Mechanical seal fault diagnosis method and device based on PINN

Abstract

The invention relates to the technical field of fluid sealing, in particular to a PINN-based mechanical sealing fault diagnosis method and device, and the method comprises the steps: embedding a multi-physical field coupling simulation model, meeting a preset condition, of a target mechanical sealing system into a loss function of a physical information neural network (PINN), so as to construct an initial PINN fault diagnosis model, and inputting actual monitoring data of a plurality of sensors of the target mechanical sealing system into the initial PINN fault diagnosis model to output a plurality of fault parameters to be identified, further updating the initial PINN fault diagnosis model, generating a target PINN fault diagnosis model, and diagnosing the sealing state of the target mechanical sealing system according to the target PINN fault diagnosis model. And generating a fault diagnosis result. Therefore, the problems that in the prior art, in the face of complex working conditions and variable environments, the sealing performance cannot be comprehensively evaluated in real time, and consequently the accuracy and timeliness of mechanical sealing fault diagnosis are insufficient are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of fluid sealing, and in particular to a mechanical seal fault diagnosis method and device based on PINN (Physics-Informed Neural Networks). Background Art

[0002] Mechanical seals are widely used in various industrial equipment, especially in rotating machinery such as pumps and compressors, where they play a critical role in preventing fluid leakage and maintaining system pressure. However, mechanical seals are prone to failure over long periods of operation due to factors such as wear, corrosion, and temperature fluctuations, which can affect the safety, stability, and efficiency of the equipment.

[0003] In related technologies, mechanical seal fault diagnosis methods usually include vibration monitoring, temperature monitoring, and visual inspection. Although they have achieved certain results in some scenarios, they still have certain limitations when facing complex working conditions and changing environments. These methods often require a lot of manual operations and cannot comprehensively evaluate the sealing performance in real time, resulting in difficulties in early identification of faults, reducing the accuracy and timeliness of mechanical seal fault diagnosis, and urgently need to be solved. Summary of the Invention

[0004] The present invention provides a mechanical seal fault diagnosis method and device based on PINN to solve the problem in related technologies that, when faced with complex working conditions and changing environments, the sealing performance cannot be comprehensively evaluated in real time, resulting in difficulty in early identification of faults, and causing insufficient accuracy and timeliness in mechanical seal fault diagnosis.

[0005] A first aspect of the present invention provides a PINN-based mechanical seal fault diagnosis method, comprising the following steps: establishing a multi-physics field coupling simulation model of a target mechanical seal system that meets preset conditions; embedding the multi-physics field coupling simulation model into the loss function of a physical information neural network PINN to construct an initial PINN fault diagnosis model, and inputting actual monitoring data of multiple sensors of the target mechanical seal system into the initial PINN fault diagnosis model to output multiple fault parameters to be identified; based on the multiple fault parameters to be identified, updating the initial PINN fault diagnosis model and generating a target PINN fault diagnosis model to diagnose the sealing state of the target mechanical seal system according to the target PINN fault diagnosis model, so that a fault diagnosis result of the target mechanical seal system is generated according to the sealing state.

[0006] Optionally, in one embodiment of the present invention, updating the initial PINN fault diagnosis model based on the multiple fault parameters to be identified includes: inputting the multiple fault parameters to be identified into the multi-physics field coupling simulation model to output the simulated monitoring data of the multiple sensors; determining the physical information loss function in the loss function of the PINN based on the difference between the simulated monitoring data and the actual monitoring data; combining the physical information loss function with the boundary loss function in the loss function of the PINN to train the initial PINN fault diagnosis model to obtain the target PINN fault diagnosis model.

[0007] Optionally, in one embodiment of the present invention, before combining the physical information loss function with the boundary loss function in the loss function of the PINN, it also includes: determining the target boundary of the fault parameter to be identified; and determining the boundary loss function in the loss function of the PINN based on the target boundary and the multiple fault parameters to be identified.

[0008] Optionally, in one embodiment of the present invention, the boundary loss function is calculated as follows:

[0009]

[0010] Among them, Loss b is the boundary loss function, Loss bi is the i-th boundary loss function; x ipre is x i The output during the training process of the PINN; x imin is x i The lower boundary of x imax is x i The upper boundary of x i is the fault parameter to be identified.

[0011] Optionally, in one embodiment of the present invention, the physical information loss function is calculated as follows:

[0012]

[0013] Among them, Loss r is the physical information loss function, Loss ri is the physical information loss function of the i-th item; For sensor Y i Monitoring vector, N i is the vector length; Based on the multi-physics coupling simulation model, Y i estimated value.

[0014] Optionally, in one embodiment of the present invention, diagnosing the sealing state of the target mechanical sealing system according to the target PINN fault diagnosis model includes: determining target parameters of the multi-physics field coupling simulation model that satisfy target constraint conditions based on the target PINN fault diagnosis model; and diagnosing the sealing state of the target mechanical sealing system using the target parameters.

