A mechanical equipment fault diagnosis method and system using correlation analysis

Abstract

The present invention discloses a mechanical equipment fault diagnosis method and system using correlation analysis, which relates to the field of industrial automation and intelligent manufacturing technology, including: constructing a spatiotemporal correlation model, using graph neural network technology to capture the complex correlation of sensor data in time and space from the spatiotemporal model, and outputting a feature vector that integrates spatiotemporal information; using phase space reconstruction technology, mapping one-dimensional time series data to multidimensional phase space, calculating the maximum Lyapunov exponent, and obtaining nonlinear dynamic characteristics; combining the spatiotemporal feature vector and the nonlinear dynamic characteristics, using a deep learning model to predict the future state of mechanical equipment. Nonlinear dynamic characteristics are extracted by phase space reconstruction technology and calculating the maximum Lyapunov exponent. This comprehensive analysis method can identify the complex behavior of nonlinear systems, which is of great significance for early fault warning and predictive maintenance strategy formulation, thereby reducing unplanned downtime, extending equipment life, and reducing maintenance costs.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation and intelligent manufacturing, and in particular to a mechanical equipment fault diagnosis method and system using correlation analysis. Background Art

[0002] In recent years, with the rapid development of Industry 4.0 and Internet of Things technologies, intelligent monitoring and fault diagnosis of mechanical equipment have become an important research direction in the field of industrial automation and intelligent manufacturing. Although traditional mechanical equipment fault diagnosis methods can provide information on the health status of equipment to a certain extent, they lack the ability to comprehensively analyze multi-source heterogeneous data.

[0003] Although cross-modal data analysis technology has made significant progress in the field of mechanical equipment fault diagnosis, the existing technology still has several limitations. First, it fails to fully explore the deep correlation between modes, resulting in the omission of key fault information. Second, the traditional fusion method ignores the complex correlation of data in time and space, and it is difficult to capture the dynamic evolution of the equipment's operating status. In addition, the identification and application of nonlinear dynamic characteristics are not yet mature in the existing technology, and the prediction accuracy and stability of the existing technology still need to be improved. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a mechanical equipment fault diagnosis method and system using correlation analysis to solve the problem that the existing mechanical equipment fault diagnosis methods fail to effectively integrate cross-modal data and deeply explore the correlation and nonlinear dynamic characteristics between modalities, resulting in limited diagnostic accuracy.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a mechanical equipment fault diagnosis method using correlation analysis, which includes:

[0008] Step 1: Cross-modal data collection and preprocessing;

[0009] Step 2: Cross-modal feature extraction and correlation analysis;

[0010] Step 3: spatiotemporal correlation modeling and feature fusion;

[0011] Step 4: Nonlinear dynamic system identification and fault prediction;

[0012] The cross-modal data collection and preprocessing in step 1 includes installing vibration sensors at multiple key locations of the mechanical equipment to capture vibration signals of the equipment, temperature sensors to monitor temperature changes at key locations, acoustic sensors to collect sound signals from equipment operation, and electrical sensors to record changes in electrical parameters;

[0013] Perform signal denoising, missing value filling and data standardization on the collected multimodal data;

[0014] The cross-modal feature extraction and correlation analysis in step 2 refers to using a multimodal variational autoencoder to capture the correlation between modalities through a shared latent space;

[0015] Use correlation analysis techniques to evaluate the relationship between different modal features and identify the most informative feature combinations in fault diagnosis;

[0016] The spatiotemporal correlation modeling and feature fusion in step 3 refers to constructing a spatiotemporal correlation model, using graph neural network technology to capture the complex correlation of sensor data in time and space from the spatiotemporal model, and outputting a feature vector that integrates spatiotemporal information;

[0017] The nonlinear dynamic system identification and fault prediction in step 4 includes mapping the one-dimensional time series data to the multi-dimensional phase space through the phase space reconstruction technology, calculating the maximum Lyapunov exponent, and obtaining the nonlinear dynamic characteristics;

[0018] Combining the spatiotemporal feature vectors and nonlinear dynamic characteristics, a deep learning model is used to predict the future state of mechanical equipment.

