Triple weighting equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature

Abstract

The invention discloses a lubricating vibration and temperature-based triple weighting equipment comprehensive state evaluation and prediction method, which comprises the following steps of: integrating data from different sources through a fusion technology, and acquiring lubricating, temperature and vibration signals in real time by using a sensor; the data is stored in a database after being processed; filtering and singular point elimination are carried out on the vibration signals, and normalization processing is carried out on all the signals; key parameters are extracted from time domain analysis, and the weight of each signal is determined by using an analytic hierarchy process; the comprehensive evaluation model adopts the data subjected to Z-score standardization processing, combines the weight to calculate an overall state evaluation value of the equipment, and compares the overall state evaluation value with a preset value to evaluate the state of the equipment; and finally, training a prediction model based on historical data by using a long-short-term memory network, optimizing model parameters through cross validation, and predicting the future state of the equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment evaluation and prediction, and more specifically, to a triple-weighted equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature. Background Art

[0002] With the rapid development of modern industry, the operating status of mechanical equipment has an increasingly significant impact on production efficiency and safety; equipment failure will not only lead to production interruptions and increase maintenance costs, but may also cause serious safety accidents; therefore, equipment status monitoring and fault diagnosis technology has become an important research direction in the industrial field.

[0003] In recent years, with the rapid development of sensor technology, data processing technology and artificial intelligence, multi-parameter comprehensive monitoring methods have gradually become a research hotspot; multi-parameter comprehensive monitoring can more comprehensively and accurately evaluate the operating status of equipment by simultaneously collecting and analyzing multiple key parameters; however, how to effectively integrate multiple parameters and conduct comprehensive evaluation remains a challenge.

[0004] However, there are still some shortcomings in its actual use. For example, traditional equipment status monitoring methods mainly rely on a single parameter. These methods have certain limitations in practical applications. The monitoring of a single parameter often cannot fully reflect the comprehensive status of the equipment, which can easily lead to misjudgment or missed judgment. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a triple-weighted equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Step A1: perform data fusion and unified processing on different systems of the existing system; for the new system, use sensors to collect lubrication signals, temperature signals and vibration signals;

[0008] Step A2: performing missing value and normalization processing on the collected lubrication signal, temperature signal and vibration signal;

[0009] Step A3: extract key features from the processed lubrication signal, temperature signal and vibration signal using time domain analysis;

[0010] Step A4: Determine the weights of the lubrication signal, the temperature signal, and the vibration signal by constructing a hierarchical model;

[0011] Step A5: construct a comprehensive evaluation model based on the monitoring data of lubrication state, vibration state and temperature state;

[0012] Step A6: Based on the comprehensive evaluation results, predict the future state of the equipment.

[0013] Preferably, in step A1, data is obtained from different systems in the existing system; the extracted data is corrected and deduplicated, and the data cleaning process includes removing duplicate data, filling missing values, and standardizing data formats; the data formats from different sources are standardized, and the standardized data are analyzed on a unified platform; this includes format conversion of structured data and parsing and extraction of unstructured data; the processed data is stored in a database; the database is a centralized data storage system for storing and managing data extracted from multiple heterogeneous data sources;

[0014] For new systems, the lubrication signal acquisition uses a capacitive lubricant real-time online monitoring sensor to obtain the lubrication signal by measuring the capacitance value of the lubricant;

[0015] Use a temperature sensor to collect temperature signals through the circuit board;

[0016] The vibration signal is collected using an eddy current sensor and a high-precision signal converter, and the vibration signal is collected in real time through the DH5928W data collector.

[0017] Preferably, in step A2, the vibration signal needs to be subjected to anti-aliasing filtering, narrow-band filtering, singular point elimination and missing value processing to improve the signal-to-noise ratio of the data; and then the vibration signal, lubrication signal and temperature signal are normalized to convert data of different dimensions into the same dimension;

[0018] Identify missing values ​​in the data and use interpolation to fill in the missing data.

