Medical rolling bearing fault diagnosis system and method

Abstract

The invention relates to the technical field of bearing fault diagnosis, and particularly discloses a medical rolling bearing fault diagnosis system and method, and the system comprises a fault signal decomposition module which carries out the decomposition processing of a large number of fault signals of a medical rolling bearing, and obtains a large number of feature vectors; the algorithm optimization module introduces an improved fruit fly algorithm to optimize a regularization coefficient and a wavelet kernel function parameter in the wavelet kernel extreme learning machine; a model construction and training module constructs a wavelet kernel extreme learning machine fault diagnosis model and trains the wavelet kernel extreme learning machine fault diagnosis model based on all the feature vectors to obtain a trained wavelet kernel extreme learning machine fault diagnosis model; the fault diagnosis and evaluation module uses the trained wavelet kernel extreme learning machine fault diagnosis model to diagnose and evaluate the fault of the to-be-detected medical rolling bearing, and outputs a fault diagnosis report; and efficient and accurate diagnosis and evaluation of the medical rolling bearing fault are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of bearing fault diagnosis, and in particular to a medical rolling bearing fault diagnosis system and method. Background Art

[0002] In the field of medical equipment, rolling bearings are key components, and their normal operation is crucial to the performance and reliability of the equipment. With the continuous advancement of medical technology, the precision and complexity of medical equipment are increasing, and higher requirements are placed on the operating stability and reliability of rolling bearings. However, medical rolling bearings are prone to failures during long-term operation due to various factors such as load changes, poor lubrication, wear, etc. Timely and accurate diagnosis of these faults is of great significance to ensuring the normal operation of medical equipment and improving the quality of medical services. Traditional fault diagnosis methods usually rely on manual detection and simple signal analysis, which are inefficient and difficult to guarantee accuracy. In recent years, with the development of information technology, data-driven fault diagnosis methods have gradually emerged.

[0003] However, existing methods may have problems with insufficient decomposition and inaccurate feature extraction in fault signal processing, which may affect the subsequent diagnosis effect. In the model construction and training stages, commonly used diagnostic models have defects such as unreasonable parameter selection and insufficient generalization ability. Moreover, when faced with complex and changeable operating conditions of medical rolling bearings, the existing diagnostic systems are not adaptable and robust enough to meet actual needs. Therefore, a more efficient, accurate and reliable medical rolling bearing fault diagnosis system is needed.

[0004] Therefore, the present invention proposes a medical rolling bearing fault diagnosis system and method. Summary of the invention

[0005] The present invention provides a medical rolling bearing fault diagnosis system and method. The fault signal decomposition and processing module can decompose complex fault signals into multi-dimensional feature vectors, providing a rich and detailed data basis for subsequent diagnostic analysis. The algorithm optimization module introduces an improved fruit fly algorithm to optimize key parameters, thereby improving the performance and accuracy of the model. The model construction and training module uses the optimized parameters to construct and train the model, making the model more targeted and adaptable. The trained model can accurately diagnose and evaluate the medical rolling bearing fault to be tested in the fault diagnosis and evaluation module, and output a detailed fault diagnosis report. Efficient and accurate diagnosis and evaluation of medical rolling bearing faults are achieved, which helps to timely discover and solve problems, ensure the normal operation of medical equipment, and improve the quality and safety of medical services.

[0006] The present invention provides a medical rolling bearing fault diagnosis system, comprising:

[0007] A fault signal decomposition and processing module is used to decompose and process a large number of fault signals of medical rolling bearings to obtain a large number of feature vectors with multi-dimensional features;

[0008] Algorithm optimization module, used to introduce the improved fruit fly algorithm to optimize the regularization coefficient and wavelet kernel function parameters in the wavelet kernel extreme learning machine;

[0009] The model building and training module is used to build a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters, and train the wavelet kernel extreme learning machine fault diagnosis model based on all feature vectors to obtain a trained wavelet kernel extreme learning machine fault diagnosis model;

[0010] The fault diagnosis and evaluation module is used to diagnose and evaluate the fault of the medical rolling bearing to be tested using the trained wavelet kernel extreme learning machine fault diagnosis model and output a fault diagnosis report.

[0011] Preferably, the fault signal decomposition and processing module comprises:

[0012] The fault signal decomposition submodule is used to decompose the fault signals of a large number of medical rolling bearings using the variational mode decomposition method to obtain a large number of effective mode components;

[0013] The modal component processing submodule is used to process all effective modal components based on the singular value method to obtain a large number of eigenvectors with multi-dimensional characteristics.

[0014] Preferably, the fault signal decomposition submodule includes:

[0015] A decomposition modal number range determination unit is used to analyze the frequency spectrum characteristics of a large number of medical rolling bearing fault signals and determine the value range of the decomposition modal number;

[0016] A parameter initialization unit, used to take the minimum value in the value range of the decomposition mode number as the initial value of the decomposition mode number, and determine the initial value of the quadratic penalty factor and the initial value of the convergence criterion;

[0017] A parameter iteration value sequence generating unit is used to determine the parameter iteration value sequence of the decomposition mode number and the quadratic penalty factor based on the value range of the decomposition mode number and the corresponding parameter iteration change principle, the initial value range of the quadratic penalty factor and the corresponding parameter iteration change principle;

[0018] A variational problem solving iteration unit is used to construct a variational problem, and solve the variational problem based on an initial value of a quadratic penalty factor and a corresponding parameter iteration value sequence, an initial value of a convergence criterion and a corresponding parameter iteration value sequence, and an optimization algorithm until a convergence condition is met, thereby obtaining multiple modal components;

[0019] The validity verification and screening unit is used to verify and screen the validity of all modal components to obtain a large number of valid modal components.

[0020] Preferably, the decomposition modal number range determination unit comprises:

[0021] The spectrum analysis subunit is used to obtain a large number of fault signals of medical rolling bearings under different working conditions as all reference fault signals, and calculate the mean, variance, and peak frequency of all reference fault signal spectra as statistical parameters of each reference fault signal spectrum;

[0022] A prominent frequency band identification subunit, used to identify frequency bands corresponding to all energy concentration peaks in the spectrum of each reference fault signal based on statistical parameters of the spectrum of each reference fault signal as all prominent frequency bands of each reference fault signal;

[0023] A main frequency band identification subunit, used for screening out all main frequency bands of each reference fault signal from all prominent frequency bands of each reference fault signal, and determining the total number of all main frequency bands of each reference fault signal;

[0024] The decomposition modal number range determination subunit is used to take the maximum total number of main frequency bands among the total number of main frequency bands corresponding to all fault types as the lower limit of the decomposition modal number, and determine the value range of the decomposition modal number in combination with the preset value range span of the decomposition modal number.

