Fault prediction and diagnosis method, system and equipment for crusher and medium

By constructing the triple correlation matrix of the crusher's working condition-fault characteristics-diagnosis model, the problem of difficulty in extracting fault characteristics under complex working conditions of the mine crusher is solved, and the accuracy of fault prediction and diagnosis and the effectiveness of equipment maintenance is improved. .

CN118965205APending Publication Date: 2024-11-15GUANGZHOU LEI MENG MASCH EQUIP CO LTD

Patent Information

Application Number
CN202411466741.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-21
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In the complex working environment of mine crushers, the characteristic data of equipment failure is flooded by environmental noise, making it difficult to extract the fault characteristics, and the fault evolution process is nonlinear and difficult to accurately describe. It is difficult to accurately extract the existing technology from massive state monitoring data. Fully reflect the characteristics of equipment failure conditions.

Method used

By obtaining the working status data of the equipment for preprocessing, extracting the working condition characteristic data and fault characteristic data, selecting the appropriate model for training, building a triple association matrix, quantifying the impact of working conditions and fault types on model output, and using triple groups The association matrix performs correlation analysis and optimizes equipment maintenance strategies.

Benefits of technology

It improves the accuracy of crusher fault prediction and diagnosis, can fully reflect equipment failure conditions, optimize equipment maintenance strategies, reduce high-risk working conditions and fault occurrence, and improve equipment reliability and safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965205A_ABST
    Figure CN118965205A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of crushers, in particular to a fault prediction and diagnosis method and system for a crusher, equipment and a medium, and the method comprises the steps: obtaining the working state data of the equipment, and carrying out the preprocessing of the working state data of the equipment, and obtaining working condition feature data; under the condition of acquiring the working state data of the equipment, acquiring equipment fault data, and performing feature extraction on the equipment fault data to obtain fault feature data; obtaining a to-be-trained model, and inputting the working condition feature data and the fault feature data into the to-be-trained model for training to obtain a diagnosis model; constructing a triple incidence matrix according to the working condition characteristic data, the fault characteristic data and the diagnosis model; correlation analysis is carried out according to the triple correlation matrix, an analysis result used for optimizing an equipment maintenance strategy is obtained, and the accuracy of fault prediction and diagnosis of the crusher is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of crushers, and in particular to a method, system, device and medium for predicting and diagnosing faults of crushers. Background Art

[0002] In mining production, crushers are widely used in the crushing and processing of ores. Crusher, also known as stone crusher, is an important part of the entire mining production line. The efficient operation of the crusher is crucial to ensure the ore crushing effect and production efficiency.

[0003] In the actual production process, the working conditions of the equipment are diverse, complex and changeable, resulting in large differences in the acquired equipment status data, poor data consistency and continuity, and large differences in structure, parameters, performance, etc. between equipment of different models and manufacturers, resulting in various manifestations of equipment failure. Due to the harsh working environment of the mining crusher, the factors affecting equipment failure are intricate, and the characteristic data related to the failure are usually submerged by environmental noise, making it difficult to accurately extract fault features from massive state monitoring data. Moreover, the evolution process of equipment failure is highly nonlinear, making it difficult to accurately characterize the mapping relationship between the extracted fault features and the fault state.

[0004] Therefore, when equipment operating conditions are diverse, how to extract features that can comprehensively reflect equipment failure conditions from complex status data is a major technical challenge faced by intelligent fault prediction and diagnosis methods, and the above problems need to be solved. Summary of the invention

[0005] In order to extract features that can fully reflect the equipment failure status and improve the accuracy of crusher failure prediction and diagnosis, the present application provides a crusher failure prediction and diagnosis method, system, device and medium, which adopts the following technical solutions: In a first aspect, the present application provides a method for predicting and diagnosing a crusher fault, comprising: Acquire equipment working status data, pre-process the equipment working status data, and obtain working condition characteristic data; When obtaining the equipment working status data, obtaining the equipment fault data, and performing feature extraction on the equipment fault data to obtain fault feature data; Obtain a model to be trained, input the operating condition characteristic data and the fault characteristic data into the model to be trained, and obtain a diagnostic model; Construct a triple association matrix based on the working condition characteristic data, fault characteristic data and the diagnosis model; Correlation analysis is performed based on the triple association matrix to obtain analysis results for optimizing equipment maintenance strategies.

[0006] Preferably, it also includes: Obtaining one set of operating condition characteristic data and fault characteristic data as verification data, evaluating the diagnosis result of the diagnosis model according to the verification data, and obtaining performance index evaluation information; The generalization ability is judged based on several performance indicator evaluation information. When the generalization ability is non-stationary, the diagnosis model is optimized based on the performance indicator evaluation information.

[0007] Preferably, it also includes: The operating condition characteristic data are divided and processed into operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics, and the operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics are input into the model to be trained for training.

[0008] Preferably, it also includes: The fault feature data is divided and processed into fault time domain features, fault frequency domain features and fault time-frequency domain features, and the fault time domain features, fault frequency domain features and fault time-frequency domain features are input into the model to be trained for training.

[0009] Preferably, the specific steps of obtaining one set of the operating condition characteristic data and the fault characteristic data as verification data and evaluating the diagnosis result of the diagnosis model according to the verification data include: The operating condition characteristic data is obtained, and the operating condition characteristic data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0010] Preferably, the specific step of obtaining one set of the operating condition characteristic data and the fault characteristic data as verification data and evaluating the diagnosis result of the diagnosis model according to the verification data further includes: Fault feature data is obtained, and the fault feature data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0011] Preferably, it also includes: The training data and the validation data are cross-validated, including, when one set of the validation data is validated, re-dividing the operating condition characteristic data or the fault characteristic data to obtain updated training data and validation data, training the training model based on the training data to obtain a diagnostic model, and evaluating the diagnostic results of the diagnostic model based on the validation data.

[0012] Preferably, it also includes: The time domain characteristics of the working condition divided by the working condition characteristic data are current and voltage signals, the frequency domain characteristics of the working condition divided by the working condition characteristic data are vibration signals, and the time and frequency domain characteristics of the working condition divided by the working condition characteristic data are load signals.

[0013] Preferably, it also includes: The fault time domain feature of the fault feature data division is a vibration signal, the fault frequency domain feature of the fault feature data division is a sound signal, and the fault time and frequency domain feature of the fault feature data division is a temperature signal.

[0014] In a second aspect, the present application provides a crusher fault prediction and diagnosis system, comprising: The first acquisition module is used to acquire the equipment working status data, pre-process the equipment working status data, and obtain working condition characteristic data; The second acquisition module is used to acquire device fault data while acquiring device working status data, and perform feature extraction on the device fault data to obtain fault feature data; The third acquisition module is used to acquire the model to be trained, input the operating condition characteristic data and the fault characteristic data into the model to be trained, and obtain the diagnosis model; An association module, used for constructing a triple association matrix according to the working condition characteristic data, the fault characteristic data and the diagnosis model; The association analysis module is used to perform association analysis based on the triple association matrix to obtain analysis results for optimizing equipment maintenance strategies.

