A method and system for predicting failure of a refrigeration circulator

Abstract

The present invention relates to the field of data processing, and in particular to a method and system for predicting faults of a refrigeration circulator. The contents include: collecting real-time operation data of a refrigeration circulator, pre-processing the real-time operation data and then performing intelligent processing, performing feature analysis on the data after intelligent processing to obtain real-time data features, and storing the real-time data features in a historical data feature library; training and optimizing based on the historical data features in the historical data feature library to obtain an optimized fault prediction model, processing the real-time data features through the optimized fault prediction model to obtain a fault prediction result; generating maintenance suggestions based on the fault prediction results, generating early warning information based on the maintenance suggestions and the fault prediction results, and notifying the maintenance personnel of the early warning information to implement predictive maintenance. The method solves the problems of poor accuracy of data processing, low prediction efficiency and inaccurate prediction results in the prior art in the process of implementing fault prediction of refrigeration circulators.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a fault prediction method and system for a refrigeration circulator. Background Art

[0002] In modern industrial and commercial fields, refrigeration circulators are key equipment and are widely used in cold chain logistics, food processing, chemical production and other industries. The normal operation of refrigeration circulators is crucial to ensure product quality and production efficiency. However, due to the complex structure and changeable operating environment of refrigeration circulators, their failure frequency is high, which often leads to production interruptions and economic losses. Therefore, how to achieve fault prediction and preventive maintenance of refrigeration circulators has become a hot issue in current research and application.

[0003] Traditional maintenance methods for refrigeration circulators mainly rely on regular inspections and passive maintenance. This method not only has high maintenance costs, but also fails to detect potential faults in a timely manner, which can easily lead to sudden equipment shutdowns and affect normal production. With the development of sensor technology, data analysis technology, and artificial intelligence technology, data-driven fault prediction and maintenance methods have gradually become the focus of research. This method collects equipment operation data in real time, combines data analysis and machine learning technology, and can provide early warning before a fault occurs, thereby achieving predictive maintenance, reducing maintenance costs, and improving equipment operation reliability.

[0004] However, the above technology has at least the following technical problems: in the process of implementing the fault prediction of the refrigeration circulator, the accuracy of data processing is poor, the prediction efficiency is low, and the prediction result is inaccurate. Summary of the invention

[0005] The present invention provides a method and system for predicting a fault of a refrigeration circulator, so as to solve the problems of poor accuracy of data processing, low prediction efficiency and inaccurate prediction results in the prior art during the process of realizing the fault prediction of the refrigeration circulator. The method and system for predicting a fault of a refrigeration circulator of the present invention specifically include the following technical solutions:

[0006] A method for predicting failure of a refrigeration cycler comprises the following steps:

[0007] S1. Collect the real-time operation data of the refrigeration circulator, pre-process the real-time operation data, and then use the intelligent enhancement processing algorithm to perform intelligent processing; in the implementation of the intelligent enhancement processing algorithm, under the premise of retaining the characteristics of the pre-processed real-time operation data, by introducing sine wave modulation and adding composite noise, the enhanced real-time operation data is obtained; the enhanced real-time operation data is offset transformed, and the offset real-time operation data is obtained by introducing cosine wave modulation; the offset real-time operation data is rotated to obtain the rotated real-time operation data; the rotated real-time operation data is expanded to generate new data samples to obtain the expanded real-time operation data; the extended real-time operation data is feature analyzed to obtain the real-time data features, and the real-time data features are stored in the historical data feature library;

[0008] S2. Train and optimize the historical data features in the historical data feature library to obtain an optimized fault prediction model; introduce a robust feature filtering algorithm in the fault prediction process; process the real-time data features through the optimized fault prediction model to obtain fault prediction results; generate maintenance suggestions based on the fault prediction results, generate warning information based on the maintenance suggestions and fault prediction results, and notify the maintenance personnel of the warning information to implement predictive maintenance.