[0025] Preferably, the main frequency band identification subunit includes:

[0026] A prominent frequency band combination terminal is used to arbitrarily combine all prominent frequency bands of all reference fault signals to obtain all prominent frequency band sets of all reference fault signals;

[0027] An exhaustive co-occurrence rate calculation terminal, used to calculate the exhaustive co-occurrence rate of each prominent frequency band set in all reference fault signals based on the occurrence rate of each prominent frequency band in all reference fault signals;

[0028] An effective prominent frequency band screening end is used to treat the prominent frequency bands of all reference fault signals belonging to each fault type contained in the prominent frequency band set with the maximum exhaustive co-occurrence rate as the effective prominent frequency bands of each fault type;

[0029] The frequency band verification and screening end is used to verify and screen all valid prominent frequency bands of each fault type, analyze all main frequency bands corresponding to each fault type, and determine the total number of all main frequency bands of each reference fault signal.

[0030] Preferably, the frequency band verification screening end includes:

[0031] A co-occurrence statistics sub-terminal, used to calculate the exhaustive co-occurrence rate of all valid prominent frequency bands of each fault type in all reference fault signals of the corresponding fault type based on the occurrence rate of each valid prominent frequency band of each fault type in all reference fault signals of the corresponding fault type;

[0032] The frequency band verification and screening sub-end is used to determine whether the exhaustive co-occurrence rate of all valid prominent frequency bands of each fault type in all reference fault signals of the corresponding fault type is not less than a preset exhaustive co-occurrence rate threshold. If so, all valid prominent frequency bands of the corresponding fault type are regarded as all main frequency bands of the corresponding fault type. Otherwise, it is determined whether the exhaustive co-occurrence rate of all valid prominent frequency bands of the corresponding fault type except the minimum co-occurrence rate in all reference fault signals of the corresponding fault type is not less than a preset exhaustive co-occurrence rate threshold, until it is determined that the exhaustive co-occurrence rate of all the latest remaining valid prominent frequency bands of the corresponding fault type in all reference fault signals of the corresponding fault type is not less than the preset exhaustive co-occurrence rate threshold, then all the latest remaining valid prominent frequency bands of the corresponding fault type are regarded as all main frequency bands of the corresponding fault type.

[0033] The frequency band total number counting sub-terminal is used to determine the total number of all main frequency bands of each reference fault signal.

[0034] Preferably, the algorithm optimization module includes:

[0035] The parameter initialization submodule is used to initialize the number of parameters, search coefficients, initial weights and weight coefficients of the improved fruit fly algorithm, and determine the initial fruit fly group position x′ axis , and gives a random direction and distance x for a single fruit fly to find food i ′=x′ axis +ω*rand(domainofdefinition)andω=ω0*a ∞ ;

[0036] The algorithm optimization submodule is used to replace the nonlinear generation mechanism in the fruit fly algorithm with a new linear generation mechanism for candidate solutions, so that the odor concentration judgment value S i ′=x i ′, when S′ i When the value of is fixed, S i ′ obeys uniform distribution, through input S i ′ to the odor concentration judgment equation to calculate the odor concentration Smell′ at each fruit fly's location i , and introduce inertia weight to find the fruit fly with the maximum odor concentration, keep the maximum odor concentration value and coordinates, and the fruit fly group flies to this position through vision, repeating the above steps until the odor concentration reaches the maximum value or the number of iterations reaches the set maximum number of iterations.

[0037] Preferably, the model building and training module includes:

[0038] A model building submodule is used to build a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters;

[0039] The model training submodule is used to divide all normalized feature vectors into a training set and a test set to train the wavelet kernel extreme learning machine fault diagnosis model to obtain a trained wavelet kernel extreme learning machine fault diagnosis model.

[0040] Preferably, the fault diagnosis and evaluation module includes:

[0041] A fault signal decomposition and processing submodule is used to decompose and process the fault signal of the medical rolling bearing to be tested, and obtain a feature vector with multi-dimensional characteristics of the medical rolling bearing to be tested;

[0042] The fault diagnosis and evaluation submodule is used to input the feature vector of the medical rolling bearing to be tested into the trained wavelet kernel extreme learning machine fault diagnosis model to obtain a fault diagnosis report of the medical rolling bearing to be tested.

[0043] The present invention provides a medical rolling bearing fault diagnosis method, which is applied to any of the above medical rolling bearing fault diagnosis systems, comprising:

[0044] S1: Decompose and process a large number of fault signals of medical rolling bearings to obtain a large number of feature vectors with multi-dimensional features;

[0045] S2: Introduce the improved fruit fly algorithm to optimize the regularization coefficient and wavelet kernel function parameters in the wavelet kernel extreme learning machine;

[0046] S3: construct a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters, and train the wavelet kernel extreme learning machine fault diagnosis model based on all eigenvectors to obtain a trained wavelet kernel extreme learning machine fault diagnosis model;

[0047] S4: Use the trained wavelet kernel extreme learning machine fault diagnosis model to diagnose and evaluate the medical rolling bearing fault to be tested, and output a fault diagnosis report.

[0048] The beneficial effects of the present invention compared to the prior art are as follows: the fault signal decomposition and processing module can decompose complex fault signals into multi-dimensional feature vectors, providing a rich and detailed data basis for subsequent diagnostic analysis. The algorithm optimization module introduces an improved fruit fly algorithm to optimize key parameters, thereby improving the performance and accuracy of the model. The model construction and training module uses the optimized parameters to build and train the model, making the model more targeted and adaptable. The trained model can accurately diagnose and evaluate the medical rolling bearing fault to be tested in the fault diagnosis and evaluation module, and output a detailed fault diagnosis report. Efficient and accurate diagnosis and evaluation of medical rolling bearing faults are achieved, which helps to promptly discover and solve problems, ensure the normal operation of medical equipment, and improve the quality and safety of medical services.

[0049] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in this application document.

[0050] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0052] Figure 1 It is a schematic diagram of functional modules of a medical rolling bearing fault diagnosis system in an embodiment of the present invention;

[0053] Figure 2 A schematic diagram of the hierarchical relationship of all detailed functional modules included in the medical rolling bearing fault diagnosis system in an embodiment of the present invention;

[0054] Figure 3 This is a flow chart of a medical rolling bearing fault diagnosis method in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0056] Embodiment 1:

[0057] The present invention provides a medical rolling bearing fault diagnosis system, referring to Figure 1 and Figure 2 ,include:

[0058] A fault signal decomposition and processing module is used to decompose and process a large number of fault signals of medical rolling bearings to obtain a large number of feature vectors with multi-dimensional features;

[0059] Algorithm optimization module, used to introduce the improved fruit fly algorithm to optimize the regularization coefficient and wavelet kernel function parameters in the wavelet kernel extreme learning machine;

[0060] The model building and training module is used to build a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters, and train the wavelet kernel extreme learning machine fault diagnosis model based on all feature vectors to obtain a trained wavelet kernel extreme learning machine fault diagnosis model;

[0061] The fault diagnosis and evaluation module is used to diagnose and evaluate the fault of the medical rolling bearing to be tested using the trained wavelet kernel extreme learning machine fault diagnosis model and output a fault diagnosis report.

