An online monitoring and fault diagnosis system for motors based on the Internet of Things

Abstract

The present invention relates to the technical field of motor fault diagnosis, specifically to an online motor monitoring and fault diagnosis system based on the Internet of Things, including a three-dimensional visualization unit, a first feature extraction unit, a second feature extraction unit, a fault diagnosis unit and a data analysis unit, and also includes: a three-dimensional modeling unit, the three-dimensional modeling unit is used to perform geometric modeling on multiple target motors in a target device through three-dimensional modeling technology to obtain multiple target motor geometric models, and to perform texture rendering on the multiple target motor geometric models to obtain multiple target motor three-dimensional models, and to integrate the multiple target motor three-dimensional models according to the actual positions of the target motors in the target device to obtain a scene model of the target device. The present invention uses three-dimensional modeling and texture rendering technology through the three-dimensional modeling unit to truly and intuitively display the geometric structure of the target device and its internal motor.

Description

Technical Field

[0001] The present invention relates to the technical field of motor fault diagnosis, and in particular to an online motor monitoring and fault diagnosis system based on the Internet of Things. Background Art

[0002] The Internet of Things refers to a network that connects any object to the Internet through various information sensing devices and in accordance with agreed protocols to exchange and communicate information in order to achieve intelligent identification, positioning, tracking, monitoring and management.

[0003] Traditional systems usually lack intuitive 3D modeling and texture rendering technology, and cannot display the equipment and its internal structure in a realistic and intuitive way, which may lead to inefficiency of maintenance personnel in locating and identifying motor components in the equipment. In addition, the data acquisition method of traditional systems may not be comprehensive enough, and the vibration signal and three-phase stator current signal of the motor cannot be collected in real time. Therefore, there are deficiencies in providing comprehensive information on the motor's operating status. Traditional systems may lack the ability to decompose the vibration signal sequence and extract useful feature information, which will lead to the inability to fully capture the subtle features of the motor state changes, thereby affecting the sensitivity and accuracy of fault detection. Traditional systems may lack efficient feature extraction and fault diagnosis algorithms, and cannot fully utilize feature vectors for accurate fault classification, resulting in low diagnostic accuracy. Traditional systems usually lack intuitive 3D visualization display methods, making it difficult for users to intuitively view the motor and its fault information. At the same time, there is a lack of convenient ways to quickly obtain detailed fault information, which reduces the user-friendliness and ease of use of the system. Summary of the invention

[0004] The purpose of the present invention is to solve the problems existing in the background technology and to propose an online monitoring and fault diagnosis system for motors based on the Internet of Things.

[0005] The technical solution of the present invention is: a motor online monitoring and fault diagnosis system based on the Internet of Things, including a three-dimensional visualization unit, a first feature extraction unit, a second feature extraction unit, a fault diagnosis unit and a data analysis unit, and also includes:

[0006] A three-dimensional modeling unit, the three-dimensional modeling unit is used to perform geometric modeling on multiple target motors in a target device by using a three-dimensional modeling technology to obtain multiple target motor geometric models, and perform texture rendering on the multiple target motor geometric models to obtain multiple target motor three-dimensional models, integrate the multiple target motor three-dimensional models according to the actual positions of the target motors in the target device to obtain a scene model of the target device, and transmit the scene model of the target device to a three-dimensional visualization unit;

[0007] A data acquisition unit, wherein the data acquisition unit is used to acquire a vibration signal of the target motor by using a built-in acceleration sensor in the target motor to obtain a vibration signal sequence of the target motor, and to acquire a three-phase stator current signal of the target motor by using a built-in current sensor in the target motor to obtain a three-phase stator current signal sequence of the target motor, and transmit the vibration signal sequence to a signal decomposition unit, and transmit the three-phase stator current signal sequence of the target motor to a data analysis unit;

[0008] A signal decomposition unit receives the vibration signal sequence transmitted by the data acquisition unit, and performs signal decomposition on the vibration signal sequence to obtain all modal components and residual components corresponding to the vibration signal sequence, and transmits all the modal components to the first feature extraction unit and the second feature extraction unit.

[0009] Preferably, the first feature extraction unit receives all the modal components transmitted by the signal decomposition unit, calculates the energies corresponding to all the modal components, and sums all the energies to obtain the total energy of all the modal components, calculates the energy proportion of each modal component based on the total energy and energy, and calculates the energy entropy of each modal component based on the energy proportion, uses the energy entropy as the first eigenvector, and transmits the first eigenvector to the fault diagnosis unit.