[0015] Preferably, it also includes: A model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data, evaluate the diagnosis result of the diagnosis model according to the verification data, and obtain performance index evaluation information; The generalization ability is judged based on several performance indicator evaluation information. When the generalization ability is non-stationary, the diagnosis model is optimized based on the performance indicator evaluation information.

[0016] Preferably, it also includes: The third acquisition module is also used to divide and process the operating condition characteristic data into operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics, and input the operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics into the model to be trained for training.

[0017] Preferably, it also includes: The third acquisition module is also used to divide the fault feature data into fault time domain features, fault frequency domain features and fault time-frequency domain features, and input the fault time domain features, fault frequency domain features and fault time-frequency domain features into the model to be trained for training.

[0018] Preferably, the model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data, and the specific steps of evaluating the diagnostic result of the diagnostic model according to the verification data include: The operating condition characteristic data is obtained, and the operating condition characteristic data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0019] Preferably, the model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data, and the specific step of evaluating the diagnosis result of the diagnosis model according to the verification data also includes: Fault feature data is obtained, and the fault feature data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0020] Preferably, it also includes: The third acquisition module is also used to cross-validate the training data and the verification data, including, when one set of verification data is verified, re-dividing the operating condition characteristic data or the fault characteristic data to obtain updated training data and verification data, training the training model according to the training data to obtain a diagnostic model, and evaluating the diagnostic results of the diagnostic model according to the verification data.

[0021] Preferably, it also includes: The working condition time domain characteristics of the working condition characteristic data divided by the third acquisition module are current and voltage signals, the working condition frequency domain characteristics of the working condition characteristic data are vibration signals, and the working condition time and frequency domain characteristics of the working condition characteristic data are load signals.

[0022] Preferably, it also includes: The fault time domain feature of the fault feature data divided by the third acquisition module is a vibration signal, the fault frequency domain feature of the fault feature data divided by the fault feature data is a sound signal, and the fault time and frequency domain feature of the fault feature data divided by the fault feature data is a temperature signal.

[0023] In a third aspect, the present application provides a crusher fault prediction and diagnosis device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the crusher fault prediction and diagnosis method as described above.

[0024] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the above-mentioned crusher fault prediction and diagnosis method when running.

[0025] In summary, compared with the prior art, the technical solution provided by this application has at least the following beneficial effects: The present application obtains equipment working status data and equipment fault data, extracts features from them respectively, pre-processes the equipment working status data to obtain working condition feature data, extracts features from the equipment fault data to obtain fault feature data, selects a suitable model type as the model to be trained, inputs the working condition feature data and the fault feature data into the model to be trained for training to obtain a diagnosis model, and then constructs a triple association matrix based on the working condition feature data, the fault feature data and the diagnosis model, calculates the impact of different working conditions and fault types on the model output, quantifies the strength of the association between the three, uses the triple association matrix to perform association analysis, and obtains analysis results for optimizing equipment maintenance strategies, thereby extracting features that can comprehensively reflect the equipment fault status and improving the accuracy of crusher fault prediction and diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of a method for predicting and diagnosing a crusher fault described in an embodiment of the present application.

[0027] Figure 2 It is a module schematic diagram of a crusher fault prediction and diagnosis system described in an embodiment of the present application.

[0028] Description of reference numerals: 1. First acquisition module; 2. Second acquisition module; 3. Third acquisition module; 4. Association module; 5. Association analysis module. DETAILED DESCRIPTION

[0029] The following combination Figure 1-Figure 2 The present application is described in further detail. The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to be limiting.

[0030] In the actual production process of mining crushers, due to the diverse, complex and changeable working conditions of the equipment, the acquired equipment status data is very different, and the consistency and continuity of the data are poor. In addition, there are large differences in the structure, parameters, performance, etc. of equipment of different models and manufacturers, which makes the manifestations of equipment failures varied. This requires that when conducting intelligent fault prediction and diagnosis, the constructed diagnostic model must have strong generalization capabilities and be able to adapt to the personalized characteristics of different equipment. However, improving the generalization ability of the diagnostic model often reduces the model's ability to characterize the personalized characteristics of the equipment, and cannot accurately reflect the fault status of the equipment, resulting in a decrease in diagnostic accuracy.

[0031] Due to the harsh working environment of mining crushers, the factors affecting equipment failure are complicated, and the characteristic data related to the fault is usually submerged by environmental noise, making it difficult to accurately extract fault features from massive state monitoring data. Moreover, the evolution process of equipment failure is highly nonlinear, making it difficult to accurately characterize the mapping relationship between the extracted fault features and the fault state. Therefore, in the case of diverse equipment working conditions, how to extract features that can fully reflect the equipment failure status from complex state data is a major technical challenge faced by intelligent fault prediction and diagnosis methods.

[0032] There is a positive correlation between the contradiction between fault feature extraction and the diversity of equipment operating conditions and the contradiction between the generalization ability of the diagnostic model and personalized needs. If the diagnostic model can adapt well to the personalized characteristics and working conditions of different equipment, then even in the case of diverse equipment working conditions, it can extract features that fully reflect the fault condition, thereby improving the accuracy of diagnosis. On the contrary, if the generalization ability of the diagnostic model is insufficient and cannot meet personalized needs, then in the case of diverse equipment working conditions, the extracted fault features may not accurately reflect the fault condition, affecting the diagnostic effect. The present application provides a method for constructing an equipment working condition-fault feature-diagnostic model triple association matrix based on attributes such as equipment working conditions, fault features, diagnostic models, and generalization capabilities, quantifying the strength of association between the diversity of equipment working conditions, the difference in fault features, and the generalization ability of diagnostic models, to improve the accuracy of crusher fault prediction and diagnosis.

[0033] Reference Figure 1 The present application relates to a method for predicting and diagnosing a crusher fault, which specifically includes: Step S1: Acquire equipment working status data, pre-process the equipment working status data, and obtain working condition characteristic data; Step S2: when obtaining the equipment working status data, obtaining the equipment fault data, and extracting the features of the equipment fault data to obtain the fault feature data; Step S3: obtaining a model to be trained, inputting the operating condition characteristic data and the fault characteristic data into the model to be trained, and obtaining a diagnostic model; Step S4: constructing a triple association matrix according to the operating condition characteristic data, the fault characteristic data and the diagnosis model; Step S5: performing association analysis based on the triple association matrix to obtain analysis results for optimizing equipment maintenance strategies.