[0019] Preferably, in step A3, a signal to be analyzed is selected from the processed lubrication, temperature and vibration signals, and then basic statistical characteristics, absolute fluctuation amplitude and root mean square value are calculated, and key characteristics of the lubrication signal, temperature signal and vibration signal are extracted by calculation through the basic statistical characteristics, absolute fluctuation amplitude and root mean square value;

[0020] The key features extracted from lubrication signals include the viscosity, particle size and moisture content of the lubricating oil; the key features extracted from temperature signals include the temperature change rate and fluctuation range; the key features extracted from vibration signals include the spectrum, amplitude and phase of the vibration.

[0021] Preferably, in step A4, the problem is decomposed into three levels, namely, a target level, a criterion level and a solution level;

[0022] Among them, the target layer determines the importance assessment of the signal;

[0023] The criterion layer includes lubrication signal, temperature signal, and vibration signal;

[0024] If there are multiple plans to be compared in the plan layer, you can list the different plans at this layer;

[0025] Then compare the signals in the criterion layer in pairs and construct a judgment matrix; use the 1-9 notation method to determine the degree of comparison;

[0026] Among them, 1 means that both are equally important, 3 means that one is 60% important, 5 means that one is 70% important, 7 means that one is 80% important, and 9 means that one is 90% important; 2, 4, 6, and 8 are between the above two scores; 2 is between 1 and 3; 4 is between 3 and 5, 6 is between 5 and 7, and 8 is between 7 and 9;

[0027] The calculation method for constructing the judgment matrix is ​​as follows:

[0028] Among them, A represents the judgment matrix, a ij It is expressed as the importance score of signal i relative to signal j;

[0029] After constructing the judgment matrix, the sum of the i-th row in the judgment matrix is ​​calculated. The specific calculation method of the sum of the i-th row in the judgment matrix is:

[0030] Among them, B i It is represented as the sum of the i-th row in the judgment matrix, a ij It is represented as the importance score of signal i relative to signal j, and b is represented as the total number of signals;

[0031] Then calculate the normalized judgment matrix. The calculation method of the normalized judgment matrix is ​​as follows:

[0032] Among them, C ij Represented as the normalized judgment matrix, a ij It is expressed as the importance score of signal i relative to signal j, B i It is represented as the sum of the i-th row in the judgment matrix;

[0033] The average value of each row of the normalized judgment matrix is ​​the weight of each factor, and the specific method for calculating the weight is:

[0034] Among them, w i Represented as the weight value of the i-th signal, C ij It is represented as the normalized judgment matrix, and b is represented as the total number of signals;

[0035] After the weights are calculated, the judgment matrix is ​​checked for consistency. The calculation method of the consistency ratio is as follows:

[0036] Among them, D represents the consistency ratio, E represents the consistency index, and F represents the random consistency index;

[0037] Among them, the random consistency index is obtained by looking up the matrix order table;

[0038] The calculation method of consistency index is as follows:

[0039] Among them, E represents the consistency index, α max It is represented as the maximum eigenvalue of the judgment matrix, and b is represented as the total number of signals;

[0040] If the calculated consistency ratio is less than 0.1, the judgment matrix is ​​of acceptable consistency; if the calculated consistency ratio is greater than 0.1, the judgment matrix is ​​of unacceptable consistency and needs to be readjusted.

[0041] Preferably, in step A5, each state indicator is standardized; since the monitoring data units and dimensions of the lubrication state, vibration state and temperature state are different, directly performing weighted calculation may lead to distorted results; therefore, each state indicator needs to be Z-score standardized, and the Z-score standardization method converts the value of each state indicator into a standard normal distribution;

[0042] The standardized state index values ​​are multiplied by the corresponding weighting coefficients, and then weighted summed to obtain the overall state evaluation value of the equipment. The specific calculation method of the overall state evaluation value is:

[0043] Q=w1G+w2H+w3K, where Q represents the overall state evaluation value, G represents the standardized lubrication state, H represents the standardized vibration state, K represents the standardized temperature state, w1 represents the weight of the lubrication state, w2 represents the weight of the vibration state, and w3 represents the weight of the temperature state.