[0062] In the embodiment, the medical rolling bearing is a bearing device used in the medical field and operates through rolling elements, such as a rolling bearing used to support rotating parts in a nuclear magnetic resonance device in a hospital.

[0063] In this embodiment, the fault signal of the medical rolling bearing is a change signal of various physical quantities generated when the medical rolling bearing fails and can reflect the fault characteristics and status, such as an abnormal vibration signal generated when the rolling bearing in the medical device is worn.

[0064] In this embodiment, the feature vector with multi-dimensional features is a set of numerical combinations containing multiple different feature dimensions, which is used to comprehensively describe the specific state or attribute of the medical rolling bearing, and may include a feature vector composed of multiple dimensional numerical values ​​such as the frequency, amplitude, and temperature change rate of the bearing vibration.

[0065] In this embodiment, the improved fruit fly algorithm is an improvement and optimization of the traditional fruit fly algorithm, making it more suitable for processing specific optimization problems, such as parameter optimization in medical rolling bearing fault diagnosis. The improvement may be to adjust the fruit fly search range, food concentration judgment method, etc.

[0066] In this embodiment, the regularization coefficient and wavelet kernel function parameter in the wavelet kernel extreme learning machine are the regularization coefficient used to control the complexity of the wavelet kernel extreme learning machine model and avoid overfitting, and the wavelet kernel function parameter determines the characteristics of the wavelet kernel function and the model's ability to fit the data. The regularization coefficient may be a value such as 0.05, and the wavelet kernel function parameter, such as the scale parameter, is 2.

[0067] In this embodiment, the wavelet kernel extreme learning machine fault diagnosis model is a model built based on the wavelet kernel extreme learning machine algorithm and is specifically used to diagnose whether a medical rolling bearing has a fault and the type of fault. The relevant operating data of the bearing is input, and the results such as whether there is a fault and whether the fault type is wear or crack can be output.

[0068] In this embodiment, the fault diagnosis report is a detailed record and description of the medical rolling bearing fault diagnosis results, covering the specific circumstances of the fault, evaluation and related suggestions. The report indicates that the bearing has a slight wear fault and recommends timely replacement of related parts.

[0069] The beneficial effects of the above technologies are as follows: the fault signal decomposition and processing module can decompose complex fault signals into multi-dimensional feature vectors, providing a rich and detailed data basis for subsequent diagnostic analysis. The algorithm optimization module introduces the improved fruit fly algorithm to optimize key parameters, thereby improving the performance and accuracy of the model. The model construction and training module uses the optimized parameters to build and train the model, making the model more targeted and adaptable. The trained model can accurately diagnose and evaluate the medical rolling bearing faults to be tested in the fault diagnosis and evaluation module, and output a detailed fault diagnosis report. It achieves efficient and accurate diagnosis and evaluation of medical rolling bearing faults, which helps to discover and solve problems in a timely manner, ensure the normal operation of medical equipment, and improve the quality and safety of medical services.

[0070] Embodiment 2:

[0071] Based on Example 1, the fault signal decomposition processing module refers to Figure 2 ,include:

[0072] The fault signal decomposition submodule is used to decompose the fault signals of a large number of medical rolling bearings using the variational mode decomposition method to obtain a large number of effective mode components;

[0073] The modal component processing submodule is used to process all effective modal components based on the singular value method to obtain a large number of eigenvectors with multi-dimensional characteristics.

[0074] In this embodiment, the effective modal component refers to the modal component that can truly reflect the fault characteristics and has actual diagnostic value after the fault signal of the medical rolling bearing is decomposed by a specific method (such as variational modal decomposition). For example, after the bearing vibration signal is decomposed, those modal components that are highly correlated with specific fault types (such as wear, cracks, etc.) can be regarded as effective modal components.

[0075] The beneficial effects of the above technical solution include: the fault signal decomposition submodule can effectively decompose the complex medical rolling bearing fault signal into a large number of effective modal components by adopting the variational modal decomposition method, which provides a basis for subsequent analysis and processing. The modal component processing submodule uses the singular value method to process the effective modal components to obtain a feature vector with multi-dimensional characteristics, making the characteristics of the fault signal richer and more accurate. The combination of the two improves the comprehensiveness and accuracy of the fault signal feature extraction, and provides high-quality data for the subsequent construction and training of the fault diagnosis model. It helps to more accurately diagnose the fault type of medical rolling bearings and evaluate the degree of faults. Overall, the performance and reliability of the fault diagnosis system are improved.

[0076] Embodiment 3:

[0077] Based on Example 2, the fault signal decomposition submodule, refer to Figure 2 ,include:

[0078] A decomposition modal number range determination unit is used to analyze the frequency spectrum characteristics of a large number of medical rolling bearing fault signals and determine the value range of the decomposition modal number;

[0079] A parameter initialization unit, used to take the minimum value in the value range of the decomposition mode number as the initial value of the decomposition mode number, and determine the initial value of the quadratic penalty factor and the initial value of the convergence criterion;

[0080] A parameter iteration value sequence generating unit is used to determine the parameter iteration value sequence of the decomposition mode number and the quadratic penalty factor based on the value range of the decomposition mode number and the corresponding parameter iteration change principle, the initial value range of the quadratic penalty factor and the corresponding parameter iteration change principle;

[0081] A variational problem solving iteration unit is used to construct a variational problem, and solve the variational problem based on an initial value of a quadratic penalty factor and a corresponding parameter iteration value sequence, an initial value of a convergence criterion and a corresponding parameter iteration value sequence, and an optimization algorithm until a convergence condition is met, thereby obtaining multiple modal components;

[0082] The validity verification and screening unit is used to verify and screen the validity of all modal components to obtain a large number of valid modal components.

[0083] In this embodiment, the value range of the decomposition modal number refers to the possible value interval of the decomposition modal number when decomposing the medical rolling bearing fault signal, for example, it may be [5, 10].

[0084] In this embodiment, the minimum value in the range of the decomposed modal number is used as the initial value of the decomposed modal number, and the initial value of the quadratic penalty factor and the initial value of the convergence criterion are determined, that is, the minimum number in the range is used as the value when the decomposed modal number is calculated, and the values ​​of the quadratic penalty factor and the convergence criterion are determined at the beginning. For example, if the range of the decomposed modal number is [3,8], 3 is used as the initial value, which controls the bandwidth constraint of the signal. If the value is too large, the signal may be over-decomposed, and if it is too small, the mode may not be effectively separated. It is usually adjusted between 1000 and 10000, and the middle value of 5000 can be taken first. The convergence criterion is generally set to a smaller value, such as 1e-6.