[0019] A preferred solution of the mechanical equipment fault diagnosis method using correlation analysis of the present invention, wherein: the use of a multimodal variational autoencoder to capture the correlation between modes by sharing a latent space includes the following steps:

[0020] Design and build a multimodal variational autoencoder model, train it using the preprocessed multimodal dataset, and discover common features between modalities in the latent space;

[0021] For each modality of input data, the encoder of the multimodal variational autoencoder model will output a latent variable;

[0022] The latent variables of all modalities are collected to form a latent variable set.

[0023] As a preferred solution of the mechanical equipment fault diagnosis method using correlation analysis of the present invention, wherein: the use of correlation analysis technology to evaluate the relationship between different modal features and identify the most informative feature combination in fault diagnosis includes the following steps:

[0024] For each pair of features in the latent variable set, the Pearson correlation coefficient and mutual information between them are calculated;

[0025] The expression of Pearson correlation coefficient is:

[0026]

[0027] in, is the Pearson correlation coefficient, is the mean value of zm, is the mean of zn, N is the number of samples, and a is the sample index variable;

[0028] The expression of mutual information is:

[0029]

[0030] Among them, I(z m ; z n ) is the mutual information, p(z m ,z n ) is z m and z n The joint probability density function, p(z m ) is z m The marginal probability density function, p(z n ) is z n The marginal probability density function of ;

[0031] According to the calculation results of Pearson correlation coefficient and mutual information, the Pearson correlation coefficient and mutual information results between all modal latent variables are summarized into Pearson correlation coefficient matrix and mutual information matrix;

[0032] Analyze the Pearson correlation coefficient matrix and mutual information matrix to identify which latent variable combinations show high correlation and mutual information values;

[0033] Based on the results of the relevance evaluation, the most informative feature combination in fault diagnosis is selected.

[0034] As a preferred solution of the mechanical equipment fault diagnosis method using correlation analysis described in the present invention, the construction of the spatiotemporal correlation model, using the graph neural network technology, capturing the complex association of sensor data in time and space from the spatiotemporal model, and outputting the feature vector of the fused spatiotemporal information includes the following steps:

[0035] The sensor data points collected from the mechanical equipment are used as nodes in the graph that carry the characteristics of the latent variables;

[0036] Establish edge sets based on the relationship between sensor data points in time and space;

[0037] The weights of the edges are set by the similarity of the time series and the proximity of the spatial layout;

[0038] The latent variables carried by each node are used as the initial feature vector of each node;

[0039] Apply the GCN iterative update formula to update the feature vector of each node, and the output feature vector of each layer is used as the input of the next layer until the predetermined number of layers is reached;

[0040] In the last layer of GCN, global average pooling is used to aggregate the hidden state vectors of all nodes into a feature vector that integrates spatiotemporal information.

[0041] As a preferred solution of the mechanical equipment fault diagnosis method using correlation analysis described in the present invention, the mapping of one-dimensional time series data to multi-dimensional phase space by phase space reconstruction technology includes the following steps:

[0042] Based on time series data, the mutual information is calculated under a series of delay times and the optimal delay time is found when the mutual information begins to drop to the local minimum.

[0043] Initialize the embedding dimension from the one-dimensional time series data obtained from the phase space reconstruction;

[0044] Find the nearest neighbor of each phase space vector at the current embedding dimension;

[0045] For each point and its nearest neighbor, calculate the rate of change of distance after increasing the embedding dimension to determine the most appropriate embedding dimension;

[0046] Based on the optimal delay time and the most appropriate embedding dimension, the one-dimensional time series is converted into a vector in the multi-dimensional phase space.

[0047] As a preferred solution of the mechanical equipment fault diagnosis method using correlation analysis described in the present invention, wherein: the calculation of the maximum Lyapunov exponent to obtain the nonlinear dynamic characteristics includes the following steps:

[0048] In the vector of the reconstructed multidimensional phase space, a starting point is selected as a reference point, the Euclidean distance between the reference point and all other points in the phase space is calculated, and the point with the closest distance is found as the nearest neighbor point;

[0049] Record the information of the nearest neighbor point and the distance between it and the reference point and find the nearest neighbor point for each point in the phase space;

[0050] Track the evolution path of the reference point and its nearest neighbor over time. For each reference point and its nearest neighbor, calculate the rate of change of the distance between them over time.