[0009] Preferably, the S1 specifically includes:

[0010] In the process of implementing the intelligent enhancement processing algorithm, the specific implementation formula for generating enhanced real-time operation data is as follows:

[0011] ,

[0012] in, It is the enhanced real-time operation data; It is the real-time running data after preprocessing; is the number of noise samples; It is The standard deviation of the noise samples; is a normal distribution with a mean of 0 and a standard deviation of 1, used to generate random noise; Used to prevent zero value problems in logarithmic calculations; It is a sine wave modulation.

[0013] Preferably, in said S1, it further includes:

[0014] In the process of implementing the intelligent enhancement processing algorithm, the specific implementation process of offset transformation of the enhanced real-time operation data is as follows:

[0015] ,

[0016] in, It is the real-time operation data after the offset; offset vector; is the amount of enhanced real-time running data; Indicates Enhanced real-time operation data; It is cosine wave modulation.

[0017] Preferably, in said S1, it further includes:

[0018] In the process of implementing the intelligent enhancement processing algorithm, the specific formula for rotating the real-time running data after the offset is as follows:

[0019] ,

[0020] in, It is the real-time running data after rotation transformation; is the rotation matrix; is the rotation angle.

[0021] Preferably, in said S1, it further includes:

[0022] In the process of implementing the intelligent enhancement processing algorithm, the specific implementation process of expanding the rotated real-time running data is as follows:

[0023] ,

[0024] in, It is the expanded real-time operation data; It is Real-time operation data after rotation; is the number of samples to expand the rotated real-time running data; T is the transpose; It is The standard deviation of the real-time running data after rotation; is the identity matrix; is a natural constant used for exponential transformation; It is The L2 norm of the rotated real-time running data plus the unit matrix; It is cosine wave modulation.

[0025] Preferably, the S2 specifically includes:

[0026] The historical data features in the historical data feature library are divided into a training set and a validation set. A multi-layer neural network model is used for training. The training set is normalized and then feature derivation is performed. The derived features are merged with the normalized training set to obtain a fused comprehensive feature set. The comprehensive feature set is processed through an input layer, multiple hidden layers, and an output layer to obtain a fault prediction model. The validation set is used to evaluate the model performance of the fault prediction model.

[0027] Preferably, in said S2, it further includes:

[0028] The robust feature filtering algorithm is introduced to perform robustness calculation and noise filtering on the fused comprehensive feature set; the specific implementation process is as follows:

[0029] Calculate the robustness of each feature in the comprehensive feature set:

[0030] ,

[0031] in, represents a comprehensive feature set; Represents the mean vector of the comprehensive feature set; represents the standard deviation vector of the comprehensive feature set; represents the feature robustness matrix;

[0032] Filter each feature in the comprehensive feature set to remove noise features:

[0033] ,

[0034] in, represents the filtered comprehensive feature set; represents the Hadamard product;

[0035] The filtered comprehensive feature set is used to train the fault prediction model to obtain an optimized fault prediction model.

[0036] A fault prediction system for a refrigeration circulator is applied to a fault prediction method for a refrigeration circulator, and includes the following parts:

[0037] Data acquisition module, data intelligent processing module, data analysis and storage module, fault prediction model module, maintenance decision support module, early warning notification module;

[0038] A data acquisition module collects real-time operation data of the refrigeration circulator and sends the real-time operation data to the data intelligent processing module;

[0039] The data intelligent processing module pre-processes the real-time operation data and then performs intelligent processing, and sends the intelligently processed data to the data analysis and storage module;

[0040] The data analysis and storage module performs feature analysis on the intelligently processed data to obtain real-time data features, stores the real-time data features in a historical data feature library, and sends them to the fault prediction model module;

[0041] The fault prediction model module is trained and optimized based on the historical data features in the historical data feature library to obtain an optimized fault prediction model, and the real-time data features are processed by the optimized fault prediction model to obtain a fault prediction result, and the fault prediction result is sent to the maintenance decision support module and the early warning notification module;

[0042] The maintenance decision support module generates maintenance suggestions based on the fault prediction results and sends the maintenance suggestions to the early warning notification module;

[0043] The early warning notification module generates early warning information based on fault prediction results and maintenance suggestions, notifies maintenance personnel of the early warning information, and implements predictive maintenance.

