Method and system for detecting early failure of rotary machinery based on von Neumann diagram entropy

Abstract

The invention provides a rotary machine early fault detection method and system based on von Neumann diagram entropy, and relates to the technical field of rotary machine fault detection. Comprising the following steps: acquiring a vibration signal of a rotating machine, performing Gabor transformation to obtain a series of short-time power spectrums, mapping the short-time power spectrums, and obtaining a plurality of undirected weighted neighbor graph models; calculating a Laplacian matrix of the undirected weighted neighbor graph model, constructing a graph density vector associated with the corresponding undirected weighted neighbor graph model, and combining the graph density vector with von Neumann entropy to obtain von Neumann graph entropy; ranking von Neumann diagram entropies of each undirected weighted neighbor graph model to obtain a von Neumann diagram entropy set; and when the von Neumann diagram entropy in the judgment set is abnormally increased, judging that the rotating machine has a fault. According to the method, the sensitivity to the change of an abnormal state and the robustness to noise interference are improved, and the occurrence of a fault state can be timely and accurately captured in an early stage.

Description

Technical Field

[0001] The present invention belongs to the technical field of rotating machinery fault detection, and in particular relates to a rotating machinery early fault detection method and system based on von Neumann graph entropy. Background Art

[0002] Rotating machinery is a very important equipment in the industrial field. As a vital component in rotating machinery, rolling bearings can withstand axial and radial loads during rotation, thereby ensuring efficient and stable operation of rotating parts. With the development of diversified and complex industrial applications, these parts are prone to damaging failures and serious accidents due to long-term operation under harsh working conditions such as high speed, high temperature, and heavy load. According to statistics, 70% of rolling failures in rotating machinery are caused by rolling bearings. Therefore, early status detection of rotating machinery is important for understanding the performance status of bearings and timely discovering potential failures.

[0003] At present, the fault diagnosis methods of rolling bearings can be mainly divided into three categories: model-based, knowledge-based and signal processing-based. Among them, the fault diagnosis method based on signal processing has been widely studied and applied because it does not require fault samples and accurate mathematical models. This method extracts data features during the operation of the equipment and identifies them to determine the operating status of the equipment.

[0004] However, the state changes of rolling bearings in the early degradation stage are very weak. In actual industrial scenarios, accurately extracting effective information and state characteristics from signals is still a challenging task. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a rotating machinery early fault detection method and system based on von Neumann graph entropy, which uses a neighbor graph to establish a dynamic model of the time-frequency analysis results of vibration signals, thereby improving the sensitivity to changes in abnormal states. By changing the structure of the von Neumann graph entropy quantization model, the robustness to noise interference is improved, thereby being able to capture the occurrence of early fault states in a timely and accurate manner at an early stage.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] A first aspect of the present invention provides a method for detecting early faults of rotating machinery based on von Neumann graph entropy.

[0008] The method for detecting early faults of rotating machinery based on von Neumann graph entropy comprises the following steps:

[0009] Acquire a vibration signal of a rotating machine, pre-process the vibration signal, and obtain vibration signals corresponding to a plurality of time series segments;

[0010] Gabor transform is performed on the vibration signals of multiple time series segments to obtain a series of short-time power spectra, and the series of short-time power spectra are mapped respectively to obtain multiple undirected weighted neighbor graph models;

[0011] Calculate the Laplace matrix of the undirected weighted neighbor graph model, construct a graph density vector associated with the corresponding undirected weighted neighbor graph model based on the eigenvalue of the Laplace matrix, combine the graph density vector with the von Neumann entropy, and calculate the von Neumann graph entropy describing the health status of the rotating machinery; sort the von Neumann graph entropies of each undirected weighted neighbor graph model in sequence according to the graph model serial number to obtain a von Neumann graph entropy set;

[0012] The changing pattern of the von Neumann diagram entropy in the judgment set is determined. When the von Neumann diagram entropy increases abnormally, it is determined that a fault has occurred in the rotating machinery.

[0013] A second aspect of the present invention provides a rotating machinery early fault detection system based on von Neumann graph entropy.