[0015] The second aspect of the present invention provides a PINN-based mechanical seal fault diagnosis device, including: an establishment module for establishing a multi-physics field coupling simulation model of a target mechanical seal system that meets preset conditions; a processing module for embedding the multi-physics field coupling simulation model into the loss function of the physical information neural network PINN to construct an initial PINN fault diagnosis model, and inputting the actual monitoring data of multiple sensors of the target mechanical seal system into the initial PINN fault diagnosis model to output multiple fault parameters to be identified; a diagnosis module for updating the initial PINN fault diagnosis model based on the multiple fault parameters to be identified, generating a target PINN fault diagnosis model, and diagnosing the sealing state of the target mechanical seal system according to the target PINN fault diagnosis model, so that a fault diagnosis result of the target mechanical seal system is generated according to the sealing state.

[0016] Optionally, in one embodiment of the present invention, the diagnosis module includes: an acquisition unit, used to input the multiple fault parameters to be identified into the multi-physics field coupling simulation model to output the simulated monitoring data of the multiple sensors; a first determination unit, used to determine the physical information loss function in the PINN loss function based on the difference between the simulated monitoring data and the actual monitoring data; a training unit, used to combine the physical information loss function with the boundary loss function in the PINN loss function to train the initial PINN fault diagnosis model to obtain the target PINN fault diagnosis model.

[0017] Optionally, in one embodiment of the present invention, the device of the embodiment of the present invention further includes: a first determination module, used to determine the target boundary of the fault parameter to be identified; and a second determination module, used to determine the boundary loss function in the loss function of the PINN based on the target boundary and the multiple fault parameters to be identified.

[0018] Optionally, in one embodiment of the present invention, the boundary loss function is calculated as follows:

[0019]

[0020] Among them, Loss b is the boundary loss function, Loss biis the i-th boundary loss function; x ipre is x i The output during the training process of the PINN; x imin is x i The lower boundary of x imax is x i The upper boundary of x i is the fault parameter to be identified.

[0021] Optionally, in one embodiment of the present invention, the physical information loss function is calculated as follows:

[0022]

[0023] Among them, Loss r is the physical information loss function, Loss ri is the physical information loss function of the i-th item; For sensor Y i Monitoring vector, N i is the vector length; Based on the multi-physics coupling simulation model, Y i estimated value.

[0024] Optionally, in one embodiment of the present invention, the diagnostic module includes: a second determination unit, used to determine the target parameters of the multi-physics field coupling simulation model that meet the target constraint conditions based on the target PINN fault diagnosis model; and a diagnostic unit, used to diagnose the sealing status of the target mechanical sealing system using the target parameters.

[0025] A third aspect of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the PINN-based mechanical seal fault diagnosis method as described in the above embodiment.

[0026] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which implements the above-mentioned PINN-based mechanical seal fault diagnosis method when executed by a processor.

[0027] A fifth aspect of the present invention provides a computer program product, including a computer program. When the computer program is executed, it is used to implement the above-mentioned PINN-based mechanical seal fault diagnosis method.

[0028] The embodiment of the present invention can embed the established multi-physics field coupling simulation model of the target mechanical seal system into the loss function of the physical information neural network PINN to construct an initial PINN fault diagnosis model, and input the actual monitoring data of multiple sensors of the target mechanical seal system into the initial PINN fault diagnosis model to output multiple fault parameters to be identified, thereby updating the initial PINN fault diagnosis model and generating a target PINN fault diagnosis model. The sealing state of the target mechanical seal system is diagnosed according to the target PINN fault diagnosis model, and a fault diagnosis result is generated, which effectively improves the accuracy and timeliness of mechanical seal fault diagnosis. Thus, the problem in the related art that the sealing performance cannot be evaluated in real time and comprehensively, resulting in insufficient accuracy and poor timeliness of mechanical seal fault diagnosis, is solved.

[0029] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0031] Figure 1 A flowchart of a PINN-based mechanical seal fault diagnosis method according to an embodiment of the present invention;

[0032] Figure 2 A framework diagram of mechanical seal fault diagnosis based on PINN according to a specific embodiment of the present invention;

[0033] Figure 3 A schematic diagram of a main pump three-stage machine sealing system and sensor measuring points according to a specific embodiment of the present invention;

[0034] Figure 4 Schematic diagram of the dynamic operation state DE of a single-pole mechanical seal ring according to a specific embodiment of the present invention;

[0035] Figure 5 This is a flow chart of a PINN-based mechanical seal fault diagnosis method according to a specific embodiment of the present invention;

[0036] Figure 6 A schematic structural diagram of a PINN-based mechanical seal fault diagnosis device according to an embodiment of the present invention;

