A rotating machinery state monitoring method based on performance degradation space fluctuation evaluation
By extracting multi-domain state features and performing unconstrained orthogonal optimization on the vibration data of rotating machinery, a performance degradation space is constructed. This solves the problems of limited single-indicator capability and data sample size in the existing technology for rotating machinery condition monitoring, and achieves efficient and interpretable condition monitoring results.
Patent Information
- Application Number
- CN202210365305.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-07
- Publication Date
- 2026-02-13
- Estimated Expiration
- 2042-04-07
AI Technical Summary
Existing methods for monitoring the condition of rotating machinery suffer from limitations in characterizing the operating status of rotating machinery due to the limitations of single statistical indicators and the need for large amounts of data samples and computational time for machine learning methods, as well as a lack of interpretability.
By collecting vibration data of rotating machinery, multi-domain state features are extracted, and an unconstrained orthogonal optimization method is used to construct a multi-domain intrinsic state feature matrix. The performance degradation space of rotating machinery is constructed, and the state fluctuation rate is quantified through orthogonal matrix transformation to achieve state monitoring of rotating machinery equipment.
It enables accurate monitoring of the operating status of rotating machinery, improves fault sensitivity and robustness against transient disturbances, and has good engineering application value.
Smart Images
Figure CN115238726B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of mechanical equipment health monitoring and intelligent operation and maintenance, and particularly relates to a rotating machinery state monitoring method based on performance degradation space fluctuation evaluation. BACKGROUND
[0002] Rotating machinery is related to aerospace, high-speed railway, wind power generation and other fields that can affect the national economy and people's livelihood, and its health condition is directly related to the stability of the production process, and even determines the safety of people's life and property. Passive post-maintenance cannot avoid the serious consequences caused by failure, and will also disrupt the existing production rhythm; blind preventive maintenance will lead to unnecessary shutdown inspection, causing excessive maintenance and resource waste. Therefore, it is very important to effectively monitor the operating condition of rotating machinery, and early weak faults can be found in time through condition monitoring, so that a reasonable maintenance plan can be made, which is crucial for the safe operation of the whole machine and production operation.
[0003] It is worth noting that as the service conditions of rotating machinery become more and more complex, it becomes very difficult to establish an accurate and sensitive condition monitoring system. In view of this problem, experts and scholars have proposed a state index system to evaluate the stability of the operation of mechanical equipment. According to the different construction strategies, these state indexes can be divided into two categories: health indexes based on statistics and signal processing, and health indexes based on machine learning. From the construction method, the health indexes based on statistics and signal processing have clear physical meaning, and the calculation is also relatively convenient, and they have a very wide application in the industry; however, since they only consider the statistical information of a certain aspect, the representation ability of the operating state of rotating machinery equipment that has experienced different degradation stages is limited. In order to solve this problem, the construction of health indexes based on machine learning has gradually become a research hotspot, and through powerful feature extraction and information fusion, this kind of health index can more comprehensively represent the operating condition of mechanical equipment. However, the machine learning related algorithm does not have good physical interpretability, and the generalization ability is not strong, and the long time and data sample size problems still need to be further discussed and optimized.
[0004] Based on the above analysis, it is very urgent to establish an efficient, sensitive and interpretable rotating machinery equipment operating state monitoring index, SUMMARY
[0005] The purpose of the present application is to provide a rotating machinery state monitoring method based on performance degradation space fluctuation evaluation, comprising the following steps:
[0006] 1) Collecting rotating machinery vibration data.
[0007] The equipment for collecting rotating machinery vibration data includes a sensor.
[0008] 2) Multi-domain state feature extraction is performed on the rotating machinery vibration data, so as to obtain a multi-domain state feature matrix in the rotating machinery operation process.
[0009] The multi-domain state feature comprises rotating machinery state feature indexes in a time domain, a frequency domain and a time-frequency domain.
[0010] 3) An unconstrained orthogonal optimization method is adopted to optimize the multi-domain state feature matrix, so as to obtain a multi-domain intrinsic state feature matrix Y in the rotating machinery operation process.
