Mechanical equipment fault monitoring system and method based on multi-dimensional data driving

Abstract

This invention discloses a mechanical equipment fault monitoring system and method based on multidimensional data-driven approaches, belonging to the field of fault prediction technology. The method specifically includes: collecting multidimensional data of the mechanical equipment, preprocessing the collected multidimensional data, extracting features from the preprocessed multidimensional data, introducing a time-frequency locality adjustment mechanism to obtain first local weight coefficients, weighting the first local weight coefficients using a likelihood function to obtain second local weight coefficients, weighting and fusing the extracted features of the preprocessed multidimensional data, training a fault prediction model using common fault data of the mechanical equipment, and predicting faults of the mechanical equipment based on the trained fault prediction model and the features of the weighted and fused multidimensional data. This invention, through continuous training and optimization of the model, adapts to the operating characteristics of the equipment under different working conditions, automatically adjusts the fault prediction model, and possesses strong adaptive capabilities.

Description

Technical Field

[0001] This invention belongs to the field of fault prediction technology, specifically a mechanical equipment fault monitoring system and method based on multi-dimensional data-driven approaches. Background Technology

[0002] Fault monitoring and prediction for mechanical equipment has always been a core issue in the industrial sector, especially with the increasing development of smart manufacturing and the Industrial Internet of Things (IIoT). Equipment health management and fault prediction have become important tasks for improving production efficiency, reducing maintenance costs, and extending equipment life. Traditional fault detection methods typically rely on experience, periodic inspections, and single sensor data. These methods have certain limitations and cannot comprehensively reflect the health status of equipment. As industrial equipment develops towards intelligence and automation, fault monitoring methods based on multi-dimensional data are gradually becoming a new research hotspot.

[0003] Furthermore, traditional fault monitoring technologies often rely on expert experience and preset rules, such as threshold-based judgments. These methods are generally unable to adapt to complex and dynamic industrial environments. Moreover, traditional methods are mostly based on human experience or designed for specific operating conditions, lacking universality and automated intelligent capabilities.

[0004] While there are some fault prediction methods based on machine learning and deep learning in the existing technology, most of these methods are limited to using a single type of sensor data or rely on traditional feature extraction and selection methods. They lack a global perspective for multi-dimensional data integration and often ignore the nonlinear and time-varying characteristics that may exist during equipment operation. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a mechanical equipment fault monitoring system and method based on multi-dimensional data-driven approaches. By utilizing multi-dimensional sensor data fusion, machine learning algorithms, and multi-level feature extraction, the system can effectively improve fault prediction accuracy and adapt to various changes in the operating conditions of mechanical equipment.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Multidimensional data-driven methods for mechanical equipment fault monitoring include:

[0008] Collect multidimensional data of mechanical equipment, including vibration, temperature, pressure and current data, and preprocess the collected multidimensional data of mechanical equipment;

[0009] Features of the preprocessed multidimensional data of mechanical equipment are extracted, a time-frequency locality adjustment mechanism is introduced to obtain the first local weight coefficient, the first local weight coefficient is weighted using the likelihood function to obtain the second local weight coefficient, and the features of the extracted preprocessed multidimensional data of mechanical equipment are weighted and fused.

[0010] Train a fault prediction model using public data on mechanical equipment faults;

[0011] Based on the trained fault prediction model and combined with the features of the weighted fusion of multidimensional data of mechanical equipment, faults of mechanical equipment are predicted.

[0012] When new working environments or operating conditions occur, the fault prediction model is retrained or its coefficients are adjusted to ensure that the fault prediction model always maintains its real-time predictive capability.

[0013] Specifically, the features extracted from the preprocessed multidimensional data of the mechanical equipment include:

[0014] Let the preprocessed multidimensional dataset of mechanical equipment be X. i (t), where i = 1, 2, ..., m, i represents the sensor index, m represents the number of sensors, and t represents the time step;

[0015] Features of the preprocessed multidimensional data of mechanical equipment are extracted, including vibration data features, temperature and pressure data features, and current data features. Vibration data features include time domain features and frequency domain features. Time domain features include mean, standard deviation, skewness, and kurtosis. Frequency domain features include dominant frequency and spectral energy. Temperature and pressure data features include maximum value, minimum value, and average rate of change. Current data features include peak factor, root mean square, and harmonic variation.