[0037] Figure 7 A schematic structural diagram of an electronic device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0039] The following describes a PINN-based mechanical seal fault diagnosis method and device according to an embodiment of the present invention with reference to the accompanying drawings. In response to the problem that the related art mentioned in the background art center cannot comprehensively evaluate sealing performance in real time, resulting in insufficient accuracy and timeliness of mechanical seal fault diagnosis, the present invention provides a PINN-based mechanical seal fault diagnosis method. In this method, a multi-physics field coupled simulation model of a target mechanical seal system can be embedded into the loss function of a physical information neural network (PINN) to construct an initial PINN fault diagnosis model. Actual monitoring data from multiple sensors of the target mechanical seal system is input into the initial PINN fault diagnosis model to output multiple fault parameters to be identified. The initial PINN fault diagnosis model is then updated to generate a target PINN fault diagnosis model. The sealing state of the target mechanical seal system is diagnosed based on the target PINN fault diagnosis model, and a fault diagnosis result is generated, effectively improving the accuracy and timeliness of mechanical seal fault diagnosis. Thus, the problem that the related art cannot comprehensively evaluate sealing performance in real time, resulting in insufficient accuracy and timeliness of mechanical seal fault diagnosis, is solved.

[0040] Specifically, Figure 1 A schematic flow chart of a PINN-based mechanical seal fault diagnosis method provided in an embodiment of the present invention.

[0041] like Figure 1 As shown, the PINN-based mechanical seal fault diagnosis method includes the following steps:

[0042] In step S101 , a multi-physics field coupling simulation model of a target mechanical sealing system that meets preset conditions is established.

[0043] In the embodiment of the present invention, the target mechanical seal system is a system currently diagnosing mechanical seal failure.

[0044] It can be understood that the embodiment of the present invention can establish a multi-physics field coupling simulation model of the target mechanical seal system that meets certain conditions. For example, the embodiment of the present invention can establish a high-fidelity numerical calculation model based on the existing sealing structure design and operating parameters, that is, a multi-physics field coupling simulation model, and adopt an accelerated calculation method to improve the simulation calculation efficiency. Among them, the accelerated calculation method can adopt model simplification, efficient numerical methods or proxy models, etc., which effectively improves the feasibility of mechanical seal fault diagnosis.

[0045] It should be noted that the physical loss function of the generalized PINN is provided by a partial differential equation. However, in real physical scenarios, especially multi-physics field coupling scenarios such as mechanical seals, it is difficult to construct a reasonable and simple partial differential equation to describe the entire sealing system. Therefore, the embodiment of the present invention provides a basis for constructing the physical loss function of the PINN in the following steps by establishing a high-efficiency and high-fidelity multi-physics field coupling simulation model.

[0046] In step S102, the multi-physics field coupling simulation model is embedded into the loss function of the physical information neural network PINN to construct an initial PINN fault diagnosis model, and the actual monitoring data of multiple sensors of the target mechanical sealing system are input into the initial PINN fault diagnosis model to output multiple fault parameters to be identified.

[0047] It is understandable that the embodiment of the present invention can embed the multi-physics field coupling simulation model into the loss function of the physical information neural network PINN to construct an initial PINN fault diagnosis model, for example, Figure 2 As shown, the basic network architecture of the initial PINN fault diagnosis model in the embodiment of the present invention can adopt the classic MLP (Multilayer Perceptron), CNN (Convolutional Neural Network), or RNN (Recurrent Neural Network), etc., depending on the data type of the sensor. For example, CNN and its variants are recommended for periodic data, RNN and its variants are recommended for time series data, and MLP and its variants are recommended for other data. The input of the initial PINN fault diagnosis model is the sensor monitoring data in the mechanical seal system, including leakage monitoring, thermocouple temperature measurement, pressure sensing, acoustic emission, and eddy current displacement measurement.

[0048] It should be noted that the sensor monitoring data in the embodiments of the present invention must be calculated by a multi-physics field coupling simulation model; the output of the initial PINN fault diagnosis model is various fault parameters to be identified, that is, the actual sensor monitoring data of multiple sensors can be input into the initial PINN fault diagnosis model to output multiple fault parameters to be identified, including factors such as dynamic anomalies and sealing end face defects, and these factors must be used as inputs to the multi-physics field coupling simulation model to effectively improve the accuracy and interpretability of mechanical seal fault diagnosis.

[0049] In step S103, based on multiple fault parameters to be identified, the initial PINN fault diagnosis model is updated to generate a target PINN fault diagnosis model, so as to diagnose the sealing state of the target mechanical sealing system according to the target PINN fault diagnosis model, so that a fault diagnosis result of the target mechanical sealing system is generated according to the sealing state.

[0050] In the embodiment of the present invention, the target PINN fault diagnosis model is a trained and optimized PINN fault diagnosis model.

[0051] It can be understood that the embodiment of the present invention can perform iterative training and parameter update on the initial PINN fault diagnosis model based on the multiple fault parameters to be identified in the following steps, combined with the parameters constrained by the multi-physics field coupling simulation model, to generate a trained and optimized PINN fault diagnosis model. In other words, the embodiment of the present invention can select a reasonable PINN parameter update strategy and set a reasonable stopping threshold. When the PINN training is completed, the output of the PINN training process obtained for the last time, such as x ipre This is the i-th fault parameter to be identified, which meets the constraints of the multi-physics field coupling simulation model.

