An intelligent diagnosis method for subway fans based on multi-classification SVM information perception

Abstract

This invention discloses an intelligent diagnosis method for subway fans based on multi-classification SVM information perception. The steps are as follows: vibration signals are collected via vibration sensors installed on fans within stations; the vibration signals are subjected to Fourier analysis to obtain a spectrum; fuzzy methods are used to further divide the signal frequencies and extract the frequency distribution based on expert knowledge; a support vector machine model is trained using the generated historical data; and the vibration frequency signals are monitored in real time using the support vector machine model. This method enables online monitoring of subway fans, quickly identifying fault types and providing alarms, providing maintenance personnel with maintenance and repair advice, and, to a certain extent, reducing unplanned downtime in rail transit, increasing the safety and cost-effectiveness of equipment operation, and improving equipment utilization.

Description

Technical Field

[0001] The present invention relates to the practical application of machine learning, and in particular to an intelligent diagnosis method for subway fans based on multi-classification SVM information perception. Background Art

[0002] In recent years, my country has actively promoted the process of industrialization, and the transportation industry has flourished. Major cities across the country are building modern transportation projects such as subways and light rail. Among them, subways, as the main means of transportation for urban residents, play an important role in the entire transportation system.

[0003] To avoid congestion on the urban surface and maximize space, subways operate mostly underground, often in confined spaces. Therefore, for successful operation, the subway's ventilation system must maintain a regular exchange of air between the underground space and the outside world, ensuring timely ventilation and proper equipment operation. Subway fans are the core equipment in the ventilation system, ensuring safe and reliable operation. They are primarily used for exhaust in subway tunnels, station halls, and platforms, as well as for fire smoke extraction. They are a crucial component in ensuring safe subway operation and a key component in achieving environmental control, playing a crucial role in safeguarding safe subway operations. The operational requirements of subway fans necessitate their high reliability. With increasing demands for equipment safety and reliability, and the increasing complexity of the structures and functions of modern machinery, the demand for equipment fault diagnosis is also increasing. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to provide an intelligent diagnosis method for subway fans based on multi-classification SVM information perception, so as to quickly confirm the fault type and provide an alarm, thereby reducing the unplanned downtime of rail transit to a certain extent, increasing the safety and economy of equipment operation and improving the effective utilization rate of equipment.

[0005] Technical solution: The present invention provides a subway fan intelligent diagnosis method based on multi-classification SVM information perception, comprising the following steps:

[0006] (1) Extract vibration signals through vibration sensors installed on fan bearings in the station;

[0007] (2) The vibration signal is analyzed by Fourier analysis to obtain the spectrum. The vibration signal f(t) can be expressed as:

[0008]

[0009] Where ω is the frequency, t represents the time, and e -2πikt is a complex function;

[0010] (3) Use fuzzy methods to further divide the signal frequency and extract the frequency distribution, and use prior expert knowledge to process the data to extract more accurate and effective information;

[0011] (4) Use the generated historical data to train the support vector machine model:

[0012] To transform the solution of linear separable problem into the solution of constrained optimization problem, input the training sample set:

[0013] (x1,y1),(x2,y2),…,(x l ,y l )

[0014] in is the m-dimensional input vector, y i ={-1,1} is the corresponding sample label, l is the number of samples, i=1,2,…,l;

[0015] Constructing the optimization problem:

[0016]

[0017] Where: ω is the normal vector of the hyperplane (ω, b); b is the hyperplane intercept; C is the penalty factor; ξ i is a slack variable; further solving gives the classification function expression:

[0018]

[0019] For nonlinear problems, we introduce a kernel function to transform them into linear separable problems in high-dimensional space, and keep the inner product unchanged before and after the transformation; the kernel function K(x,z) satisfies:

[0020] K(x,z)=φ(x)×φ(z)

[0021] Where φ(x) is the mapping function and φ(x)×φ(z) is the inner product;

[0022] The most commonly used kernel function is the Gaussian radial basis function (RBF):

[0023]

[0024] The final classification function expression is transformed into:

[0025]

[0026] (5) Use the support vector machine model to monitor the vibration frequency signal in real time.