[0010] Preferably, the second feature extraction unit receives all the modal components transmitted by the signal decomposition unit, extracts time-frequency domain features of all the modal components, uses the time-frequency domain features as a second feature vector, and transmits the second feature vector to the fault diagnosis unit.

[0011] Preferably, the fault diagnosis unit receives the first feature vector transmitted by the first feature extraction unit and the second feature vector transmitted by the second feature extraction unit, and inputs the first feature vector and the second feature vector into a trained fault diagnosis model, and outputs the diagnosis type corresponding to the target motor through the trained fault diagnosis model. The fault diagnosis model includes a support vector machine algorithm and a swarm intelligence optimization algorithm. The support vector machine algorithm is used to classify the fault type of the target motor to obtain the diagnosis type corresponding to the target motor. The swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, and the parameters include penalty parameters and kernel function parameters, and the diagnosis type corresponding to the target motor is transmitted to the three-dimensional visualization unit.

[0012] Preferably, the data analysis unit receives the three-phase stator current signal sequence of the target motor transmitted by the data acquisition unit, and performs three-layer wavelet decomposition on the three-phase stator current signal sequence to obtain wavelet decomposition coefficients, wherein the wavelet decomposition coefficients include 1 high-frequency coefficient and 3 low-frequency coefficients, and calculates the binary norm of the wavelet decomposition coefficients, and uses the binary norm as the ordinate, and uses the time of receiving the three-phase stator current signal sequence of the target motor as the abscissa to construct a time series graph, and transmit the time series graph to the three-dimensional visualization unit.

[0013] Preferably, the three-dimensional visualization unit receives the scene model of the target device transmitted by the three-dimensional modeling unit, the diagnosis type corresponding to the target motor transmitted by the fault diagnosis unit, and the time series diagram transmitted by the data analysis unit, and displays the scene model of the target device in the intelligent terminal, hides the diagnosis type corresponding to the target motor and the time series diagram in the corresponding three-dimensional model of the target motor in the scene model of the target device, and when the target user clicks on the corresponding three-dimensional model of the target motor in the scene model of the target device, the diagnosis type corresponding to the target motor and the time series diagram pop up in the form of a pop-up window.

[0014] Preferably, performing signal decomposition on the vibration signal sequence to obtain a modal component and a residual component corresponding to the vibration signal sequence comprises the following steps:

[0015] A1. Adding Gaussian white noise to the vibration signal sequence to obtain a vibration signal sequence containing controllable noise. The vibration signal sequence containing controllable noise is expressed as follows:

[0016] ;

[0017] in, represents a vibration signal with controllable noise, Indicates vibration signal, Indicates the initial value of the amplitude coefficient of the added Gaussian white noise, Indicates Gaussian white noise is added, Indicates The first-order modal component of the Gaussian white noise added is generated by the empirical mode decomposition;

[0018] A2. Performing a first-order decomposition of the vibration signal containing controllable noise by empirical mode decomposition to obtain a first-order residual component and a first-order modal component. The expression of the first-order residual component is as follows:

[0019] ;

[0020] in, represents the first-order residual component, represents an operator for solving the local mean of a vibration signal containing controllable noise;

[0021] The expression of the first-order modal component is as follows:

[0022] ;

[0023] in, represents the first-order modal component;

[0024] A3. Performing a k-th order decomposition on the vibration signal containing controllable noise by empirical mode decomposition to obtain a k-th order residual component and a k-th order modal component. The expression of the k-th order residual component is as follows:

[0025] ;

[0026] in, represents the k-th order residual component, represents the k-1th order residual component, represents the amplitude coefficient of the k-1th addition of Gaussian white noise, Indicates The k-th modal component generated by the empirical mode decomposition of the Gaussian white noise added for the second time;

[0027] The expression of the k-th order modal component is as follows:

[0028] ;

[0029] in, represents the k-th order modal component;

[0030] A4. Repeat step A3 until all vibration signals containing controllable noise in the vibration signal sequence containing controllable noise are traversed to obtain all modal components and residual components corresponding to the vibration signal sequence.

[0031] Preferably, extracting the time-frequency domain features of all the modal components comprises the following steps:

[0032] B1. Calculate the correlation coefficient between each modal component and the vibration signal respectively, and compare the correlation coefficient with a preset correlation coefficient threshold, and take the modal component greater than the preset correlation coefficient threshold as a valid modal component;

[0033] B2. Performing Hilbert transform on the effective modal components to obtain complex analytical signals corresponding to the effective modal components, and calculating the instantaneous frequency of each of the complex analytical signals. The complex analytical signal expression is as follows:

[0034] ;

[0035] in, represents the complex analytical signal, Indicates time, represents the order of the effective modal component;

[0036] The instantaneous frequency expression is as follows:

[0037] ;

[0038] in, represents the instantaneous frequency;

[0039] B3. Calculate the fractal box dimension of the instantaneous frequency of the effective modal component, and construct a fractal box dimension feature vector, and use the fractal box dimension feature vector as the second feature vector. The fractal box dimension expression is as follows:

[0040] ;

[0041] in, represents the fractal box dimension, represents any constant term, The maximum diameter representing the minimum mesh count, represents the minimum grid count, represents the end point of the interval with good linearity, Indicates the starting point of the interval with good linearity.