[0052] For example, an embodiment of the present invention can analyze the operating status of the mechanical sealing system through the target PINN fault diagnosis model, and output key dynamic parameters and fault characteristic parameters related to its sealing performance, so that the sealing status of the mechanical sealing system can be judged according to the sealing performance and fault characteristic parameters, and the corresponding mechanical sealing system fault diagnosis results can be generated based on the sealing status, thereby realizing accurate assessment of the health status of the sealing system and identification of fault types.

[0053] Among them, in one embodiment of the present invention, based on multiple fault parameters to be identified, the initial PINN fault diagnosis model is updated, including: inputting the multiple fault parameters to be identified into a multi-physics field coupling simulation model to output simulated monitoring data of multiple sensors; determining the physical information loss function in the PINN loss function based on the difference between the simulated monitoring data and the actual monitoring data; combining the physical information loss function with the boundary loss function in the PINN loss function to train the initial PINN fault diagnosis model to obtain the target PINN fault diagnosis model.

[0054] In the actual implementation process, Figure 2As shown, the embodiment of the present invention can input multiple fault parameters to be identified in the above steps into the multi-physics field coupling simulation model, and output simulated monitoring data of multiple sensors, such as leakage monitoring, thermocouple temperature measurement, pressure sensing, acoustic emission or eddy current displacement measurement sensors; then, the embodiment of the present invention can calculate the difference between the actual monitoring data and the simulated monitoring data of multiple sensors to determine the physical information loss function in the PINN loss function, and secondly, combine the boundary loss function in the PINN loss function to train the initial PINN fault diagnosis model, and gradually reduce the loss function during the training process to obtain the final optimized PINN fault diagnosis model and parameters that meet the constraints of the multi-physics field coupling simulation model, so that the health status of the mechanical seal can be judged according to the size of the corresponding parameters, effectively improving the accuracy and timeliness of mechanical seal fault diagnosis.

[0055] Optionally, in one embodiment of the present invention, before combining the physical information loss function with the boundary loss function in the PINN loss function, it also includes: determining the target boundary of the fault parameter to be identified; and determining the boundary loss function in the PINN loss function based on the target boundary and multiple fault parameters to be identified.

[0056] In the embodiment of the present invention, the target boundary includes an upper boundary and a lower boundary of the fault parameter to be identified.

[0057] As a possible implementation method, the embodiment of the present invention can set the boundaries of the fault parameters to be identified in advance. If the output parameters of the network exceed the boundary range during the network update process, the network parameters are updated. Assume that {x1, x2, ..., x n}These fault parameters to be identified, the boundary loss function is calculated as follows:

[0058]

[0059] Among them, Loss b is the boundary loss function, Loss bi is the i-th boundary loss function; x ipre is x i Output during PINN training; x imin is x i The lower boundary of x imax is x i The upper boundary of x i is the fault parameter to be identified.

[0060] Assume that multiple sensor parameters include {Y1, Y2, ..., Y n}, the signal length of each sensor parameter is {N1, N2,…, N n}, then the calculation method of the physical information loss function is:

[0061]

[0062] Among them, Loss r is the physical information loss function, Loss ri is the physical information loss function of the i-th item; For sensor Y i Monitoring vector, N i is the vector length; To simulate Y based on multi-physics coupling model i estimated value.

[0063] in, To simulate Y based on multi-physics coupling model i The estimated value is obtained by converting the x generated during the PINN training process ipre The result is obtained by substituting it into the efficient and high-fidelity calculation model established in advance.

[0064] Optionally, in one embodiment of the present invention, the sealing state of the target mechanical sealing system is diagnosed according to the target PINN fault diagnosis model, including: determining the target parameters of the multi-physics field coupling simulation model that meet the target constraint conditions based on the target PINN fault diagnosis model; and using the target parameters to diagnose the sealing state of the target mechanical sealing system.

[0065] In some embodiments, the embodiments of the present invention can determine the target parameters of the multi-physics field coupling simulation model that meet the target constraints, such as dynamic parameters, based on the trained and optimized PINN fault diagnosis model. Then, the target parameters can be used to diagnose the sealing status of the mechanical sealing system, effectively improving the accuracy of sealing fault diagnosis.

[0066] For example, two embodiments are provided below. Example 1 establishes a PINN fault diagnosis model for the multi-stage mechanical seal of the main pump of a nuclear power plant, uses pressure and flow sensor data for machine learning model training, and performs fault identification on the sealing status of each stage of the main pump shaft seal; Example 2 establishes a PINN fault diagnosis model for a non-single set mechanical seal, uses mechanical sensor data for machine learning model training, and the parameters to be identified are various dynamic parameters.