[0051] The relationship between the distance change rate and time is fitted into an exponential function.

[0052] The exponential growth rates of all reference points are averaged to obtain the average exponential growth rate;

[0053] By calculating the maximum Lyapunov exponent in a limiting manner, the nonlinear dynamic characteristics are obtained, and the expression is:

[0054]

[0055] where λ is the maximum Lyapunov exponent, T is the time span, ΔX(0) and ΔX(T) are the phase space distances between the nearest neighbors at the initial time and after the time span T, respectively.

[0056] As a preferred solution of the mechanical equipment fault diagnosis method using correlation analysis described in the present invention, the future state of the mechanical equipment is predicted by combining the spatiotemporal feature vector and the nonlinear dynamic characteristics and using a deep learning model, and the expression is:

[0057]

[0058] in, is the predicted state of the mechanical equipment at time t, c is the index variable of the historical observation time point, C is the total number of historical observation points, To calculate the spatiotemporal feature vector F at the current time point * (t) and the past time point t i The space-time characteristic vector F * (t i ), γ is the parameter that controls the decay speed of nonlinear dynamic characteristics, t c is the past time point, t is the current time point, and s is the integral variable;

[0059] Integrate the mechanical equipment status prediction results into the real-time monitoring system to continuously monitor the operating status of the mechanical equipment;

[0060] When a failure is predicted, the system immediately generates an alert and provides maintenance recommendations based on the failure prediction results.

[0061] In a second aspect, the present invention provides a mechanical equipment fault diagnosis system using correlation analysis, including a cross-modal data acquisition and preprocessing module: vibration sensors are installed at multiple key parts of the mechanical equipment to capture the vibration signals of the equipment, temperature sensors monitor the temperature changes of key parts, acoustic sensors collect sound signals of equipment operation, electrical sensors record changes in electrical parameters, and the collected multi-modal data are subjected to signal denoising, missing value filling and data standardization processing;

[0062] Cross-modal feature extraction and correlation analysis module: uses multimodal variational autoencoders to capture the correlation between modalities through a shared latent space, uses correlation analysis techniques to evaluate the relationship between different modal features, and identifies the most informative feature combination in fault diagnosis;

[0063] Spatiotemporal correlation modeling and feature fusion module: Build a spatiotemporal correlation model, use graph neural network technology to capture the complex temporal and spatial correlation of sensor data from the spatiotemporal model, and output a feature vector that integrates spatiotemporal information;

[0064] Nonlinear dynamic system identification and fault prediction module: Through phase space reconstruction technology, one-dimensional time series data is mapped to multidimensional phase space, the maximum Lyapunov exponent is calculated, and nonlinear dynamic characteristics are obtained. The future state of mechanical equipment is predicted by combining the space-time feature vector and nonlinear dynamic characteristics, and using deep learning models.

[0065] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the mechanical equipment fault diagnosis method using correlation analysis as described in the first aspect of the present invention is implemented.

[0066] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the mechanical equipment fault diagnosis method using correlation analysis as described in the first aspect of the present invention is implemented.

[0067] The beneficial effects of the present invention are as follows: through comprehensive collection and high-quality preprocessing of cross-modal data, the reliability of subsequent analysis is ensured. Multimodal variational autoencoders and correlation analysis are used to deeply explore the correlation between modes, identify high-value fault features, and enhance diagnostic accuracy. Spatiotemporal correlation modeling and feature fusion, with the help of graph neural networks, integrate complex spatiotemporal information to improve fault mode understanding. Nonlinear dynamic system identification combined with deep learning prediction can achieve early warning and predictive maintenance, effectively reduce unplanned downtime, extend equipment life, reduce maintenance costs, and provide strong support for industrial safe operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0069] Figure 1 This is a flow chart of the mechanical equipment fault diagnosis method using correlation analysis in Example 1.

