Fault detection method, device, equipment and medium based on physical information neural network model

Abstract

The present application relates to the technical field of mechanical fault diagnosis, and specifically provides a fault detection method, device, equipment and medium based on a physical information neural network model. The method of the present application includes: obtaining the external force of the target device and the actual displacement at each time point within a preset time period and the actual detection parameters of at least one target part at each time point within a preset time period, based on the external force and each actual detection parameter, using a first detection model to obtain the first predicted displacement at each time point, based on the external force and the actual displacement corresponding to each time point, using a second detection model to obtain the second predicted displacement and predicted detection parameters at each time point, if any displacement difference between the second predicted displacement at each time point and the first predicted displacement exceeds the preset safety displacement value range and / or if any parameter difference between the predicted detection parameter at each time point and the actual detection parameter exceeds the preset safety parameter difference range, then the target device has a fault. Through the above method, the accuracy of fault detection is improved.

Description

Technical Field

[0001] The present application relates to the technical field of mechanical fault diagnosis, and in particular to a fault detection method, device, equipment and medium based on a physical information neural network model. Background Art

[0002] Machinery troubleshooting is the process of detecting potential faults and problems by analyzing and evaluating the performance and status of mechanical equipment. It involves monitoring key parameters, observing equipment operating conditions, and using a variety of tools and techniques for data collection and analysis to identify faults in mechanical systems.

[0003] In the prior art, vibration diagnosis is an important method. Vibration diagnosis focuses on identifying potential faults and problems by analyzing and interpreting vibration signals in mechanical systems. However, the vibration models involved in actual industrial production are usually complex, involving limitations such as the nonlinear behavior of the vibration system, the coupling effects of multiple degrees of freedom, and various complex boundary conditions. Traditional parameter estimation algorithms often fail to fully consider these factors, resulting in the accuracy and robustness of the estimation results being limited. In addition, the acquisition process of vibration signals is often interfered by noise or data is missing, which brings huge challenges to parameter estimation. Summary of the invention

[0004] The application aims to solve the above-mentioned technical problem, namely, the vibration model is complex, and the traditional parameter estimation algorithm cannot fully consider factors such as the nonlinear behavior of the vibration system, the coupling effect of multiple degrees of freedom, and various complex boundary conditions, resulting in poor accuracy of the estimation results.

[0005] In a first aspect, an embodiment of the present application provides a fault detection method based on a physical information neural network model, including: obtaining an external force of a target device, and an actual displacement corresponding to each time point within a preset time period of the target device, and an actual detection parameter corresponding to each time point within a preset time period of at least one target part of the target device, the actual detection parameter including an actual mass, an actual damping coefficient, and an actual elastic coefficient; based on the external force and the actual detection parameter corresponding to each time point, using a pre-trained first detection model, obtaining a first predicted displacement corresponding to each time point; based on the external force and the actual displacement corresponding to each time point, using a pre-trained second detection model, obtaining a second predicted displacement and predicted detection parameters corresponding to each time point, the predicted detection parameters including a predicted mass, a predicted damping coefficient, and a predicted elastic number; if the displacement difference between the second predicted displacement corresponding to each time point and any one of the first predicted displacements exceeds a preset safety displacement value range, and / or if the parameter difference between the predicted detection parameter corresponding to each time point and any one of the actual detection parameters exceeds a preset safety parameter difference range, then the target device has a fault.

[0006] In a second aspect, an embodiment of the present application provides a fault detection device based on a physical information neural network model, including: an acquisition module, used to acquire an external force of a target device, and an actual displacement corresponding to each time point within a preset time period of the target device, and an actual detection parameter corresponding to each time point within a preset time period of at least one target part of the target device, wherein the actual detection parameter includes an actual mass, an actual damping coefficient, and an actual elastic coefficient; a first prediction module, used to obtain a first predicted displacement corresponding to each time point based on the external force and the actual detection parameter corresponding to each time point, using a pre-trained first detection model; a second prediction module, used to obtain a second predicted displacement and predicted detection parameters corresponding to each time point based on the external force and the actual displacement corresponding to each time point, using a pre-trained second detection model, wherein the predicted detection parameters include a predicted mass, a predicted damping coefficient, and a predicted elastic number; a judgment module, used to determine if the displacement difference between the second predicted displacement corresponding to each time point and the first predicted displacement exceeds a preset safe displacement value range, and / or if the parameter difference between the predicted detection parameter corresponding to each time point and the actual detection parameter exceeds a preset safe parameter difference range.