[0014] The rotating machinery early fault detection system based on von Neumann graph entropy includes:

[0015] The vibration signal acquisition module is configured to: acquire the vibration signal of the rotating machinery, pre-process the vibration signal, and obtain vibration signals corresponding to a plurality of time-series segment segments;

[0016] The neighbor graph modeling module is configured to: perform Gabor transformation on the vibration signals of multiple time series segments to obtain a series of short-time power spectra, and respectively map the series of short-time power spectra to obtain multiple undirected weighted neighbor graph models;

[0017] The von Neumann graph entropy index calculation module is configured to: calculate the Laplace matrix of the undirected weighted neighbor graph model, construct a graph density vector associated with the corresponding undirected weighted neighbor graph model based on the eigenvalue of the Laplace matrix, combine the graph density vector with the von Neumann entropy, and calculate the von Neumann graph entropy describing the health status of the rotating machinery; sort the von Neumann graph entropy of each undirected weighted neighbor graph model in sequence according to the graph model sequence number to obtain a von Neumann graph entropy set;

[0018] The judgment decision module is configured to: judge the change law of the von Neumann graph entropy in the set, and when the von Neumann graph entropy increases abnormally, it is judged that the rotating machinery has a fault.

[0019] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the method for early fault detection of rotating machinery based on von Neumann graph entropy as described in the first aspect of the present invention.

[0020] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for early fault detection of rotating machinery based on von Neumann graph entropy as described in the first aspect of the present invention are implemented.

[0021] One or more of the above technical solutions have the following beneficial effects:

[0022] 1. The present invention provides a method and system for detecting early faults of rotating machinery based on von Neumann graph entropy. Through the neighbor graph model, the correlation relationship of local frequency components in the spectrum is established to characterize the weak changes of the signal. By using the von Neumann graph entropy to quantify the structural changes between graphs, a nonlinear description of signal changes is achieved, so that the occurrence of early fault states can be captured in a timely and accurate manner at an early stage.

[0023] 2. Considering the local changes in the spectrum under fault conditions, the nearest neighbor graph is used to extract effective information from the signal and use less data to describe the bearing state changes. Compared with the complete graph model, the modeling time complexity of the nearest neighbor graph is reduced from O(N^ 2 ) is reduced to O(N), which improves the computational efficiency.

[0024] 3. The points in the graph represent frequency component data, and the weighted edges represent the correlation between component data. The contribution of random noise to the change of von Neumann graph entropy is almost negligible. Therefore, the structural similarity between graphs is measured by von Neumann graph entropy, which improves the robustness to noise. In addition, when the machine operation state changes, the changes between some strong connections in the model will cause significant differences in the structure of the graph model, so that regularity and noise can be distinguished in noisy data, which improves the sensitivity of early change detection.

[0025] 4. In the field of fault diagnosis, measuring the differences between graphs is a challenging task. The traditional measurement is a distance measurement from the perspective of geometric relationships. When outliers appear in the data set, the distance measurement of the graph is instable, which may cause a large difference between the measurement result and the expected result. In the quantum field, von Neumann entropy quantifies the complexity of the quantum system through the density matrix. The present invention combines it with graph theory and proposes von Neumann graph entropy to evaluate the evolution state of continuous dynamic graphs. When the machine state changes abnormally, the graph structure will change differentially. By extracting the local structural changes of the Laplace density operator to measure the graph, the nonlinear description of the fault state based on the graph theory method is improved, which opens up a new direction for the application of graph theory in the field of fault diagnosis.

[0026] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0028] Figure 1 The figure is a schematic diagram of a method flow of a first embodiment of the present invention.

[0029] Figure 2 Schematic diagram of the neighbor graph model.

[0030] Figure 3 Schematic diagram of the graph modeling process flow.

[0031] Figure 4 Comparison chart of time complexity for graph modeling.

[0032] Figure 5 This is a schematic diagram of the test signal detection results.

[0033] Figure 6 Schematic diagram of the test signal detection results with unknown interference.