[0070] Figure 2 This is the maximum Lyapunov exponent determination diagram in Example 1. DETAILED DESCRIPTION

[0071] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0072] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0073] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0074] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a mechanical equipment fault diagnosis method using correlation analysis, comprising the following steps:

[0075] S1. Cross-modal data acquisition and preprocessing include installing vibration sensors at multiple key locations of mechanical equipment to capture vibration signals of the equipment, temperature sensors to monitor temperature changes at key locations, acoustic sensors to collect sound signals from equipment operation, and electrical sensors to record changes in electrical parameters;

[0076] Perform signal denoising, missing value filling and data standardization on the collected multimodal data;

[0077] S2. Using a multimodal variational autoencoder to capture the correlation between modalities through a shared latent space includes designing and building a multimodal variational autoencoder model, which contains multiple input ports, each port corresponding to a modal data input, such as vibration signals, temperature changes, acoustic signals, and electrical parameters, etc. The multimodal variational autoencoder model is trained using a preprocessed multimodal dataset. During the training process, the model will try to learn how to reconstruct the input data of various modalities in the latent space and discover the common features between modalities in the latent space; for each modal input data, the encoder of the multimodal variational autoencoder model will output a latent variable; the latent variables of all modalities are collected to form a latent variable set.

[0078] Furthermore, the encoding process of the multimodal variational autoencoder is as follows:

[0079]

[0080] Among them, q φ (z|x) is the posterior probability distribution of the latent variable z under the input data x given the parameter φ. x is the input multimodal data, z is the latent variable, which is located in the latent space of the model and usually has a lower dimension and can capture the key features and patterns of the input data. M is the number of modes, m is the modal index variable, and x is the latent variable. m is the input data of the mth mode, z m is the latent variable corresponding to the mth mode.

[0081] S3. Use correlation analysis techniques to evaluate the relationship between different modal features and identify the most informative feature combination in fault diagnosis, including, for each pair of features in the latent variable set, calculating the Pearson correlation coefficient and mutual information between them;

[0082] The expression of Pearson correlation coefficient is:

[0083]

[0084] in, is the Pearson correlation coefficient, is the mean value of zm, is the mean of zn, N is the number of samples, and a is the sample index variable;

[0085] The Pearson correlation coefficient measures the degree of linear correlation between two variables. Its value ranges from -1 to 1. The closer the value is to 1 or -1, the stronger the correlation. Mutual information is a non-negative value that measures the amount of information between two random variables. The larger its value, the stronger the dependence between the two variables.

[0086] The expression of mutual information is:

[0087]

[0088] Among them, I(z m ; z n ) is the mutual information, p(z m ,z n ) is z m and z n The joint probability density function, p(z m ) is z m The marginal probability density function, p(z n ) is z n The marginal probability density function of .

[0089] According to the calculation results of Pearson correlation coefficient and mutual information, the Pearson correlation coefficient and mutual information results between all modal latent variables are summarized into Pearson correlation coefficient matrix and mutual information matrix. Generally speaking, feature pairs with high correlation or large mutual information are more likely to carry key information about the health status of the equipment; analyze the Pearson correlation coefficient matrix and mutual information matrix to identify which latent variable combinations show higher correlation and mutual information values; based on the results of correlation evaluation, select the feature combination with the most information value in fault diagnosis.

[0090] S4. Construct a spatiotemporal correlation model, and use graph neural network technology to capture the complex temporal and spatial correlation of sensor data from the spatiotemporal model, and output a feature vector that integrates spatiotemporal information. This includes taking sensor data points collected from mechanical equipment as nodes in the graph that carry latent variable features; establishing edge sets based on the relationship between sensor data points in time and space; setting edge weights based on the similarity of time series and the proximity of spatial layout; taking the latent variables carried by each node as the initial feature vector of each node; applying the GCN iterative update formula to update the feature vector of each node, and using the output feature vector of each layer as the input of the next layer until the predetermined number of layers is reached; in the last layer of GCN, use global average pooling to aggregate the hidden state vectors of all nodes into a feature vector that integrates spatiotemporal information.