[0016] Specifically, the fusion of features from the extracted preprocessed multidimensional data of mechanical equipment includes:

[0017] A time-frequency locality adjustment mechanism is introduced to assign a first local weight coefficient to the features of the preprocessed multidimensional data of mechanical equipment. The specific formula is as follows:

[0018]

[0019] Where, ω s,freq (f) represents the first local weighting coefficient of the s-th preprocessed multidimensional data feature of the mechanical equipment at frequency f, where T represents the time window length, F represents the frequency window length, and W represents the frequency window length. s (t,f) represents the wavelet transform coefficients of the s-th preprocessed multidimensional data feature of mechanical equipment at time t and frequency f;

[0020] The first local weight coefficients of the preprocessed multidimensional data features of mechanical equipment are incorporated into the likelihood function calculation. For each first local weight coefficient, the likelihood function is used to perform weighted adjustment to obtain the second local weight coefficients of the preprocessed multidimensional data features of mechanical equipment.

[0021] Based on the second local weight coefficients of the preprocessed multidimensional data features of the mechanical equipment, the features of the preprocessed multidimensional data of the mechanical equipment are fused. The specific formula is as follows:

[0022]

[0023] Among them, F fused (t,f) represents the fused multidimensional data feature values ​​of the mechanical equipment, ts represents the total number of multidimensional data features of the mechanical equipment, and ω s,freq,sr (f) represents the second local weighting coefficient of the s-th preprocessed multidimensional data feature of the mechanical equipment at frequency f, F s (t,f) represents the values ​​of the features of the s-th preprocessed multidimensional data of mechanical equipment at time t and frequency f.

[0024] Specifically, training a fault prediction model using public data on mechanical equipment faults includes:

[0025] We selected a machine learning algorithm to build a fault prediction model, using a labeled dataset D. train The fault prediction model is trained by setting the training objective as minimizing the loss function, the specific formula of which is:

[0026]

[0027] Where L represents the loss function that minimizes the training objective of the fault prediction model, Γ(·) represents the loss function, and xl represents the labeled dataset D. train The number of data in y a Represents the labeled dataset D train Labeling of data in China This indicates that the fault prediction model is applicable to D. train The prediction result for the a-th data point;

[0028] Train the fault prediction model until the loss function that minimizes the training objective of the model converges and remains constant, then stop training to obtain the trained fault prediction model.

[0029] Specifically, the step of predicting mechanical equipment faults based on the trained fault prediction model and the features of the weighted fused multidimensional data of the mechanical equipment includes:

[0030] The fused features F fused(t,f) is input into the trained fault prediction model to predict the probability of mechanical equipment failure. The specific formula is as follows:

[0031]

[0032] Wherein, P(y=c|F fused (t,f) represents the condition where, given feature F fused Given (t,f), the probability of type c fault occurring, where y represents the fault category. The model parameters represent the c-th type of fault, k represents the index of all possible fault categories, and Tgz represents the total number of fault categories.

[0033] Set the fault warning threshold Ψ according to actual needs, if P(y=c|F fused If (t,f))>Ψ, then an alarm is triggered; otherwise, no alarm is triggered.

[0034] Specifically, the preprocessing includes: denoising, interpolation filling, and normalization;

[0035] The noise reduction is used to remove noise from the vibration data of mechanical equipment.

[0036] The interpolation filling is used to fill in missing values ​​in the temperature and pressure data of mechanical equipment;

[0037] The standardization is used to standardize the multidimensional data of mechanical equipment.

[0038] A multi-dimensional data-driven mechanical equipment fault monitoring system is used to implement the multi-dimensional data-driven mechanical equipment fault monitoring method, including: a data acquisition module, a feature fusion module, a model training module, a fault prediction module, and an adjustment and optimization module.