[0085] In this embodiment, the iterative change principle of the parameters of the decomposition mode number and the quadratic penalty factor refers to the rule of how these two parameters are gradually changed during the calculation process, such as increasing by 1 or multiplying by 1.2 each time.

[0086] In this embodiment, based on the value range of the decomposed modal number and the corresponding parameter iteration change principle, the initial value range of the quadratic penalty factor and the corresponding parameter iteration change principle, the parameter iteration value sequence of the decomposed modal number and the quadratic penalty factor is determined, that is, according to the above range, initial value and change principle, the sequence formed by the specific values ​​of the two parameters at each iteration is calculated. For example, the sequence of the decomposed modal number is [3,4,5,6], and the sequence of the quadratic penalty factor is [5000,5001,5002].

[0087] In this embodiment, the parameter iteration value sequence of the decomposition mode number and the quadratic penalty factor refers to a sequence formed by arranging specific numerical values ​​obtained by calculating the above two parameters each time during the iteration process in order.

[0088] In this embodiment, constructing a variational problem means describing and modeling the actual task of decomposing and processing the fault signal of a medical rolling bearing in mathematical language and formulas according to the principle of the variational modal decomposition algorithm. Specifically, the decomposition process of the fault signal is represented as an objective function that needs to be optimized, and combined with some constraints to form a complete mathematical optimization problem. For example, an objective function may be defined to measure the difference or similarity between the decomposed result and the original fault signal, and some constraints may be set, such as certain properties of the decomposed signal (such as energy conservation, frequency range limitation, etc.). By solving this constructed variational problem, that is, finding a solution that makes the objective function optimal (minimum or maximum) and satisfies the constraints, the decomposition method and parameters are determined, thereby achieving effective and accurate decomposition of the fault signal. Such mathematical transformation and solution process can be completed with the help of various optimization algorithms and tools, so that the decomposition process is scientific and repeatable. For each modal function, its analytical signal is obtained by Hilbert transform, and then its unilateral spectrum is calculated, and the center frequency is estimated, and the spectrum is modulated to the vicinity of the baseband, and finally the variational problem objective function is constructed.

[0089] In this embodiment, the variational problem is solved based on the initial value of the quadratic penalty factor and the corresponding parameter iteration value sequence, the initial value of the convergence criterion and the corresponding parameter iteration value sequence, and the optimization algorithm (such as the alternating direction multiplier method) until the convergence condition is met, and multiple modal components are obtained. This means that according to the values ​​determined at the beginning and the subsequent iterative values, a suitable optimization algorithm is used to solve the constructed variational problem, and the calculation is continued until the convergence condition set in advance is met, thereby obtaining multiple modal components for fault diagnosis. For example, [modal component 1, modal component 2, modal component 3] is obtained. In the iterative process, the center frequency is first fixed, the modal function is updated, and the updated modal function is obtained by solving a quadratic programming problem. Then the updated modal function is fixed, the center frequency is updated, and the center frequency is updated by calculating the spectral center of gravity of the current modal function. These two steps are repeated until the convergence condition is met, that is, the change of the modal function and the center frequency in two adjacent iterations is less than the set threshold.

[0090] The beneficial effects of the above technical solutions include: the decomposition modal number range determination unit determines a reasonable range of decomposition modal numbers by analyzing the fault signal and spectrum characteristics, providing a basis for subsequent parameter settings and improving the accuracy of decomposition. The parameter initialization unit determines the initial values ​​of the decomposition modal number, the quadratic penalty factor and the convergence criterion, preparing for subsequent iterative calculations. The parameter iteration value sequence generation unit determines the iteration value sequence of the parameter according to the value range and the iteration principle, making the decomposition process more systematic and scientific. The variational problem solving iteration unit obtains multiple modal components by constructing a variational problem and solving it until the convergence conditions are met, thereby ensuring the reliability of the decomposition. The validity verification and screening unit screens the modal components to obtain a large number of valid modal components, thereby improving the data quality of subsequent processing. The scientificity, accuracy and effectiveness of the fault signal decomposition are improved, providing more valuable data for subsequent fault diagnosis.

[0091] Embodiment 4:

[0092] Based on Example 3, the modal number range determination unit is decomposed, referring to Figure 2 ,include:

[0093] The spectrum analysis subunit is used to obtain a large number of fault signals of medical rolling bearings under different working conditions as all reference fault signals, and calculate the mean, variance, and peak frequency of all reference fault signal spectra as statistical parameters of each reference fault signal spectrum;

[0094] A prominent frequency band identification subunit, used to identify frequency bands corresponding to all energy concentration peaks in the spectrum of each reference fault signal based on statistical parameters of the spectrum of each reference fault signal as all prominent frequency bands of each reference fault signal;

[0095] A main frequency band identification subunit, used for screening out all main frequency bands of each reference fault signal from all prominent frequency bands of each reference fault signal, and determining the total number of all main frequency bands of each reference fault signal;

[0096] The decomposition modal number range determination subunit is used to take the maximum total number of main frequency bands among the total number of main frequency bands corresponding to all fault types as the lower limit of the decomposition modal number, and determine the value range of the decomposition modal number in combination with the preset value range span of the decomposition modal number.

[0097] In this embodiment, different working conditions refer to various working conditions and environments in which the medical rolling bearing is located during operation, such as different load sizes, rotation speeds, lubrication conditions, etc.

[0098] In this embodiment, the mean, variance, and peak frequency of all reference fault signal spectra are calculated, that is, the mean, dispersion (variance), and maximum frequency value of the spectra of the numerous fault signals collected for reference are calculated respectively.

[0099] In this embodiment, the frequency segments corresponding to all energy concentration peaks in each reference fault signal spectrum are identified based on the statistical parameters of each reference fault signal spectrum, and the frequency range corresponding to the peak with high energy concentration is found out according to the relevant statistical data of each reference fault signal spectrum. The following steps may be followed: First, the spectrum of each reference fault signal is analyzed in detail to obtain its statistical parameters, such as mean, variance, peak frequency, etc. Then, by observing the spectrum graph or using a specific algorithm, the area with relatively high energy value in the spectrum is found. An energy threshold may be set, and when the energy of a certain frequency band in the spectrum exceeds this threshold, it is considered to be an energy concentration peak. Next, the position of this energy concentration peak in the spectrum is determined to obtain the corresponding frequency value, and the frequency band where this frequency value is located is the frequency band to be found.

[0100] In this embodiment, all main frequency bands of each reference fault signal refer to frequency ranges having important characteristics and significant influences in each fault signal used for reference.

[0101] In this embodiment, the preset value range of the decomposition mode number is a preset interval size of possible values ​​of the decomposition mode number.