[0007] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described in the first aspect above is implemented.

[0008] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect above is implemented.

[0009] The technical solution provided in the embodiments of the present application has at least the following technical effects or advantages:

[0010] In an embodiment of the present application, an external force of a target device and an actual displacement corresponding to each time point within a preset time period of the target device and an actual detection parameter corresponding to each time point within a preset time period of at least one target part of the target device are obtained, wherein the actual detection parameter includes an actual mass, an actual damping coefficient and an actual elastic coefficient. Then, based on the external force and the actual detection parameter corresponding to each time point, a first predicted displacement corresponding to each time point is obtained by using a pre-trained first detection model. Then, based on the external force and the actual displacement corresponding to each time point, a second predicted displacement and predicted detection parameter corresponding to each time point are obtained by using a pre-trained second detection model. The predicted detection parameter includes a predicted mass, a predicted damping coefficient and a predicted elastic number. Finally, if the displacement difference between the second predicted displacement corresponding to each time point and the first predicted displacement exceeds a preset safety displacement value range, and / or if the parameter difference between the predicted detection parameter corresponding to each time point and the actual detection parameter exceeds a preset safety parameter difference range, then the target device has a fault, which can help engineers and technicians understand the operating status of the target device, determine whether there is a fault or abnormality, and provide accurate fault diagnosis results.

[0011] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0013] Figure 1 A flowchart of a fault detection method based on a physical information neural network model provided in an embodiment of the present application is shown;

[0014] Figure 2 A schematic diagram of the structure of a fault detection device based on a physical information neural network model provided in an embodiment of the present application is shown;

[0015] Figure 3A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0016] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0017] In the prior art, vibration diagnosis is an important method. Vibration diagnosis focuses on identifying potential faults and problems by analyzing and interpreting vibration signals in mechanical systems. However, the vibration models involved in actual industrial production are usually complex, involving limitations such as the nonlinear behavior of the vibration system, the coupling effects of multiple degrees of freedom, and various complex boundary conditions. Traditional parameter estimation algorithms often fail to fully consider these factors, resulting in the accuracy and robustness of the estimation results being limited. In addition, the acquisition of vibration signals is often subject to noise interference or data missing, which brings huge challenges to parameter estimation.

[0018] Based on this, the embodiment of the present application provides a fault detection method based on a physical information neural network model. The solution of the embodiment of the present application is described in detail below in conjunction with the accompanying drawings.

[0019] See also Figure 1 A flow chart of a fault detection method based on a physical information neural network model is shown, and the method specifically includes the following steps:

[0020] Step 101: Obtain the external force of the target device, the actual displacement corresponding to each time point within a preset time period of the target device, and the actual detection parameters corresponding to each time point within a preset time period of at least one target part of the target device, wherein the actual detection parameters include the actual mass, the actual damping coefficient and the actual elastic coefficient.

[0021] Step 102: Based on the external force and the actual detection parameters corresponding to each time point, a first predicted displacement corresponding to each time point is obtained using a pre-trained first detection model.

[0022] Step 103: Based on the external force and the actual displacement corresponding to each time point, the second predicted displacement and predicted detection parameters corresponding to each time point are obtained using a pre-trained second detection model. The predicted detection parameters include predicted mass, predicted damping coefficient and predicted elasticity times.

[0023] Step 104: If the difference between any one of the second predicted displacements corresponding to each time point and the first predicted displacement exceeds a preset safety displacement value range, and / or if the difference between any one of the predicted detection parameters corresponding to each time point and the actual detection parameters exceeds a preset safety parameter value range, the target device is faulty.

[0024] In the embodiment of the present application, the target device can be vibrated for faults through vibration diagnosis. Vibration diagnosis focuses on identifying potential faults and problems by analyzing and interpreting the vibration signals of the target device in the mechanical system. By performing spectral analysis and time domain analysis on the vibration signal, we can determine the type, location and severity of the fault to achieve accurate fault diagnosis.