[0034] Figure 7 This is a schematic diagram of the actual test signal detection results of a pulp mill. DETAILED DESCRIPTION

[0035] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0036] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0037] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0038] Embodiment 1

[0039] Based on the above-mentioned background technology, the state change of rolling bearings in the early degradation stage is very weak. In actual industrial scenarios, accurately extracting effective information and state characteristics from signals is still a challenging task. In this regard, the inventors found that the fault frequency of the time-frequency energy spectrum is closely related to the health state of the rolling bearing. Extracting these components and constructing them into a graph model can map these local related frequency changes into graph topology changes, thereby enhancing and amplifying the complex dynamic changes of the frequency through the graph structure, and improving the sensitivity of the description of the health state.

[0040] In addition, Von Neumann entropy is a measure of the topological structure of a graph, which can be used to quantify the degree of mixing between nodes in the graph, reflecting the structural distribution and complexity of the graph. Since the graph structure represents the difference between components, it has a certain robustness to noise interference. Therefore, Von Neumann entropy can accurately describe the nonlinear changes in the bearing state. This method effectively improves the sensitivity and accuracy of early state change recognition.

[0041] Based on this, this embodiment proposes a method for early fault detection of rotating machinery based on von Neumann graph entropy. The method uses a neighbor graph to establish a dynamic model of the vibration signal spectrum analysis results, thereby improving the sensitivity to changes in abnormal conditions. By quantifying the structure of the von Neumann graph entropy model, the robustness to noise interference is improved, thereby being able to capture the occurrence of fault conditions in a timely and accurate manner at an early stage.

[0042] like Figure 1 As shown, the early fault detection method of rotating machinery based on von Neumann graph entropy includes the following steps:

[0043] Acquire a vibration signal of a rotating machine, pre-process the vibration signal, and obtain vibration signals corresponding to a plurality of time series segments;

[0044] Gabor transform is performed on the vibration signals of multiple time series segments to obtain a series of short-time power spectra, and the series of short-time power spectra are mapped respectively to obtain multiple undirected weighted neighbor graph models;

[0045] Calculate the Laplace matrix of the undirected weighted neighbor graph model, construct a graph density vector associated with the corresponding undirected weighted neighbor graph model based on the eigenvalue of the Laplace matrix, combine the graph density vector with the von Neumann entropy, and calculate the von Neumann graph entropy describing the health status of the rotating machinery; sort the von Neumann graph entropies of each undirected weighted neighbor graph model in sequence according to the graph model serial number to obtain a von Neumann graph entropy set;

[0046] The changing pattern of the von Neumann diagram entropy in the judgment set is determined. When the von Neumann diagram entropy increases abnormally, it is determined that a fault has occurred in the rotating machinery.

[0047] Further:

[0048] The rotating machinery early fault detection method based on von Neumann graph entropy proposed in this embodiment mainly includes three parts: the first part is dynamic modeling based on the neighbor graph, the second part is the extraction of von Neumann graph entropy indicators, and the third part is making decisions based on monitoring data to determine the health status of rolling bearings.

[0049] like Figure 1As shown, a horizontal or vertical acceleration sensor is set to collect the horizontal vibration or vertical vibration signal of the bearing. The discrete original vibration sequence processed by the data acquisition card (DAQ) is read for preprocessing to obtain the vibration signals of multiple time series segments after segmentation, and the short-time power spectrum corresponding to each time series segment vibration signal is obtained after Gabor transformation; based on each short-time power spectrum, a neighbor graph model is constructed, thereby obtaining multiple neighbor graph models, and these multiple neighbor graph models constitute a graph model sequence; the node connection weights of each neighbor graph model in the graph model sequence are calculated and constructed as a Laplace matrix;

[0050] The Laplace matrix of each undirected weighted neighbor graph model is calculated, and the Laplace matrix is ​​decomposed to obtain the eigenvector of the matrix. Based on the eigenvalues ​​in the eigenvector, the von Neumann graph entropy corresponding to each undirected weighted neighbor graph model is calculated; a set of von Neumann graph entropies corresponding to multiple neighbor graph models is obtained;

[0051] Based on the piecewise linear regression method, the threshold decision is made on the change of von Neumann graph entropy to determine whether the machine has a fault. If it exceeds the threshold, it is determined that there is a fault. If not, it indicates that there is no fault and continues to observe at the next moment.