[0044] The beneficial effects of the technical solution of the present invention are:

[0045] 1. Through the intelligent enhancement processing algorithm, a small amount of composite noise is added to the preprocessed real-time operation data for enhancement, the enhanced real-time operation data is offset transformed, the offset real-time operation data is rotated, and the rotated real-time operation data is expanded to improve the quality and diversity of the real-time operation data so that it can be better generalized during training while maintaining the basic trends and basic patterns of the real-time operation data; by expanding the rotated real-time operation data, the basic structure and characteristics of the real-time operation data are maintained.

[0046] 2. Based on the data in the historical data feature library, through training and optimization of the multi-layer neural network model, it can accurately predict faults, generate maintenance suggestions and warning information, and promptly notify maintenance personnel to implement predictive maintenance; divide the training set data into multiple small batches and process them in parallel to improve the efficiency of fault prediction model training, introduce a parallel computing framework to achieve rapid processing of large-scale data, and significantly shorten the training time of the fault prediction model; the use of parallel computing enables the fault prediction system to maintain high efficiency when processing big data and meet the needs of real-time prediction; introduce a robust feature filtering algorithm to effectively measure the robustness of each feature, perform robustness calculation and noise filtering on the comprehensive feature set, remove noise features, ensure the data quality of the input fault prediction model, solve the data noise problem in the prediction process, and improve the accuracy of the fault prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A structural diagram of a fault prediction system for a refrigeration circulator according to the present invention;

[0048] Figure 2 The present invention is a flowchart of a method for predicting a fault of a refrigeration circulator. DETAILED DESCRIPTION

[0049] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0051] The specific scheme of a fault prediction method and system for a refrigeration circulator provided by the present invention is described in detail below with reference to the accompanying drawings.

[0052] Refer to the attached Figure 1 , which shows a structural diagram of a fault prediction system for a refrigeration cycler provided by an embodiment of the present invention, the system includes the following parts:

[0053] Data acquisition module, data intelligent processing module, data analysis and storage module, fault prediction model module, maintenance decision support module, early warning notification module;

[0054] The data acquisition module collects the real-time operation data of the refrigeration circulator, including temperature data, pressure data, vibration data, and current data, and sends the real-time operation data to the data intelligent processing module;

[0055] The data intelligent processing module pre-processes the real-time operation data and then performs intelligent processing, and sends the intelligently processed data to the data analysis and storage module;

[0056] The data analysis and storage module performs feature analysis on the intelligently processed data to obtain real-time data features, stores the real-time data features in a historical data feature library, and sends them to the fault prediction model module;

[0057] The fault prediction model module is trained and optimized based on the historical data features in the historical data feature library to obtain an optimized fault prediction model, and the real-time data features are processed by the optimized fault prediction model to obtain a fault prediction result, and the fault prediction result is sent to the maintenance decision support module and the early warning notification module;

[0058] The maintenance decision support module generates maintenance suggestions based on the fault prediction results, and sends the maintenance suggestions to the early warning notification module; the fault prediction results are such as fault type, fault probability, and expected time of occurrence; the early warning notification module generates early warning information based on the fault prediction results and maintenance suggestions, and notifies the maintenance personnel of the early warning information through SMS, email, mobile application push, etc., to implement predictive maintenance.

[0059] See attached Figure 2 , which shows a flow chart of a method for predicting a failure of a refrigeration cycler provided by an embodiment of the present invention, the method comprising the following steps:

[0060] S1. Collect the real-time operation data of the refrigeration circulator, pre-process the real-time operation data and then perform intelligent processing, perform feature analysis on the data after intelligent processing to obtain real-time data features, and store the real-time data features in a historical data feature library to provide data support for fault prediction;

[0061] Temperature sensors, pressure sensors, vibration sensors and current sensors are installed on key components of the refrigeration circulator (such as compressor, evaporator, condenser, etc.). Real-time operation data of the refrigeration circulator (such as temperature data, pressure data, vibration data, current data) is collected through sensors to obtain real-time operation data.