[0025] The vibration equation is a mathematical equation that describes the vibration behavior of the target device in the mechanical system. By considering the mass, elasticity, damping and other parameters of the target device, as well as the external excitation conditions (external force), we can incorporate the vibration equation into the neural network as a constraint equation, establish the first detection model and the second detection model of the target device, and then use the PINN (Physics-Informed Neural Network) algorithm to predict the vibration signal of the target device based on the first detection model and the second detection model. Finally, we can determine whether the predicted vibration signal and the actual vibration signal are within the safe value range, and then determine whether the target device has a fault.

[0026] It should be noted that the vibration signal may refer to the displacement of the target device at each time point, and parameter information such as the mass, damping coefficient, elastic coefficient, etc. of at least one target part of the target device at each time point. This embodiment of the present application is not specifically limited.

[0027] Specifically, an implementation manner of the present application can obtain the external force of the target device, the actual displacement corresponding to each time point within a preset time period of the target device, and the actual detection parameters corresponding to each time point within the preset time period of at least one target part of the target device, the actual detection parameters including the actual mass, the actual damping coefficient and the actual elastic coefficient. Then, based on the external force and the actual detection parameters corresponding to each time point, a PINN algorithm is used to obtain the first predicted displacement corresponding to each time point based on a pre-trained first detection model. Based on the external force and the actual displacement corresponding to each time point, a PINN algorithm is used to obtain the second predicted displacement and predicted detection parameters corresponding to each time point based on a pre-trained second detection model. The predicted detection parameters include predicted mass, predicted damping coefficient and predicted elastic number. Finally, it is determined whether the second predicted displacement corresponding to each time period is within a preset safety displacement value range and the first predicted displacement, and / or whether the predicted detection parameter and the actual detection parameter are within a safety parameter value range. If any displacement difference between the second predicted displacement corresponding to each time point and the first predicted displacement exceeds the preset safety displacement value range, and / or if any parameter difference between the predicted detection parameter corresponding to each time point and the actual detection parameter exceeds the preset safety parameter value range, then the target device has a fault.

[0028] As an emerging parameter estimation method, the PINN algorithm does not require a complete and accurate mathematical model. The algorithm can obtain the unknown displacement and detection parameters in the constraint equation by learning the patterns and laws in the data, overcoming the dependence of the algorithm based on the mathematical model on the accurate model and expanding the scope of application of the algorithm; moreover, the PINN algorithm combines the characteristics of knowledge-driven and data-driven, and realizes the modeling of physical laws by incorporating physical prior knowledge (actual displacement and actual detection parameters) into the structure and loss function of the neural network of the first detection model and the second detection model, so that the PINN algorithm can flexibly use the actual displacement and actual detection parameters to predict the fault of the target device, and improve the accuracy and robustness of the fault detection of the target device through the model; finally, the PINN algorithm has excellent anti-noise performance and can effectively handle noise under different distribution conditions. The PINN algorithm uses physical prior knowledge to integrate physical laws into the first detection model and the second detection model, thereby constraining and guiding the noise, and improving the accuracy and robustness of the first predicted displacement, the second predicted displacement and the predicted detection parameter estimation.

[0029] In an embodiment of the present application, an external force of a target device and an actual displacement corresponding to each time point within a preset time period of the target device and an actual detection parameter corresponding to each time point within a preset time period of at least one target part of the target device are obtained, wherein the actual detection parameter includes an actual mass, an actual damping coefficient and an actual elastic coefficient. Then, based on the external force and the actual detection parameter corresponding to each time point, a first predicted displacement corresponding to each time point is obtained by using a pre-trained first detection model. Then, based on the external force and the actual displacement corresponding to each time point, a second predicted displacement and predicted detection parameter corresponding to each time point are obtained by using a pre-trained second detection model. The predicted detection parameter includes a predicted mass, a predicted damping coefficient and a predicted elastic number. Finally, if the displacement difference between the second predicted displacement corresponding to each time point and the first predicted displacement exceeds a preset safety displacement value range, and / or if the parameter difference between the predicted detection parameter corresponding to each time point and the actual detection parameter exceeds a preset safety parameter difference range, then the target device has a fault, which can help engineers and technicians understand the operating status of the target device, determine whether there is a fault or abnormality, and provide accurate fault diagnosis results.