[0027] The step (3) is specifically as follows:

[0028] (3.1) Based on field experience and theoretical analysis, the frequency fuzzy processing of spectrum analysis is divided into 9 levels: 0~0.39f, 0.4~0,49f, 0.5f, 0.51~0.99f, 1f, 2f, 3~5f, odd multiples of f, >5f;

[0029] (3.2) Conduct diagnostic research on common faults of wind turbines, including mass imbalance, rotor misalignment, axial friction, shaft cracks, loose bearing seats, bearing damage, coupling damage, and vibration caused by clearance;

[0030] (3.3) Through the study of the mechanism of common wind turbine faults and expert and field experience, through signal analysis and fuzzy processing, the characteristic vectors of the above common faults are obtained. Some results of each type of fault are shown in Table 1.

[0031] Table 1 Common fault samples of fans Fault types

[0032]

[0033] In the step (4), a one-vs-all support vector machine (One vs All SVM) is used to construct a hyperplane between one type of samples and the remaining samples to achieve the purpose of multi-classification.

[0034] A computer storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned intelligent diagnosis method for subway fans based on multi-classification SVM information perception.

[0035] A computer device includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the above-mentioned intelligent diagnosis method for subway fans based on multi-classification SVM information perception is implemented.

[0036] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0037] 1. The present invention can monitor subway fans online, quickly identify fault types and provide alarms, and provide maintenance and repair advice to maintenance personnel. This, to a certain extent, reduces unplanned downtime of rail transit, increases the safety and economy of equipment operation, and improves the effective utilization rate of equipment.

[0038] 2. Under the condition of only a small number of training samples, the accuracy of the present invention is superior to that of the support regression model method, which is more suitable for actual engineering needs, lowers the usage threshold, and improves the applicability of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0040] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0041] like Figure 1 As shown, a subway fan intelligent diagnosis method based on multi-classification SVM information perception includes the following steps:

[0042] (1) Extract vibration signals through vibration sensors installed on fan bearings in the station;

[0043] (2) The vibration signal is analyzed by Fourier analysis to obtain the spectrum. The vibration signal f(t) can be expressed as:

[0044]

[0045] Where ω is the frequency, t represents the time, and e -2πikt is a complex function;

[0046] (3) Use fuzzy methods to further divide the signal frequency and extract the frequency distribution, and use prior expert knowledge to process the data to extract more accurate and effective information;

[0047] Based on field experience and theoretical analysis, the frequency fuzzy processing of spectrum analysis is divided into 9 levels, namely: 0~0.39f, 0.4~0,49f, 0.5f, 0.51~0.99f, 1f, 2f, 3~5f, odd multiples of f, >5f;

[0048] Subway fans often malfunction due to long-term high-load operation. This example mainly diagnoses and studies common fan faults, including eight types of faults: mass imbalance, rotor misalignment, axial friction, shaft cracks, loose bearing seats, bearing damage, coupling damage, and vibration caused by clearance.

[0049] Through the study of the mechanism of common wind turbine faults and expert and field experience, through signal analysis and fuzzy processing, the characteristic vectors of the above common faults are obtained. Some results of each type of fault are shown in Table 1. The vectors are used as fault training samples for one-to-many support vector machine (OVR) and support vector regression (SVR). The network target output is shown in Table 2.

[0050] Table 1 Common fault samples of fans Fault types

[0051]

[0052]

[0053] Table 2 Grid target output

[0054]

[0055] (4) Use the generated historical data to train the support vector machine model:

[0056] To transform the solution of linear separable problem into the solution of constrained optimization problem, input the training sample set:

[0057] (x1,y1),(x2,y2),…,(x l ,y l )

[0058] in is the m-dimensional input vector, y i ={-1,1} is the corresponding sample label, l is the number of samples, i=1,2,…,l;

[0059] Constructing the optimization problem:

[0060]

[0061] Where: ω is the normal vector of the hyperplane (ω, b); b is the hyperplane intercept; C is the penalty factor; ξ i is a slack variable; further solving gives the classification function expression:

[0062]

[0063] For nonlinear problems, we introduce a kernel function to transform them into linear separable problems in high-dimensional space, and keep the inner product unchanged before and after the transformation; the kernel function K(x,z) satisfies:

[0064] K(x,z)=φ(x)×φ(z)

[0065] Where φ(x) is the mapping function and φ(x)×φ(z) is the inner product;

[0066] The most commonly used kernel function is the Gaussian radial basis function (RBF):

[0067]

[0068] The final classification function expression is transformed into:

[0069]

[0070] The one-to-many support vector model was trained using the aforementioned fault sample library, and the unknown fault samples in Table 3 were diagnosed using the LinearSVC model and the LogisticRegression model, respectively. The diagnostic results of the LinearSVC model were determined by the maximum distance from the decision plane, as shown in Figure 4. The LogisticRegression model was determined by the maximum likelihood of each classification, as shown in Figure 5.