[0042] Preferably, the swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, comprising the following steps:

[0043] C1. Randomly initialize the particle population, where the particle represents a combination of the penalty parameter and the kernel function parameter. The random initialization expression is as follows:

[0044] ;

[0045] in, Indicates Particle No. The location of the dimension variable, Indicates that in the interval The random number inside represents the maximum diameter of the minimum mesh count, and Respectively represent Upper and lower bounds on dimensional variables;

[0046] The matrix expression of the particle population is as follows:

[0047] ;

[0048] in, is the matrix representing the particle population, Indicates The position of a particle;

[0049] C2. Construct the objective function of the particle based on the matrix of the particle population. The matrix expression of the objective function is as follows:

[0050] ;

[0051] in, The matrix representing the objective function of the particle, Indicates The objective function value of a particle.

[0052] Preferably, the swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, and further comprises the following steps:

[0053] C3. Based on the objective function of the particle, the particle is moved to obtain a new position of the particle, and it is determined whether the objective function value of the new position of the particle is greater than the objective function value of the original position of the particle. If it is greater than the objective function value of the original position of the particle, the original position of the particle is updated to the new position. The new position movement expression is as follows:

[0054] ;

[0055] in, Indicates Particle No. The new position of the dimension variable, Indicates that the target particle is The location of the dimension variable, represents the objective function value of the target particle;

[0056] The new position update expression is as follows:

[0057] ;

[0058] in, Indicates The updated position of the particle, Indicates The new position of the particle, Indicates The original position of the particle, Indicates The objective function value of the new position of the particle;

[0059] C4. Move the new position of the particle to obtain the final position of the particle, and judge whether the objective function value of the final position of the particle is greater than the objective function value of the new position of the particle. If it is greater than the objective function value of the new position of the particle, update the new position of the particle to the final position. The final position movement expression is as follows:

[0060] ;

[0061] in, Indicates Particle No. The final position of the dimensional variable, Represents a constant parameter, represents the maximum number of iterations, Indicates the current iteration number;

[0062] The final position update expression is as follows:

[0063] ;

[0064] in, Indicates The final position of the particle after update, Indicates The final position of a particle, Indicates The objective function value of the final position of the particle.

[0065] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0066] 1. The present invention utilizes 3D modeling and texture rendering technology through a 3D modeling unit to truly and intuitively display the geometric structure of the target device and its internal motor. This visualization method facilitates maintenance personnel to quickly locate and identify motor components in the equipment, thereby improving inspection and maintenance efficiency. The data acquisition unit collects the vibration signal and three-phase stator current signal of the motor in real time, thereby providing comprehensive information on the motor's operating status, laying a foundation for subsequent analysis and diagnosis. The signal decomposition unit decomposes the vibration signal sequence to obtain modal components and residual components, thereby helping to extract useful feature information and better capture subtle features of motor state changes, thereby improving the sensitivity and accuracy of fault detection. The first feature extraction unit calculates the energy of the modal component and its proportion, further calculates the energy entropy, and uses it as the first feature vector. These energy features can effectively characterize the motor's operating status and help identify potential fault types.

[0067] 2. The present invention extracts the time-frequency domain features of the modal component through the second feature extraction unit. As the second feature vector, the time-frequency domain features can reflect the local changes of the signal, provide richer fault feature information, and enhance the accuracy and reliability of fault diagnosis. The fault diagnosis unit comprehensively utilizes the first and second feature vectors, and classifies faults through the trained fault diagnosis model. The combination of the support vector machine algorithm and the swarm intelligence optimization algorithm ensures the high accuracy and efficiency of the diagnosis model, so that the fault type of the motor can be quickly and accurately identified. The three-phase stator current signal is decomposed by wavelet through the data analysis unit, the wavelet decomposition coefficients are extracted and their second norms are calculated, and a time series diagram is constructed. These analysis results can provide detailed changes in the motor current signal, providing strong data support for fault diagnosis. The scene model, diagnosis type and time series diagram of the target device are displayed through the three-dimensional visualization unit, so that the user can intuitively view the motor and its fault information. The pop-up window display method enables the user to quickly obtain detailed fault information when needed, greatly improving the user-friendliness and ease of use of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A schematic diagram of the overall system flow in an embodiment of the present invention;

[0069] Figure 2 A schematic diagram of a flow chart of signal decomposition in an embodiment of the present invention;

[0070] Figure 3 A schematic diagram of a flow chart of time-frequency domain feature extraction in an embodiment of the present invention;

[0071] Figure 4 A schematic diagram of a process for parameter optimization in an embodiment of the present invention.