[0030] On the basis of the above implementation manner, in some modified implementation manners, based on the external force and the actual detection parameters corresponding to each time point, the first detection model can be trained before obtaining the first predicted displacement corresponding to each time point using a pre-trained first detection model. Specifically, the actual detection parameters and actual displacements of a first number of first time points can be selected from each time point, wherein the first number of first time points at least includes the starting time point of a preset time period, and then training is performed based on the actual detection parameters and actual displacements of the first number of first time points to obtain a first loss value. If the first loss value satisfies the first preset loss value, the first detection model is obtained.

[0031] Furthermore, training based on the actual detection parameters and actual displacements of the first number of first time points can include solving the constraint equations of the first detection model to be trained based on the actual detection parameters corresponding to the starting time point of at least one target part to obtain a first training displacement corresponding to the starting time point of at least one target part, and then calculating the first loss value of the first detection model to be trained based on the first training displacement, the actual displacement corresponding to the starting time point, and the actual displacements corresponding to the first number of first time points excluding the starting time point, using a pre-constructed virtual function.

[0032] It should be noted that the constraint equation is obtained by transforming the vibration equation in the form of an ordinary differential equation into a constraint equation in the form of a partial differential equation. For example, if the second-order ordinary differential vibration equation is as shown in the following formula (1):

[0033]

[0034] in, is with The displacement of the relevant target device, is the quality of the target device, is the damping coefficient of the target device, is the elastic coefficient of the target device, It is an external force applied from the outside.

[0035] Our goal in training the first detection model is to obtain and a range of quality Displacement under , that is, the goal of training the first detection model is to enable the first detection model to solve the constraint equation in the form of partial differential equations using the PINN algorithm. We assume that the parameters Known and Fixed, define the equation input as Based on these assumptions and known conditions, we can obtain the constraint equation of the first detection model in the form of partial differential equation as shown in the following formula (2):

[0036]

[0037] in, represents the first training displacement corresponding to the starting time point of at least one target part, Indicates the actual mass corresponding to the starting time point of at least one target part, Indicates the actual damping coefficient corresponding to the starting time point of at least one target part, Indicates the actual elastic coefficient corresponding to the starting time point of at least one target part, Represents the external force on the target device.

[0038] After clarifying the input and output of the first detection model, the constructed virtual function can be shown as the following formula (3):

[0039]

[0040] in, represents the first loss value, N represents the number of target parts, represents the first training displacement corresponding to the starting time point of the Nth target component, represents the actual displacement corresponding to the starting time point of the Nth target component, O represents the number of all first time points except the starting time point of at least one target part, represents the actual displacement corresponding to the Oth first time point, represents the actual detection parameters corresponding to the starting time point of the Nth target component, Indicates the starting time point of the Nth target component, represents the actual detection parameter corresponding to the Oth first time point, Represents the Oth first time point.

[0041] On the basis of the above-mentioned implementation manner, in some changed implementation manner, based on the external force and the actual displacement corresponding to each time point, the second detection model can be trained before obtaining the second predicted displacement and predicted detection parameters corresponding to each time point using a pre-trained second detection model. Specifically, the actual detection parameters and actual displacement of a second number of second time points and the actual detection parameters and actual displacement of a third number of third time points can be selected from each time point. Then, the second detection model to be trained is trained based on the actual detection parameters and actual displacement of the second number of second time points and the actual detection parameters and actual displacement of the third number of third time points to obtain a second loss value. If the second loss value satisfies the second preset loss value, the second detection model is obtained.

[0042] Furthermore, training the second detection model to be trained based on the actual detection parameters and actual displacements of the second number of second time points, and the actual detection parameters and actual displacements of the third number of third time points can include: solving the constraint equations of the second detection model to be trained based on the actual displacements corresponding to the second number of second time points to obtain the first training detection parameters and the second training displacements corresponding to the second number of second time points, and calculating the second loss value of the second detection model to be trained based on the second training displacements and actual displacements corresponding to the second number of second time points, and the actual displacements corresponding to the third number of third time points, using a pre-constructed virtual function.