[0062] Preprocess the real-time operation data, which is implemented as follows:

[0063] The real-time operation data is subjected to outlier detection using statistical methods such as the interquartile range method, and the detected outliers are deleted, interpolated, or replaced. The interpolation may be performed using linear interpolation or spline interpolation. The real-time operation data after outlier detection is subjected to filtering techniques (such as low-pass filtering, high-pass filtering, and band-pass filtering) to remove noise, and then smoothing techniques such as Kalman filtering and average filtering are selected. Finally, the real-time operation data after smoothing is normalized to obtain pre-processed real-time operation data. The above-mentioned pre-processing techniques are well known to those skilled in the art and will not be elaborated here.

[0064] The pre-processed real-time operation data is intelligently processed using an intelligent enhancement processing algorithm to improve the quality and diversity of the data and provide richer data for subsequent feature extraction and analysis. The specific implementation process is as follows:

[0065] On the premise of retaining the basic characteristics of the preprocessed real-time running data, a small amount of composite noise is added to enhance the diversity of the preprocessed real-time running data so that it can be better generalized during training; the specific implementation formula is as follows:

[0066] ,

[0067] in, is the enhanced real-time running data, which means the enhanced data obtained after adding noise; It is the preprocessed real-time running data, which means the data obtained after outlier detection, smoothing and normalization; is the number of noise samples, which indicates the total number of noise samples used to enhance the data; It is The standard deviation of noise samples indicates the intensity of the noise; is a normal distribution with a mean of 0 and a standard deviation of 1, used to generate random noise; is a very small constant to prevent zero value problems in logarithmic calculations; It is a sine wave modulation used to increase the complexity of the data.

[0068] Furthermore, by performing an offset transformation on the enhanced real-time operation data, the diversity of real-time operation data samples is increased while maintaining the basic trend of the real-time operation data. The specific implementation process is as follows:

[0069] ,

[0070] in: It is the real-time running data after migration, which means the data obtained after migration transformation; An offset vector, which represents a vector for performing an offset transformation on the enhanced real-time operation data, and its elements may be random values ​​or specific values ​​within a data range set according to an expert experience method; is the number of enhanced real-time running data, indicating the total number of data involved in the calculation; Indicates Enhanced real-time operation data; It is a cosine wave modulation used to increase the complexity of the data.

[0071] Through rotation transformation, the diversity of real-time operation data is further increased while maintaining the basic pattern of real-time operation data. The specific formula is as follows:

[0072] ,

[0073] in: is the real-time operation data after the rotation transformation, indicating the real-time operation data obtained after the rotation transformation; is the rotation matrix, which is used to rotate the real-time running data after the offset. is the rotation angle.

[0074] Finally, the rotated real-time operation data is expanded to generate new data samples, maintaining the basic structure and characteristics of the real-time operation data. The specific implementation process is as follows:

[0075] ,

[0076] in, It is the expanded real-time running data, which means the new data samples generated after data expansion processing, that is, the data after intelligent processing; It is The real-time running data after rotation indicates the data samples; is the number of samples used to expand the rotated real-time running data, indicating the total number of data used for expansion; T is the transpose; It is The standard deviation of the real-time running data after rotation; is the identity matrix, used to ensure stability in matrix operations; is a natural constant used for exponential transformation; It is The L2 norm of the rotated real-time running data plus the unit matrix; It is a cosine wave modulation used to increase the complexity of the data.

[0077] The existing feature engineering technology is used to extract features from the data after intelligent processing, and the extracted features are analyzed using existing statistical analysis methods (such as principal component analysis and factor analysis algorithms) to obtain real-time data features. The real-time data features are stored in the historical data feature library to provide data support for fault prediction.

[0078] S2. Perform training and optimization based on the historical data features in the historical data feature library to obtain an optimized fault prediction model, process the real-time data features through the optimized fault prediction model to obtain fault prediction results, generate maintenance suggestions based on the fault prediction results, generate warning information based on the maintenance suggestions and fault prediction results, and notify the maintenance personnel of the warning information to implement predictive maintenance.