[0034] Specifically, the method of the present application is configured in the fault prediction and diagnosis equipment of the crusher. The equipment working status data, equipment fault data and other data acquired by the system of the present application are acquired by data acquisition device modules such as sensors and data acquisition systems. The working status data, equipment fault data and other data can be stored in the storage module of the crusher after acquisition and retrieved when fault prediction and diagnosis are needed.

[0035] This application obtains equipment working status data and equipment failure data and extracts features from them respectively. The equipment working status data is preprocessed to obtain working condition feature data. Through these features, the actual working environment of the equipment can be described more accurately. The equipment failure data is feature extracted to obtain failure feature data, which helps to understand the dynamic process of failure development.

[0036] Select the appropriate model type as the model to be trained, input the working condition characteristic data and fault characteristic data into the model to be trained, and obtain the diagnostic model to achieve accurate classification of fault type and degree and prediction of remaining life. Then, construct a triple association matrix based on the working condition characteristic data, fault characteristic data and diagnostic model, calculate the impact of different working conditions and fault types on the model output, quantify the strength of the association between the three, and use the triple association matrix for association analysis. Through association analysis, it can be found which working conditions and fault types have the greatest impact on the diagnostic results, and obtain analysis results for optimizing equipment maintenance strategies. By adjusting operating parameters or improving maintenance plans, the occurrence of high-risk working conditions and faults can be reduced, the reliability and safety of equipment can be improved, and features that can fully reflect the equipment fault status can be extracted to improve the accuracy of crusher fault prediction and diagnosis.

[0037] As one implementation mode, it also includes: The operating condition characteristic data are divided and processed into operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics, and the operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics are input into the model to be trained for training.

[0038] Specifically, the present application obtains equipment working status data, including environmental parameters, working condition type and duration, etc. Sensors and data acquisition systems are used to record the working status of the equipment under different operating conditions, analyze the collected working condition data, and extract key features by using data preprocessing techniques such as denoising and normalization to obtain data such as temperature peak value and pressure change rate.

[0039] In the case of crusher data analysis, the working condition characteristic data extracted from the equipment working status data is the equipment working condition data, which is usually the raw data collected by sensors or monitoring systems. These data include various environmental parameters, such as temperature, humidity, pressure, etc., working condition types, such as equipment operating status, load conditions, etc., and equipment fault information, such as fault type, frequency of occurrence, impact, etc.

[0040] Among them, the time domain characteristics of the working condition refer to the data characteristics in the time domain, which are obtained by statistics and analysis of the equipment working condition data. For example, the mean, variance, maximum value, minimum value and other statistics of the working condition data are calculated. These characteristics reflect the typical behavior and fluctuation of the equipment under different working conditions. Among them, in the time domain, the data is represented in the form of a time series. Common time domain characteristics include: mean and standard deviation: describe the stability of the trend and distribution in the data set. Maximum and minimum values: reflect the extreme conditions of the data. Waveform peak: describes the peak characteristics in the data. Time series statistics: such as skewness, kurtosis, etc., are used to describe the shape of the data distribution.

[0041] The frequency domain characteristics of the operating conditions are obtained by converting the time domain data into the frequency domain, usually through Fourier transform or other frequency domain analysis methods. Frequency domain characteristics can tell us the frequency response of the equipment in response to external excitation or at different frequencies. For example, for vibration signals, the spectrum density or main frequency components of the vibration signal are obtained through frequency domain analysis. Frequency domain analysis converts data into the frequency domain. Common frequency domain features include: Spectrum: shows the energy distribution of the signal at different frequencies. Frequency components: such as main frequency, frequency bandwidth, etc. Spectral peak: indicates a significant peak in the frequency domain, which may correspond to fault frequency or other key features.

[0042] The time-frequency domain characteristics of the working condition combine the information of the time domain and the frequency domain, and are usually obtained through time-frequency analysis methods such as wavelet transform. These characteristics can simultaneously reflect the changing characteristics of the signal in time and frequency, and are usually aimed at complex signal analysis and fault diagnosis.

[0043] The operating condition characteristic data provides the basic raw data of the signal, including the status and environmental parameters of the equipment under different operating conditions. These data are the basis for obtaining time domain, frequency domain or time-frequency domain features. The extraction of these features can better understand the operating status of the equipment, predict possible failures and optimize equipment maintenance plans. Time-frequency domain analysis combines the advantages of time domain and frequency domain. The conversion methods include wavelet transform and time-frequency analysis technology, which can capture the time-varying characteristics of the signal, such as transient response, frequency change, etc.

[0044] Dividing the crusher's operating data and fault data into time domain, frequency domain or time-frequency domain features has the following functions and effects: First, time domain features, such as mean, standard deviation, peak value, etc., can directly reflect the time-varying characteristics of the signal, and are suitable for describing the changing law of data on the time axis, and capturing the overall changing trend of operating data such as current and voltage. Frequency domain features, such as main frequency components, spectrum energy distribution, etc., can convert signals to the frequency domain through methods such as Fourier transform, revealing the signal energy distribution and frequency components at different frequencies, which helps to detect specific frequency components in vibration signals, thereby analyzing vibration conditions and mechanical failures. Time-frequency domain features, such as the time-frequency spectrum obtained by wavelet transform, can capture the changing relationship between signals in time and frequency, and analyze the time-varying characteristics of load signals or temperature signals, especially when involving large frequency changes or non-stationary signals. This improves the diversity and comprehensiveness of features.

[0045] Then, different feature extraction methods can reveal the characteristics of the data from different angles and levels. Time domain features mainly focus on the statistical characteristics of the data, frequency domain features focus on the frequency distribution and energy characteristics of the data, and time-frequency domain features can simultaneously consider the time changes and frequency components of the data, providing a more comprehensive and detailed data description. Multi-angle feature extraction can increase the model's understanding and perception of data, help to more accurately distinguish different working conditions or fault states, thereby improving the accuracy of fault diagnosis and prediction.

[0046] Secondly, different feature extraction methods are suitable for different types of data and application scenarios. For example, time domain features are very effective for static signals or slowly changing signals, while frequency domain features are suitable for capturing rapidly changing or periodic signal components. Time-frequency domain features perform well when both time and frequency changes need to be considered simultaneously. This flexibility allows the most appropriate feature extraction method to be selected based on specific data characteristics and application requirements, thereby optimizing the performance and predictive ability of the model.

[0047] In summary, dividing the crusher's operating data and fault data into time domain, frequency domain or time-frequency domain features can not only improve the efficiency and accuracy of data analysis, but also provide a more representative and informative feature set for subsequent machine learning or statistical modeling, thereby better supporting the application of equipment health monitoring, predictive maintenance and performance optimization.