[0039] The data acquisition module is used to collect multidimensional data of mechanical equipment, including vibration, temperature, pressure and current data, and to preprocess the collected multidimensional data of mechanical equipment.

[0040] The feature fusion module is used to extract features from the preprocessed multidimensional data of mechanical equipment, introduce a time-frequency locality adjustment mechanism to obtain a first local weight coefficient, use a likelihood function to weight the first local weight coefficient to obtain a second local weight coefficient, and perform weighted fusion on the extracted features of the preprocessed multidimensional data of mechanical equipment.

[0041] The model training module uses common data on mechanical equipment failures to train a fault prediction model.

[0042] The fault prediction module predicts faults of mechanical equipment based on the trained fault prediction model and the features of the weighted and fused multidimensional data of the mechanical equipment.

[0043] The adjustment and optimization module retrains or adjusts the coefficients of the fault prediction model based on changes in the working environment or operating conditions.

[0044] Specifically, the feature fusion module includes: a feature extraction unit and a feature fusion unit;

[0045] The feature extraction unit is used to extract features from the preprocessed multidimensional data of mechanical equipment;

[0046] The feature fusion unit is used to introduce a time-frequency locality adjustment mechanism to obtain a first local weight coefficient, and to use a likelihood function to weight the first local weight coefficient to obtain a second local weight coefficient, thereby performing weighted fusion of the features of the extracted preprocessed multidimensional mechanical equipment data.

[0047] Specifically, the fault prediction module includes: a fault prediction unit and an early warning unit;

[0048] The fault prediction unit predicts the faults of the mechanical equipment based on the features of the weighted and fused multidimensional data of the mechanical equipment and the trained fault prediction model.

[0049] The early warning unit sets a fault warning threshold according to actual needs to determine whether a fault in the mechanical equipment triggers an alarm.

[0050] Compared with the prior art, the beneficial effects of the present invention are:

[0051] 1. This invention proposes a mechanical equipment fault monitoring method and system based on multi-dimensional data-driven approach. It can automatically learn the normal operation mode and fault mode of equipment in a big data environment, without relying on human experience. Through continuous training and optimization of the model, it can adapt to the operating characteristics of the equipment under different working conditions, automatically adjust the fault prediction model, and has strong adaptive capabilities.

[0052] 2. This invention proposes a mechanical equipment fault monitoring method based on multi-dimensional data-driven analysis. It can monitor the health status of equipment in real time and identify potential faults in advance by analyzing historical and real-time data and using fault prediction models. This allows for timely prediction and warning of equipment faults before they occur, avoiding production downtime and significant losses caused by sudden faults. The accuracy of fault prediction is improved, and the failure rate and maintenance costs of equipment are reduced.

[0053] 3. This invention proposes a mechanical equipment fault monitoring method based on multi-dimensional data-driven approach. By combining data from multiple sensors, it can identify and classify various fault modes. Through hierarchical feature extraction and fusion, it can achieve multi-level and multi-angle fault mode identification, accurately identify various types of faults, and provide comprehensive protection for equipment health management. Attached Figure Description

[0054] Figure 1 A flowchart of the mechanical equipment fault monitoring method based on multidimensional data-driven approach provided by the present invention;

[0055] Figure 2 The diagram shows the architecture of a mechanical equipment fault monitoring system based on multidimensional data-driven principles provided by this invention. Detailed Implementation

[0056] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0058] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0059] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0060] Example 1

[0061] Please see Figure 1 The present invention provides an embodiment of a mechanical equipment fault monitoring method based on multidimensional data, comprising the following specific steps:

[0062] Step S1: Collect multi-dimensional data of mechanical equipment using sensors, including but not limited to: vibration, temperature, pressure and current data, and preprocess the collected multi-dimensional data of mechanical equipment;

[0063] The preprocessing includes: denoising, interpolation imputation, and standardization. The denoising is used to remove noise from the vibration data of the mechanical equipment (e.g., using wavelet transform for denoising). The interpolation imputation is used to fill in missing values ​​in the temperature and pressure data of the mechanical equipment. The standardization is used to standardize the multidimensional data of the mechanical equipment to ensure data comparison across different scales.