[0071] Table 3 Fault samples to be identified

[0072]

[0073] Table 4 Distances between fault samples and each hyperplane

[0074]

[0075] Table 5 The probability of each fault in the fault sample

[0076]

[0077] As shown in Table 4, for the first set of input data, the output fault 4 has a maximum distance of -0.546, indicating a shaft crack; for the second set of input data, the output fault 7 has a maximum distance of -0.543, indicating a coupling failure; and for the final set of input data, the output fault 3 has a maximum value of -0.536, indicating an axial friction failure. The diagnostic results are consistent with the actual inspection results. Furthermore, as shown in Table 5, for the first set of data, the output fault 4 has a maximum probability of 0.1297, indicating a coupling failure; for the second set of data, the output fault 2 has a maximum probability of 0.1301, indicating a rotor misalignment; and for the third set of data, the output fault 3 has a maximum probability of 0.1282, indicating an axial friction failure. With the exception of the second set, the diagnostic results are consistent with the actual inspection results.

[0078] The above research results show that the LinearSVC model and the LogisticRegression model can achieve a mapping relationship between fault characteristics and fault types. However, the accuracy of multi-fault diagnosis using support vector regression models such as LogisticRegression is not as good as that of the LinearSVC model. This is because the accuracy of regression models increases with the amount of data, while the accuracy of support vector machine models is only determined by the support vector absolutes and is less affected by the amount of data. Therefore, when only a small number of training samples are available, the accuracy of support vector machine models is superior to that of support vector regression models.

[0079] (5) Use the support vector machine model to monitor the vibration frequency signal in real time.

Claims

1. An intelligent diagnosis method for subway fans based on multi-classification SVM information perception, characterized in that: The following steps are involved: (1) Extract vibration signals through vibration sensors installed on fan bearings in the station; (2) The vibration signal is analyzed by Fourier analysis to obtain the spectrum. Expressed as: ； Where t represents time, is a complex function; (3) Use fuzzy methods to further divide the signal frequency and extract the frequency distribution, use prior expert knowledge to process the data, and extract more accurate and effective information; Specifically: (3.1) Based on field experience and theoretical analysis, the frequency fuzzy processing of spectrum analysis is divided into 9 levels: 0~0.39f, 0.4~0,49f, 0.5f, 0.51~0.99f, 1f, 2f, 3~5f, odd multiples of f, >5f; (3.2) Conduct diagnostic research on common faults of fans, including mass imbalance, rotor misalignment, axial friction, shaft cracks, loose bearing seats, bearing damage, coupling damage, and vibration caused by clearance; (3.3) Through the study of the mechanism of common wind turbine faults and the experience of experts and field workers, the characteristic vectors of the above common faults are obtained through signal analysis and fuzzy processing; (4) Use the generated historical data to train the support vector machine model: To transform the solution of linear separable problem into the solution of constrained optimization problem, input the training sample set: ； in, is an m-dimensional input vector, is the corresponding sample label, is the number of samples, ; Construct the optimization problem: ； in, is a hyperplane The normal vector of the hyperplane; b is the hyperplane intercept; C is the penalty factor; is a slack variable; further solving gives the classification function expression: ； For nonlinear problems, the kernel function is introduced to transform them into linear separable problems in high-dimensional space, and the inner product remains unchanged before and after the transformation; the kernel function satisfy: ； in, is the mapping function, is the inner product; The most commonly used kernel function is the Gaussian radial basis function: ； The final classification function expression is transformed into: ； (5) Use the support vector machine model to monitor the vibration frequency signal in real time.

2. The intelligent diagnosis method for subway fans based on multi-classification SVM information perception according to claim 1 is characterized in that: In the step (4), a one-to-many support vector machine is used to construct a hyperplane between one type of samples and the remaining samples to achieve the purpose of multi-classification.

3. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements a subway fan intelligent diagnosis method based on multi-classification SVM information perception as described in any one of claims 1-2.

4. A computer 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, it implements an intelligent diagnosis method for subway fans based on multi-classification SVM information perception according to any one of claims 1-2.

Images