shown in the lower right, reflecting the influence of parameter changes on the evolution of roughness standard deviation. Figure 4 The parameter perturbation shown is only a simulation example of the perturbation of one parameter. The actual perturbation should consider all model parameters and is not limited to this parameter.

[0111] In addition, based on Figures 2 to 4The simulation results in the paper are used to fit the initial state and parameters of the dynamic model of wear particle generation through the wear particle time series of the specific monitoring object. The wear particle sensor obtains The time series of wear particles is divided into clusters; then randomly select from the actual possible state space and parameter space samples ,in ; Secondly, construct the fitness function based on the least quadratic method , and calculate the fitness of each sample for the detected wear particle time series. is the wear particle generation rate obtained based on the wear particle time series, is the abrasive particle generation rate calculated based on the dynamic model; finally, through advanced machine learning algorithms such as gradient descent method, maximum likelihood estimation or regularization method, the solution that satisfies The best estimate of Since the optimal estimate Contained in Characterizes the state of the friction pair, and as the wear time series is updated, the corresponding , thus realizing the observation of the wear state of the friction pair, such as Figure 5 shown. is the observation node at the early stage of wear, and the actual number of wear particles at this time is , is the observation node in the middle stage of wear, and the actual number of wear particles at this time is , is the observation node at the late stage of wear, and the actual number of wear particles at this time is In actual working conditions, actual wear data should be recorded in three periods as much as possible. The actual number of abrasive particles With the continuous increase in the number of samples, an optimal parameter trajectory can be obtained from many parameter trajectories based on the actual wear particle data, thereby realizing the optimal parameter prediction.

[0112] Finally, the mechanical life is investigated through simulation to verify the accuracy and prediction performance of the model.

[0113] In summary, compared with the prior art, this application has the following advantages: This application proposes a method and system for observing the wear state of friction pairs based on online abrasive particle detection. Focusing on the conservation of energy and mass at a macroscopic time scale, the application avoids the complexity of real-world wear by abstracting core state variables. Furthermore, differential equations are used to characterize the dynamic relationship between core state variables, ultimately solving the problem of describing the wear evolution behavior at a macroscopic time scale. Drawing on currently popular machine learning ideas, the error between the model output and the actual detection data is used as the evaluation function. Advanced machine learning algorithms are used to iteratively search for model parameters and initial states in the real-world possible solution space, ultimately obtaining the optimal model parameters and enabling diagnosis of the wear process of a specific object.

[0114] This application proposes a wear state observation method based on parameter identification. Through dynamic modeling at a macro scale, the evolution law of the wear system is reflected. Further, through data-driven parameter estimation, the initial state and model parameters that are consistent with the online data law are obtained. Compared with traditional methods, this method not only avoids the time lag of offline detection methods, but also breaks through the limitations of existing online detection that has no theoretical model support or is prone to display complexity. This method uses a combination of dynamic modeling and data-driven parameter identification. It can reveal the generation law of abrasive particles under complex wear conditions and realize the wear state observation of friction pairs, thereby more comprehensively and accurately predicting the service life of mechanical equipment.

[0115] It should be understood that the above-mentioned device is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the device are similar to those described in the above-mentioned method. The working process of the device can refer to the corresponding process in the above-mentioned method and will not be repeated here.

[0116] Based on the methods in the above embodiments, embodiments of the present application provide an electronic device that may include: a processor, a communications interface, a memory, and a communication bus, wherein the processor, the communications interface, and the memory communicate with each other via the communication bus. The processor may invoke logic instructions in the memory to execute the methods in the above embodiments.

[0117] In addition, the logical instructions in the above-mentioned memory can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0118] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0119] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0120] It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0121] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC.

[0122] The above embodiments can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions. When loaded and executed on a computer, the computer program instructions fully or partially produce the processes or functions described in the embodiments of this application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state drive (SSD)).

[0123] It will be understood that the various numerical numbers involved in the embodiments of the present application are merely distinctions for the convenience of description and are not intended to limit the scope of the embodiments of the present application.

[0124] It is easy for those skilled in the art to understand that the above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for observing the wear state of a friction pair based on online wear particle detection, characterized in that: include: The initial state and parameters of the kinetic model of wear particle generation are dynamically fitted through the real-time online detected wear particle number sequence; A plurality of samples are randomly selected in the actual state space and the actual parameter space, each sample is composed of a state variable and a parameter, and the wear state evolution pattern existing in the actual parameter space is determined; Calculate the fitness of each sample for the wear particle number sequence detected online, where the fitness reflects the degree of consistency between the wear particle generation rates of the two; Find the optimal sample that satisfies the highest degree of compliance; Based on the parameters of the optimal sample, critical parameter perturbation is used to analyze the influence of each parameter on the wear evolution pattern.