[0052] 1. Construction of neighbor graph model:

[0053] The rotating machinery in this embodiment uses bearings. In the first part, it is necessary to first obtain the vibration signal of the bearing, pre-process the vibration signal, and obtain vibration signals corresponding to multiple time series segments. Gabor transform is performed on the vibration signals of multiple time series segments to obtain a series of short-time power spectra corresponding to multiple time series segments. A series of short-time power spectra are mapped respectively to obtain multiple undirected weighted neighbor graph models, and the undirected weighted neighbor graph models are represented by corresponding weight storage matrices.

[0054] Specifically include:

[0055] Step 1:

[0056] A horizontal or vertical acceleration sensor is installed on the bearing seat to collect the horizontal vibration or vertical vibration signal of the bearing. The signal is digitized by a data acquisition card (DAQ) and the continuous time series of the vibration signal is discretized into a sequence of sampling points, denoted as N = {x[1], x[2], ..., x[n], ..., x[L]}, where x[n] represents the signal amplitude at the nth sampling point; L represents the total length of the signal.

[0057] Step 2:

[0058] The original signal is segmented using a sliding window of length T and overlap degree 0 to monitor the health status of the period. The mth segment is recorded as N m ={x[mT+1],…,x[(m+1)T]}. Calculate the Gabor power spectrum of the original signal in each segment to obtain the energy distribution of the original signal in the segment. The calculation formula is:

[0059]

[0060] Where Q(k,m) is the output of the short-time power spectrum at time mΔt and frequency kΔfHz; g[] is a Gaussian function; Δt and Δf are the time interval and frequency interval, respectively; N is the number of sampling points of the signal; T is the length of the sliding window function, which is set according to the sampling conditions and data size; x[n] represents the signal amplitude at the nth sampling point; m is the time index; and k is the frequency index.

[0061] The above calculations yield a series of short-time power spectra {Q1, ..., Q m , …, Q M}, where Q m The power spectrum calculated for the mth segment corresponds to each Q m Contains T Q(k,m) components. Q(k,m) is Q m One of the components in , there are T in total.

[0062] Step 3:

[0063] Power spectrum Q m Mapping is performed to construct an undirected weighted neighbor graph model G m It consists of three components: nodes, connecting edges and weights:

[0064] Each power spectrum Q m The component Q(k,m) in is regarded as a node, with a total of T nodes;

[0065] Connect each node with its neighboring nodes in pairs, denoted as l i,j , represents the weighted edge between points v(i) and v(j), and its mathematical description can be expressed as:

[0066]

[0067] Where c represents the neighborhood of node i, and the value of c is set according to the frequency range; i,j represents the i,jth nodes of T nodes in the graph model constructed by the power spectrum Qm.

[0068] like Figure 2As shown, it is a schematic diagram of a neighbor graph model with the number of nodes N=12 and the number of neighbors c=3, where the points represent nodes and the lines between the points represent connecting edges.

[0069] Step 4: Use the N×N square matrix W to construct the neighbor graph model Gm. m To store:

[0070]

[0071] W m Storage matrix for weights.

[0072] Among them, l i,j The corresponding weight is denoted as d i,j , the calculation formula is:

[0073]

[0074] Among them, i,j represents the i,jth node of T nodes in the graph model constructed by the power spectrum Qm.

[0075] 2. Von Neumann graph entropy feature extraction:

[0076] In the second part of this embodiment, the Laplace matrix of the undirected weighted neighbor graph model is calculated, and based on the eigenvalues ​​of the Laplace matrix, a graph density vector associated with the corresponding undirected weighted neighbor graph model is constructed, and the graph density vector is combined with the von Neumann entropy to calculate a health index describing the health status of the rotating machinery, which is called the von Neumann graph entropy;

[0077] The von Neumann graph entropy of each undirected weighted neighbor graph model is sorted in sequence according to the graph model sequence number to obtain a von Neumann graph entropy set.