[0079] Based on the historical data features in the historical data feature library, the existing multi-layer neural network model is used for training to obtain a fault prediction model. The specific training process includes dividing the historical data features into a training set and a validation set, using the training set as the input of the multi-layer neural network model, first normalizing the training set to obtain a training set that conforms to the standard normal distribution, and then using the existing feature derivation technology (such as logarithmic transformation, square root transformation) to obtain the derived features, and merging the derived features with the training set to obtain a fused comprehensive feature set, and processing the comprehensive feature set through the input layer, multiple hidden layers, and output layer to finally obtain the fault prediction results, including the fault type, fault probability, and expected time of occurrence. The specific number of hidden layers can be set according to the specific implementation scenario, and is not limited here. At the same time, the gradient is calculated through back propagation, and the parameters are updated using the existing gradient descent method. In the back propagation process, an existing loss function such as the cross entropy loss function is used. Finally, the fault prediction model is evaluated for model performance using the validation set to verify the accuracy of the fault prediction model.

[0080] In order to improve the efficiency of fault prediction model training, the training set data is divided into L batches and processed in parallel to improve the efficiency of fault prediction model training. In the fault prediction model prediction process, the existing parallel computing framework is introduced to realize the rapid processing of large-scale historical data feature data. The specific batch can be set according to the specific implementation scenario and is not limited here.

[0081] In order to reduce the data noise problem in the fault prediction process, a robust feature filtering algorithm is introduced to solve the data noise problem in fault prediction; the specific implementation process is as follows:

[0082] By calculating the robustness of the comprehensive feature set and filtering the noise, the noise problem of the comprehensive feature set in the fault prediction process is solved and the accuracy of the fault prediction model is improved.

[0083] Calculate the robustness of each feature in the comprehensive feature set:

[0084] ,

[0085] in, represents a comprehensive feature set; Represents the mean vector of the comprehensive feature set; represents the standard deviation vector of the comprehensive feature set; Represents the feature robustness matrix, which measures the robustness of each feature and filters out noise features;

[0086] Filter each feature in the comprehensive feature set to remove noise features:

[0087] ,

[0088] in, represents the filtered comprehensive feature set, removing the noise features; represents Hadamard product (element-by-element multiplication); the filtered comprehensive feature set is used to train the fault prediction model to obtain the optimized fault prediction model; the real-time data features are processed by the optimized fault prediction model to obtain the fault prediction results, which may include fault type, fault probability, and expected occurrence time.

[0089] Generate maintenance recommendations based on fault prediction results, that is, provide corresponding solutions for different types of faults and put forward preventive maintenance recommendations, such as regular inspection and replacement of wearing parts. Break down maintenance recommendations into specific operation steps, ensure that each step has detailed instructions, and facilitate maintenance personnel to perform. At the same time, the fault prediction system will compare with historical maintenance records to verify the effectiveness of the recommendations, collect maintenance feedback, and continuously optimize maintenance recommendations.

[0090] Generate early warning information based on fault prediction results and maintenance recommendations, such as:

[0091] Fault type: Compressor failure

[0092] Failure probability: 85%

[0093] Estimated time of occurrence: 2024-06-20 14:30

[0094] Maintenance advice: Please check the operating status of the compressor and be prepared to replace any parts that may be faulty.

[0095] According to the maintenance personnel’s contact information, select the appropriate notification channel, including SMS, email, and mobile application push, to notify the maintenance personnel of the warning information and implement predictive maintenance.

[0096] In summary, a fault prediction method and system for a refrigeration circulator are completed.

[0097] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0098] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0099] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such 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 invention, and should be included in the protection scope of the present invention.