[0067] Example 1:

[0068] like Figure 3 The figure shows the main pump's three-stage mechanical seal system and sensor measurement points. The inlet pressures of the first, second, and third stage mechanical seals are represented by p1, p2, and p3, respectively. The leakage of each stage mechanical seal is represented by q1, q2, and q3. The flow rate of the four coils is represented by q 12,q 13 ,q 25 ,q 35 .q H and q L Respectively represent the high-pressure leakage and low-pressure leakage of the system. Figure 3 The black box part in the figure is the sensor monitoring point. In order to describe the health status of the seal and coil, the concept of health status factor is defined. The health status factors of the first-stage, second-stage, and third-stage mechanical seals are r1, r2, and r3 respectively. The health status factors of the four coils can be expressed as r 12 , r 13 , r 25 , r 35 If the seal and coil are in a faulty state, the health factor will decrease and the seal and coil leakage will increase.

[0069] In order to establish the physical information model of the sealing system, we must first establish a pressure and flow calculation model for each stage of mechanical seal and each coil. The calculation model can be a numerical calculation model such as finite element, finite difference, finite volume, or an empirical formula fitting model. Here, taking the empirical formula as an example, each stage of seal and each coil are locally approximated to the following linear model near the design state, namely:

[0070] q i r i =k i Δp i +b i ,i∈{1,2,3,12,13,25,35}

[0071] Among them, 1, 2 and 3 represent the first, second and third stage mechanical seals respectively; 12 and 13 represent the coils leading to the second and third stage seals upstream of the first stage seal respectively; 25 and 35 represent the coils leading to the high pressure leakage line upstream of the second and third stage seals respectively; q i is the leakage of each stage of mechanical seal; Δp i is the pressure difference; k i and b i is the component parameter in the design state; r i Indicates the health factor of the seal and coil.

[0072] Next, the hydraulic model of the three-stage mechanical seal is established, and the pressure-flow relationship of the entire system satisfies the following equation:

[0073] q1+q 12 -q2-q 25 =0

[0074] q2+q 13 -q3-q 35 =0

[0075] from Figure 3 It can be seen that the high-pressure leakage flow rate q H is monitored, and the low pressure leakage q L Not monitored, q H is the sum of the flow rates of the two coils:

[0076] q H =q 25 +q 35

[0077] The above equations are combined to obtain the physical model of the three-stage mechanical seal system. The input of the model is r i , the output is p i and q i Due to the use of linearized calculation method, the calculation speed is extremely high and no acceleration is required.

[0078] The input of PINN is p1, p2, p3 and q H Four sensor data, the output is r i The overall architecture adopts MLP, and the gradient descent method is used to update the correlation coefficient of MLP. In each round of update, the r output by PINN i All of them will be substituted into the physical model of the three-stage mechanical seal system for calculation to obtain p1, p2, p3 and q H The estimated value is taken as the difference between the estimated value and the true value to obtain the physical information loss function. Combined with the boundary loss function, the loss function is gradually reduced during the training process, and finally the r that meets the physical model constraints is obtained. i , according to r i The size of the coil determines the health of the seal and coil.

[0079] Example 2:

[0080] like Figure 4 The figure shows the dynamic operation state of a single-pole mechanical seal ring. In Example 1, an approximate linear model is used to construct the physical model, and in Example 2, a finite element method is used to construct the physical model. The following is the model building process:

[0081] (1) Film thickness equation, film thickness distribution will directly affect the sealing medium pressure and end face contact pressure distribution. The static ring has three degrees of freedom due to the floating installation method (see Figure 4 ), the dynamic ring has a certain movement form due to the rigid installation method, and the static ring movement U s and dynamic ring motion U r can be expressed as:

[0082]

[0083] Among them, z sis the movement of the static ring along the z-axis; γ sx and γ sy are the angular deflections of the static surrounding x-axis and y-axis respectively; z r is the axial movement of the dynamic ring, which is approximately zero; γ rx and γ ry γ is the angular deflection of the moving ring r The projection frequency ω on the x-axis and y-axis is consistent with the rotational speed frequency.

[0084] The surface profile of the sealing ring, manufacturing errors, and ring body movement all affect the sealing end face film thickness h, which can be expressed as:

[0085] h=h0+h gr +h mo +h pl +[1y-x](U s -U r )

[0086] Among them, U s is the static ring motion, U r is the motion of the dynamic ring, h0 is the equilibrium film thickness; h gr is the spiral groove depth; h pl is the flatness, including the waviness and taper produced during the machining process and the secondary deformation produced during the operation; h mo To monitor the groove depth; (x, y) covers the sealing ring range, which is defined as:

[0087]

[0088] Among them, r i is the inner diameter of the outlet, r o The outer diameter of the entrance.