[0078] Further, including:

[0079] Step 5:

[0080] Compute the Laplacian matrix L of the graph model m :

[0081] L m =D m -W m (6)

[0082] Among them, D m is the degree matrix of the graph, which is calculated from the weight storage matrix. The calculation formula is:

[0083]

[0084] Therefore, the Laplace matrix L m It can be expressed as:

[0085]

[0086] like Figure 3 As shown in the figure, it is a schematic diagram of the graph modeling process. After the graph model is built, Figure 3 The diagram of the leftmost position in the figure is shown below. Then the weight storage matrix is ​​calculated to obtain Figure 3 The schematic diagram of the middle position in the figure is shown in Figure 1. Finally, the Laplace matrix is ​​calculated to obtain Figure 3 Schematic diagram of the rightmost position.

[0087] Step 6:

[0088] For the Laplace matrix L m Decompose and extract eigenvalues:

[0089] L=UΛU T (9)

[0090] In the formula, Λ=diag{λ1,λ2,...,λ i ,...,λ T}, eigenvalue λ i Arrange in ascending order. U={u1,u2,...,u T} is an orthogonal matrix composed of eigenvectors corresponding to T eigenvalues.

[0091] Step 7:

[0092] The Laplacian matrix L of the graph n It is a matrix that represents the structure of a graph. The smallest eigenvalue and the second smallest eigenvalue represent the connectivity of the graph structure, and the other eigenvalues ​​reflect the distance between nodes. Therefore, the 3rd to Nth eigenvalues ​​are selected as {λ1,λ2,...,λN} to construct a graph density vector ρ associated with the graph, which reflects the distribution and complexity of the structure in the graph. Its calculation formula is:

[0093]

[0094] Among them, {λ1,λ2,...,λN} is the eigenvector obtained by Laplace matrix decomposition in the mth graph model, and the vector contains T eigenvalues. ρ is the corresponding graph density vector, and ρ i is the i-th density in the graph density vector ρ.

[0095] Step 8:

[0096] Calculate the von Neumann graph entropy of the graph density vector ρ. When the local changes in the relevant frequencies cause the graph structure to change, the ρ at the relevant position in the graph density vector will be prominent, causing the distribution of the graph density vector to diverge and the von Neumann graph entropy to increase.

[0097] Specifically, the calculation formula of von Neumann graph entropy is:

[0098]

[0099] The above steps result in the von Neumann graph entropy set H(G)={H(G i )}, i=1,2,···,M.

[0100] 3. Analytical decision-making:

[0101] In the third part of this embodiment, the changing rules in the graph entropy set are determined, and when the von Neumann graph entropy increases abnormally, it is determined that a fault has occurred in the rotating machinery.

[0102] Further, it specifically includes:

[0103] Step 9:

[0104] The von Neumann graph entropy describes the health status of the bearing, so the occurrence of bearing failure can be monitored by the change in entropy value. Since the graph structure represents the difference between components, it has a certain robustness to noise interference, and the von Neumann graph entropy is not sensitive to the structural changes of noise interference. Therefore, in a healthy state, the entropy value does not change significantly, but when a fault occurs, the entropy value increases abnormally.

[0105] According to this change rule, the slope change of the linear regression model of the indicator sequence is used to monitor the fault state of the bearing. The sliding window w is set to fit the linear regression model of the entropy value in the window. The selection of w is related to the sequence length. When the current slope k changes significantly from the historical value, it indicates that the von Neumann graph entropy has an increasing trend, which means that the current state is faulty.

[0106] The formula is:

[0107]

[0108] Where ξ is the threshold value, which is set according to the operating conditions; k is the slope of the fitted linear regression model of the von Neumann diagram entropy value data in the sliding window; kt is the slope of the regression model of the von Neumann diagram entropy fitting in the t-th sliding window; kt-1 is the slope of the regression model of the von Neumann diagram entropy fitting in the t-1th sliding window; H0 is the no-fault symbol; H1 is the fault symbol.

[0109] The calculation formula for k is:

[0110]

[0111] Where w is the sliding window length.

[0112] See also Figure 4 , Figure 5 , Figure 6 and Figure 7 As shown in FIG. , it is a simulation schematic diagram of this embodiment. Figure 4 Comparison of time consumption for graph modeling. Figure 5 This is a schematic diagram of the test signal detection results. Figure 6 This is a schematic diagram of the test signal detection results with strong interference. Figure 7 This is a schematic diagram of the test signal detection results in an actual factory.