[0091] Furthermore, the expression of GCN iterative update is:

[0092]

[0093] Among them, f i (l+1) For node v i The updated feature vector at layer l+1, F(i) is v i Neighbor set, W (l) is the weight matrix of the first layer, σ is the activation function, w ijFor the edge (v i ,v j ), f j (l) For node v j In the feature vector of the lth layer, i is the node index variable, and j is any node in the space-time graph;

[0094] The expression of the fused feature vector is:

[0095]

[0096] Among them, F * is the fused feature vector, f (l) is the feature vector set of all nodes in the last layer, and N is the total number of nodes.

[0097] S5. Mapping one-dimensional time series data to multi-dimensional phase space through phase space reconstruction technology includes: based on time series data, calculating the mutual information under a series of delay times and finding the optimal delay time when the mutual information begins to drop to the local minimum; initializing the embedding dimension from the one-dimensional time series data obtained from the phase space reconstruction; finding the nearest neighbor point of each phase space vector under the current embedding dimension; for each point and its nearest neighbor point, calculating the rate of change of distance after increasing the embedding dimension to determine the most appropriate embedding dimension; based on the optimal delay time and the most appropriate embedding dimension, converting the one-dimensional time series into a vector in the multi-dimensional phase space.

[0098] S6. Calculate the maximum Lyapunov exponent to obtain the nonlinear dynamic characteristics, including selecting a starting point as a reference point in the vector of the reconstructed multidimensional phase space, calculating the Euclidean distance between the reference point and all other points in the phase space, and finding the point with the closest distance as the nearest neighbor point; record the information of the nearest neighbor point and the distance between it and the reference point and find the nearest neighbor point for each point in the phase space; track the evolution path of the reference point and its nearest neighbor point over time, and for each reference point and its nearest neighbor point, calculate the rate of change of the distance between the two as time goes by; fit the relationship between the rate of change of distance and time into the form of an exponential function, average the exponential growth rates of all reference points, and obtain the average exponential growth rate; calculate the maximum Lyapunov exponent by seeking the limit to obtain the nonlinear dynamic characteristics, which is expressed as follows:

[0099]

[0100] Where λ is the maximum Lyapunov exponent, which quantifies the degree of chaos in the system, T is the time span, which represents the length of time the system evolves, and ΔX(0) and ΔX(T) are the phase space distances between the nearest neighbors at the initial moment and after a time span T, respectively.

[0101] When λ>0, it indicates that the system is chaotic because the distance between the nearest neighbor points increases exponentially with time; when λ=0, it indicates that the system is stable and the distance between the nearest neighbor points remains unchanged; when λ<0, it indicates that the system is convergent and the distance between the nearest neighbor points decreases with time.

[0102] S7. Combining the spatiotemporal feature vector and nonlinear dynamic characteristics, using the deep learning model, predict the future state of the mechanical equipment. The expression is:

[0103]

[0104] in, is the predicted state of the mechanical equipment at time t, and its value range is usually [0,1]. A value close to 1 indicates a high risk of failure, and a value close to 0 indicates that the mechanical equipment is operating normally. c is the index variable of the historical observation time point, and C is the total number of historical observation points. To calculate the spatiotemporal feature vector F at the current time point * (t) and the past time point t i The space-time characteristic vector F * (t i ), γ is the parameter that controls the decay speed of nonlinear dynamic characteristics, t c is the past time point, t is the current time point, and s is the integral variable;

[0105] Integrate the mechanical equipment status prediction results into the real-time monitoring system to continuously monitor the operating status of the mechanical equipment; when a fault is predicted, the system immediately generates an alarm and provides maintenance recommendations based on the fault prediction results.

[0106] This embodiment also provides a mechanical equipment fault diagnosis system using correlation analysis, including a cross-modal data acquisition and preprocessing module: installing vibration sensors at multiple key locations of the mechanical equipment to capture vibration signals of the equipment, temperature sensors to monitor temperature changes at key locations, acoustic sensors to collect sound signals of equipment operation, electrical sensors to record changes in electrical parameters, and performing signal denoising, missing value filling and data standardization on the collected multi-modal data;

[0107] Cross-modal feature extraction and correlation analysis module: uses multimodal variational autoencoders to capture the correlation between modalities through a shared latent space, uses correlation analysis techniques to evaluate the relationship between different modal features, and identifies the most informative feature combination in fault diagnosis;