[0043] It should be noted that the constraint equation in the second detection model is the same as the constraint equation in the first detection model. However, when the constraint equation represented by the above formula (2) is used as the constraint equation in the second detection model, represents the second training displacement corresponding to the second number of second time points, represents the actual mass corresponding to the second number of second time points, represents the actual damping coefficient corresponding to the second number of second time points, represents the actual elastic coefficient corresponding to the second number of second time points, Represents the external force of the target device. At this time, our goal of training the second detection model is to enable the second detection model to predict the detection parameters in the constraint equation in the form of partial differential equations using the PINN algorithm.

[0044] Similarly, after clarifying the input and output of the second detection model, the constructed virtual function is also as shown in the above formula (3). However, when the virtual function represented by the above formula (3) is the virtual function of the second detection model, represents the second loss value, N represents the second quantity, represents the second training displacement corresponding to the Nth second time point, represents the actual displacement corresponding to the Nth second time point, O represents the third quantity, represents the actual displacement corresponding to the Oth third time point, represents the actual detection parameter corresponding to the Nth second time point, represents the Nth second time point, represents the actual detection parameter corresponding to the Oth third time point, Indicates the Oth third time point.

[0045] The above implementation method describes what parameters to input and what data to output for training the first detection model and the second detection model respectively. Then, the steps for training the first detection model and the second detection model are the same, except that the input parameters and the output data are different. In order to avoid repetition, only the unified training steps are described here. For the input and output of the relevant parameters, please refer to the above implementation method.

[0046] Typically, model training involves two steps: forward propagation and back propagation.

[0047] In the forward propagation phase, we feed the input data into the model's network, and then use the current weights and biases to calculate the network output and calculate the loss function. Specifically, for each sample, the following operations are performed: the input data is passed to the first layer of neurons, and the weights and biases are multiplied and summed to calculate the weighted sum of the first layer of neurons; an activation function (such as hyperbolic tangent) is applied to the weighted sum function) to calculate the output of the first layer of neurons; pass the output of the first layer of neurons to the next layer of neurons, and repeat the above steps until the output of the network is calculated; calculate the loss function to measure the error between the output of the network and the actual output.

[0048] In the back-propagation phase, we use automatic differentiation techniques to calculate the derivatives of the loss function with respect to weights and biases, and use optimization algorithms (such as Adam and L-BFGS algorithms) to use the derivatives to update weights and biases. Specifically, for each sample, the following operations are performed: the derivative of the loss function with respect to the network output is calculated; the derivative of the loss function with respect to weights and biases is calculated using the chain rule as follows ; Use an optimization algorithm to update weights and biases using derivatives.

[0049] Forward propagation and backpropagation are repeated until the network converges or the maximum number of iterations is reached.

[0050] In the above-mentioned changed implementation methods, in some changed implementation methods, in addition to using the PINN algorithm, the UOPINN (Universal Oil Products (UOP) and Physics-informed Neural Networks (PINN) combined algorithm) algorithm can also be used: UOPINN can handle various partial differential equations; DD (Data-driven discretization) algorithm: DD solves partial differential equations through time-space discretization technology. It uses PINN instead of polynomial functions in the finite difference method, and it can solve various partial differential equations; fPINN (fractional Physics-Informed Neural Networks) algorithm: Combining PINN and time discretization technology, fPINN is used to solve time fractional partial differential equations; PPINN (Physics-informed Probabilistic Neural Networks) algorithm: Combining PINN and time discretization technology, fPINN is used to solve time fractional partial differential equations; PPINN (Physics-Informed Neural Networks) algorithm: By decomposing a long-term problem into multiple independent short-term problems to reduce the algorithm's training cost, PPINN can handle long-term problems more efficiently than PINN; vPINN (Variational Physics-Informed Neural Networks) algorithm: By integrating the integrand in the variational, the vPINN algorithm reduces the order of the differential operator, thereby effectively reducing the training cost; B-PINNs (Bayesian Physics-Informed Neural Networks) algorithm: In the Bayesian framework, the Bayesian neural network takes the neural network representing the unknown function of the partial differential square as the prior, and uses Hamiltonian Monte Carlo inference or variational inference for posterior estimation. B-PINNs use physically determined and dispersed noise measurements to provide predictions and quantify the uncertainty caused by noisy data in the Bayesian framework.