[0102] In this embodiment, the maximum total number of main frequency bands among the total numbers of main frequency bands corresponding to all fault types is used as the lower limit of the decomposition modal number, and the value range of the decomposition modal number is determined in combination with the preset value range span of the decomposition modal number. This means that the value with the largest number of main frequency bands in various fault types is first found, and it is used as the lower limit of the decomposition modal number, and then the possible value interval of the decomposition modal number is finally determined based on the preset value range span.

[0103] The beneficial effects of the above technical solutions include: the spectrum analysis subunit provides basic data for subsequent frequency band analysis by calculating the statistical parameters of the reference fault signal spectrum. The prominent frequency band identification subunit can identify the frequency band with concentrated energy, that is, the prominent frequency band, according to the statistical parameters, which helps to focus on the key frequency band. The main frequency band identification subunit screens out the main frequency bands from the prominent frequency bands and determines the total number, which improves the pertinence of the frequency band analysis. The decomposition modal number range determination subunit takes the maximum total number of main frequency bands as the lower limit value, and determines the value range in combination with the preset span, so that the determination of the decomposition modal number is more reasonable and accurate. The ability to more accurately determine the value range of the decomposition modal number provides a scientific basis for the subsequent fault signal decomposition and improves the accuracy and reliability of fault diagnosis.

[0104] Embodiment 5:

[0105] Based on Example 4, the main frequency band identification subunit, refer to Figure 2 ,include:

[0106] A prominent frequency band combination terminal is used to arbitrarily combine all prominent frequency bands of all reference fault signals to obtain all prominent frequency band sets of all reference fault signals;

[0107] An exhaustive co-occurrence rate calculation terminal, used to calculate the exhaustive co-occurrence rate of each prominent frequency band set in all reference fault signals based on the occurrence rate of each prominent frequency band in all reference fault signals;

[0108] An effective prominent frequency band screening end is used to treat the prominent frequency bands of all reference fault signals belonging to each fault type contained in the prominent frequency band set with the maximum exhaustive co-occurrence rate as the effective prominent frequency bands of each fault type;

[0109] The frequency band verification and screening end is used to verify and screen all valid prominent frequency bands of each fault type, analyze all main frequency bands corresponding to each fault type, and determine the total number of all main frequency bands of each reference fault signal.

[0110] In this embodiment, the set of all prominent frequency bands of all reference fault signals refers to a set formed by combining the respective prominent frequency bands of all medical rolling bearing fault signals used for reference.

[0111] In this embodiment, the occurrence rate of each prominent frequency band in all reference fault signals refers to the ratio of the number of times a specific prominent frequency band appears in all reference fault signals to the total number of reference fault signals.

[0112] In this embodiment, based on the occurrence rate of each prominent frequency band in all reference fault signals, the exhaustive co-occurrence rate of each prominent frequency band set in all reference fault signals is calculated, which means that based on the occurrence rate of each prominent frequency band itself, the probability of a frequency band set composed of multiple prominent frequency bands co-appearing in all reference fault signals is further calculated. The following steps can be followed: First, determine the number of times each prominent frequency band appears in all reference fault signals, and calculate its occurrence rate. Then, for the prominent frequency band set to be analyzed, consider the multiple prominent frequency bands contained therein. By traversing all reference fault signals, determine whether all the prominent frequency bands in the prominent frequency band set appear simultaneously in each signal. Count the number of times these prominent frequency bands appear simultaneously. Finally, divide the number of simultaneous occurrences by the total number of reference fault signals, and the result is the exhaustive co-occurrence rate of the prominent frequency band set in all reference fault signals.

[0113] In this embodiment, the fault type refers to various different types of faults that may occur in medical rolling bearings, such as inner ring fault (wear, crack, deformation), outer ring fault (wear, crack, deformation), rolling element fault (wear, crack, deformation) and normal state, etc.

[0114] The beneficial effects of the above technical solutions include: the prominent frequency band combination end provides a comprehensive set of frequency bands for subsequent analysis by any combination of prominent frequency bands. The exhaustive co-occurrence rate calculation end calculates the co-occurrence rate of each prominent frequency band set and quantifies the frequency characteristics of the frequency bands. The effective prominent frequency band screening end determines the effective prominent frequency bands according to the maximum co-occurrence rate, thereby improving the accuracy of screening. The frequency band verification and screening end further verifies and screens the effective prominent frequency bands to determine the main frequency bands and the total number, making the identification of the frequency bands more accurate and reliable. It can more accurately identify the main frequency bands, provide more valuable feature information for fault diagnosis, and improve the accuracy and reliability of diagnosis.

[0115] Embodiment 6:

[0116] Based on Example 5, the frequency band verification screening end refers to Figure 2 ,include:

[0117] A co-occurrence statistics sub-terminal, used to calculate the exhaustive co-occurrence rate of all valid prominent frequency bands of each fault type in all reference fault signals of the corresponding fault type based on the occurrence rate of each valid prominent frequency band of each fault type in all reference fault signals of the corresponding fault type;

[0118] The frequency band verification and screening sub-end is used to determine whether the exhaustive co-occurrence rate of all valid prominent frequency bands of each fault type in all reference fault signals of the corresponding fault type is not less than a preset exhaustive co-occurrence rate threshold. If so, all valid prominent frequency bands of the corresponding fault type are regarded as all main frequency bands of the corresponding fault type. Otherwise, it is determined whether the exhaustive co-occurrence rate of all valid prominent frequency bands of the corresponding fault type except the minimum co-occurrence rate in all reference fault signals of the corresponding fault type is not less than a preset exhaustive co-occurrence rate threshold, until it is determined that the exhaustive co-occurrence rate of all the latest remaining valid prominent frequency bands of the corresponding fault type in all reference fault signals of the corresponding fault type is not less than the preset exhaustive co-occurrence rate threshold, then all the latest remaining valid prominent frequency bands of the corresponding fault type are regarded as all main frequency bands of the corresponding fault type.

[0119] The frequency band total number counting sub-terminal is used to determine the total number of all main frequency bands of each reference fault signal.

[0120] In this embodiment, the occurrence rate of each effective prominent frequency band of each fault type in all reference fault signals of the corresponding fault type refers to the frequency ratio of a certain prominent frequency band identified as effective in all reference fault signals belonging to a specific fault type.

[0121] In this embodiment, based on the occurrence rate of each valid prominent frequency band of each fault type in all reference fault signals of the corresponding fault type, the exhaustive co-occurrence rate of all valid prominent frequency bands of each fault type in all reference fault signals of the corresponding fault type is calculated. This means that based on the above-mentioned occurrence rate, the probability of all prominent frequency bands that are recognized as valid to co-occur in all reference fault signals to which they belong under this fault type is further calculated.