[0072] Figure numerals: 1. three-dimensional modeling unit; 2. three-dimensional visualization unit; 3. data acquisition unit; 4. signal decomposition unit; 5. first feature extraction unit; 6. second feature extraction unit; 7. fault diagnosis unit; 8. data analysis unit. DETAILED DESCRIPTION

[0073] Embodiment 1, as Figure 1 As shown, the present invention proposes an online monitoring and fault diagnosis system for a motor based on the Internet of Things, comprising a three-dimensional visualization unit 2, a first feature extraction unit 5, a second feature extraction unit 6, a fault diagnosis unit 7 and a data analysis unit 8, and also includes:

[0074] A three-dimensional modeling unit 1, the three-dimensional modeling unit 1 is used to perform geometric modeling on multiple target motors in a target device through a three-dimensional modeling technology to obtain multiple target motor geometric models, and to perform texture rendering on the multiple target motor geometric models to obtain multiple target motor three-dimensional models, and to integrate the multiple target motor three-dimensional models according to the actual positions of the target motors in the target device to obtain a scene model of the target device, and to transmit the scene model of the target device to the three-dimensional visualization unit 2;

[0075] The data acquisition unit 3 is used to collect the vibration signal of the target motor by using a built-in acceleration sensor in the target motor to obtain a vibration signal sequence of the target motor, and to collect the three-phase stator current signal of the target motor by using a built-in current sensor in the target motor to obtain a three-phase stator current signal sequence of the target motor, and transmit the vibration signal sequence to the signal decomposition unit 4, and transmit the three-phase stator current signal sequence of the target motor to the data analysis unit 8;

[0076] The signal decomposition unit 4 receives the vibration signal sequence transmitted by the data acquisition unit 3, and performs signal decomposition on the vibration signal sequence to obtain all modal components and residual components corresponding to the vibration signal sequence, and transmits all modal components to the first feature extraction unit 5 and the second feature extraction unit 6.

[0077] In the present invention, three-dimensional modeling technology is used to geometrically model multiple motors in a target device and generate three-dimensional models of these motors; integrated modeling is to combine multiple independent three-dimensional models into a complete scene model according to actual positions and relationships; it is a common practice to use an acceleration sensor to obtain vibration signals. When the acceleration sensor is installed on the device, it moves with the vibration of the device, and the mass block inside the sensor also moves accordingly, thereby generating electrical signals. These signals reflect the vibration characteristics of the device, including information such as amplitude and frequency; modal components are also called vibration modes, which refer to the various frequency components in the vibration signal caused by different vibration modes. A vibration signal can usually be represented by the superposition of multiple modal components. Each modal component corresponds to an inherent vibration mode of the system or structure, and its frequency and amplitude are related to the characteristics of the vibration mode; the residual component refers to the part of the vibration signal that cannot be explained by modal decomposition, that is, the component that cannot be attributed to a specific vibration mode. The residual component may contain noise, high-order modes or other nonlinear effects.

[0078] In an optional embodiment, the first feature extraction unit 5 receives all modal components transmitted by the signal decomposition unit 4, calculates the energies corresponding to all modal components, and sums all the energies to obtain the total energy of all modal components, calculates the energy proportion of each modal component based on the total energy and energy, and calculates the energy entropy of each modal component based on the energy proportion, uses the energy entropy as the first eigenvector, and transmits the first eigenvector to the fault diagnosis unit 7.

[0079] In an optional embodiment, the second feature extraction unit 6 receives all modal components transmitted by the signal decomposition unit 4, extracts time-frequency domain features of all modal components, uses the time-frequency domain features as a second feature vector, and transmits the second feature vector to the fault diagnosis unit 7.

[0080] It should be noted that time-frequency domain features refer to the characteristics of signals that take into account both the time and frequency dimensions. Traditional signal analysis methods are usually only performed in the time domain or the frequency domain. However, the characteristics of signals in many practical applications change over time. Analysis simply in the time domain or the frequency domain may not be able to fully reveal all the information of the signal. Therefore, the introduction of time-frequency domain features can describe the signal more comprehensively.