[0108] Spatiotemporal correlation modeling and feature fusion module: Build a spatiotemporal correlation model, use graph neural network technology to capture the complex temporal and spatial correlation of sensor data from the spatiotemporal model, and output a feature vector that integrates spatiotemporal information;

[0109] Nonlinear dynamic system identification and fault prediction module: Through phase space reconstruction technology, one-dimensional time series data is mapped to multidimensional phase space, the maximum Lyapunov exponent is calculated, and nonlinear dynamic characteristics are obtained. The future state of mechanical equipment is predicted by combining the space-time feature vector and nonlinear dynamic characteristics, and using deep learning models.

[0110] This embodiment also provides a computer device, which is suitable for the case of a mechanical equipment fault diagnosis method using correlation analysis, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the mechanical equipment fault diagnosis method using correlation analysis as proposed in the above embodiment.

[0111] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0112] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the mechanical equipment fault diagnosis method using correlation analysis as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination of them, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0113] In summary, the present invention ensures the reliability of subsequent analysis through comprehensive collection and high-quality preprocessing of cross-modal data. Multimodal variational autoencoders and correlation analysis are used to deeply explore the correlation between modes, identify high-value fault features, and enhance diagnostic accuracy. Spatiotemporal correlation modeling and feature fusion use graph neural networks to integrate complex spatiotemporal information and improve fault mode understanding. Nonlinear dynamic system identification combined with deep learning prediction can achieve early warning and predictive maintenance, effectively reduce unplanned downtime, extend equipment life, reduce maintenance costs, and provide strong support for industrial safe operation.

[0114] Example 2

[0115] Referring to Table 1, which is the second embodiment of the present invention, in order to further verify the advancement of the present invention, experimental simulation data of a mechanical equipment fault diagnosis method using correlation analysis is provided.

[0116] The key parts of the rotary compressor are equipped with a variety of sensors, including vibration sensors, temperature sensors, acoustic sensors, and electrical sensors, to monitor the equipment status in real time. The data collection frequency is set to 10 times per second, and data is collected continuously for two weeks. In order to ensure data quality, signal denoising, missing value filling, and data standardization are performed.

[0117] Next, a multimodal variational autoencoder model was designed and trained, which successfully captured the correlation between different modal data. The most informative feature combinations were identified by performing correlation analysis on the latent variables output by the model. Based on these features, a spatiotemporal correlation model was constructed, and the graph neural network technology was used to integrate sensor data from a spatiotemporal perspective.

[0118] In order to quantify the equipment status, the phase space reconstruction technology was used to calculate the maximum Lyapunov exponent and reveal the nonlinear dynamic characteristics of the equipment. Finally, the spatiotemporal feature vector and nonlinear dynamic characteristics were input into the deep learning model to predict the future state of the rotary compressor and achieve early warning of faults, as shown in Table 1:

[0119] Table 1 Experimental record table

[0120]

[0121] By analyzing the above data, it can be observed that as the equipment ages, the vibration amplitude, temperature, acoustic level and electrical parameters all show an upward trend. This is directly related to the degree of wear of the equipment. At the same time, the correlation score shows that the relationship between different modal data is getting closer and closer, especially near the time point when the failure prediction score increases significantly.

[0122] The increase in the Lyapunov exponent indicates that the equipment is gradually entering an unstable state, which is consistent with the theory of nonlinear dynamics. It is worth noting that the fault prediction score rises sharply in the later stage of the experiment, gradually increasing from 0.1 to 0.99, clearly indicating the risk of impending equipment failure. Compared with traditional monitoring methods, this result provides earlier warning, allowing maintenance teams to intervene in advance and avoid potential production interruptions and high maintenance costs.