[0064] Step S2: Extract the features of the preprocessed multidimensional data of mechanical equipment, introduce a time-frequency locality adjustment mechanism to obtain the first local weight coefficient, use the likelihood function to weight the first local weight coefficient to obtain the second local weight coefficient, and perform weighted fusion on the extracted features of the preprocessed multidimensional data of mechanical equipment.

[0065] The specific steps of step S2 are as follows:

[0066] Step S201: Set the preprocessed multidimensional data set of mechanical equipment as X i (t), where i = 1, 2, ..., m, i represents the sensor index, m represents the number of sensors, and t represents the time step;

[0067] Step S202: Extract the features of the preprocessed multidimensional data of the mechanical equipment, including vibration data features, temperature and pressure data features, and current data features. Among them, vibration data features include time domain features and frequency domain features; time domain features include mean, standard deviation, skewness, and kurtosis; frequency domain features include dominant frequency and spectral energy; temperature and pressure data features include maximum value, minimum value, and average rate of change; current data features include peak factor, root mean square, and harmonic variation.

[0068] The specific formula for obtaining the mean feature in vibration data is as follows: Where μ zd The mean characteristic of the vibration data is represented by N, which represents the total number of sampling points, and x. zd (j) represents the j-th sampling time point of the vibration data. The mean feature is used to measure the average vibration level of the data. The specific formula for the standard deviation feature in the vibration data features is: Where σ zdThe standard deviation characteristic of vibration data describes the amplitude of data fluctuation. The specific formula for the skewness characteristic in vibration data is as follows: Ske zd The skewness characteristic of vibration data measures the degree of skewness in the data distribution, reflecting the asymmetry of the data. The specific formula for the kurtosis characteristic in vibration data is as follows: Among them Kur zd The kurtosis characteristic of vibration data measures the sharpness of a signal and is often used to identify potential sudden faults in mechanical equipment. The specific formula for the dominant frequency characteristic in vibration data is: f dom =argmax(|F(x) zd (t))|), where f dom The dominant frequency characteristic of the vibration data is represented by argmax(·), which represents the function to find the maximum value, and |·| represents the function to find the absolute value. F(x) zd (t) represents the Fourier transform of vibration data. The dominant frequency is the main frequency component during the operation of mechanical equipment, reflecting the rotational characteristics of the equipment. The specific formula for the spectral energy characteristic in vibration data features is: Where E freq f represents the spectral energy characteristics of vibration data. max The maximum frequency of the vibration data is represented by f, where f represents the frequency, and |F(x) zd (t))| 2 It represents the energy of vibration data in the frequency domain. The distribution of frequency domain signal energy can reveal the intensity of frequency components;

[0069] The specific formula for the average rate of change of temperature and pressure data is as follows: Here, Rate represents the average rate of change characteristic of temperature or pressure data, x(t) represents the temperature or pressure data at time t, x(t-1) represents the temperature or pressure data at time t-1, and Δt represents the unit time. The average rate of change characteristic describes the rate of change of temperature or pressure and is used to determine whether mechanical equipment is overloaded or operating abnormally.

[0070] The specific formula for the peak factor characteristic of current data is as follows: Where Peal represents the peak factor characteristic of the current data, x dl (t) represents the current data at time t, μ dl The mean of the current data is represented by the peak factor feature, which describes the ratio of the maximum value to the mean of the current data. The specific formula for the root mean square feature of the current data is as follows: Where RMS represents the root mean square characteristic of the current data, x dl(j) represents the j-th sampling time point of the current data. The root mean square feature reflects the overall energy of the current data. The harmonic changes in the current data are captured by short-time Fourier transform (STFT) and wavelet transform to capture the time-varying characteristics of the current data.