[0081] In an optional embodiment, the fault diagnosis unit 7 receives the first feature vector transmitted by the first feature extraction unit 5 and the second feature vector transmitted by the second feature extraction unit 6, and inputs the first feature vector and the second feature vector into a trained fault diagnosis model, and outputs the diagnosis type corresponding to the target motor through the trained fault diagnosis model. The fault diagnosis model includes a support vector machine algorithm and a swarm intelligence optimization algorithm. The support vector machine algorithm is used to classify the fault type of the target motor to obtain the diagnosis type corresponding to the target motor. The swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, and the parameters include penalty parameters and kernel function parameters, and the diagnosis type corresponding to the target motor is transmitted to the three-dimensional visualization unit 2.

[0082] It should be noted that support vector machine is a supervised learning model, mainly used for classification and regression analysis. Support vector machine classifies data by finding an optimal hyperplane to maximize the interval between the classification boundary and the data points, thereby improving the accuracy and robustness of classification; the penalty parameter is used to control the degree of penalty for misclassification in the training data. A larger penalty parameter will try to make all training data points correctly classified, but may cause overfitting; a smaller penalty parameter allows some misclassifications to exist, which may improve generalization ability; the kernel function parameter determines the distance between the mapped data points in the high-dimensional space; the swarm intelligence optimization algorithm is a type of optimization algorithm based on simulating group behavior in nature, which achieves the purpose of overall optimization through information exchange and collaboration between individuals.

[0083] In an optional embodiment, the data analysis unit 8 receives the three-phase stator current signal sequence of the target motor transmitted by the data acquisition unit 3, and performs three-layer wavelet decomposition on the three-phase stator current signal sequence to obtain wavelet decomposition coefficients, the wavelet decomposition coefficients include 1 high-frequency coefficient and 3 low-frequency coefficients, calculates the binary norm of the wavelet decomposition coefficients, and uses the binary norm as the vertical coordinate, and uses the time of receiving the three-phase stator current signal sequence of the target motor as the horizontal coordinate to construct a time series graph, and transmits the time series graph to the three-dimensional visualization unit 2.

[0084] It should be noted that wavelet decomposition is a signal processing technology that can decompose signals into components at different scales. Three-layer wavelet decomposition means decomposing the signal three times, each decomposition decomposing the signal into a high-frequency part and a low-frequency part. The low-frequency part obtained after each decomposition can continue to be decomposed next time until the required number of decomposition layers is reached; the bi-norm is a norm measure of a vector. Calculating the bi-norm of a wavelet decomposition coefficient means square-summing the elements in each wavelet decomposition coefficient vector, and then taking the square root to obtain a value as the bi-norm of the wavelet decomposition coefficient.

[0085] In an optional embodiment, the three-dimensional visualization unit 2 receives the scene model of the target device transmitted by the three-dimensional modeling unit 1, the diagnosis type corresponding to the target motor transmitted by the fault diagnosis unit 7, and the time series diagram transmitted by the data analysis unit 8, and displays the scene model of the target device in the intelligent terminal, and hides the diagnosis type and the time series diagram corresponding to the target motor in the corresponding three-dimensional model of the target motor in the scene model of the target device. When the target user clicks on the corresponding three-dimensional model of the target motor in the scene model of the target device, the diagnosis type and the time series diagram corresponding to the target motor will pop up in the form of a pop-up window.

[0086] Embodiment 2, as Figure 2-Figure 4 As shown, the present invention proposes an online monitoring and fault diagnosis system for a motor based on the Internet of Things. Compared with the first embodiment, this embodiment further includes: performing signal decomposition on the vibration signal sequence to obtain the modal component and residual component corresponding to the vibration signal sequence, including the following steps:

[0087] A1. Add Gaussian white noise to the vibration signal sequence to obtain a vibration signal sequence containing controllable noise. The expression of the vibration signal containing controllable noise is as follows:

[0088] ;

[0089] in, represents a vibration signal with controllable noise, Indicates vibration signal, Indicates the initial value of the amplitude coefficient of the added Gaussian white noise, Indicates Gaussian white noise is added, Indicates The first-order modal component of the Gaussian white noise added is generated by the empirical mode decomposition;

[0090] A2. Perform the first-order decomposition of the vibration signal containing controllable noise by empirical mode decomposition to obtain the first-order residual component and the first-order modal component. The expression of the first-order residual component is as follows:

[0091] ;

[0092] in, represents the first-order residual component, represents an operator for solving the local mean of a vibration signal containing controllable noise;

[0093] The expression of the first-order modal component is as follows:

[0094] ;

[0095] in, represents the first-order modal component;