[0011] The method for optimizing the multi-domain state feature matrix by adopting the unconstrained orthogonal optimization method comprises: unconstrained orthogonal optimization is performed on the multi-domain state feature matrix of adjacent time points [N-N0, N], so as to obtain the multi-domain intrinsic state feature matrix Y. N is a current time data sample serial number, and N0 is an adjacent sample number.
[0012] The step of establishing the multi-domain intrinsic state feature matrix Y comprises:
[0013] 3.1) An objective function of unconstrained orthogonal optimization is established, that is:
[0014]
[0015] In the formula, W is a weight matrix, X is a multi-domain state feature matrix of adjacent time points; X i represents the i-th column of the matrix X; n is the number of multi-domain features; λ is a regularization coefficient; Θ(x) = 1 / 2log[cosh(2x)] is an activation function; m is the number of columns of the weight matrix W; W j represents the j-th column of the weight matrix W;
[0016] 3.2) A weight matrix W satisfying the objective function of unconstrained orthogonal optimization is constructed.
[0017] 3.3) A multi-domain intrinsic state feature matrix Y is established.
[0018] 4) Orthogonal matrix transformation is performed on the multi-domain intrinsic state feature matrix Y, so as to construct a rotating machinery performance degradation space.
[0019] The step of constructing the rotating machinery performance degradation space comprises:
[0020] 4.1) F-norm of the multi-domain intrinsic state feature matrix Y is solved that is:
[0021]
[0022] 4.2) A matrix A = W T W is obtained, and F-norm
[0023] 4.3) Solve the eigenvalues and eigenvectors of matrix A, and arrange the eigenvalues and eigenvectors in ascending order to obtain:
[0024]
[0025] wherein Φ, Υ are the eigenvector matrix and the eigenvalue matrix; υ n , λ n are the eigenvector and the eigenvalue.
[0026] 4.4) Orthogonally transform the multi-domain state feature matrix X to obtain an orthogonal matrix X * , that is:
[0027] X * = Φ T X (4)
[0028] 4.5) Update the F-norm of the multi-domain eigenstate feature matrix according to the orthogonal matrix X * , to obtain:
[0029]
[0030] 4.6) Transform the F-norm of the multi-domain eigenstate feature matrix according to formula (5) to obtain:
[0031]
[0032] For the actual operation of the rotating machinery equipment, then:
[0033]
[0034] 4.7) Set the eigenvalue constraint condition, that is:
[0035]
[0036] wherein λ is the optimized eigenvalue.
[0037] 4.8) Construct the rotating machinery performance degradation space , that is:
[0038]
[0039] 5) Calculate the rotating machinery state fluctuation rate based on the rotating machinery performance degradation space.
[0040] The steps for calculating the rotating machinery state fluctuation rate index include:
[0041] 5.1) According to the performance degradation space of rotating machinery Calculate the current operating state f(X * ,Φ,Υ,n) of the rotating machinery, that is:
[0042]
[0043] In the formula, is the i-th fluctuation direction of the performance degradation space of the rotating machinery; Ω is the n-dimensional integral domain after eigenvalue constraint, and is expressed as:
[0044]
[0045] 5.2) Calculate the state fluctuation rate C FR of the rotating machinery under the current operating state, that is:
[0046]
[0047] In the formula, is the initial operating state of the rotating machinery.
[0048] 6) Analyze the state fluctuation rate of the rotating machinery to obtain the operating state of the rotating machinery equipment.
[0049] The step of analyzing the state fluctuation rate of the rotating machinery to obtain the operating state of the rotating machinery equipment includes: judging whether the state fluctuation rate C FR of the rotating machinery under the current operating state is less than a preset state fluctuation rate threshold 3σ, if yes, the operating state of the rotating machinery equipment is normal, otherwise, it is abnormal.
[0050] If the operating state of the rotating machinery equipment is abnormal, then perform square envelope spectrum analysis on the abnormal vibration data collected in the current period to determine the fault position of the rotating machinery.