[0122] In this embodiment, the preset exhaustive co-occurrence rate threshold is a pre-set standard value used to determine whether the probability of co-occurrence of effective prominent frequency bands in all reference fault signals of the corresponding fault type meets the standard. If the calculated exhaustive co-occurrence rate reaches or exceeds this threshold, these effective prominent frequency bands are considered to be of great significance; otherwise, the effectiveness of the relevant frequency bands may need to be re-evaluated or adjusted.

[0123] The beneficial effects of the above technical solutions include: the co-occurrence statistics sub-terminal provides a quantitative basis for the verification and screening of frequency bands by calculating the exhaustive co-occurrence rate of effective prominent frequency bands. The frequency band verification and screening sub-terminal gradually screens the effective prominent frequency bands through preset thresholds to ensure that the determined main frequency bands are sufficiently representative and important. The total number of frequency bands statistics sub-terminal can accurately determine the total number of main frequency bands for each reference fault signal, providing key data for subsequent operations such as determining the number of decomposed modes. Through precise statistics, screening and calculation, the accuracy and reliability of main frequency band identification are improved, laying a solid foundation for the effective decomposition and diagnosis of fault signals.

[0124] Embodiment 7:

[0125] Based on Example 1, the algorithm optimization module refers to Figure 2 ,include:

[0126] The parameter initialization submodule is used to initialize the number of parameters, search coefficients, initial weights and weight coefficients of the improved fruit fly algorithm, and determine the initial position of the fruit fly group: x′ axis = n*rand(domainofdefinition), and gives the random direction and distance x for a single fruit fly to find food i ′=x′ axis +ω*rand(domainofdefinition)andω=ω0*a ∞ ;

[0127] The algorithm optimization submodule is used to replace the nonlinear generation mechanism in the fruit fly algorithm with a new linear generation mechanism for candidate solutions, so that the odor concentration judgment value S i ′=x i ′, such S′ i The range can cover the entire range of the domain, overcoming the limitation of the original algorithm in the negative domain. i When the value of ′ is fixed, S i ′ follows a uniform distribution, thus allowing the search to be performed uniformly in the domain, enhancing the ability of the fruit fly swarm to find the global optimal solution. By inputting S′ i The odor concentration judgment equation is used to calculate the odor concentration Smell at each fruit fly's location. i ′, and introduce inertia weight to find the fruit fly with the maximum odor concentration, keep the maximum odor concentration value and coordinates, and the fruit fly group flies to this position through vision, repeating the above steps until the odor concentration reaches the maximum value or the number of iterations reaches the set maximum number of iterations. By introducing inertia weight to balance global search and local search, it tends to have higher global searchability at the beginning of the run, and has more local search capabilities at the end of the run, thereby improving the performance of the algorithm in complex optimization problems.

[0128] In this embodiment, initializing the number and scale of parameters, search coefficients, initial weights and weight coefficients of the improved fruit fly algorithm means setting the initial values ​​of the parameters involved in the algorithm before starting to use the improved fruit fly algorithm for calculation, including the number of parameters involved in the calculation, the search coefficient used to control the search range and direction, the initial weights of the fruit fly individuals and the weight coefficient used to adjust the calculation.

[0129] In this embodiment, the odor concentration judgment equation is a mathematical expression for calculating the odor concentration corresponding to the position of the fruit fly based on input related data (such as the position of the fruit fly, related parameters, etc.).

[0130] In this embodiment, the odor concentration judgment value is input into the odor concentration judgment equation to calculate the odor concentration of each fruit fly's location, find the fruit fly with the maximum odor concentration, keep the maximum odor concentration value and coordinates, and the fruit fly group flies to the location through vision, repeating the above steps until the odor concentration reaches the maximum value or the number of iterations reaches the set maximum number of iterations, which means that first a value for judging the odor concentration is given, and it is substituted into the odor concentration judgment equation to calculate the odor concentration of each fruit fly's location, find the fruit fly with the highest concentration, and record its concentration value and coordinates. Then other fruit flies move to this location, and the process of calculation, comparison, and movement is repeated until the calculated odor concentration no longer increases (reaches the maximum value) or has been repeated the set maximum number of times.

[0131] In this embodiment, in the improved fruit fly algorithm, the inertia weight is an important parameter for balancing global search and local search. It is usually associated with the search step length or direction adjustment of the fruit fly individual. In the iterative process of the algorithm, the inertia weight will affect the movement mode of the fruit fly individual in the search space. For example, when the inertia weight is large, the fruit fly individual will be more inclined to maintain the previous movement direction and larger step length when updating the position, which helps the algorithm to quickly explore the entire search space in the early stage of the search, thereby enhancing the global search ability and giving the algorithm a greater chance of finding a better area. As the iteration proceeds, when it is necessary to conduct a more detailed search on the found better area to determine the global optimal solution, the inertia weight will gradually decrease. At this time, the movement step length of the fruit fly individual will be reduced accordingly, and more attention will be paid to conducting local search near the current better area, thereby increasing the probability of finding an accurate optimal solution. In the improved fruit fly algorithm, the specific calculation or adjustment method of the inertia weight may be determined based on factors such as the number of iterations and the current search state. For example, a linear or nonlinear decreasing method may be used to adjust the inertia weight so that it gradually becomes smaller during the operation of the algorithm to achieve an effective transition from global search to local search.

[0132] The beneficial effects of the above technical solutions include: the parameter initialization submodule prepares for the operation of the algorithm by initializing the relevant parameters of the improved fruit fly algorithm, ensuring that the algorithm has a reasonable starting state. The algorithm optimization submodule replaces the nonlinear generation mechanism in the fruit fly algorithm with a new linear generation mechanism of candidate solutions, and continuously calculates the odor concentration at the fruit fly position to find the optimal position, thereby realizing the optimization search for parameters. By repeating the optimization steps until the maximum odor concentration or the maximum number of iterations is reached, the accuracy and efficiency of parameter optimization are improved. It can effectively optimize the regularization coefficient and wavelet kernel function parameters in the wavelet kernel extreme learning machine and improve the performance of the model. It enhances the optimization ability and adaptability of the algorithm in the fault diagnosis system, and provides strong support for more accurate fault diagnosis.

[0133] Embodiment 8:

[0134] Based on Example 1, the model construction and training module, refer to Figure 2 ,include:

[0135] A model building submodule is used to build a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters;

[0136] The model training submodule is used to divide all normalized feature vectors into a training set and a test set to train the wavelet kernel extreme learning machine fault diagnosis model to obtain a trained wavelet kernel extreme learning machine fault diagnosis model.

[0137] In this embodiment, a wavelet kernel extreme learning machine fault diagnosis model is constructed using the optimized regularization coefficient and wavelet kernel function parameters. This means that the optimal values ​​of the regularization coefficient and wavelet kernel function parameters obtained by the previous algorithm optimization are applied to the architecture of the wavelet kernel extreme learning machine to establish an initial model that can be used to diagnose medical rolling bearing faults.