[0096] A3. Perform k-th order decomposition on the vibration signal containing controllable noise by empirical mode decomposition to obtain k-th order residual component and k-th order modal component. The expression of k-th order residual component is as follows:

[0097] ;

[0098] in, represents the k-th order residual component, represents the k-1th order residual component, represents the amplitude coefficient of the k-1th addition of Gaussian white noise, Indicates The k-th modal component generated by the empirical mode decomposition of the Gaussian white noise added for the second time;

[0099] The expression of the k-th modal component is as follows:

[0100] ;

[0101] in, represents the k-th order modal component;

[0102] A4. Repeat step A3 until all vibration signals containing controllable noise in the vibration signal sequence containing controllable noise are traversed to obtain all modal components and residual components corresponding to the vibration signal sequence.

[0103] In this embodiment, Gaussian white noise is often used to simulate random interference in a real environment, and its statistical characteristics make it a widely used noise model; empirical mode decomposition is a data decomposition method used to decompose nonlinear and non-stationary signals into a superposition of several intrinsic mode functions.

[0104] In an optional embodiment, extracting the time-frequency domain features of all modal components includes the following steps:

[0105] B1. Calculate the correlation coefficient between each modal component and the vibration signal respectively, and compare the correlation coefficient with a preset correlation coefficient threshold, and take the modal component greater than the preset correlation coefficient threshold as the effective modal component;

[0106] B2. Perform Hilbert transform on the effective modal components to obtain the complex analytical signals corresponding to the effective modal components, and calculate the instantaneous frequency of each complex analytical signal. The complex analytical signal expression is as follows:

[0107] ;

[0108] in, represents the complex analytical signal, Indicates time, represents the order of the effective modal component;

[0109] The instantaneous frequency expression is as follows:

[0110] ;

[0111] in, represents the instantaneous frequency;

[0112] B3. Calculate the fractal box dimension of the instantaneous frequency of the effective modal component, and construct the fractal box dimension eigenvector. The fractal box dimension eigenvector is used as the second eigenvector. The fractal box dimension expression is as follows:

[0113] ;

[0114] in, represents the fractal box dimension, represents any constant term, The maximum diameter representing the minimum mesh count, represents the minimum grid count, represents the end point of the interval with good linearity, Indicates the starting point of the interval with good linearity.

[0115] It should be noted that the Hilbert transform is a mathematical tool for signal analysis, which can convert a real signal into its corresponding complex analytical signal; the fractal box dimension is a geometric feature used to describe the complexity of a signal, which characterizes the complexity of the signal by dividing the signal into grids and counting the number of grids occupied by the signal at different scales.

[0116] In an optional embodiment, a swarm intelligence optimization algorithm is used to optimize the parameters of a support vector machine algorithm, comprising the following steps:

[0117] C1. Randomly initialize the particle population. Particles represent a combination of penalty parameters and kernel function parameters. The random initialization expression is as follows:

[0118] ;

[0119] in, Indicates Particle No. The location of the dimension variable, Indicates that in the interval The random number inside represents the maximum diameter of the minimum mesh count, and Respectively represent Upper and lower bounds on dimensional variables;

[0120] The matrix expression of particle population is as follows:

[0121] ;

[0122] in, is the matrix representing the particle population, Indicates The position of a particle;

[0123] C2. Construct the objective function of the particle based on the matrix of the particle population. The matrix expression of the objective function is as follows:

[0124] ;

[0125] in, The matrix representing the objective function of the particle, Indicates The objective function value of a particle.

[0126] In an optional embodiment, the swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, and further includes the following steps:

[0127] C3. Based on the objective function of the particle, move the particle to obtain the new position of the particle, and judge whether the objective function value of the new position of the particle is greater than the objective function value of the original position of the particle. If it is greater than the objective function value of the original position of the particle, the original position of the particle is updated to the new position. The new position movement expression is as follows:

[0128] ;

[0129] in, Indicates Particle No. The new position of the dimension variable, Indicates that the target particle is The location of the dimension variable, represents the objective function value of the target particle;

[0130] The new position update expression is as follows:

[0131] ;

[0132] in, Indicates The updated position of the particle, Indicates The new position of the particle, Indicates The original position of the particle, Indicates The objective function value of the new position of the particle;

[0133] C4. Move the new position of the particle to obtain the final position of the particle, and judge whether the objective function value of the final position of the particle is greater than the objective function value of the new position of the particle. If it is greater than the objective function value of the new position of the particle, the new position of the particle is updated to the final position. The final position movement expression is as follows:

[0134] ;

[0135] in, Indicates Particle No. The final position of the dimensional variable, Represents a constant parameter, represents the maximum number of iterations, Indicates the current iteration number;

[0136] The final position update expression is as follows:

[0137] ;

[0138] in, Indicates The final position of the particle after update, Indicates The final position of a particle, Indicates The objective function value of the final position of the particle.