[0051] The method for determining the fault position of the rotating machinery includes: if the fault characteristic frequency and its multiple frequency components of the rotating machinery parts appear in the square envelope spectrum, then the parts are faulty.
[0052] It is worth mentioning that the application firstly extracts multi-domain state features from the collected rotating machinery vibration data, establishes high-dimensional statistical feature expression of the running process, then adopts unconstrained orthogonal optimization method to perform weighted processing on the established high-dimensional statistical features, thereby obtaining intrinsic state features of the mechanical equipment running process, and on this basis, a mechanical performance degradation space expressed by integrated multi-domain intrinsic state features is constructed through orthogonal matrix transformation, and effective state monitoring of the mechanical equipment is realized by quantifying the fluctuation of the performance degradation space. The state monitoring method based on performance degradation space fluctuation evaluation provided by the application overcomes the problems of limited representation ability of traditional single statistical index and large sample size and long time consumption of machine learning method, can accurately monitor the state of the mechanical equipment, and has good engineering application value.
[0053] The technical effect of the application is self-evident. For the running process of rotating machinery equipment, the application provides a rotating machinery state monitoring method based on performance degradation space fluctuation evaluation, constructs a performance degradation space capable of accurately reflecting the running state of the mechanical equipment through orthogonal integrated transformation of multi-domain intrinsic state features, and overcomes the problem of limited representation ability of traditional single index based on statistical information, and does not need to train a large number of data samples, has the advantages of convenient calculation and interpretability; in addition, since the multi-domain intrinsic state features of adjacent time points are considered, the fault sensitivity and transient interference robustness are improved, and the method has good engineering application value.
[0054] The application fully utilizes the multi-domain state features of rotating machinery equipment to establish a performance degradation space capable of accurately reflecting the running state change of the rotating machinery equipment, and realizes effective state monitoring of the rotating machinery equipment by quantifying the fluctuation of the performance degradation space. BRIEF DESCRIPTION OF DRAWINGS
[0055] Figure 1 A performance degradation space fluctuation evaluation based rotating machinery state monitoring method provided by the application;
[0056] Figure 2 A bearing data acquisition test bed in an embodiment of the application;
[0057] Figure 3 A bearing vibration signal time domain graph in an embodiment of the application;
[0058] Figure 4 A bearing state monitoring result based on performance degradation space fluctuation evaluation in an embodiment of the application;
[0059] Figure 5 An abnormal data square envelope spectrum analysis result in an embodiment of the application; Figure 5(a)-(d) are abnormal data square envelope spectrum analysis results of health state, data file 526, data file 527 and data file 535 respectively;
[0060] Figure 6 For the bearing multiple index state monitoring comparison results in the embodiment of the application;
[0061] Figure 7 For the monotonicity and trend measurement results of the constructed index in the embodiment of the application. DETAILED DESCRIPTION
[0062] The application will be further described below with reference to the embodiments, but should not be understood as limiting the above-mentioned subject matter of the application to the following embodiments. Various substitutions and changes can be made according to ordinary technical knowledge and conventional means in the art without departing from the above-mentioned technical ideas of the application, and all should be included in the protection scope of the application.
[0063] Embodiment 1:
[0064] Reference Figures 1 to 7 A rotating machine state monitoring method based on performance degradation space fluctuation evaluation, comprising the following steps:
[0065] 1) Collecting rotating machine vibration data.
[0066] The device for collecting rotating machine vibration data includes a sensor.
[0067] 2) Extracting multi-domain state features from the rotating machine vibration data to obtain a multi-domain state feature matrix during the operation of the rotating machine.
[0068] The multi-domain state features include rotating machine state feature indicators in time domain, frequency domain and time-frequency domain.
[0069] 3) Optimizing the multi-domain state feature matrix using an unconstrained orthogonal optimization method to obtain a multi-domain intrinsic state feature matrix Y during the operation of the rotating machine.