[0048] As one implementation mode, it also includes: The fault feature data is divided and processed into fault time domain features, fault frequency domain features and fault time-frequency domain features, and the fault time domain features, fault frequency domain features and fault time-frequency domain features are input into the model to be trained for training.

[0049] Specifically, the equipment failure data obtained in this application provides specific information about failures that occur during equipment operation, including failure type, frequency, etc. This information can be used as an important basis or condition for feature selection during feature extraction. Detailed information related to the failure is obtained through historical maintenance records and real-time monitoring. Feature extraction is performed on the failure data, and a quantitative description of the failure characteristics is obtained through statistical analysis methods such as frequency analysis and trend prediction. The data helps to understand the dynamic process of failure development.

[0050] Among them, fault time domain characteristics refer to data characteristics in the time domain, which are obtained by statistics and analysis of equipment operating data. For example, the mean, variance, maximum value, minimum value and other statistics of the operating data are calculated. These characteristics reflect the typical behavior and fluctuation of the equipment under different working conditions. Among them, in the time domain, the data is represented in the form of time series. Common time domain characteristics include: mean and standard deviation: describe the stability of trends and distributions in the data set. Maximum and minimum values: reflect the extreme conditions of the data. Waveform peak: describes the peak characteristics in the data. Time series statistics: such as skewness, kurtosis, etc., are used to describe the shape of the data distribution.

[0051] The frequency domain features of a fault are obtained by converting the time domain data into the frequency domain, usually through Fourier transform or other frequency domain analysis methods. Frequency domain features can tell us the frequency response of the device in response to external excitation or at different frequencies. For example, for a vibration signal, the spectral density or main frequency components of the vibration signal are obtained through frequency domain analysis. Frequency domain analysis converts the data into the frequency domain. Common frequency domain features include: Spectrum: shows the energy distribution of the signal at different frequencies. Frequency components: such as main frequency, frequency bandwidth, etc. Spectral peak: indicates a significant peak in the frequency domain, which may correspond to the fault frequency or other key features.

[0052] Fault time-frequency domain features combine information in the time domain and frequency domain, and are usually obtained through time-frequency analysis methods such as wavelet transform. These features can simultaneously reflect the changing characteristics of the signal in time and frequency, and are usually used for complex signal analysis and fault diagnosis.

[0053] After extracting the features, the model is trained, such as a machine learning model or a statistical model, to identify and classify different failure modes or predict the life of the equipment. For example, the vibration sensor data of the crusher is obtained. First, the time domain features such as mean, standard deviation and peak are extracted from the original time series data. Then, the data is converted to the frequency domain through Fourier transform, and the frequency domain features such as the spectrum and the main frequency components are extracted. Finally, the wavelet transform is applied to analyze the time and frequency domain features to obtain more detailed time and frequency change information. These features are used as input features for subsequent model training to identify potential failure modes or perform predictive maintenance.

[0054] Processing the original working condition data and fault data into time domain, frequency domain or time-frequency domain features can reduce the dimension and enhance the feature expression ability. Since the original working condition data and fault data may be very large and contain a lot of details and noise, by extracting features, the data can be reduced from a high-dimensional space to a more compact feature space. This not only helps to reduce the computational complexity of the model, but also improves the model's processing efficiency and generalization ability for data.

[0055] It can better capture the essential characteristics of data. The time domain, frequency domain and time-frequency domain characteristics can more effectively express the periodicity, volatility and dynamic changes of data. For example, frequency domain analysis can reveal the main frequency components and spectrum characteristics of data, which helps to identify periodic faults or anomalies. Time domain characteristics can reflect the time series properties of data, including mean, variance, change trend, etc.

[0056] It can achieve better model interpretability and explainability. The extracted features are usually easier to interpret and understand, which helps to understand the physical or engineering meaning behind the model's prediction results. This interpretability is very important for fault diagnosis and maintenance decisions, and can guide specific maintenance actions and strategy formulation.

[0057] It can achieve better generalization and performance. Feature extraction helps the model learn key patterns and structures in the data, thereby improving the model's generalization ability on new data. That is, the model can effectively predict and identify faults in unseen situations, ensuring stable operation of the equipment under various working conditions.

[0058] In summary, converting the original operating condition data and fault data into appropriate feature representations enables the model to more effectively learn and understand the essential characteristics of the data, thereby improving the accuracy and reliability of fault prediction and diagnosis.

[0059] As one implementation mode, it also includes: The time domain characteristics of the working condition divided by the working condition characteristic data are current and voltage signals, the frequency domain characteristics of the working condition divided by the working condition characteristic data are vibration signals, and the time and frequency domain characteristics of the working condition divided by the working condition characteristic data are load signals.

[0060] Specifically, the time domain characteristics of the operating conditions of the embodiments of the present application use current and voltage signals. The motor of the crusher usually generates current and voltage signals. Various statistics are extracted from these signals in the time domain, such as mean, standard deviation, maximum value, minimum value, etc. These characteristics can reflect the operating stability and power consumption of the motor.

[0061] Time domain features refer to the performance of a signal on the time axis. In operating data, current and voltage signals are often used for analysis because they directly reflect the electrical properties of the equipment or system under different operating conditions. For example, current can be used to understand the load condition of the equipment and the stability of the current waveform, while voltage can reflect the stability of the power grid and voltage fluctuations. This information is crucial for diagnosing the operating status of the equipment, analyzing the power grid load and the stability of the power system, so it is directly used as a time domain feature to train a diagnostic model, thereby directly reflecting the operating condition of the crusher.

[0062] The frequency domain characteristics of the working condition use vibration signals. Crusher usually has vibration sensors to monitor vibration conditions. Vibration signals are extracted through Fourier transform or spectrum analysis to obtain frequency domain characteristics, such as main frequency components, spectrum energy distribution, spectrum peaks, etc. These characteristics can help analyze the vibration condition of the crusher.

[0063] Frequency domain features refer to the characteristics of signals in the frequency domain. The frequency domain features of operating data are usually analyzed using vibration signals. Vibration signals can reveal the vibrations generated by mechanical equipment during operation, which directly reflects the prediction of equipment failures, monitoring of equipment health, and preventive maintenance. The responses of different devices or systems at different frequencies can provide important information about the structure, coupling, and response frequency of the equipment. Spectral analysis of vibration signals can help engineers identify abnormal frequency components and thus locate possible failure modes, such as bearing failures, imbalance, loose parts, etc.