[0071] Step S203: Introduce a time-frequency locality adjustment mechanism to assign a first local weight coefficient to the features of the preprocessed multidimensional data of mechanical equipment. The specific formula is as follows:

[0072]

[0073] Where, ω s,freq (f) represents the first local weighting coefficient of the s-th preprocessed multidimensional data feature of the mechanical equipment at frequency f. This weighting coefficient reflects the importance of the s-th feature at frequency f and is used to measure the contribution of the signal to the fused features. T represents the time window length, F represents the frequency window length, and W... s (t,f) represents the wavelet transform coefficients of the s-th preprocessed multidimensional data feature of mechanical equipment at time t and frequency f;

[0074] The explanation and principle of the above formula: In equipment failure prediction, certain frequency bands may exhibit stronger failure characteristics (for example, bearing failures may be more significant within a specific frequency range). Therefore, the data characteristics of these frequency bands are given greater weight. The local weight of the data characteristics is determined based on the magnitude of the wavelet transform coefficients, reflecting the intensity of the data in that frequency band. Wavelet transform decomposes the original data in the time-frequency domain to reveal the local characteristics of the data. By scaling and translating the wavelet basis functions, detailed information of the data at different frequency bands and times is extracted. Especially when equipment fails, the features in the time-frequency domain can reveal the initial signs of the failure.

[0075] Step S204: Incorporate the first local weight coefficients of the preprocessed multidimensional data features of mechanical equipment into the likelihood function calculation. For each first local weight coefficient, use the likelihood function to perform weighted adjustment to obtain the second local weight coefficients of the preprocessed multidimensional data features of mechanical equipment.

[0076] In this embodiment, the likelihood function is used to describe the probability of observed data occurring under given model parameters. In the scenario of power equipment fault monitoring, different dimensions of monitoring data (such as temperature, vibration, voltage, current, etc.) have different importance in judging equipment faults. The local weights obtained by wavelet transform reflect the relative importance of each local data region in fault detection. Integrating these weights into the likelihood function makes the model pay more attention to key data regions during inference, thereby more accurately assessing the probability of equipment faults.

[0077] Step S205: Based on the second local weight coefficients of the preprocessed multidimensional data features of the mechanical equipment, the features of the preprocessed multidimensional data of the mechanical equipment are fused. The specific formula is as follows:

[0078]

[0079] Among them, F fused (t,f) represents the fused multidimensional data feature value of the mechanical equipment, which is a weighted sum of multiple data points over time t and frequency band f. By weighted fusing features from different sensors and frequency bands, a final, comprehensive equipment state feature is obtained for fault prediction. ts represents the total number of multidimensional data features of the mechanical equipment, ω s,freq,sr (f) represents the second local weighting coefficient of the s-th preprocessed multidimensional data feature of the mechanical equipment at frequency f, F s (t,f) represents the value of the feature of the s-th preprocessed multidimensional data of the mechanical equipment in time t and frequency f. It can be data from different sensors (such as vibration, temperature, current, etc.). The feature values ​​of the data in different frequency bands are obtained by frequency domain analysis (e.g., wavelet transform or Fourier transform). The data of each sensor may have different fault modes, so the extracted features may have different effects in different frequency bands.

[0080] The principle behind the above formula is as follows: The core idea of ​​this formula is to obtain a comprehensive feature vector that accurately reflects the equipment status by weighted fusion of features from different frequency bands and different sensors. The fusion process considers the contribution of each signal in each frequency band and adjusts the corresponding weight coefficients according to the magnitude of the contribution. In the fault prediction of mechanical equipment, data from different frequency bands may carry different types of fault information. By using methods such as Fourier transform and wavelet transform to convert the original time domain data to the frequency domain, the features of the data at different frequencies can be extracted. Different fault modes of equipment are often more obvious in specific frequency bands. For example, the low-frequency band of vibration data may reflect the overall operating status of the equipment, while the high-frequency band may be related to fault modes such as bearing wear.

[0081] By introducing local weighting coefficients, a dynamic weight can be assigned to the contribution of each data point in each frequency band. For example, in the case of bearing failure, the high-frequency components of vibration data may have a higher weight, while the contribution of temperature data in some frequency bands is smaller. By dynamically adjusting the weighted fusion based on the operating status of the equipment, the weighted fusion can not only respond to the different characteristics of different data in the frequency band, but also take into account the operating status and working conditions of the equipment.