Claims

1. A method for predicting failure of a refrigeration cycler, characterized in that: The following steps are involved: S1. Collect the real-time operation data of the refrigeration circulator, pre-process the real-time operation data, and then use the intelligent enhancement processing algorithm to perform intelligent processing; In the implementation of the intelligent enhancement processing algorithm, under the premise of retaining the characteristics of the pre-processed real-time operation data, the enhanced real-time operation data is obtained by introducing sinusoidal wave modulation and adding composite noise. The specific implementation formula is as follows: Among them, X enhanced It is the enhanced real-time operation data; X preprocessed is the real-time running data after preprocessing; n is the number of noise samples; σ i is the standard deviation of the ith noise sample; It is a normal distribution with a mean of 0 and a standard deviation of 1; ∈ is used to prevent the zero value problem in logarithmic calculation; It is sine wave modulation; The enhanced real-time operation data is subjected to offset transformation, and the offset real-time operation data is obtained by introducing cosine wave modulation. The specific implementation formula is as follows: Among them, X shifted is the real-time running data after migration; S is the migration vector; m is the number of enhanced real-time running data; X enhanced，j represents the real-time operation data after the jth enhancement; It is cosine wave modulation; The real-time running data after the offset is rotated to obtain the rotated real-time running data. The specific implementation formula is as follows: Among them, X rotated is the real-time running data after rotation transformation; R is the rotation matrix; θ is the rotation angle; The rotated real-time operation data is expanded to generate new data samples to obtain the expanded real-time operation data. The specific implementation formula is as follows: Among them, X augmented It is the expanded real-time operation data; is the real-time running data after the lth rotation; k is the number of samples expanded for the real-time running data after rotation; T is the transposition; σ l is the standard deviation of the real-time running data after the lth rotation; I is the unit matrix; e is a natural constant; It is the L2 norm of the real-time running data after the lth rotation plus the unit matrix; It is cosine wave modulation; Perform feature analysis on the expanded real-time operation data to obtain real-time data features, and store the real-time data features in a historical data feature library; S2. Divide the historical data features in the historical data feature library into a training set and a validation set, use a multi-layer neural network model for training, normalize the training set, and then derive features. Merge the derived features with the normalized training set to obtain a fused comprehensive feature set. Process the comprehensive feature set through an input layer, multiple hidden layers, and an output layer to obtain a fault prediction model. Use the validation set to evaluate the model performance of the fault prediction model. In the fault prediction process, introduce a robust feature filtering algorithm to perform robustness calculation and noise filtering on the fused comprehensive feature set. The specific implementation process is as follows: Calculate the robustness of each feature in the comprehensive feature set: Among them, Z h represents a comprehensive feature set; Represents the mean vector of the comprehensive feature set; Represents the standard deviation vector of the comprehensive feature set; R feature represents the feature robustness matrix; Filter each feature in the comprehensive feature set to remove noise features: Among them, Z f represents the filtered comprehensive feature set; ° represents the Hadamard product; The filtered comprehensive feature set is used for fault prediction model training to obtain an optimized fault prediction model; The real-time data features are processed through an optimized fault prediction model to obtain fault prediction results; maintenance suggestions are generated based on the fault prediction results, warning information is generated based on the maintenance suggestions and fault prediction results, and the warning information is notified to maintenance personnel to implement predictive maintenance.

2. A refrigeration cycle fault prediction system, applied to a refrigeration cycle fault prediction method as claimed in claim 1, characterized in that: Includes the following sections: Data acquisition module, data intelligent processing module, data analysis and storage module, fault prediction model module, maintenance decision support module, early warning notification module; A data acquisition module collects real-time operation data of the refrigeration circulator and sends the real-time operation data to the data intelligent processing module; The data intelligent processing module pre-processes the real-time operation data and then performs intelligent processing, and sends the intelligently processed data to the data analysis and storage module; The data analysis and storage module performs feature analysis on the intelligently processed data to obtain real-time data features, stores the real-time data features in a historical data feature library, and sends them to the fault prediction model module; The fault prediction model module is trained and optimized based on the historical data features in the historical data feature library to obtain an optimized fault prediction model, and the real-time data features are processed by the optimized fault prediction model to obtain a fault prediction result, and the fault prediction result is sent to the maintenance decision support module and the early warning notification module; The maintenance decision support module generates maintenance suggestions based on the fault prediction results and sends the maintenance suggestions to the early warning notification module; The early warning notification module generates early warning information based on fault prediction results and maintenance suggestions, notifies maintenance personnel of the early warning information, and implements predictive maintenance.

Images