[0138] In this embodiment, all normalized feature vectors (minimum-maximum normalization method or mean normalization method can be used) are divided into a training set and a test set to train the wavelet kernel extreme learning machine fault diagnosis model, and a trained wavelet kernel extreme learning machine fault diagnosis model is obtained, which means that the feature vector data that has been normalized is divided into two parts: a training set and a test set. The data of the training set is input into the constructed model, and the parameters of the model are continuously adjusted so that the corresponding results can be accurately predicted based on the input feature vectors. During the training process, the test set is also used to verify the accuracy and generalization ability of the model. After multiple training and adjustments, a trained model that can accurately diagnose faults is finally obtained.

[0139] The beneficial effects of the above technical solution include: the model building submodule uses the optimized parameters to build a fault diagnosis model to ensure that the model has better performance and adaptability. The model training submodule divides the feature vector into a training set and a test set for training, which can effectively evaluate the generalization ability and accuracy of the model. The trained model is obtained through training, which improves the accuracy and reliability of medical rolling bearing fault diagnosis. It helps to detect potential faults in time and provide guarantee for the normal operation of equipment maintenance and medical services. The process of model building and training is improved, and the practicality and effectiveness of the fault diagnosis system are improved.

[0140] Embodiment 9:

[0141] Based on Example 1, the fault diagnosis and evaluation module, refer to Figure 2 ,include:

[0142] A fault signal decomposition and processing submodule is used to decompose and process the fault signal of the medical rolling bearing to be tested, and obtain a feature vector with multi-dimensional characteristics of the medical rolling bearing to be tested;

[0143] The fault diagnosis and evaluation submodule is used to input the feature vector of the medical rolling bearing to be tested into the trained wavelet kernel extreme learning machine fault diagnosis model to obtain a fault diagnosis report of the medical rolling bearing to be tested.

[0144] In this embodiment, the fault signal of the medical rolling bearing to be tested is decomposed and processed to obtain a feature vector with multi-dimensional characteristics of the medical rolling bearing to be tested, which means that the fault signal generated by the medical rolling bearing that needs to be diagnosed whether there is a fault is converted into a numerical combination containing multiple different feature dimensions by using the decomposition processing method used when decomposing the fault signal of the medical rolling bearing in Examples 1 to 6. These feature vectors can comprehensively and meticulously describe the characteristics and properties of the fault signal of the bearing to be tested, and provide a key data basis for the subsequent accurate diagnosis of the fault type and assessment of the fault degree. For example, through decomposition processing, it is possible to obtain feature values ​​covering multiple dimensions such as vibration frequency, amplitude, and energy distribution, forming a feature vector with multi-dimensional characteristics.

[0145] The beneficial effects of the above technical solution include: the fault signal decomposition and processing submodule can decompose and process the fault signal of the medical rolling bearing to be tested, obtain the feature vector of multi-dimensional features, and provide basic data for accurate diagnosis. The fault diagnosis and evaluation submodule inputs the feature vector into the trained model to quickly obtain the fault diagnosis report, thereby improving the diagnostic efficiency. The ability to timely detect the fault condition of the bearing to be tested helps to take maintenance or replacement measures in advance to ensure the normal operation of medical equipment. Through accurate diagnosis and evaluation, the risk of medical accidents caused by faults is reduced and the safety of medical services is improved. Rapid and accurate diagnosis and evaluation of medical rolling bearing faults are achieved, improving the reliability and stability of medical equipment.

[0146] Embodiment 10:

[0147] The present invention provides a medical rolling bearing fault diagnosis method, which is applied to any one of the medical rolling bearing fault diagnosis systems in embodiments 1 to 9, referring to Figure 3 ,include:

[0148] S1: Decompose and process a large number of fault signals of medical rolling bearings to obtain a large number of feature vectors with multi-dimensional features;

[0149] S2: Introduce the improved fruit fly algorithm to optimize the regularization coefficient and wavelet kernel function parameters in the wavelet kernel extreme learning machine;

[0150] S3: construct a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters, and train the wavelet kernel extreme learning machine fault diagnosis model based on all eigenvectors to obtain a trained wavelet kernel extreme learning machine fault diagnosis model;

[0151] S4: Use the trained wavelet kernel extreme learning machine fault diagnosis model to diagnose and evaluate the medical rolling bearing fault to be tested, and output a fault diagnosis report.

[0152] The beneficial effects of the above technical solution include: Step S1 decomposes a large number of fault signals to obtain multi-dimensional feature vectors, providing a rich and accurate data basis for subsequent diagnosis. Step S2 uses the improved fruit fly algorithm to optimize key parameters, thereby improving the performance and accuracy of the model. Step S3 builds and trains the optimized model to make the model more targeted and reliable. Step S4 uses the trained model to perform diagnosis and evaluation, and outputs a diagnostic report, which can promptly detect the faults of the bearing to be tested and provide a basis for repair and maintenance. Efficient and accurate diagnosis and evaluation of medical rolling bearing faults are achieved, which helps to ensure the normal operation of medical equipment and improve the quality and safety of medical services.

[0153] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A medical rolling bearing fault diagnosis system, characterized in that: include: A fault signal decomposition and processing module is used to decompose and process a large number of fault signals of medical rolling bearings to obtain a large number of feature vectors with multi-dimensional features; Algorithm optimization module, used to introduce the improved fruit fly algorithm to optimize the regularization coefficient and wavelet kernel function parameters in the wavelet kernel extreme learning machine; The model building and training module is used to build a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters, and train the wavelet kernel extreme learning machine fault diagnosis model based on all feature vectors to obtain a trained wavelet kernel extreme learning machine fault diagnosis model; The fault diagnosis and evaluation module is used to diagnose and evaluate the fault of the medical rolling bearing to be tested using the trained wavelet kernel extreme learning machine fault diagnosis model and output a fault diagnosis report.

2. The medical rolling bearing fault diagnosis system according to claim 1, characterized in that: Fault signal decomposition and processing module, including: The fault signal decomposition submodule is used to decompose the fault signals of a large number of medical rolling bearings using the variational mode decomposition method to obtain a large number of effective mode components; The modal component processing submodule is used to process all effective modal components based on the singular value method to obtain a large number of eigenvectors with multi-dimensional characteristics.

3. The medical rolling bearing fault diagnosis system according to claim 2, characterized in that: Fault signal decomposition submodule, including: A decomposition modal number range determination unit is used to analyze the frequency spectrum characteristics of a large number of medical rolling bearing fault signals and determine the value range of the decomposition modal number; A parameter initialization unit, used to take the minimum value in the value range of the decomposition mode number as the initial value of the decomposition mode number, and determine the initial value of the quadratic penalty factor and the initial value of the convergence criterion; A parameter iteration value sequence generating unit is used to determine the parameter iteration value sequence of the decomposition mode number and the quadratic penalty factor based on the value range of the decomposition mode number and the corresponding parameter iteration change principle, the initial value range of the quadratic penalty factor and the corresponding parameter iteration change principle; A variational problem solving iteration unit is used to construct a variational problem, and solve the variational problem based on an initial value of a quadratic penalty factor and a corresponding parameter iteration value sequence, an initial value of a convergence criterion and a corresponding parameter iteration value sequence, and an optimization algorithm until a convergence condition is met, thereby obtaining multiple modal components; The validity verification and screening unit is used to verify and screen the validity of all modal components to obtain a large number of valid modal components.