[0070] The method for optimizing the multi-domain state feature matrix using an unconstrained orthogonal optimization method includes: unconstrained orthogonal optimization of the multi-domain state feature matrix of adjacent time points [N-N0, N] to obtain the multi-domain intrinsic state feature matrix Y. N is the current time data sample serial number, and N0 is the number of adjacent samples
[0071] The step of establishing the multi-domain intrinsic state feature matrix Y includes:
[0072] 3.1) Establishing an objective function for unconstrained orthogonal optimization, that is:
[0073]
[0074] where W is the weight matrix, X is the multi-domain state feature matrix at the current time; X i denotes the i-th column of matrix X; n is the number of multi-domain features; λ is the regularization coefficient; Θ(x) = 1 / 2log[cosh(2x)] is the activation function; m is the number of columns of the weight matrix W; W j denotes the j-th column of the weight matrix W;
[0075] 3.2) Construct the weight matrix W that satisfies the unconstrained orthogonal optimization objective function.
[0076] 3.3) Establish the multi-domain eigenstate feature matrix.
[0077] 4) Perform orthogonal matrix transformation on the multi-domain eigenstate feature matrix Y, thereby constructing a rotating machinery performance degradation space.
[0078] The step of constructing a rotating machinery performance degradation space includes:
[0079] 4.1) Solve the F-norm of the multi-domain eigenstate feature matrix Y That is:
[0080]
[0081] 4.2) Let matrix A = W T W, update the F-norm
[0082] 4.3) Solve the eigenvalues and eigenvectors of matrix A, and arrange the eigenvalues and eigenvectors in ascending order, to obtain:
[0083]
[0084] where Φ, Υ are the eigenvalue matrix and eigenvector matrix. υ n , λ n are the eigenvalues and eigenvectors.
[0085] 4.4) Perform orthogonal transformation on the multi-domain state feature matrix X, to obtain the orthogonal matrix X * That is:
[0086] X * = Φ T X (4)
[0087] 4.5) According to the orthogonal matrix X * , update the F-norm of the multi-domain eigenstate feature matrix to obtain:
[0088]
[0089] 4.6) According to formula (5), the F-norm of the multi-domain eigenstate feature matrix is calculated Transforming, we get
[0090]
[0091] For the actual operation of the rotating machinery equipment, Then we have
[0092]
[0093] 4.7) Set the eigenvalue constraint condition, that is,
[0094]
[0095] In the formula, is the optimized eigenvalue.
[0096] 4.8) Construct the rotating machinery performance degradation space That is,
[0097]
[0098] 5) Based on the rotating machinery performance degradation space, the rotating machinery state fluctuation rate is calculated.
[0099] The steps of calculating the rotating machinery state fluctuation rate index include:
[0100] 5.1) According to the rotating machinery performance degradation space The current operating state f(X * ,Φ,γ,n) of the rotating machinery is calculated, that is,
[0101]
[0102] In the formula, is the i-th fluctuation direction of the rotating machinery performance degradation space; Ω is the n-dimensional integral domain after eigenvalue constraint, which is expressed as:
[0103]
[0104] 5.2) The state fluctuation rate C FR of the rotating machinery under the current operating state is calculated, that is,
[0105]
[0106] In the formula, is the initial operating state of the rotating machinery.
[0107] 6) The rotating machinery state fluctuation rate is analyzed to obtain the operating state of the rotating machinery equipment.
[0108] The step of analyzing the state fluctuation rate of the rotating machinery to obtain the running state of the rotating machinery equipment comprises: judging whether the state fluctuation rate C of the rotating machinery under the current running state is less than a preset state fluctuation rate threshold 3σ FR If yes, the running state of the rotating machinery equipment is normal, otherwise, it is abnormal.
[0109] If the running state of the rotating machinery equipment is abnormal, square envelope spectrum analysis is performed on the abnormal vibration data collected in the current period to determine the fault position of the rotating machinery.
[0110] The method for determining the fault position of the rotating machinery comprises: if the fault characteristic frequency and its multiple frequency components of the rotating machinery part appear in the square envelope spectrum, the part is in fault.