[0113] Figure 4 The triangle-shaped line in the middle is a schematic diagram showing the change of modeling time with the number of nodes in the neighbor graph modeling graph; Figure 4 The line without triangles in the figure is a schematic diagram showing the change of modeling time with the number of nodes in the complete graph. Figure 4 It can be seen that the time complexity of storing the complete graph model to the weight storage matrix is ​​O(N^ 2 ); the neighbor graph is O(N).

[0114] Depend on Figure 5 , Figure 6 and Figure 7 It can be seen that the method described in this embodiment can accurately detect faults in the test signals without strong interference, the test signals with strong interference, and the actual test signals of the pulp mill.

[0115] Embodiment 2

[0116] This embodiment discloses a rotating machinery early fault detection system based on von Neumann graph entropy.

[0117] The rotating machinery early fault detection system based on von Neumann graph entropy includes:

[0118] The vibration signal acquisition module is configured to: acquire the vibration signal of the rotating machinery, pre-process the vibration signal, and obtain vibration signals corresponding to a plurality of time-series segment segments;

[0119] The neighbor graph modeling module is configured to: perform Gabor transformation on the vibration signals of multiple time series segments to obtain a series of short-time power spectra, and respectively map the series of short-time power spectra to obtain multiple undirected weighted neighbor graph models;

[0120] The von Neumann graph entropy index calculation module is configured to: calculate the Laplace matrix of the undirected weighted neighbor graph model, construct a graph density vector associated with the corresponding undirected weighted neighbor graph model based on the eigenvalue of the Laplace matrix, combine the graph density vector with the von Neumann entropy, and calculate the von Neumann graph entropy describing the health status of the rotating machinery; sort the von Neumann graph entropy of each undirected weighted neighbor graph model in sequence according to the graph model sequence number to obtain a von Neumann graph entropy set;

[0121] The judgment decision module is configured to: judge the change law of the von Neumann graph entropy in the set, and when the von Neumann graph entropy increases abnormally, it is judged that the rotating machinery has a fault.

[0122] Embodiment 3

[0123] The purpose of this embodiment is to provide a computer-readable storage medium.

[0124] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for early fault detection of rotating machinery based on von Neumann graph entropy as described in Example 1 of the present disclosure.

[0125] Embodiment 4

[0126] The purpose of this embodiment is to provide an electronic device.

[0127] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for early fault detection of rotating machinery based on von Neumann graph entropy as described in Example 1 of the present disclosure are implemented.

[0128] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0129] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0130] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for detecting early faults of rotating machinery based on von Neumann graph entropy, characterized in that: The following steps are involved: Acquire a vibration signal of a rotating machine, pre-process the vibration signal, and obtain vibration signals corresponding to a plurality of time series segments; Gabor transform is performed on the vibration signals of multiple time series segments to obtain a series of short-time power spectra, and the series of short-time power spectra are mapped respectively to obtain multiple undirected weighted neighbor graph models; Calculate the Laplace matrix of the undirected weighted neighbor graph model, construct a graph density vector associated with the corresponding undirected weighted neighbor graph model based on the eigenvalue of the Laplace matrix, combine the graph density vector with the von Neumann entropy, and calculate the von Neumann graph entropy describing the health status of the rotating machinery; sort the von Neumann graph entropies of each undirected weighted neighbor graph model in sequence according to the graph model serial number to obtain a von Neumann graph entropy set; The changing pattern of the von Neumann diagram entropy in the judgment set is determined. When the von Neumann diagram entropy increases abnormally, it is determined that a fault has occurred in the rotating machinery.

2. The method for detecting early faults of rotating machinery based on von Neumann graph entropy according to claim 1, characterized in that: Preprocess the vibration signal, including: Discretize the vibration signal of the continuous time series to obtain a sequence consisting of sampling points; The sliding window is used to segment the sequence composed of sampling points to obtain vibration signals corresponding to multiple time series segments.