[0051] For specific practical problems, since differential equations may have different characteristics, we can adopt different PINN variant algorithm variants to estimate parameters of complex differential equations in a targeted manner; in addition, some variant algorithms may improve calculation accuracy or reduce training costs in specific scenarios. We can choose the best PINN variant algorithm according to the specific situation.

[0052] On the basis of the above-mentioned implementation manner, in some modified implementation manners, if the difference between the predicted mass and the actual mass exceeds the preset safety mass value range, it can be determined that a load is attached to the target device, or that corrosion or material loss has occurred to at least one target part of the target device; if the difference between the predicted damping coefficient and the actual damping coefficient exceeds the preset safety damping coefficient value range, it can be determined that the target device is poorly lubricated or that the friction between the parts of the target device has increased, and if the difference between the predicted damping coefficient and the actual damping coefficient is less than the minimum value of the safety damping coefficient value range, it can be determined that the target device is loose or the sealing has deteriorated; if the difference between the predicted elastic coefficient and the actual elastic coefficient exceeds the preset safety elastic coefficient value range, it can be determined that the spring or supporting component of the target device is fatigued, damaged or aged.

[0053] The fault detection method based on the physical information neural network model of the embodiment of the present application overcomes the limitations of existing algorithms by applying the PINN algorithm to parameter estimation in ordinary differential equations, and exhibits strong noise resistance suitable for complex vibration models, and can be applied to more complex vibration models.

[0054] See also Figure 2 The embodiment of the present application further provides a fault detection device based on a physical information neural network model, which is used to execute the fault detection method based on a physical information neural network model described in the above embodiment, and the device includes:

[0055] An acquisition module 201 is used to acquire the external force of the target device, the actual displacement corresponding to each time point within a preset time period of the target device, and the actual detection parameters corresponding to each time point within a preset time period of at least one target part of the target device, wherein the actual detection parameters include the actual mass, the actual damping coefficient, and the actual elastic coefficient;

[0056] A first prediction module 202 is used to obtain a first predicted displacement corresponding to each time point based on the external force and the actual detection parameters corresponding to each time point using a pre-trained first detection model;

[0057] The second prediction module 203 is used to obtain the second predicted displacement and predicted detection parameters corresponding to each time point based on the external force and the actual displacement corresponding to each time point by using the pre-trained second detection model, and the predicted detection parameters include predicted mass, predicted damping coefficient and predicted elasticity number;

[0058] The judgment module 204 is used to determine that a target device has a fault if the displacement difference between the second predicted displacement corresponding to each time point and any one of the first predicted displacements exceeds a preset safety displacement value range, and / or if the parameter difference between the predicted detection parameter corresponding to each time point and any one of the actual detection parameters exceeds a preset safety parameter difference range.

[0059] The fault detection device based on the physical information neural network model provided in the embodiment of the present application and the fault detection method based on the physical information neural network model provided in the above embodiment are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.

[0060] The present application also provides an electronic device corresponding to the fault detection method based on the physical information neural network model provided in the above embodiment. Figure 3 , which shows a schematic diagram of an electronic device provided by some embodiments of the present application. Figure 3 As shown, the electronic device 30 may include: a processor 300, a memory 301, a bus 302 and a communication interface 303, wherein the processor 300, the communication interface 303 and the memory 301 are connected via the bus 302; the memory 301 stores a computer program that can be run on the processor 300, and when the processor 300 runs the computer program, the fault detection method based on the physical information neural network model provided in any of the aforementioned embodiments of the present application is executed.

[0061] The memory 301 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one physical port 303 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0062] The bus 302 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 301 is used to store programs, and the processor 300 executes the programs after receiving the execution instruction. The fault detection method based on the physical information neural network model disclosed in any implementation of the embodiment of the present application may be applied to the processor 300, or implemented by the processor 300.

[0063] The processor 300 may be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 300 or an instruction in the form of software. The above processor 300 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a readily available programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module may be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 301, and the processor 300 reads the information in the memory 301 and completes the steps of the above method in combination with its hardware.