[0082] By weighted fusion of different data in each frequency band, comprehensive features at each moment and in each frequency band can be obtained. It integrates information from different data sources (such as vibration, temperature, current, etc.) and different frequency bands. This fusion method can capture subtle changes in equipment and improve the accuracy of fault prediction.

[0083] Step S3: Train a fault prediction model using common data on mechanical equipment faults;

[0084] The public data on mechanical equipment failures is obtained from a public dataset;

[0085] The specific formula for step S3 is as follows:

[0086] Step S301: Select a machine learning algorithm to build a fault prediction model, using a labeled dataset D. train The fault prediction model is trained by setting the training objective as minimizing the loss function, the specific formula of which is:

[0087]

[0088] Where L represents the loss function that minimizes the training objective of the fault prediction model. Let xl represent the loss function and XL represent the labeled dataset D. train The number of data in y a Represents the labeled dataset D train The annotation of the data indicates whether a fault occurred at the corresponding time. This indicates that the fault prediction model is applicable to D. train The prediction result for the a-th data point;

[0089] loss function The specific formula is:

[0090] in, This represents the class weight of the a-th data point, used to handle class imbalance. For example, if a class has fewer samples, it can be given a higher weight.

[0091] Cross-entropy loss is a common method for measuring the difference between two probability distributions. In classification problems, it assesses the distance between the model's predicted distribution and the true distribution. Weights are introduced to balance the impact of class imbalance. In many practical applications, some classes may have significantly fewer samples than others, causing the model to favor the majority class. By giving higher weights to minority class samples, the model can prioritize these samples during training, thereby improving its predictive performance. The goal of minimizing this loss function is to adjust the model parameters through iterative optimization (such as gradient descent) to make the predicted results closer to the true labels. This approach not only improves the model's accuracy but also enhances its ability to identify minority class errors.

[0092] Step S302: Train the fault prediction model until the loss function that minimizes the training objective of the model converges and remains unchanged, then stop training and obtain the trained fault prediction model.

[0093] Step S4: Based on the trained fault prediction model and combined with the features of the weighted fused multidimensional data of the mechanical equipment, predict the faults of the mechanical equipment.

[0094] The specific steps of step S4 are as follows:

[0095] Step S401: Merge the features F fused (t,f) is input into the trained fault prediction model to predict the probability of mechanical equipment failure. The specific formula is as follows:

[0096]

[0097] Wherein, P(y=c|F fused (t,f) represents the condition where, given feature F fused Given (t,f), the probability of type c fault occurring, where y represents the fault category. The model parameters for the c-th type of fault are usually weight vectors obtained through training. They represent the relationship between features and fault categories and affect the model's prediction of that category. k represents the index of all possible fault categories, and Tgz represents the total number of fault categories.

[0098] In this embodiment, the principle of the formula is as follows: the formula uses a soft maximization function to linearly combine the model outputs (i.e., The function converts the predicted probabilities of all categories into a probability distribution, ensuring that the sum of the predicted probabilities of all categories is 1, so that the output can be directly interpreted as a probability.

[0099] Model parameters for each category In other words, the weights are learned through training data and reflect the importance of each feature in different fault categories. The larger the weight, the greater the influence of that feature on identifying that type of fault.

[0100] By calculating the probability of each category, the model can determine the input feature F. fused (t,f) represents the most likely fault type, and the category with the highest probability is selected as the final prediction result. This formula and its parameters are crucial in fault prediction. By analyzing real-time features and applying the corresponding model parameters, potential fault types can be effectively identified, and real-time monitoring and maintenance suggestions can be provided.

[0101] Step S402: Set the fault warning threshold Ψ according to actual needs. If P(y=c|F fused If (t,f))>Ψ, then an alarm is triggered; otherwise, no alarm is triggered.

[0102] Step S5: When new working environments or operating conditions change, retrain or adjust the coefficients of the fault prediction model to ensure that the fault prediction model always maintains the latest predictive capabilities.