[0123] In summary, the proposed fault prediction system effectively utilizes cross-modal data and significantly improves the accuracy and timeliness of fault prediction through deep learning and nonlinear dynamic analysis, demonstrating its innovation and practicality in industrial equipment health management.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A mechanical equipment fault diagnosis method using correlation analysis, characterized in that: include, Step 1: Cross-modal data collection and preprocessing; Step 2: Cross-modal feature extraction and correlation analysis; Step 3: spatiotemporal correlation modeling and feature fusion; Step 4: Nonlinear dynamic system identification and fault prediction; The cross-modal data collection and preprocessing in step 1 includes installing vibration sensors at multiple key locations of the mechanical equipment to capture vibration signals of the equipment, temperature sensors to monitor temperature changes at key locations, acoustic sensors to collect sound signals from equipment operation, and electrical sensors to record changes in electrical parameters; Perform signal denoising, missing value filling and data standardization on the collected multimodal data; The cross-modal feature extraction and correlation analysis in step 2 refers to using a multimodal variational autoencoder to capture the correlation between modalities through a shared latent space; Use correlation analysis techniques to evaluate the relationship between different modal features and identify the most informative feature combinations in fault diagnosis; The spatiotemporal correlation modeling and feature fusion in step 3 refers to constructing a spatiotemporal correlation model, using graph neural network technology to capture the complex correlation of sensor data in time and space from the spatiotemporal model, and outputting a feature vector that integrates spatiotemporal information; The nonlinear dynamic system identification and fault prediction in step 4 includes mapping the one-dimensional time series data to the multi-dimensional phase space through the phase space reconstruction technology, calculating the maximum Lyapunov exponent, and obtaining the nonlinear dynamic characteristics; Combining spatiotemporal feature vectors and nonlinear dynamic characteristics, using deep learning models to predict the future state of mechanical equipment; The future state of mechanical equipment is predicted by combining the spatiotemporal feature vector and nonlinear dynamic characteristics using a deep learning model. The expression is: in, is the predicted state of the mechanical equipment at time t, c is the index variable of the historical observation time point, C is the total number of historical observation points, To calculate the spatiotemporal feature vector F at the current time point * (t) and the past time point t i The space-time characteristic vector F * (t i ), γ is the parameter that controls the decay speed of nonlinear dynamic characteristics, t c is the past time point, t is the current time point, and s is the integral variable; Integrate the mechanical equipment status prediction results into the real-time monitoring system to continuously monitor the operating status of the mechanical equipment; When a failure is predicted, the system immediately generates an alert and provides maintenance recommendations based on the failure prediction results.

2. The mechanical equipment fault diagnosis method using correlation analysis according to claim 1, characterized in that: The method of using a multimodal variational autoencoder to capture the correlation between modalities by sharing a latent space includes the following steps: Design and build a multimodal variational autoencoder model, train it using the preprocessed multimodal dataset, and discover common features between modalities in the latent space; For each modality of input data, the encoder of the multimodal variational autoencoder model will output a latent variable; The latent variables of all modalities are collected to form a latent variable set.

3. The mechanical equipment fault diagnosis method using correlation analysis according to claim 1, characterized in that: The method of using correlation analysis technology to evaluate the relationship between different modal features and identify the most informative feature combination in fault diagnosis includes the following steps: For each pair of features in the latent variable set, the Pearson correlation coefficient and mutual information between them are calculated; The expression of Pearson correlation coefficient is: in, is the Pearson correlation coefficient, For z m The mean of For z n The mean of , N is the number of samples, and a is the sample index variable; The expression of mutual information is: Among them, I(z m ;z n ) is the mutual information, p(z m ,z n ) is z m and z n The joint probability density function, p(z m ) is z m The marginal probability density function, p(z n ) is z n The marginal probability density function of ; According to the calculation results of Pearson correlation coefficient and mutual information, the Pearson correlation coefficient and mutual information results between all modal latent variables are summarized into Pearson correlation coefficient matrix and mutual information matrix; Analyze the Pearson correlation coefficient matrix and mutual information matrix to identify which latent variable combinations show high correlation and mutual information values; Based on the results of the relevance evaluation, the most informative feature combination in fault diagnosis is selected.