[0064] The time-frequency domain characteristics of the working condition adopt the load signal. The load condition of the crusher varies with the material. This signal is analyzed by wavelet transform to obtain the time-frequency domain characteristics of the load change at different frequencies. These characteristics help to understand the performance of the crusher under different load conditions. The load signal reflects the load condition of the system or equipment, so the load signal is an important time-domain feature in the working condition data. The load signal refers to the energy consumed or the power executed by the system or equipment under different working conditions. By analyzing the time-domain characteristics of the load signal, we can understand the energy efficiency, power consumption characteristics and working state stability of the equipment under different workloads. Therefore, the diagnostic model is directly trained as a time-frequency domain feature, which directly reflects the working condition of the crusher.

[0065] For the time domain characteristics of the working condition, use statistics or basic time domain descriptions of the signal to characterize the data. For the frequency domain characteristics of the working condition, convert the signal to the frequency domain through methods such as Fourier transform or power spectral density estimation. For the time and frequency domain characteristics of the working condition, apply wavelet transform or other time-frequency analysis methods to capture the time-varying characteristics of the signal. Then, according to the specific application requirements, select the most representative features, and consider using dimensionality reduction techniques such as principal component analysis, i.e. PCA, to reduce the dimensionality and redundant information of the input features.

[0066] As one implementation mode, it also includes: The fault time domain feature of the fault feature data division is a vibration signal, the fault frequency domain feature of the fault feature data division is a sound signal, and the fault time and frequency domain feature of the fault feature data division is a temperature signal.

[0067] Specifically, the fault time domain feature of the embodiment of the present application is a vibration signal, and the fault usually causes a change in the vibration signal. Time domain features such as the mean, peak value, peak-to-valley ratio, etc. of the vibration signal can reflect the vibration mode of different types of faults.

[0068] Vibration signals are often used to analyze the time domain characteristics of fault data, because mechanical systems or equipment are often accompanied by abnormal vibrations when they fail. These vibrations can be caused by bearing failure, gear failure, imbalance or other mechanical problems. By monitoring and analyzing the time domain characteristics of vibration signals, the specific failure mode and location of mechanical equipment can be diagnosed, which helps to predict and avoid equipment downtime and damage in advance.

[0069] The fault frequency domain feature is the sound signal. Some faults may cause the crusher's sound spectrum to change. The signal obtained by the sound sensor can extract frequency domain features, such as spectrum diagrams, changes in frequency components, etc.

[0070] Since many fault modes produce sounds of specific frequencies, sound signals in the frequency domain can provide important information about fault data. For example, bearing failures, motor problems, fan blade damage, etc. can all be detected and diagnosed through frequency domain analysis of sound signals. Spectral analysis of sound signals can help determine abnormal frequency components, thereby locating and diagnosing possible fault types in equipment.

[0071] The time-frequency domain characteristics of faults are temperature signals. Faults usually cause temperature changes in crusher components. Temperature signals can be extracted through time-frequency analysis methods, such as wavelet transform, to extract the time-frequency domain characteristics of temperature changes at different frequencies, helping to detect thermal faults in equipment.

[0072] Since many devices generate abnormal heat when they fail or operate abnormally, temperature signals are a common time-frequency domain feature in fault data analysis. By monitoring and analyzing the temperature changes of equipment or systems, potential fault signs can be detected, such as circuit board overheating, motor winding problems, lubricant failure, etc. The time domain characteristics of temperature signals, such as the temperature change trend, and the frequency domain characteristics, such as the frequency components of temperature changes, can provide important references for fault prediction and diagnosis.

[0073] For fault time domain features, use statistics or basic time domain descriptions of the signal to characterize the data. For fault frequency domain features, convert the signal to the frequency domain through methods such as Fourier transform or power spectral density estimation. For fault time and frequency domain features, apply wavelet transform or other time-frequency analysis methods to capture the time-varying characteristics of the signal. Then, according to the specific application requirements, select the most representative features, and consider using dimensionality reduction techniques such as principal component analysis, i.e. PCA, to reduce the dimension and redundant information of the input features. The extracted features are used as input to a machine learning model or a statistical model for applications such as fault diagnosis, predictive maintenance or performance optimization. This application effectively utilizes the operating data and fault data of the crusher to extract appropriate time domain, frequency domain or time-frequency domain features, providing a good foundation for subsequent data analysis and model building.

[0074] As another implementation method, in addition to the operating condition time domain characteristics divided by the operating condition characteristic data as current and voltage signals, the operating condition frequency domain characteristics divided by the operating condition characteristic data as vibration signals, the operating condition time and frequency domain characteristics divided by the operating condition characteristic data as load signals, and the fault time domain characteristics divided by the fault characteristic data as vibration signals, the fault frequency domain characteristics divided by the fault characteristic data as sound signals, and the fault time and frequency domain characteristics divided by the fault characteristic data as temperature signals, other types of information are also set as operating condition characteristic data or fault characteristic data, which are suitable as time domain characteristics, frequency domain characteristics or time and frequency domain characteristics.

[0075] Specifically, it also includes pressure signals. Processing the pressure signals into time domain features can reflect the pressure changes inside the system or on the equipment, such as the average pressure value, the peak-to-peak value of the waveform, etc., which helps to evaluate the working status of the equipment and the pressure fluctuations.

[0076] The frequency domain characteristics of the pressure signal analyze the pressure fluctuations at different frequencies. For example, spectrum analysis reveals possible periodic pressure changes or frequency components.

[0077] Considering that pressure signals may vary with time and frequency, time-frequency domain features can simultaneously analyze their time-varying and frequency characteristics, such as time-frequency spectrum analysis of pressure signals.

[0078] It also includes flow signals. The time domain characteristics of flow signals include average flow rate, flow velocity change trend, etc., which are helpful to understand the fluid dynamic characteristics of the system or equipment. In the field of rock crushers, flow signals usually refer to the flow rate of feed or discharge.

[0079] The frequency domain characteristics of the flow signal will analyze the fluid flow conditions at different frequencies. For example, possible vortices or fluid instabilities can be identified through spectrum analysis of the flow signal.

[0080] Considering that flow signals vary with time and frequency, time-frequency domain features can comprehensively analyze their time-varying and frequency characteristics, such as time-frequency spectrum analysis of flow signals to detect the occurrence of abnormal flow patterns or vortices.

[0081] It also includes humidity signals. The time domain characteristics of the humidity signal include average humidity, standard deviation of humidity changes, etc., which are helpful in evaluating the humidity stability of the environment or equipment.

[0082] The frequency domain characteristics of the humidity signal analyze the humidity changes at different frequencies, such as by using spectrum analysis to detect possible humidity change patterns or periodic changes.

[0083] Considering that humidity signals may be affected by both time and frequency, time-frequency domain feature analysis can reveal their dynamic time-varying and frequency characteristics. For example, time-frequency spectrum analysis of humidity signals is used to detect abnormal patterns of ambient humidity changes.