[0064] The electronic device provided in the embodiment of the present application and the fault detection method based on the physical information neural network model provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.

[0065] An embodiment of the present application also provides a computer-readable storage medium corresponding to the fault detection method based on the physical information neural network model provided in the aforementioned embodiment, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it will execute the fault detection method based on the physical information neural network model provided in any of the aforementioned embodiments.

[0066] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0067] An embodiment of the present application also provides a computer program product corresponding to the fault detection method based on the physical information neural network model provided in the aforementioned embodiment, including a computer program, which is executed by a processor to implement the fault detection method based on the physical information neural network model provided in the aforementioned embodiments.

[0068] The computer-readable storage medium and computer program product provided in the above-mentioned embodiments of the present application are based on the same inventive concept as the fault detection method based on the physical information neural network model provided in the embodiments of the present application, and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0069] It should be noted that:

[0070] The algorithm and display provided herein are not inherently related to any particular computer, virtual device or other equipment. Various general purpose devices can also be used together with the teachings based on this. According to the above description, it is obvious to construct the structure required for this type of device. In addition, the application is not directed to any specific programming language. It should be understood that the content of the application described herein can be realized using various programming languages, and the description of the specific language above is to disclose the best mode of implementation of the application.

[0071] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0072] Similarly, it should be understood that in order to streamline the present application and help understand one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present application, the various features of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be interpreted as reflecting the following intention: the claimed application requires more features than the features clearly stated in each claim. More specifically, as reflected in the claims below, the inventive aspects are less than all the features of the single embodiment disclosed above. Therefore, the claims following the specific embodiment are hereby expressly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present application.

[0073] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0074] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present application and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0075] The various component embodiments of the present application can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all functions of some or all components in the creation device of the virtual machine according to the embodiment of the present application. The present application can also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present application can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0076] It should be noted that the above embodiments illustrate the present application rather than limit the present application, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets should not be constructed as a limitation to the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present application may be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names.

[0077] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A fault detection method based on a physical information neural network model, characterized in that: include: Acquire the external force of the target device, the actual displacement corresponding to each time point within a preset time period of the target device, and the actual detection parameters corresponding to each time point within a preset time period of at least one target part of the target device, wherein the actual detection parameters include the actual mass, the actual damping coefficient and the actual elastic coefficient; Based on the external force and the actual detection parameters corresponding to each time point, using a pre-trained first detection model, obtaining a first predicted displacement corresponding to each time point; Based on the external force and the actual displacement corresponding to each time point, using a pre-trained second detection model, obtain a second predicted displacement and predicted detection parameters corresponding to each time point, wherein the predicted detection parameters include predicted mass, predicted damping coefficient and predicted elasticity number; If the displacement difference between the second predicted displacement corresponding to each time point and any one of the first predicted displacements exceeds a preset safety displacement value range, and / or if the parameter difference between the predicted detection parameter corresponding to each time point and any one of the actual detection parameters exceeds a preset safety parameter value range, then the target device is faulty.

2. The fault detection method based on the physical information neural network model according to claim 1 is characterized in that: Before obtaining the first predicted displacement corresponding to each time point based on the external force and the actual detection parameters corresponding to each time point by using a pre-trained first detection model, the method further includes: Selecting actual detection parameters and actual displacements of a first number of first time points from each time point, wherein the first number of first time points at least includes a starting time point of the preset time period; Performing training based on the actual detection parameters and the actual displacements at the first number of first time points to obtain a first loss value; If the first loss value meets a first preset loss value, the first detection model is obtained.

3. The fault detection method based on the physical information neural network model according to claim 2 is characterized in that: The training based on the actual detection parameters and the actual displacements of the first number of first time points includes: Based on the actual detection parameters corresponding to the starting time point of at least one target part, the constraint equation of the first detection model to be trained is used to solve the first training displacement corresponding to the starting time point of at least one target part; Based on the first training displacement, the actual displacement corresponding to the starting time point and the actual displacements corresponding to the first number of first time points other than the starting time point, a first loss value of the first detection model to be trained is calculated using a pre-constructed virtual function.