[0089] (2) Lubrication equation

[0090] Assuming the gas is an ideal gas, when the film thickness is large, the absolute pressure of the fluid film is controlled by the transient Reynolds equation, and the boundary condition is the Dirichlet boundary:

[0091]

[0092]

[0093] Among them, the axis coordinate system (r, θ) is used for convenient calculation; μ is the gas dynamic viscosity coefficient; t is time; p i The inner diameter of the outlet r i Pressure at p o is the outer diameter of the inlet r o The pressure at the sealing end is h; the film thickness at the sealing end is h; To find the sign of the partial derivative; p is the flow field pressure at a certain point on the sealing end face; is the gradient of the flow field pressure at a certain point on the sealing end surface, and r is the radius at a certain point on the sealing end surface.

[0094] (3) Contact equation

[0095] Under the conditions of low-speed operation, disturbance working condition, and start-stop process, the sealing end face will have solid contact. Assuming that the contact roughness peak undergoes elastic deformation in slight contact and plastic deformation in severe contact, the roughness peak distribution obeys Gaussian distribution, and the contact pressure P c It can be solved by the Chang-Etsion-Bogy (CEB) model:

[0096]

[0097] Where y is the elastic modulus; η s is the rough peak density; R s is the average curvature radius of the roughness peak; σ s is the standard deviation of the roughness peak height; is the average height of the roughness peak; ω c is the limit deformation at which the plastic deformation of the roughness peak begins; p m is the maximum contact pressure under plastic deformation; φ(·) is the probability density function of the standard normal distribution.

[0098] (4) Three-degree-of-freedom dynamic equation

[0099] The three-degree-of-freedom motion of the static ring is expressed in the following equations, where the fluid pressure, contact pressure, external generalized equilibrium force, abnormal disturbance force, and system force are balanced:

[0100]

[0101] Among them, M s is the inertia matrix of the static ring; the auxiliary support system composed of springs and auxiliary seals is equivalent to the stiffness matrix C s and the damping matrix K s ;P f is the sealing end face fluid film pressure matrix; P c is the sealing end face contact pressure matrix; P b is the generalized equilibrium pressure matrix, which only has the axial force F b , including the compression force generated by the gas on the back of the static ring and the designed preload force; P d is the abnormal disturbance on the static ring, which is simplified into axial force and moment.

[0102] Therefore, the embodiment of the present invention can combine the above equations to establish a set of lubrication, contact and dynamic equations. The dynamic response of the sealing ring can be obtained by using the finite element method. Due to the limited efficiency of numerical simulation calculations, the embodiment of the present invention also needs to adopt an accelerated calculation method, such as an efficient numerical fall method or a proxy model, to finally obtain an efficient and high-fidelity physical model. The input of the physical model includes various operating parameters and sealing dynamic fault parameters. The dynamic fault parameters include the angular deflection γ of the dynamic ring. r , flatness h pl , static ring axial force F b The output of this model is the periodic displacement of the sealing ring and the periodic end face contact force. The periodic displacement can be monitored by an eddy current sensor, and the periodic contact force can be monitored by a torque sensor or an acoustic emission sensor.

[0103] Therefore, the inputs to the PINN are periodic displacements and periodic end-face contact forces, and the outputs are various seal dynamic fault parameters. Since the inputs are periodic vectors, the overall architecture of the PINN utilizes a CNN, with gradient descent updating the CNN's correlation coefficients. During each update, the dynamic parameters output by the PINN are substituted into the aforementioned physical model for calculation, yielding estimates of the periodic displacements and periodic end-face contact forces. The difference between the estimated values ​​and the true values ​​yields a physical information loss function. Combined with the boundary loss function, this loss function is gradually reduced during training, ultimately yielding dynamic parameters that meet the constraints of the physical model. The health of the mechanical seal can be determined based on the magnitude of these dynamic parameters.

[0104] For example, if Figure 5 As shown, the working principle of the present invention is described in detail below with a specific embodiment.

[0105] Step S501: Based on the existing sealing structure design and working conditions, a high-fidelity calculation model of the mechanical sealing system is established, and an accelerated calculation method is used to improve the simulation calculation efficiency.

[0106] Step S502: Establish a PINN network, where the input is the sensor monitoring signal and the output is the fault parameter to be identified. The PINN neural network's loss function consists of two components: boundary loss and physical information loss. The physical information loss is calculated by subtracting the simulation result from the sensor monitoring signal. After updating the network parameter weights, the loss function is reduced to below a threshold. The PINN output can be considered a reasonable fault parameter, effectively improving the accuracy and timeliness of mechanical seal fault diagnosis.

[0107] According to the PINN-based mechanical seal fault diagnosis method proposed in an embodiment of the present invention, the established multi-physics field coupling simulation model of the target mechanical seal system can be embedded into the loss function of the physical information neural network PINN to construct an initial PINN fault diagnosis model, and the actual monitoring data of multiple sensors of the target mechanical seal system are input into the initial PINN fault diagnosis model to output multiple fault parameters to be identified, thereby updating the initial PINN fault diagnosis model and generating a target PINN fault diagnosis model. The sealing state of the target mechanical seal system is diagnosed according to the target PINN fault diagnosis model, and a fault diagnosis result is generated, which effectively improves the accuracy and timeliness of mechanical seal fault diagnosis. Thus, the problem in the related art that the sealing performance cannot be evaluated in real time and comprehensively, resulting in insufficient accuracy and poor timeliness of mechanical seal fault diagnosis, is solved.