[0084] These signals and feature analyses have important application value in operating condition monitoring and fault diagnosis, and can help to more comprehensively understand the operating status of equipment or systems, thereby achieving effective predictive maintenance and fault warning.

[0085] As one implementation method, a set of data including the operating condition characteristic data and the fault characteristic data is obtained as verification data, and the diagnosis result of the diagnosis model is evaluated according to the verification data to obtain performance indicator evaluation information; The generalization ability is judged based on several performance indicator evaluation information. When the generalization ability is non-stationary, the diagnosis model is optimized based on the performance indicator evaluation information.

[0086] The specific steps of obtaining one set of operating condition characteristic data and fault characteristic data as verification data and evaluating the diagnosis result of the diagnosis model according to the verification data include: The operating condition characteristic data is obtained, and the operating condition characteristic data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0087] The specific steps of obtaining one set of the operating condition characteristic data and the fault characteristic data as verification data and evaluating the diagnosis result of the diagnosis model according to the verification data also include: Fault feature data is obtained, and the fault feature data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0088] Specifically, this embodiment first divides the data set into a training set and a test set. Usually, the ratio is 70% for training and 30% for testing, ensuring that the training set is used for model training and parameter adjustment, and the test set is used for the final evaluation of the model's performance.

[0089] Then perform cross validation and use cross validation techniques to evaluate the stability and generalization ability of the model. The methods include either K-fold cross validation or leave-one-out cross validation. In the K-fold cross validation example, the training set is divided into K subsets, one of which is used as the validation set each time, and the remaining K-1 subsets are used as the training set. Repeat K times, select a different validation set each time, and calculate the performance indicators of the model on the validation set, such as the average value of accuracy, recall, etc. as the final evaluation result.

[0090] Next, we conduct performance evaluation. During each cross-validation process, we record and calculate various performance indicators of the model on the validation set. These indicators include classification accuracy, confusion matrix, area under the ROC curve, etc. The specific selection depends on the nature and requirements of the problem.

[0091] Perform generalization analysis to analyze the performance differences of the model under different cross-validation folds. If the model performs stably on different training sets and validation sets, it means that the model has strong generalization ability; otherwise, the model structure or parameters need to be further optimized.

[0092] Adjust and optimize the model. According to the results of cross-validation, optimize the parameters and structure of the model to improve its applicability and accuracy under different equipment and working conditions. Adjustments include increasing the complexity of the model, adjusting the learning rate, regularization parameters, etc.

[0093] As one implementation mode, it also includes: The training data and the validation data are cross-validated, including, when one set of the validation data is validated, re-dividing the operating condition characteristic data or the fault characteristic data to obtain updated training data and validation data, training the training model based on the training data to obtain a diagnostic model, and evaluating the diagnostic results of the diagnostic model based on the validation data.

[0094] Specifically, the specific implementation method of this embodiment is as follows: data set division, dividing the available crusher operating data and fault data into a training set and a test set. Cross-validation, select a 5-fold cross-validation method, and divide the training set into 5 subsets. Model selection, assuming that a classification model based on a convolutional neural network is selected to identify different fault modes from vibration and current data. Performance evaluation, for each cross-validation fold, train the CNN model, and evaluate its accuracy and confusion matrix on the validation set. Generalization ability analysis, analyze the performance of each cross-validation, and compare the performance indicators of the model on different validation sets. Ensure that the model performs stably on all folds. Parameter optimization, based on the cross-validation results, the convolutional layer depth and filter size of the CNN model may be adjusted, or regularization techniques such as dropout layers may be added to improve the model's generalization ability to new data and different working conditions.

[0095] Through the above steps, the generalization ability of the crusher fault diagnosis model can be effectively evaluated and optimized, so that it can reliably predict and diagnose various fault types in practical applications.

[0096] As one implementation method, correlation analysis is performed based on a triple correlation matrix to obtain analysis results for optimizing equipment maintenance strategies.

[0097] Specifically, the embodiment of the present application first collects a large amount of equipment working state data, fault information and output data of related models. These data are organized into triples, which include working conditions, fault types and model output results. An association matrix is ​​established to record the impact of different working conditions and fault types on model output. Through mathematical methods, the degree of influence of different working conditions and fault types on model output is calculated. These influences can be quantified by statistical analysis, correlation analysis or other mathematical models, so as to quantify the strength of association between working conditions, faults and model outputs, thereby helping to determine the most significant factors. After establishing the association matrix and calculating the strength of association, an association analysis is performed. By analyzing the association matrix, it is found that which working conditions and fault types have the greatest impact on the equipment diagnosis results. It is helpful to identify key working conditions and fault types, and provide guidance for subsequent maintenance and fault prevention strategies. Combined with the results of the association analysis, a targeted equipment maintenance and fault prevention strategy is formulated, and the equipment maintenance plan is adjusted according to the key working conditions and fault types, the maintenance efficiency is improved, the frequency of equipment failures is reduced through preventive measures, and the reliability and performance of the equipment are improved.

[0098] Through the above steps, the relationship between operating conditions, faults and model outputs can be effectively analyzed, providing a scientific basis and decision-making support for equipment maintenance and fault prevention.

[0099] The specific implementation methods of the present application include but are not limited to: First, the acquired data is preprocessed, where the operating data is the operating parameters of the crusher, such as feed speed, current, vibration, etc., and environmental conditions such as temperature and humidity. The fault data is the common types of crusher faults, such as wear of wearing parts, motor failure, transmission system problems, etc.

[0100] Next, feature extraction and selection are performed: vibration features: frequency, amplitude, waveform analysis. Current features: power spectrum density, current waveform features. Temperature features: temperature distribution, temperature change trend.

[0101] Then, model selection and training are carried out. The model selects a suitable machine learning or statistical model, such as support vector machine, logistic regression or deep learning model, such as convolutional neural network, etc., and training data. Historical fault data and operating condition data are used for model training, so that the model can accurately predict the possibility or occurrence of different fault types.

[0102] The training process is then evaluated using accuracy, precision, recall, etc. Cross-validation is used to improve the generalization ability of the model and verify the performance of the model on different data sets.

[0103] A triple association matrix, namely the working condition-fault-model triple association matrix, is established. The vertical type of the matrix is ​​the working condition, and the first to third rows are low feed speed, medium feed speed and high feed speed respectively. The horizontal type of the matrix is ​​the fault type, and from left to right they are wear of wearing parts, motor fault and transmission system problems, among which the corresponding values ​​for low feed speed and wear of wearing parts are 0.7, low feed speed and motor fault are 0.1, low feed speed and transmission system are 0.3, medium feed speed and wear of wearing parts are 0.4, medium feed speed and motor fault are 0.5, medium feed speed and transmission system are 0.2, high feed speed and wear of wearing parts are 0.2, high feed speed and motor fault are 0.8, and high feed speed and transmission system are 0.6. The values ​​represent the influence of various fault types on the model output at different feed speeds. For example, motor fault has the greatest influence on the model output at high feed speed.