4. The medical rolling bearing fault diagnosis system according to claim 3, characterized in that: Decomposition mode number range determination unit, including: The spectrum analysis subunit is used to obtain a large number of fault signals of medical rolling bearings under different working conditions as all reference fault signals, and calculate the mean, variance, and peak frequency of all reference fault signal spectra as statistical parameters of each reference fault signal spectrum; A prominent frequency band identification subunit, used to identify frequency bands corresponding to all energy concentration peaks in the spectrum of each reference fault signal based on statistical parameters of the spectrum of each reference fault signal as all prominent frequency bands of each reference fault signal; A main frequency band identification subunit, used for screening out all main frequency bands of each reference fault signal from all prominent frequency bands of each reference fault signal, and determining the total number of all main frequency bands of each reference fault signal; The decomposition modal number range determination subunit is used to take the maximum total number of main frequency bands among the total number of main frequency bands corresponding to all fault types as the lower limit of the decomposition modal number, and determine the value range of the decomposition modal number in combination with the preset value range span of the decomposition modal number.

5. The medical rolling bearing fault diagnosis system according to claim 4, characterized in that: The main frequency band identification subunits include: A prominent frequency band combination terminal is used to arbitrarily combine all prominent frequency bands of all reference fault signals to obtain all prominent frequency band sets of all reference fault signals; An exhaustive co-occurrence rate calculation terminal, used to calculate the exhaustive co-occurrence rate of each prominent frequency band set in all reference fault signals based on the occurrence rate of each prominent frequency band in all reference fault signals; An effective prominent frequency band screening end is used to treat the prominent frequency bands of all reference fault signals belonging to each fault type contained in the prominent frequency band set with the maximum exhaustive co-occurrence rate as the effective prominent frequency bands of each fault type; The frequency band verification and screening end is used to verify and screen all valid prominent frequency bands of each fault type, analyze all main frequency bands corresponding to each fault type, and determine the total number of all main frequency bands of each reference fault signal.

6. The medical rolling bearing fault diagnosis system according to claim 5, characterized in that: Band verification screening end, including: A co-occurrence statistics sub-terminal, used to calculate the exhaustive co-occurrence rate of all valid prominent frequency bands of each fault type in all reference fault signals of the corresponding fault type based on the occurrence rate of each valid prominent frequency band of each fault type in all reference fault signals of the corresponding fault type; The frequency band verification and screening sub-end is used to determine whether the exhaustive co-occurrence rate of all valid prominent frequency bands of each fault type in all reference fault signals of the corresponding fault type is not less than a preset exhaustive co-occurrence rate threshold. If so, all valid prominent frequency bands of the corresponding fault type are regarded as all main frequency bands of the corresponding fault type. Otherwise, it is determined whether the exhaustive co-occurrence rate of all valid prominent frequency bands of the corresponding fault type except the minimum co-occurrence rate in all reference fault signals of the corresponding fault type is not less than a preset exhaustive co-occurrence rate threshold, until it is determined that the exhaustive co-occurrence rate of all the latest remaining valid prominent frequency bands of the corresponding fault type in all reference fault signals of the corresponding fault type is not less than the preset exhaustive co-occurrence rate threshold, then all the latest remaining valid prominent frequency bands of the corresponding fault type are regarded as all main frequency bands of the corresponding fault type. The frequency band total number counting sub-terminal is used to determine the total number of all main frequency bands of each reference fault signal.

7. The medical rolling bearing fault diagnosis system according to claim 1, characterized in that: Algorithm optimization module, including: The parameter initialization submodule is used to initialize the number of parameters, search coefficients, initial weights and weight coefficients of the improved fruit fly algorithm, and determine the initial fruit fly group position x′ axis , and gives a random direction and distance x for a single fruit fly to find food i ′=x′ axis +ω*rand(domainofdefinition)andω=ω0*a ∞ ; The algorithm optimization submodule is used to replace the nonlinear generation mechanism in the fruit fly algorithm with a new linear generation mechanism for candidate solutions, so that the odor concentration judgment value S′ i = x′ i , when S′ i When the value of is fixed, S′ i Obeying uniform distribution, by inputting S′ i The odor concentration judgment equation is used to calculate the odor concentration Smell′ at each fruit fly's location i , and introduce inertia weight to find the fruit fly with the maximum odor concentration, keep the maximum odor concentration value and coordinates, and the fruit fly group flies to this position through vision, repeating the above steps until the odor concentration reaches the maximum value or the number of iterations reaches the set maximum number of iterations.

8. The medical rolling bearing fault diagnosis system according to claim 1, characterized in that: Model building and training modules, including: A model building submodule is used to build a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters; The model training submodule is used to divide all normalized feature vectors into a training set and a test set to train the wavelet kernel extreme learning machine fault diagnosis model to obtain a trained wavelet kernel extreme learning machine fault diagnosis model.

9. The medical rolling bearing fault diagnosis system according to claim 1, characterized in that: Fault diagnosis and evaluation modules, including: A fault signal decomposition and processing submodule is used to decompose and process the fault signal of the medical rolling bearing to be tested, and obtain a feature vector with multi-dimensional characteristics of the medical rolling bearing to be tested; The fault diagnosis and evaluation submodule is used to input the feature vector of the medical rolling bearing to be tested into the trained wavelet kernel extreme learning machine fault diagnosis model to obtain a fault diagnosis report of the medical rolling bearing to be tested.

10. A method for diagnosing a medical rolling bearing fault, characterized in that: The medical rolling bearing fault diagnosis system applied to any one of claims 1 to 9 comprises: S1: Decompose and process a large number of fault signals of medical rolling bearings to obtain a large number of feature vectors with multi-dimensional features; S2: Introduce the improved fruit fly algorithm to optimize the regularization coefficient and wavelet kernel function parameters in the wavelet kernel extreme learning machine; S3: construct a wavelet kernel extreme learning machine fault diagnosis model using the optimized regularization coefficient and wavelet kernel function parameters, and train the wavelet kernel extreme learning machine fault diagnosis model based on all eigenvectors to obtain a trained wavelet kernel extreme learning machine fault diagnosis model; S4: Use the trained wavelet kernel extreme learning machine fault diagnosis model to diagnose and evaluate the medical rolling bearing fault to be tested, and output a fault diagnosis report.

Images