[0108] Next, a PINN-based mechanical seal fault diagnosis device according to an embodiment of the present invention will be described with reference to the accompanying drawings.

[0109] Figure 6 4 is a block diagram of a PINN-based mechanical seal fault diagnosis device according to an embodiment of the present invention.

[0110] like Figure 6 As shown, the PINN-based mechanical seal fault diagnosis device 10 includes: an establishment module 100 , a processing module 200 and a diagnosis module 300 .

[0111] Specifically, the establishment module 100 is used to establish a multi-physics field coupling simulation model of the target mechanical sealing system that meets preset conditions.

[0112] The processing module 200 is used to embed the multi-physics field coupling simulation model into the loss function of the physical information neural network PINN to construct an initial PINN fault diagnosis model, and input the actual monitoring data of multiple sensors of the target mechanical sealing system into the initial PINN fault diagnosis model to output multiple fault parameters to be identified.

[0113] The diagnostic module 300 is used to update the initial PINN fault diagnosis model based on multiple fault parameters to be identified, generate a target PINN fault diagnosis model, and diagnose the sealing state of the target mechanical seal system according to the target PINN fault diagnosis model, so as to generate a fault diagnosis result of the target mechanical seal system according to the sealing state.

[0114] Optionally, in one embodiment of the present invention, the diagnosis module 300 includes: an acquisition unit, a first determination unit, and a training unit.

[0115] The acquisition unit is used to input multiple fault parameters to be identified into the multi-physics field coupling simulation model to output simulated monitoring data of multiple sensors.

[0116] The first determining unit is configured to determine a physical information loss function in a loss function of the PINN based on a difference between the simulated monitoring data and the actual monitoring data.

[0117] The training unit is used to combine the physical information loss function with the boundary loss function in the PINN loss function to train the initial PINN fault diagnosis model to obtain the target PINN fault diagnosis model.

[0118] Optionally, in one embodiment of the present invention, the apparatus 10 of the embodiment of the present invention further includes: a first determining module and a second determining module.

[0119] The first determination module is used to determine the target boundary of the fault parameter to be identified.

[0120] The second determining module is configured to determine a boundary loss function in the PINN loss function based on a target boundary and a plurality of fault parameters to be identified.

[0121] Optionally, in one embodiment of the present invention, the boundary loss function is calculated as follows:

[0122]

[0123] Among them, Loss b is the boundary loss function, Loss bi is the i-th boundary loss function; x ipre is x i Output during PINN training; x imin is x i The lower boundary of x imax is x i The upper boundary of x i is the fault parameter to be identified.

[0124] Optionally, in one embodiment of the present invention, the physical information loss function is calculated as follows:

[0125]

[0126] Among them, Loss r is the physical information loss function, Loss ri is the physical information loss function of the i-th item; For sensor Y i Monitoring vector, N i is the vector length; To simulate Y based on multi-physics coupling model i estimated value.

[0127] Optionally, in one embodiment of the present invention, the diagnosis module 300 includes: a second determination unit and a diagnosis unit.

[0128] The second determining unit is configured to determine target parameters of the multi-physics field coupling simulation model that meet target constraint conditions based on the target PINN fault diagnosis model.

[0129] The diagnosis unit is used for diagnosing a sealing state of a target mechanical sealing system using target parameters.

[0130] It should be noted that the above explanation of the embodiment of the PINN-based mechanical seal fault diagnosis method is also applicable to the PINN-based mechanical seal fault diagnosis device of this embodiment, and will not be repeated here.

[0131] According to the PINN-based mechanical seal fault diagnosis device proposed in an embodiment of the present invention, the established multi-physics field coupling simulation model of the target mechanical seal system can be embedded into the loss function of the physical information neural network PINN to construct an initial PINN fault diagnosis model, and the actual monitoring data of multiple sensors of the target mechanical seal system can be input into the initial PINN fault diagnosis model to output multiple fault parameters to be identified, thereby updating the initial PINN fault diagnosis model and generating a target PINN fault diagnosis model to diagnose the sealing state of the target mechanical seal system according to the target PINN fault diagnosis model, generate a fault diagnosis result, and effectively improve the accuracy and timeliness of mechanical seal fault diagnosis. Thus, the problem in the related art that the sealing performance cannot be evaluated in real time and comprehensively, resulting in insufficient accuracy and poor timeliness of mechanical seal fault diagnosis, is solved.

[0132] Figure 7 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device may include:

[0133] Memory 701 , processor 702 , and computer programs stored in the memory 701 and executable on the processor 702 .

[0134] When the processor 702 executes the program, the PINN-based mechanical seal fault diagnosis method provided in the above embodiment is implemented.