[0103] Example 2

[0104] Please see Figure 2 Another embodiment of the present invention provides a mechanical equipment fault monitoring system based on multidimensional data, comprising: a data acquisition module, a feature fusion module, a model training module, a fault prediction module, and an adjustment and optimization module;

[0105] The data acquisition module is used to collect multidimensional data of mechanical equipment, including vibration, temperature, pressure and current data, and to preprocess the collected multidimensional data of mechanical equipment.

[0106] The feature fusion module is used to extract features from the preprocessed multidimensional data of mechanical equipment, introduce a time-frequency locality adjustment mechanism to obtain a first local weight coefficient, use a likelihood function to weight the first local weight coefficient to obtain a second local weight coefficient, and perform weighted fusion on the extracted features of the preprocessed multidimensional data of mechanical equipment.

[0107] The model training module uses common data on mechanical equipment failures to train a fault prediction model.

[0108] The fault prediction module predicts faults of mechanical equipment based on the trained fault prediction model and the features of the weighted and fused multidimensional data of the mechanical equipment.

[0109] The adjustment and optimization module is used to retrain or adjust the coefficients of the fault prediction model when new working environments or working conditions occur, so as to ensure that the fault prediction model always maintains the latest prediction capabilities.

[0110] The feature fusion module includes: a feature extraction unit and a feature fusion unit;

[0111] The feature extraction unit is used to extract features from the preprocessed multidimensional data of mechanical equipment;

[0112] The feature fusion unit is used to introduce a time-frequency locality adjustment mechanism to obtain a first local weight coefficient, and to use a likelihood function to weight the first local weight coefficient to obtain a second local weight coefficient, thereby performing weighted fusion of the features of the extracted preprocessed multidimensional mechanical equipment data.

[0113] The fault prediction module includes: a fault prediction unit and an early warning unit;

[0114] The fault prediction unit predicts the faults of the mechanical equipment based on the features of the weighted and fused multidimensional data of the mechanical equipment and the trained fault prediction model.

[0115] The early warning unit sets a fault warning threshold according to actual needs to determine whether a fault in the mechanical equipment triggers an alarm.

[0116] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0117] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring mechanical equipment failure based on multi-dimensional data driving, characterized in that, include: Collect multidimensional data of mechanical equipment, including vibration, temperature, pressure and current data, and preprocess the collected multidimensional data of mechanical equipment; Features of the preprocessed multidimensional data of mechanical equipment are extracted, a time-frequency locality adjustment mechanism is introduced to obtain the first local weight coefficient, the first local weight coefficient is weighted using the likelihood function to obtain the second local weight coefficient, and the features of the extracted preprocessed multidimensional data of mechanical equipment are weighted and fused. Train a fault prediction model using public data on mechanical equipment faults; Based on the trained fault prediction model and combined with the features of the weighted fusion of multidimensional data of mechanical equipment, faults of mechanical equipment are predicted. The weighted fusion of features from the extracted preprocessed multidimensional data of mechanical equipment includes: A time-frequency locality adjustment mechanism is introduced, which assigns the first local weight coefficient to the preprocessed multidimensional data features of mechanical equipment based on the wavelet transform coefficients, time window length, and frequency window length at time t and frequency f. The first local weight coefficients of the preprocessed multidimensional data features of mechanical equipment are incorporated into the likelihood function calculation. For each first local weight coefficient, the likelihood function is used to perform weighted adjustment to obtain the second local weight coefficients of the preprocessed multidimensional data features of mechanical equipment. Based on the second local weight coefficient of the preprocessed multidimensional data features of mechanical equipment, the features of the preprocessed multidimensional data of mechanical equipment are weighted and fused.

2. The mechanical equipment fault monitoring method based on multi-dimensional data-driven as described in claim 1, characterized in that, The features extracted from the preprocessed multidimensional data of the mechanical equipment include: Set the pre-processed mechanical equipment multi-dimensional data set as X i (t), wherein i = 1, 2, …, m, i represents the index of the sensor, m represents the number of sensors, and t represents the time step; Features of the preprocessed multidimensional data of mechanical equipment are extracted, including vibration data features, temperature and pressure data features, and current data features. Vibration data features include time domain features and frequency domain features. Time domain features include mean, standard deviation, skewness, and kurtosis. Frequency domain features include dominant frequency and spectral energy. Temperature and pressure data features include maximum value, minimum value, and average rate of change. Current data features include peak factor, root mean square, and harmonic variation.