4. The mechanical equipment fault diagnosis method using correlation analysis as claimed in claim 3, characterized in that: The construction of the spatiotemporal correlation model, using the graph neural network technology, capturing the complex temporal and spatial correlation of sensor data from the spatiotemporal model, and outputting a feature vector that integrates spatiotemporal information includes the following steps: The sensor data points collected from the mechanical equipment are used as nodes in the graph that carry the characteristics of the latent variables; Establish edge sets based on the relationship between sensor data points in time and space; The weights of the edges are set by the similarity of the time series and the proximity of the spatial layout; The latent variables carried by each node are used as the initial feature vector of each node; Apply the GCN iterative update formula to update the feature vector of each node, and the output feature vector of each layer is used as the input of the next layer until the predetermined number of layers is reached; In the last layer of GCN, global average pooling is used to aggregate the hidden state vectors of all nodes into a feature vector that integrates spatiotemporal information.

5. The mechanical equipment fault diagnosis method using correlation analysis as claimed in claim 4, characterized in that: The method of mapping one-dimensional time series data to multi-dimensional phase space by phase space reconstruction technology includes the following steps: Based on time series data, the mutual information is calculated under a series of delay times and the optimal delay time is found when the mutual information begins to drop to the local minimum. Initialize the embedding dimension from the one-dimensional time series data obtained from the phase space reconstruction; Find the nearest neighbor of each phase space vector at the current embedding dimension; For each point and its nearest neighbor, calculate the rate of change of distance after increasing the embedding dimension to determine the most appropriate embedding dimension; Based on the optimal delay time and the most appropriate embedding dimension, the one-dimensional time series is converted into a vector in the multi-dimensional phase space.

6. The mechanical equipment fault diagnosis method using correlation analysis as claimed in claim 5, characterized in that: The calculation of the maximum Lyapunov exponent to obtain the nonlinear dynamic characteristics comprises the following steps: In the vector of the reconstructed multidimensional phase space, a starting point is selected as a reference point, the Euclidean distance between the reference point and all other points in the phase space is calculated, and the point with the closest distance is found as the nearest neighbor point; Record the information of the nearest neighbor point and the distance between it and the reference point and find the nearest neighbor point for each point in the phase space; Track the evolution path of the reference point and its nearest neighbor over time. For each reference point and its nearest neighbor, calculate the rate of change of the distance between them over time. The relationship between the distance change rate and time is fitted into an exponential function. The exponential growth rates of all reference points are averaged to obtain the average exponential growth rate; By calculating the maximum Lyapunov exponent in a limiting manner, the nonlinear dynamic characteristics are obtained, and the expression is: where λ is the maximum Lyapunov exponent, T is the time span, ΔX(0) and ΔX(T) are the phase space distances between the nearest neighbors at the initial time and after the time span T, respectively.

7. A mechanical equipment fault diagnosis system using correlation analysis, based on the mechanical equipment fault diagnosis method using correlation analysis according to any one of claims 1 to 6, characterized in that: include, Cross-modal data acquisition and preprocessing module: Vibration sensors are installed at multiple key locations of mechanical equipment to capture vibration signals of the equipment, temperature sensors monitor temperature changes at key locations, acoustic sensors collect sound signals from equipment operation, electrical sensors record changes in electrical parameters, and the collected multi-modal data are subjected to signal denoising, missing value filling, and data standardization processing; Cross-modal feature extraction and correlation analysis module: uses multimodal variational autoencoders to capture the correlation between modalities through a shared latent space, uses correlation analysis techniques to evaluate the relationship between different modal features, and identifies the most informative feature combination in fault diagnosis; Spatiotemporal correlation modeling and feature fusion module: Build a spatiotemporal correlation model, use graph neural network technology to capture the complex temporal and spatial correlation of sensor data from the spatiotemporal model, and output a feature vector that integrates spatiotemporal information; Nonlinear dynamic system identification and fault prediction module: Through phase space reconstruction technology, one-dimensional time series data is mapped to multidimensional phase space, the maximum Lyapunov exponent is calculated, and nonlinear dynamic characteristics are obtained. The future state of mechanical equipment is predicted by combining the space-time feature vector and nonlinear dynamic characteristics, and using deep learning models.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the mechanical equipment fault diagnosis method using correlation analysis described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the mechanical equipment fault diagnosis method using correlation analysis described in any one of claims 1 to 6 are implemented.

Images