[0104] Finally, the analysis and optimization strategy are carried out. The correlation analysis is based on the correlation matrix to analyze which operating conditions and fault types have the greatest impact on the prediction of the crusher's operating status. The optimization strategy formulates more effective maintenance plans and fault prevention strategies based on the analysis results. For example, the frequency of tool inspections can be adjusted according to the feed speed, or the motor operating status at high feed speeds can be regularly checked.

[0105] As another implementation, a triple association matrix, i.e., a working condition-fault-model triple association matrix, is established. The vertical type of the matrix is ​​the working condition, and the first to third rows are low feed speed, medium feed speed, and high feed speed, respectively. The horizontal type of the matrix is ​​the fault type, and from left to right, they are wear of consumables, motor faults, and transmission system problems, respectively, wherein the corresponding low feed speed and wear of consumables is 0.8, the low feed speed and motor fault is 0.2, the low feed speed and transmission system is 0.1, the medium feed speed and wear of consumables is 0.3, the medium feed speed and motor fault is 0.7, the medium feed speed and transmission system is 0.4, the high feed speed and wear of consumables is 0.1, the high feed speed and motor fault is 0.5, and the high feed speed and transmission system is 0.9. The numerical value represents the influence of various fault types on the model output at different feed speeds. For example, the wear of consumables has the greatest influence on the model output at high feed speed.

[0106] Through the above steps, the operating condition-fault-model triple association matrix method is used to quantify the impact of different operating conditions and fault types on crusher performance and fault prediction, thereby optimizing the equipment maintenance and fault prevention strategies and improving the equipment reliability and operation efficiency.

[0107] Reference Figure 2 , an embodiment of the present application provides a crusher fault prediction and diagnosis system, the system includes a first acquisition module, a second acquisition module, a third acquisition module, a correlation module and a correlation analysis module.

[0108] The first acquisition module is used to acquire the equipment working status data, pre-process the equipment working status data, and obtain working condition characteristic data; The second acquisition module is used to acquire device fault data while acquiring device working status data, and perform feature extraction on the device fault data to obtain fault feature data; The third acquisition module is used to acquire the model to be trained, input the operating condition characteristic data and the fault characteristic data into the model to be trained, and obtain the diagnosis model; An association module, used for constructing a triple association matrix according to the working condition characteristic data, the fault characteristic data and the diagnosis model; The association analysis module is used to perform association analysis based on the triple association matrix to obtain analysis results for optimizing equipment maintenance strategies.

[0109] Specifically, the present application obtains equipment working status data and equipment fault data, extracts features from them respectively, preprocesses the equipment working status data to obtain working condition feature data, extracts features from the equipment fault data to obtain fault feature data, selects a suitable model type as the model to be trained, inputs the working condition feature data and the fault feature data into the model to be trained for training, obtains a diagnostic model, and then constructs a triple association matrix based on the working condition feature data, fault feature data and the diagnostic model, calculates the impact of different working conditions and fault types on the model output, quantifies the strength of the association between the three, uses the triple association matrix to perform association analysis, and obtains analysis results for optimizing equipment maintenance strategies, thereby extracting features that can comprehensively reflect the equipment fault status and improving the accuracy of crusher fault prediction and diagnosis.

[0110] As one implementation mode, it also includes: A model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data, evaluate the diagnosis result of the diagnosis model according to the verification data, and obtain performance index evaluation information; The generalization ability is judged based on several performance indicator evaluation information. When the generalization ability is non-stationary, the diagnosis model is optimized based on the performance indicator evaluation information.

[0111] As one implementation mode, it also includes: The third acquisition module is also used to divide and process the operating condition characteristic data into operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics, and input the operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics into the model to be trained for training.

[0112] As one implementation mode, it also includes: The third acquisition module is also used to divide the fault feature data into fault time domain features, fault frequency domain features and fault time-frequency domain features, and input the fault time domain features, fault frequency domain features and fault time-frequency domain features into the model to be trained for training.

[0113] As one implementation mode, the model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data, and the specific steps of evaluating the diagnosis result of the diagnosis model according to the verification data include: The operating condition characteristic data is obtained, and the operating condition characteristic data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0114] As one implementation mode, the model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data, and the specific steps of evaluating the diagnosis result of the diagnosis model according to the verification data also include: Fault feature data is obtained, and the fault feature data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

[0115] As one implementation mode, it also includes: The third acquisition module is also used to cross-validate the training data and the verification data, including, when one set of verification data is verified, re-dividing the operating condition characteristic data or the fault characteristic data to obtain updated training data and verification data, training the training model according to the training data to obtain a diagnostic model, and evaluating the diagnostic results of the diagnostic model according to the verification data.

[0116] As one implementation mode, it also includes: The working condition time domain characteristics of the working condition characteristic data divided by the third acquisition module are current and voltage signals, the working condition frequency domain characteristics of the working condition characteristic data are vibration signals, and the working condition time and frequency domain characteristics of the working condition characteristic data are load signals.

[0117] As one implementation mode, it also includes: The fault time domain feature of the fault feature data divided by the third acquisition module is a vibration signal, the fault frequency domain feature of the fault feature data divided by the fault feature data is a sound signal, and the fault time and frequency domain feature of the fault feature data divided by the fault feature data is a temperature signal.

[0118] An embodiment of the present application provides a crusher fault prediction and diagnosis device, including a memory and a processor, the memory storing a computer program, and the processor being configured to run the computer program to execute the crusher fault prediction and diagnosis method as described above.

[0119] An embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned crusher fault prediction and diagnosis method when running.

[0120] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described device and product can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0121] In several embodiments provided in this application, it should be understood that the disclosed methods, systems, devices and program products may be implemented in other ways.

[0122] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0123] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting and diagnosing a crusher fault, characterized in that: include: Acquire equipment working status data, pre-process the equipment working status data, and obtain working condition characteristic data; When obtaining the equipment working status data, obtaining the equipment fault data, and performing feature extraction on the equipment fault data to obtain fault feature data; Obtain a model to be trained, input the operating condition characteristic data and the fault characteristic data into the model to be trained, and obtain a diagnostic model; Construct a triple association matrix based on the working condition characteristic data, fault characteristic data and the diagnosis model; Correlation analysis is performed based on the triple association matrix to obtain analysis results for optimizing equipment maintenance strategies.

2. The method for predicting and diagnosing a crusher fault according to claim 1, characterized in that: Also includes: Obtaining one set of operating condition characteristic data and fault characteristic data as verification data, evaluating the diagnosis result of the diagnosis model according to the verification data, and obtaining performance index evaluation information; The generalization ability is judged based on several performance indicator evaluation information. When the generalization ability is non-stationary, the diagnosis model is optimized based on the performance indicator evaluation information.