[0111] Embodiment 2
[0112] A rotating machinery state monitoring method based on performance degradation space fluctuation evaluation, the specific steps are as follows:
[0113] Step 1: multi-domain state feature extraction is performed on the collected rotating machinery vibration data, and a high-dimensional statistical feature expression of the running process is established;
[0114] Step 2: the high-dimensional statistical features constructed in step 1 are weighted by using an unconstrained orthogonal optimization method to obtain the intrinsic state features of the mechanical equipment running process; wherein, the adjacent time is set as [N-N0, N], N is the current time data sample serial number, N0 is the adjacent sample number, and the unconstrained orthogonal optimization mode is:
[0115]
[0116] In the formula, W is a weight matrix, X is a multi-domain state feature matrix of adjacent time, n is the number of multi-domain features, λ is a regularization coefficient, Θ(x) is an activation function, and Θ(x) = 1 / 2log[cosh(2x)].
[0117] Step 3: the mechanical performance degradation space integrated by the multi-domain intrinsic state features in step 2 is constructed by orthogonal matrix transformation, and the specific steps are as follows:
[0118] Step 301: the intrinsic state feature matrix established by the unconstrained orthogonal optimization is represented as Y = W T ·X;
[0119] Step 302: the F-norm of the intrinsic state feature matrix is solved:
[0120]
[0121] Step 303: let A = W T W, then
[0122] Step 304: Solve the eigenvalues and eigenvectors of matrix A, arrange its eigenvalues in ascending order, and rearrange the corresponding eigenvectors accordingly, denoted as:
[0123]
[0124] Step 305: Perform an orthogonal transformation by the orthogonal matrix Φ:
[0125] X * = Φ T X
[0126] Step 306: Further transform the eigenstate characteristic matrix in step 302 to:
[0127]
[0128] Step 307: Utilize the matrix transformation principle, further transform the eigenstate characteristic matrix in step 306 to:
[0129]
[0130] Step 308: For the actual operation of the rotating machinery equipment, then:
[0131]
[0132] Step 309: In order to eliminate the abnormal influence of the extremely small eigenvalues, set the following constraint condition:
[0133]
[0134] Step 310: Through the above orthogonal transformation of the eigenstate characteristic matrix, the performance degradation space is constructed as:
[0135]
[0136] Step 4: According to the mechanical performance degradation space established in step 3, construct the state fluctuation rate index, and thus perform the rotating machinery equipment operation state monitoring, the specific steps are:
[0137] Step 401: According to the performance degradation space constructed in step 3, calculate its volume:
[0138]
[0139] where, is the i-th fluctuation direction of the rotating machinery performance degradation space; Ω is the n-dimensional integral domain after eigenvalue constraint.
[0140] Step 402: the representation of Ω in step 401 is:
[0141]
[0142] Step 403: calculate the state fluctuation rate of the mechanical equipment under the current running state:
[0143]
[0144] In the formula, is the initial running state of the mechanical equipment.
[0145] Step 404: set the state fluctuation rate 3σ threshold value through the normal running state data of the mechanical equipment;
[0146] Step 405: through the analysis of the relationship between the state fluctuation rate index in the actual running process of the mechanical equipment and the 3σ threshold value established in step 404, the abnormal state monitoring of the equipment is carried out.
[0147] Step 406: square envelope spectrum analysis is carried out on the detected abnormal state data, so as to realize accurate tracing of the mechanical equipment fault.
[0148] Example 3:
[0149] This embodiment is based on the full life cycle data set of rolling bearing provided by the intelligent operation and maintenance system center of the University of Cincinnati, USA. The rolling bearing running process vibration data acquisition test bed is as shown in Figure 2 During the test, the main shaft speed is 2000 revolutions per minute, the rolling bearing vibration data is collected every 10 minutes, the sampling frequency is 20 kHz, each data file contains 20480 data points, and a total of 984 data files are generated during the whole test process, and the time domain waveform of the vibration signal is as shown in Figure 3 After the test, it is confirmed that the bearing outer ring fault occurs, and the fault characteristic frequency is 236.4 Hz.