2. The wear state observation method according to claim 1, characterized in that: The kinetic model of abrasive particle generation is constructed in the following way: (1) Based on the conservation of energy and mass on a macroscopic time scale, the state variables of the wear process are abstracted from three aspects: power loss, friction pair characteristics, and wear debris characteristics. The state variables include the mean value of the friction pair roughness, the standard deviation of the friction pair roughness, the mass of the abrasive particles, and the power loss of each type of wear behavior; (2) Abstracting multiple factors affecting wear into model parameters, including the mean effect coefficient, standard deviation effect coefficient, mass effect coefficient, and loss power correlation coefficient of each type of wear behavior; (3) Based on the linear damage mechanism and using differential equations to describe the dynamic relationship between state variables and model parameters, a dynamic model of wear particle generation is established.

3. The wear state observation method according to claim 1, characterized in that: When only rough wear, fatigue wear, and adhesive wear are considered, the state space is as follows: in, is the adhesion loss power, is the fatigue loss power, is the roughness loss power, and the three are state variables of wear. is the parameter matrix of the state equation, , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, , , are the roughness mean effect coefficients of rough wear, fatigue wear and adhesive wear, , , They represent the standard deviation effect coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the fatigue loss power correlation coefficient, is the roughness loss power correlation coefficient.

4. The wear state observation method according to claim 1, wherein: Multiple samples are randomly drawn using the Monte Carlo method.

5. The wear state observation method according to claim 1, characterized in that: The fitness function is constructed based on the least squares method, as follows: in, The abrasive particle generation rate is obtained based on the abrasive particle number sequence and is directly calculated from the online detected data. is the wear particle generation rate calculated based on the dynamic model, , is the adhesion loss power, is the fatigue loss power, is the roughness loss power, and the three together constitute the state variable of wear , , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the roughness loss power correlation coefficient, is the morphology factor of the abrasive particle, which is the ratio of the bottom area to the height of the abrasive particle. is the time length of the wear particle number sequence.

6. The wear state observation method according to claim 1, characterized in that: The influence of each parameter on the wear evolution pattern is analyzed by using critical parameter perturbation, as follows: Put the system in a critical state and perturb the model parameters in sequence to analyze the effect of single parameter change on power loss, friction pair roughness characteristics, and wear particle characteristics; Based on the perturbation results, the effects of the mean action coefficient of rough wear, the mean action coefficient of fatigue wear, the mean action coefficient of adhesive wear, the standard deviation action coefficient of rough wear, the standard deviation action coefficient of adhesive wear, the fatigue wear power coefficient, the mass action coefficient, the standard deviation action coefficient of fatigue wear, the rough wear power factor coefficient and the adhesive wear power factor coefficient on the development direction and development rate of the loss power and the roughness characteristics of the friction pair are analyzed.

7. The wear state observation method according to any one of claims 1 to 6, characterized in that: The method also includes the determination of the stability of the mechanical system, as follows: (1) Obtain the dynamic model parameters of abrasive particle generation and generate the closed-loop transfer function characteristic equation of the system; (2) Construct the corresponding Routh array according to the characteristic equation; (3) Determine the system stability based on Routh criterion.

8. The wear state observation method according to claim 7, characterized in that: The characteristic equation is expressed as follows: in, in, , , are the mass action coefficients of rough wear, fatigue wear and adhesive wear, , , are the roughness mean effect coefficients of rough wear, fatigue wear and adhesive wear, , , They represent the standard deviation effect coefficients of rough wear, fatigue wear and adhesive wear, is the adhesion loss power correlation coefficient, is the fatigue loss power correlation coefficient, is the roughness loss power correlation coefficient, is the Laplace operator, is the coefficient of the characteristic equation.

9. The wear state observation method according to claim 8, characterized in that: The conditions that the state parameters need to meet when the system is stable are: When the system is unstable, the parameters must meet any of the following eleven conditions: The system is in a critically stable state. There are two possible combinations of model parameters: when hour: when and hour: in, Represents a small positive number, used to replace the value 0.

10. A friction pair wear state observation system based on online wear particle detection, characterized in that: include: at least one memory for storing a program; At least one processor is configured to enter the program stored in the memory, and when the program stored in the memory is entered, the processor is configured to enter the state observation method according to any one of claims 1 to 9.

Images