3. The mechanical equipment fault monitoring method based on multi-dimensional data-driven as described in claim 2, characterized in that, The method of training a fault prediction model using public data on mechanical equipment faults includes: selecting a machine learning algorithm to build a failure prediction model, using a labeled dataset D train training the failure prediction model, setting a training objective of the model to minimize a loss function, and setting the loss function; Train the fault prediction model until the loss function that minimizes the training objective of the model converges and remains constant, then stop training to obtain the trained fault prediction model.

4. The mechanical equipment fault monitoring method based on multi-dimensional data-driven as described in claim 3, characterized in that, The step of predicting mechanical equipment faults based on a trained fault prediction model and the features of weighted fused multidimensional data of the mechanical equipment includes: The weighted and fused multidimensional data features of the mechanical equipment are input into the trained fault prediction model to predict the fault probability of the mechanical equipment and obtain the fault probability value of the mechanical equipment. Set fault warning thresholds according to actual needs. If the predicted failure probability of mechanical equipment is > If the condition is met, an alarm will be triggered; otherwise, no alarm will be triggered.

5. The mechanical equipment fault monitoring method based on multi-dimensional data-driven as described in claim 4, characterized in that, The preprocessing includes: denoising, interpolation filling, and normalization; The noise reduction is used to remove noise from the vibration data of mechanical equipment. The interpolation filling is used to fill in missing values ​​in the temperature and pressure data of mechanical equipment; The standardization is used to standardize the multidimensional data of mechanical equipment.

6. A mechanical equipment fault monitoring system based on multidimensional data-driven methods, used to implement the mechanical equipment fault monitoring method based on multidimensional data-driven methods according to any one of claims 1-5, characterized in that, include: The system includes a data acquisition module, a feature fusion module, a model training module, a fault prediction module, and an adjustment and optimization module. The data acquisition module is used to collect multidimensional data of mechanical equipment, including vibration, temperature, pressure and current data, and to preprocess the collected multidimensional data of mechanical equipment. The feature fusion module is used to extract features from the preprocessed multidimensional data of mechanical equipment, introduce a time-frequency locality adjustment mechanism to obtain a first local weight coefficient, use a likelihood function to weight the first local weight coefficient to obtain a second local weight coefficient, and perform weighted fusion on the extracted features of the preprocessed multidimensional data of mechanical equipment. The model training module uses common data on mechanical equipment failures to train a fault prediction model. The fault prediction module predicts faults of mechanical equipment based on the trained fault prediction model and the features of the weighted and fused multidimensional data of the mechanical equipment. The adjustment and optimization module retrains or adjusts the coefficients of the fault prediction model based on changes in the working environment or operating conditions.

7. The mechanical equipment fault monitoring system based on multi-dimensional data-driven as described in claim 6, characterized in that, The feature fusion module includes: a feature extraction unit and a feature fusion unit; The feature extraction unit is used to extract features from the preprocessed multidimensional data of mechanical equipment; The feature fusion unit is used to introduce a time-frequency locality adjustment mechanism to obtain a first local weight coefficient, and to use a likelihood function to weight the first local weight coefficient to obtain a second local weight coefficient, thereby performing weighted fusion of the features of the extracted preprocessed multidimensional mechanical equipment data.

8. The mechanical equipment fault monitoring system based on multi-dimensional data-driven as described in claim 7, characterized in that, The fault prediction module includes: a fault prediction unit and an early warning unit; The fault prediction unit predicts the faults of the mechanical equipment based on the features of the weighted and fused multidimensional data of the mechanical equipment and the trained fault prediction model. The early warning unit sets a fault warning threshold according to actual needs to determine whether a fault in the mechanical equipment triggers an alarm.

Images