4. The fault detection method based on physical information neural network model according to claim 1 is characterized in that: Before obtaining the second predicted displacement and predicted detection parameters corresponding to each time point based on the external force and the actual displacement corresponding to each time point by using a pre-trained second detection model, the method further includes: Selecting actual detection parameters and actual displacements of a second number of second time points and actual detection parameters and actual displacements of a third number of third time points from each time point; Training the second detection model to be trained based on the actual detection parameters and the actual displacement of the second number of second time points and the actual detection parameters and the actual displacement of the third number of third time points to obtain a second loss value; If the second loss value meets the second preset loss value, the second detection model is obtained.

5. The fault detection method based on physical information neural network model according to claim 4 is characterized in that: The training of the second detection model to be trained based on the actual detection parameters and the actual displacement of the second number of second time points and the actual detection parameters and the actual displacement of the third number of third time points includes: Based on the actual displacements corresponding to the second number of second time points, solving the constraint equations of the second detection model to be trained to obtain first training detection parameters and second training displacements corresponding to the second number of second time points; Based on the second training displacements and actual displacements corresponding to the second number of second time points, and the actual displacements corresponding to the third number of third time points, a second loss value of the second detection model to be trained is calculated using a pre-constructed virtual function.

6. The fault detection method based on the physical information neural network model according to claim 3 or 5, characterized in that: The pre-built virtual functions include: Among them, when When represents the first loss value, N represents the number of target parts, represents the first training displacement corresponding to the starting time point of the Nth target component, represents the actual displacement corresponding to the starting time point of the Nth target component, O represents the number of all first time points except the starting time point of at least one target part, represents the actual displacement corresponding to the Oth first time point, represents the actual detection parameters corresponding to the starting time point of the Nth target component, Indicates the starting time point of the Nth target component, represents the actual detection parameter corresponding to the Oth first time point, represents the Oth first time point; when When represents the second loss value, N represents the second quantity, represents the second training displacement corresponding to the Nth second time point, represents the actual displacement corresponding to the Nth second time point, O represents the third quantity, represents the actual displacement corresponding to the Oth third time point, represents the actual detection parameter corresponding to the Nth second time point, represents the Nth second time point, represents the actual detection parameter corresponding to the Oth third time point, Indicates the Oth third time point.

7. The fault detection method based on the physical information neural network model according to claim 3 or 5, characterized in that: The constraint equations include: Wherein, when the constraint equation represents the constraint equation of the first detection model, represents the first training displacement corresponding to the starting time point of at least one target part, represents the actual mass corresponding to the starting time point of at least one target part, represents the actual damping coefficient corresponding to the starting time point of at least one target part, represents the actual elastic coefficient corresponding to the starting time point of at least one target part, represents the external force of the target device; when the constraint equation represents the constraint equation of the second detection model, represents the second training displacement corresponding to the second number of second time points, represents the actual mass corresponding to the second number of second time points, represents the actual damping coefficient corresponding to the second number of second time points, represents the actual elastic coefficient corresponding to the second number of second time points, Represents the external force of the target device.

8. A fault detection device based on a physical information neural network model, characterized in that: include: An acquisition module, used to acquire the external force of the target device, the actual displacement corresponding to each time point within a preset time period of the target device, and the actual detection parameters corresponding to each time point within a preset time period of at least one target part of the target device, wherein the actual detection parameters include the actual mass, the actual damping coefficient and the actual elastic coefficient; A first prediction module, configured to obtain a first predicted displacement corresponding to each time point based on the external force and the actual detection parameters corresponding to each time point by using a pre-trained first detection model; A second prediction module is used to obtain a second predicted displacement and predicted detection parameters corresponding to each time point based on the external force and the actual displacement corresponding to each time point by using a pre-trained second detection model, wherein the predicted detection parameters include a predicted mass, a predicted damping coefficient and a predicted elasticity number; A judgment module is used to determine that if the displacement difference between the second predicted displacement corresponding to each time point and any one of the first predicted displacements exceeds a preset safety displacement value range, and / or if the parameter difference between the predicted detection parameter corresponding to each time point and any one of the actual detection parameters exceeds a preset safety parameter difference range, then the target device has a fault.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images