[0150] Step 1: the collected rolling bearing vibration data is characterized by using the multi-domain state characteristics shown in Table 1 as follows:
[0151] Table 1 expression of multi-domain state characteristics of rolling bearing
[0152]
[0153] Then the high-dimensional statistical feature matrix of the rolling bearing running process is represented as:
[0154] X=[X1,X2,…,X i ,…,X 56 ]
[0155] Step 2: the high-dimensional statistical features at the adjacent time points constructed in step 1 are weighted by using an unconstrained orthogonal optimization method, to obtain intrinsic state features of the rolling bearing operation state; wherein the adjacent time points are set as the previous 50 adjacent data collection time periods including the current time point, that is, the dimension of the high-dimensional statistical feature X matrix is 50*56.
[0156] Step 3: a rolling bearing performance degradation space expressed by the multi-domain intrinsic state features integrated in step 2 is constructed by orthogonal matrix transformation.
[0157] Step 4: according to the rolling bearing performance degradation space established in step 3, a running process state fluctuation rate index is constructed, and thus the rolling bearing operation state monitoring is performed. The previous 100 state fluctuation rate indexes are selected as the monitoring results of the rolling bearing in a normal state, and a 3σ threshold interval is set, and the result is as shown in Figure 4 It can be seen that the state monitoring method provided by the present application has good sensitivity and monotonicity, and according to the 3σ criterion, the running state of the rolling bearing starts to abnormally fluctuate at the 526th data collection time period, which indicates that the rolling bearing is likely to have an early weak fault at this time.
[0158] In order to verify whether the fault exists, the square envelope spectrum analysis is performed on the vibration data at the 526th and adjacent time periods, and the result is as shown in Figure 5 It can be seen that from the 526th data collection time period, the rolling bearing outer ring fault characteristic frequency f o and its double frequency component start to appear with relatively weak energy, and at the 535th data collection time period, the fault characteristic frequency and its multiple frequency components have been highlighted in the square envelope spectrum, and the energy is relatively large, at this time, the local fault of the outer ring has been generated. It can be seen from this that the state monitoring method based on performance degradation space fluctuation evaluation provided by the present application has strong sensitivity to early weak faults.
[0159] In order to verify the superior performance of the method provided by the present application, the method is compared with six commonly used feature indexes, including kurtosis, root mean square (RMS), Hoyer measure, L2 / L1 norm, negative entropy and smoothness index, and the normalized results of the rolling bearing state monitoring are as shown in Figure 6 .
[0160] It can be seen from the comparison results that the state monitoring method provided by the present application has better sensitivity and monotonicity than the other six commonly used feature indexes. Figure 6It can be seen that when the local fault of the bearing outer ring begins to appear and gradually aggravates, the above six characteristic indexes all appear different degrees of sharp fluctuations, the monotonicity and trend of the indexes are not ideal, which is relatively unfavorable for subsequent residual life prediction and maintenance planning based on state indexes. In order to verify the sensitivity of the six characteristic indexes to early faults, the threshold interval is set by 3σ criterion, and the detected initial points of faults are shown in Table 2. It can be seen that the detection results of the initial points of faults of the above six characteristic indexes are all lagged behind the detection results of the method provided by the present application. It is verified that the method provided by the present application has stronger sensitivity to early weak faults, and has good state monitoring effect.
[0161] Table 2: Initial points of faults detected by different methods
[0162]
[0163] In addition to the fault sensitivity, the monotonicity and trend of different characteristic indexes are quantitatively analyzed, and the results are shown in Table 3. Figure 7 It can be seen that the running state monitoring results of the rolling bearing obtained by the method provided by the present application have good monotonicity and trend, which can provide feasible reference basis for subsequent residual life prediction and maintenance planning of the rolling bearing based on state indexes.