[0135] Furthermore, the electronic device further includes:

[0136] The communication interface 703 is used for communication between the memory 701 and the processor 702 .

[0137] The memory 701 is used to store computer programs that can be run on the processor 702 .

[0138] The memory 701 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0139] If the memory 701, processor 702, and communication interface 703 are implemented independently, the communication interface 703, memory 701, and processor 702 can be interconnected via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0140] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can communicate with each other through an internal interface.

[0141] The processor 702 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0142] This embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the above-mentioned PINN-based mechanical seal fault diagnosis method is implemented.

[0143] This embodiment further provides a computer program product, including a computer program. When the computer program is executed, it is used to implement the above-mentioned PINN-based mechanical seal fault diagnosis method.

[0144] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples without contradiction.

[0145] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "N" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0146] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or N executable instructions for implementing a custom logical function or step of a process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0147] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or N wires (electronic devices), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program can be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or processing it in other suitable ways as necessary, and then storing it in a computer memory.

[0148] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, it can be implemented using any one or a combination of the following technologies known in the art: a discrete logic circuit having logic gate circuits for implementing logical functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0149] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0150] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0151] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and are not to be construed as limiting the present invention. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A mechanical seal fault diagnosis method based on PINN, characterized in that: The following steps are involved: Establish a multi-physics field coupling simulation model of the target mechanical seal system that meets preset conditions; Embedding the multi-physics coupling simulation model into the loss function of a physical information neural network (PINN) to construct an initial PINN fault diagnosis model, and inputting actual monitoring data from multiple sensors of the target mechanical seal system into the initial PINN fault diagnosis model to output multiple fault parameters to be identified; Based on the multiple fault parameters to be identified, the initial PINN fault diagnosis model is updated to generate a target PINN fault diagnosis model, so as to diagnose the sealing state of the target mechanical sealing system according to the target PINN fault diagnosis model, so that a fault diagnosis result of the target mechanical sealing system is generated according to the sealing state.

2. The mechanical seal fault diagnosis method based on PINN according to claim 1, characterized in that: The updating of the initial PINN fault diagnosis model based on the multiple fault parameters to be identified includes: Inputting the plurality of fault parameters to be identified into the multi-physics field coupling simulation model to output simulated monitoring data of the plurality of sensors; determining a physical information loss function in a loss function of the PINN based on a difference between the simulated monitoring data and the actual monitoring data; The physical information loss function is combined with a boundary loss function in the PINN loss function to train the initial PINN fault diagnosis model to obtain the target PINN fault diagnosis model.

3. The mechanical seal fault diagnosis method based on PINN according to claim 2, characterized in that: Before combining the physical information loss function with the boundary loss function in the PINN loss function, the method further includes: Determining target boundaries of the fault parameters to be identified; A boundary loss function in a loss function of the PINN is determined based on the target boundary and the multiple fault parameters to be identified.

4. The mechanical seal fault diagnosis method based on PINN according to claim 2, characterized in that: The boundary loss function is calculated as follows: Among them, Loss b is the boundary loss function, Loss bi is the i-th boundary loss function; x ipre is x i The output during the training process of the PINN; x imin is x i The lower boundary of x imax is x i The upper boundary of x i is the fault parameter to be identified.

5. The PINN-based mechanical seal fault diagnosis method according to claim 2, characterized in that: The physical information loss function is calculated as follows: Among them, Loss r is the physical information loss function, Loss ri is the physical information loss function of the i-th item; For sensor Y i Monitoring vector, N i is the vector length; Based on the multi-physics coupling simulation model, Y i estimated value.

6. The PINN-based mechanical seal fault diagnosis method according to claim 1, characterized in that: The diagnosing the sealing state of the target mechanical sealing system according to the target PINN fault diagnosis model includes: Determining target parameters of the multi-physics field coupling simulation model that meet target constraint conditions based on the target PINN fault diagnosis model; The sealing state of the target mechanical sealing system is diagnosed using the target parameters.

7. A mechanical seal fault diagnosis device based on PINN, characterized in that: include: Establishing a module for establishing a multi-physics field coupling simulation model of a target mechanical seal system that meets preset conditions; a processing module, configured to embed the multi-physics field coupling simulation model into a loss function of a physical information neural network (PINN) to construct an initial PINN fault diagnosis model, and input actual monitoring data of multiple sensors of the target mechanical seal system into the initial PINN fault diagnosis model to output multiple fault parameters to be identified; A diagnostic module is configured to update the initial PINN fault diagnostic model based on the multiple fault parameters to be identified, generate a target PINN fault diagnostic model, and diagnose the sealing state of the target mechanical sealing system according to the target PINN fault diagnostic model, so as to generate a fault diagnosis result of the target mechanical sealing system according to the sealing state.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the PINN-based mechanical seal fault diagnosis method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the PINN-based mechanical seal fault diagnosis method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that The computer program is executed by a processor to implement the PINN-based mechanical seal fault diagnosis method according to any one of claims 1 to 6.

Images