[0139] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A motor online monitoring and fault diagnosis system based on the Internet of Things, comprising a three-dimensional visualization unit (2), a first feature extraction unit (5), a second feature extraction unit (6), a fault diagnosis unit (7) and a data analysis unit (8), characterized in that: A three-dimensional modeling unit (1), the three-dimensional modeling unit (1) being used to perform geometric modeling on a plurality of target motors in a target device using a three-dimensional modeling technique to obtain a plurality of target motor geometric models, and to perform texture rendering on the plurality of target motor geometric models to obtain a plurality of target motor three-dimensional models, and to integrate the plurality of target motor three-dimensional models according to actual positions of the target motors in the target device to obtain a scene model of the target device, and to transmit the scene model of the target device to a three-dimensional visualization unit (2); A data acquisition unit (3), the data acquisition unit (3) being used to acquire a vibration signal of the target motor by using an acceleration sensor built into the target motor to obtain a vibration signal sequence of the target motor, and to acquire a three-phase stator current signal of the target motor by using a current sensor built into the target motor to obtain a three-phase stator current signal sequence of the target motor, and to transmit the vibration signal sequence to a signal decomposition unit (4), and to transmit the three-phase stator current signal sequence of the target motor to a data analysis unit (8); A signal decomposition unit (4), the signal decomposition unit (4) receiving the vibration signal sequence transmitted by the data acquisition unit (3), performing signal decomposition on the vibration signal sequence to obtain all modal components and residual components corresponding to the vibration signal sequence, and transmitting all the modal components to a first feature extraction unit (5) and a second feature extraction unit (6); The first feature extraction unit (5) receives all the modal components transmitted by the signal decomposition unit (4), calculates the energy corresponding to all the modal components, and sums all the energies to obtain the total energy of all the modal components, calculates the energy proportion of each modal component based on the total energy and the energy, and calculates the energy entropy of each modal component based on the energy proportion, uses the energy entropy as a first feature vector, and transmits the first feature vector to the fault diagnosis unit (7); The second feature extraction unit (6) receives all the modal components transmitted by the signal decomposition unit (4), extracts time-frequency domain features of all the modal components, uses the time-frequency domain features as a second feature vector, and transmits the second feature vector to the fault diagnosis unit (7); The fault diagnosis unit (7) receives the first feature vector transmitted by the first feature extraction unit (5) and the second feature vector transmitted by the second feature extraction unit (6), and inputs the first feature vector and the second feature vector into a trained fault diagnosis model, and outputs the diagnosis type corresponding to the target motor through the trained fault diagnosis model, wherein the fault diagnosis model includes a support vector machine algorithm and a swarm intelligence optimization algorithm, wherein the support vector machine algorithm is used to classify the fault type of the target motor to obtain the diagnosis type corresponding to the target motor, and the swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, wherein the parameters include a penalty parameter and a kernel function parameter, and the diagnosis type corresponding to the target motor is transmitted to the three-dimensional visualization unit (2); The data analysis unit (8) receives the three-phase stator current signal sequence of the target motor transmitted by the data acquisition unit (3), and performs three-layer wavelet decomposition on the three-phase stator current signal sequence to obtain wavelet decomposition coefficients, wherein the wavelet decomposition coefficients include 1 high-frequency coefficient and 3 low-frequency coefficients, calculates the binary norm of the wavelet decomposition coefficients, and uses the binary norm as the ordinate, and uses the time of receiving the three-phase stator current signal sequence of the target motor as the abscissa to construct a time series graph, and transmits the time series graph to the three-dimensional visualization unit (2); Performing signal decomposition on the vibration signal sequence to obtain a modal component and a residual component corresponding to the vibration signal sequence comprises the following steps: A1. Add Gaussian white noise to the vibration signal sequence to obtain a vibration signal sequence containing controllable noise. The vibration signal sequence containing controllable noise is expressed as follows: ； in, represents a vibration signal with controllable noise, Indicates vibration signal, Indicates the initial value of the amplitude coefficient of the added Gaussian white noise, Indicates Gaussian white noise is added, Indicates The first-order modal component of the Gaussian white noise added is generated by the empirical mode decomposition; A2. Performing a first-order decomposition of the vibration signal containing controllable noise by empirical mode decomposition to obtain a first-order residual component and a first-order modal component. The expression of the first-order residual component is as follows: ； in, represents the first-order residual component, represents an operator for solving the local mean of a vibration signal containing controllable noise; The expression of the first-order modal component is as follows: ； in, represents the first-order modal component; A3. Performing a k-th order decomposition on the vibration signal containing controllable noise by empirical mode decomposition to obtain a k-th order residual component and a k-th order modal component. The expression of the