Claims
1. A method for condition monitoring of rotating machinery based on performance degradation spatial fluctuation assessment, characterized in that, The method comprises the following steps: 1) collecting the rotating machinery vibration data; 2) extracting multi-domain state features from the rotating machinery vibration data, thereby obtaining a multi-domain state feature matrix during the operation of the rotating machinery; 3) performing weighted processing on the high-dimensional statistical features of the adjacent time points constructed in step 1 by using an unconstrained orthogonal optimization method, thereby obtaining a multi-domain intrinsic state feature matrix Y of the rotating machinery during the operation of the rotating machinery; wherein the adjacent time points are set as [N-N0, N], N is the current time data sample serial number, and N0 is the number of adjacent samples, 4) performing orthogonal matrix transformation on the multi-domain intrinsic state feature matrix Y, thereby constructing a rotating machinery performance degradation space; 5) calculating the rotating machinery state fluctuation rate based on the rotating machinery performance degradation space; 6) analyzing the rotating machinery state fluctuation rate to obtain the operating state of the rotating machinery equipment; The device for collecting the rotating machinery vibration data comprises a sensor; If the operating state of the rotating machinery equipment is abnormal, then square envelope spectrum analysis is performed on the abnormal vibration data collected in the current period to determine the rotating machinery fault position; The method for determining the rotating machinery fault position comprises: if the fault characteristic frequency and its multiple frequency components of the rotating machinery parts appear in the square envelope spectrum, then the parts are in fault; The multi-domain state features comprise rotating machinery state feature indexes in the time domain, the frequency domain and the time-frequency domain; The method for optimizing the multi-domain state feature matrix by using the unconstrained orthogonal optimization method comprises: performing unconstrained orthogonal optimization on the multi-domain state feature matrix of the adjacent time points [N-N0, N] to obtain the multi-domain intrinsic state feature matrix Y; N is the current time data sample serial number, and N0 is the number of adjacent samples; The step of establishing the multi-domain intrinsic state feature matrix Y comprises: 3.1) establishing an objective function of the unconstrained orthogonal optimization, namely: In the formula, W is a weight matrix, X is a multi-domain state feature matrix at a nearby time; X i represents the ith column of the matrix X; n is the number of multi-domain features; λ is a regularization coefficient; Θ(x) = 1 / 2log[cosh(2x)] is an activation function; m is the number of columns of the weight matrix W; W j represents the jth column of the weight matrix W; 3.2) constructing a weight matrix W satisfying the objective function of the unconstrained orthogonal optimization; 3.3) establishing the multi-domain intrinsic state feature matrix; The step of constructing the rotating machinery performance degradation space comprises: 4.1) Solving the F-norm of the multi-domain eigenstate feature matrix Y i.e.: 4.2) Let A = W T W, update F-norm 4.3) solving the eigenvalues and eigenvectors of the matrix A, and arranging the eigenvalues and eigenvectors in ascending order, thereby obtaining: In the formula, Φ, Υ are the eigenvector matrix, the eigenvalue matrix; υ n , λ n are the eigenvector, the eigenvalue; 4.4) Orthogonal transformation is performed on the multi-domain state feature matrix X to obtain an orthogonal matrix X * That is: X * = Φ T X (4) 4.5) According to the orthogonal matrix X * , update the F-norm of the multi-domain eigenstate feature matrix We get: 4.6) The F-norm of the multi-domain eigenstate feature matrix is calculated according to equation (5) Transforming, we get: For the actual operation of the rotating machinery equipment, Then, 4.7) setting the eigenvalue constraint condition, namely: In the formula, is the optimized characteristic value; 4.8) Constructing a rotating machinery performance degradation space That is: The step of calculating the rotating machinery state fluctuation rate index comprises: 5.1) According to the space of performance degradation of rotating machinery Calculate the current operating state f(X * ,Φ,Υ,n) of rotating machinery, that is: In the formula, is the i-th fluctuation direction of the rotating machinery performance degradation space; Ω is the n-dimensional integral domain after eigenvalue constraint, which is represented as: 5.2) Calculate the state fluctuation rate C of the rotating machinery under the current operating state FR i.e.: In the formula, is the initial running state of the rotating machinery; The steps for analyzing the state fluctuation rate of the rotating machinery and obtaining the running state of the rotating machinery equipment include: judging whether the state fluctuation rate C of the rotating machinery under the current running state is less than a preset state fluctuation rate threshold 3σ FR If yes, the running state of the rotating machinery equipment is normal, otherwise, it is abnormal.