3. The method for detecting early faults of rotating machinery based on von Neumann graph entropy according to claim 1, characterized in that: Gabor transform is performed on the vibration signals of multiple time series segments to obtain a series of short-time power spectra, specifically: Where Q(k,m) is the output of the short-time power spectrum at time mΔt and frequency kΔfHz; g[] is a Gaussian function; Δt and Δf are the time interval and frequency interval, respectively; N is the number of sampling points of the signal; T is the length of the sliding window function; x[n] represents the signal amplitude at the nth sampling point; m is the time index; and k is the frequency index.

4. The method for detecting early faults of rotating machinery based on von Neumann graph entropy according to claim 3, characterized in that: A series of short-time power spectra are mapped separately to obtain multiple undirected weighted neighbor graph models, specifically: The power spectrum Q of the mth segment m The component Q(k,m) in is regarded as a node, with a total of N nodes; Each node is connected to the nodes in its respective neighborhood in pairs to obtain multiple undirected weighted neighbor graph models.

5. The method for detecting early faults of rotating machinery based on von Neumann entropy according to claim 4, characterized in that: Calculate the Laplacian matrix of the undirected weighted neighbor graph model, including: The undirected weighted neighbor graph model is represented by the corresponding weight storage matrix; Based on the weight storage matrix of the undirected weighted neighbor graph model, the degree matrix of the undirected weighted neighbor graph model is calculated; The difference between the degree matrix and the weight storage matrix is ​​calculated to obtain the Laplacian matrix of the undirected weighted neighbor graph model.

6. The method for detecting early faults of rotating machinery based on von Neumann graph entropy according to claim 1, characterized in that: Based on the eigenvalues ​​of the Laplace matrix, the graph density vector associated with the corresponding undirected weighted neighbor graph model is constructed, specifically: Decompose the Laplace matrix and extract the eigenvalues; Sort the eigenvalues ​​from small to large, select the eigenvalues ​​from the 3rd to the Tth as {λ1,λ2,...,λN}, and construct a graph density vector associated with the undirected weighted neighbor graph model: Among them, ρ i is the i-th density in the graph density vector ρ; The graph density vector is combined with the von Neumann entropy to calculate the von Neumann graph entropy describing the health status of the rotating machinery, which is: Where H(G) represents the von Neumann graph entropy set, H(G) = {H(G i )}, i=1,2,···,M.

7. The method for detecting early faults of rotating machinery based on von Neumann graph entropy according to claim 1, characterized in that: The changing law of the von Neumann graph entropy in the judgment set is determined. When the von Neumann graph entropy increases abnormally, it is determined that the rotating machinery has a fault, specifically: The slope change of the index sequence linear regression model is used to monitor the fault status of the bearing; Set the sliding window w to fit the linear regression model of the von Neumann graph entropy value within the window; When the current slope k changes significantly compared with the historical value, it indicates that the von Neumann graph entropy has an increasing trend, which means that the current state is faulty.

8. A rotating machinery early fault detection system based on von Neumann graph entropy, characterized in that: include: The vibration signal acquisition module is configured to: acquire the vibration signal of the rotating machinery, pre-process the vibration signal, and obtain vibration signals corresponding to a plurality of time-series segment segments; The neighbor graph modeling module is configured to: perform Gabor transformation on the vibration signals of multiple time series segments to obtain a series of short-time power spectra, and respectively map the series of short-time power spectra to obtain multiple undirected weighted neighbor graph models; The von Neumann graph entropy index calculation module is configured to: calculate the Laplace matrix of the undirected weighted neighbor graph model, construct a graph density vector associated with the corresponding undirected weighted neighbor graph model based on the eigenvalue of the Laplace matrix, combine the graph density vector with the von Neumann entropy, and calculate the von Neumann graph entropy describing the health status of the rotating machinery; sort the von Neumann graph entropy of each undirected weighted neighbor graph model in sequence according to the graph model sequence number to obtain a von Neumann graph entropy set; The judgment decision module is configured to: judge the change law of the von Neumann graph entropy in the set, and when the von Neumann graph entropy increases abnormally, it is judged that the rotating machinery has a fault.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for detecting early faults of rotating machinery based on von Neumann graph entropy as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for detecting early faults of rotating machinery based on von Neumann graph entropy as described in any one of claims 1 to 7 are implemented.

Images