3. The method for predicting and diagnosing a crusher fault according to claim 1, characterized in that: Also includes: The operating condition characteristic data are divided and processed into operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics, and the operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics are input into the model to be trained for training.

4. The method for predicting and diagnosing a crusher fault according to claim 3, characterized in that: Also includes: The fault feature data is divided and processed into fault time domain features, fault frequency domain features and fault time-frequency domain features, and the fault time domain features, fault frequency domain features and fault time-frequency domain features are input into the model to be trained for training.

5. The method for predicting and diagnosing a crusher fault according to claim 2, characterized in that: The specific steps of obtaining one set of operating condition characteristic data and fault characteristic data as verification data and evaluating the diagnosis result of the diagnosis model according to the verification data include: The operating condition characteristic data is obtained, and the operating condition characteristic data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

6. The method for predicting and diagnosing a crusher fault according to claim 5, characterized in that: The specific steps of obtaining one set of the operating condition characteristic data and the fault characteristic data as verification data and evaluating the diagnosis result of the diagnosis model according to the verification data also include: Fault feature data is obtained, and the fault feature data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

7. The method for predicting and diagnosing a crusher fault according to claim 6, characterized in that: Also includes: The training data and the validation data are cross-validated, including, when one set of the validation data is validated, re-dividing the operating condition characteristic data or the fault characteristic data to obtain updated training data and validation data, training the training model based on the training data to obtain a diagnostic model, and evaluating the diagnostic results of the diagnostic model based on the validation data.

8. The method for predicting and diagnosing a crusher fault according to claim 3, characterized in that: Also includes: The time domain characteristics of the working condition divided by the working condition characteristic data are current and voltage signals, the frequency domain characteristics of the working condition divided by the working condition characteristic data are vibration signals, and the time and frequency domain characteristics of the working condition divided by the working condition characteristic data are load signals.

9. The method for predicting and diagnosing a crusher fault according to claim 4, characterized in that: Also includes: The fault time domain feature of the fault feature data division is a vibration signal, the fault frequency domain feature of the fault feature data division is a sound signal, and the fault time and frequency domain feature of the fault feature data division is a temperature signal.

10. A crusher fault prediction and diagnosis system, characterized in that: include: The first acquisition module is used to acquire the equipment working status data, pre-process the equipment working status data, and obtain working condition characteristic data; The second acquisition module is used to acquire device fault data while acquiring device working status data, and perform feature extraction on the device fault data to obtain fault feature data; The third acquisition module is used to acquire the model to be trained, input the operating condition characteristic data and the fault characteristic data into the model to be trained, and obtain the diagnosis model; An association module, used for constructing a triple association matrix according to the working condition characteristic data, the fault characteristic data and the diagnosis model; The association analysis module is used to perform association analysis based on the triple association matrix to obtain analysis results for optimizing equipment maintenance strategies.

11. The crusher fault prediction and diagnosis system according to claim 10, characterized in that: Also includes: A model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data, evaluate the diagnosis result of the diagnosis model according to the verification data, and obtain performance index evaluation information; The generalization ability is judged based on several performance indicator evaluation information. When the generalization ability is non-stationary, the diagnosis model is optimized based on the performance indicator evaluation information.

12. The crusher fault prediction and diagnosis system according to claim 10, characterized in that: Also includes: The third acquisition module is also used to divide the operating condition characteristic data into operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics, and input the operating condition time domain characteristics, operating condition frequency domain characteristics and operating condition time-frequency domain characteristics into the model to be trained for training.

13. The crusher fault prediction and diagnosis system according to claim 12, characterized in that: Also includes: The third acquisition module is also used to divide the fault feature data into fault time domain features, fault frequency domain features and fault time-frequency domain features, and input the fault time domain features, fault frequency domain features and fault time-frequency domain features into the model to be trained for training.

14. The crusher fault prediction and diagnosis system according to claim 11, characterized in that: The model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data. The specific steps of evaluating the diagnosis result of the diagnosis model according to the verification data include: The operating condition characteristic data is obtained, and the operating condition characteristic data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

15. The crusher fault prediction and diagnosis system according to claim 14, characterized in that: The model optimization module is used to obtain one set of operating condition characteristic data and fault characteristic data as verification data. The specific steps of evaluating the diagnosis result of the diagnosis model according to the verification data also include: Fault feature data is obtained, and the fault feature data is divided into training data and verification data. The training model is trained according to the training data to obtain a diagnostic model, and the diagnostic result of the diagnostic model is evaluated according to the verification data.

16. The crusher fault prediction and diagnosis system according to claim 15, characterized in that: Also includes: The third acquisition module is also used to cross-validate the training data and the verification data, including, when one set of verification data is verified, re-dividing the operating condition characteristic data or the fault characteristic data to obtain updated training data and verification data, training the training model according to the training data to obtain a diagnostic model, and evaluating the diagnostic results of the diagnostic model according to the verification data.

17. The crusher fault prediction and diagnosis system according to claim 12, characterized in that: Also includes: The working condition time domain characteristics of the working condition characteristic data divided by the third acquisition module are current and voltage signals, the working condition frequency domain characteristics of the working condition characteristic data are vibration signals, and the working condition time and frequency domain characteristics of the working condition characteristic data are load signals.

18. The crusher fault prediction and diagnosis system according to claim 13, characterized in that: Also includes: The fault time domain feature of the fault feature data divided by the third acquisition module is a vibration signal, the fault frequency domain feature of the fault feature data divided by the fault feature data is a sound signal, and the fault time and frequency domain feature of the fault feature data divided by the fault feature data is a temperature signal.

19. A crusher fault prediction and diagnosis device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the fault prediction and diagnosis method for a crusher according to any one of claims 1 to 9.

20. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the crusher fault prediction and diagnosis method according to any one of claims 1 to 9 when running.

Citation Information

Patent Citations

  • Photovoltaic inverter fault prediction method

    CN117318614A

  • Bearing fault identification method and device, electronic equipment and storage medium

    CN117540267A

  • Water chilling unit fault diagnosis feature selection method and system based on interpretable analysis

    CN118035713A

Cited By

  • Dynamic equipment intelligent monitoring and maintenance system and method based on vibration signal fusion multi-source information

    CN119720052A

  • Intelligent Monitoring and Maintenance System and Method for Dynamic Equipment Based on Vibration Signal Fusion of Multi-Source Information

    CN119720052B

  • Method and system for monitoring working state of solid waste crushing equipment

    CN120771993A

  • A solid waste crushing equipment working state monitoring method and system

    CN120771993B