k-th order residual component is as follows: ； in, represents the k-th order residual component, represents the k-1th order residual component, represents the amplitude coefficient of the k-1th addition of Gaussian white noise, Shidi The k-th modal component generated by the empirical mode decomposition of the Gaussian white noise added for the second time; The expression of the k-th order modal component is as follows: ； in, represents the k-th order modal component; A4, repeating step A3 until all vibration signals containing controllable noise in the vibration signal sequence containing controllable noise are traversed to obtain all modal components and residual components corresponding to the vibration signal sequence; Extracting the time-frequency domain features of all the modal components comprises the following steps: B1. Calculate the correlation coefficient between each modal component and the vibration signal respectively, and compare the correlation coefficient with a preset correlation coefficient threshold, and take the modal component greater than the preset correlation coefficient threshold as a valid modal component; B2. Performing Hilbert transform on the effective modal components to obtain complex analytical signals corresponding to the effective modal components, and calculating the instantaneous frequency of each of the complex analytical signals. The complex analytical signal expression is as follows: ； in, represents the complex analytical signal, Indicates time, represents the order of the effective modal component; The instantaneous frequency expression is as follows: ； in, represents the instantaneous frequency; B3. Calculate the fractal box dimension of the instantaneous frequency of the effective modal component, and construct a fractal box dimension feature vector, and use the fractal box dimension feature vector as the second feature vector. The fractal box dimension expression is as follows: ； in, represents the fractal box dimension, represents any constant term, The maximum diameter representing the minimum mesh count, represents the minimum grid count, represents the end point of the interval with good linearity, Indicates the starting point of the interval with better linearity; The swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, and includes the following steps: C1. Randomly initialize the particle population, where the particle represents a combination of the penalty parameter and the kernel function parameter. The random initialization expression is as follows: ； in, Indicates Particle No. The location of the dimension variable, Indicates that in the interval The random number inside, represents the maximum diameter of the minimum mesh count, and Respectively represent Upper and lower bounds on dimensional variables; The matrix expression of the particle population is as follows: ； in, is the matrix representing the particle population, Indicates The position of a particle; C2. Construct the objective function of the particle based on the matrix of the particle population. The matrix expression of the objective function is as follows: ； in, The matrix representing the objective function of the particle, Indicates The objective function value of each particle; The swarm intelligence optimization algorithm is used to optimize the parameters of the support vector machine algorithm, and further comprises the following steps: C3. Based on the objective function of the particle, the particle is moved to obtain a new position of the particle, and it is determined whether the objective function value of the new position of the particle is greater than the objective function value of the original position of the particle. If it is greater than the objective function value of the original position of the particle, the original position of the particle is updated to the new position. The new position movement expression is as follows: ； in, Indicates Particle No. New position of the dimension variable Indicates that the target particle is The location of the dimension variable, represents the objective function value of the target particle; The new position update expression is as follows: ； in, Indicates The updated position of the particle, Indicates The new position of the particle, Indicates The original position of the particle, Indicates The objective function value of the new position of the particle; C4. Move the new position of the particle to obtain the final position of the particle, and judge whether the objective function value of the final position of the particle is greater than the objective function value of the new position of the particle. If it is greater than the objective function value of the new position of the particle, update the new position of the particle to the final position. The final position movement expression is as follows: ； in, Indicates Particle No. The final position of the dimensional variable, Represents a constant parameter, represents the maximum number of iterations, Indicates the current iteration number; The final position update expression is as follows: ； in, Indicates The final position of the particle after update, Indicates The final position of a particle, Indicates The objective function value of the final position of the particle.

2. The motor online monitoring and fault diagnosis system based on the Internet of Things according to claim 1 is characterized in that: The three-dimensional visualization unit (2) receives the scene model of the target device transmitted by the three-dimensional modeling unit (1), the diagnosis type corresponding to the target motor transmitted by the fault diagnosis unit (7), and the time series diagram transmitted by the data analysis unit (8), and displays the scene model of the target device in the intelligent terminal, hides the diagnosis type corresponding to the target motor and the time series diagram in the corresponding three-dimensional model of the target motor in the scene model of the target device, and when the target user clicks on the corresponding three-dimensional model of the target motor in the scene model of the target device, the diagnosis type corresponding to the target motor and the time series diagram are popped up in the form of a pop-up window.

Images