[0044] Preferably, in step A6, a prediction model is constructed using a long short-term memory network method; parameters need to be continuously adjusted during model training to improve prediction accuracy; the performance of the model is evaluated by cross-validation and other methods, and indicators such as accuracy and recall are calculated; the model is optimized based on the evaluation results, the algorithm parameters are adjusted, and the model structure is replaced to improve the accuracy and stability of the prediction; the future state of the equipment is predicted based on the output results of the prediction model; and corresponding maintenance plans and operation plans are formulated in combination with the operating conditions and maintenance strategies of the equipment to ensure that the equipment operates in the best condition.

[0045] Technical effects and advantages of the present invention:

[0046] The present invention first obtains and fuses data from different systems, uses sensors to collect lubrication, temperature and vibration signals, and stores them in a database after cleaning, deduplication and standardization. Then, the missing value processing and normalization of the signal are performed to extract key features such as viscosity, temperature change rate and vibration spectrum. Then, the weight of each signal is determined by the hierarchical analysis method, and a comprehensive evaluation model is constructed. The overall status evaluation value of the equipment is calculated through the data after Z-score standardization. Finally, a prediction model is constructed based on a long short-term memory network, and the future status of the equipment is predicted after evaluating the model performance, and a maintenance plan is formulated. Through the above method to monitor the three types of parameters and adopt weighted processing and time series analysis methods, this method can achieve a comprehensive and integrated evaluation of the equipment status, improve the operating efficiency and safety of the equipment, reduce misjudgment and predict future development. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of 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.

[0049] See also Figure 1 As shown, the present invention provides a triple weighted equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature, and the method is:

[0050] Step A1: perform data fusion and unified processing on different systems of the existing system; for the new system, use sensors to collect lubrication signals, temperature signals and vibration signals;

[0051] In the step A1, data is obtained from different systems in the existing system; the extracted data is corrected and deduplicated, and the data cleaning process includes removing duplicate data, filling missing values, and standardizing data formats; the data formats from different sources are standardized, and the standardized data are analyzed on a unified platform; this includes format conversion of structured data and parsing and extraction of unstructured data; the processed data is stored in a database; the database is a centralized data storage system for storing and managing data extracted from multiple heterogeneous data sources;

[0052] For new systems, the lubrication signal acquisition uses a capacitive lubricant real-time online monitoring sensor to obtain the lubrication signal by measuring the capacitance value of the lubricant;

[0053] Use a temperature sensor to collect temperature signals through the circuit board;

[0054] The vibration signal is collected using an eddy current sensor and a high-precision signal converter, and the vibration signal is collected in real time through the DH5928W data collector.

[0055] Step A2: performing missing value and normalization processing on the collected lubrication signal, temperature signal and vibration signal;

[0056] In step A2, the vibration signal needs to be subjected to anti-aliasing filtering, narrow-band filtering, singular point elimination and missing value processing to improve the signal-to-noise ratio of the data; the vibration signal, lubrication signal and temperature signal are then normalized to convert data of different dimensions into the same dimension;

[0057] Identify missing values ​​in the data and use interpolation to fill in the missing data.

[0058] Step A3: extract key features from the processed lubrication signal, temperature signal and vibration signal using time domain analysis;

[0059] In the step A3, a signal to be analyzed is selected from the processed lubrication, temperature and vibration signals, and then basic statistical characteristics, absolute fluctuation amplitude and root mean square value are calculated, and key features of the lubrication signal, temperature signal and vibration signal are extracted by calculation through the basic statistical characteristics, absolute fluctuation amplitude and root mean square value;

[0060] The key features extracted from lubrication signals include the viscosity, particle size and moisture content of the lubricating oil; the key features extracted from temperature signals include the temperature change rate and fluctuation range; the key features extracted from vibration signals include the spectrum, amplitude and phase of the vibration;

[0061] For housing vibration, narrow-band feature extraction is used according to the type of components under test, and the alarm threshold of each narrow-band vibration energy is defined;

[0062] For relative vibration, the characteristic frequency vector and precession component are used for feature extraction, and the alarm threshold of each characteristic value is defined; for the axis position, the eccentricity and azimuth are used as characteristic parameters, and their alarm thresholds are defined;

[0063] Step A4: Determine the weights of the lubrication signal, the temperature signal, and the vibration signal by constructing a hierarchical model;

[0064] In step A4, the problem is decomposed into three levels, namely, the target level, the criterion level and the solution level;

[0065] Among them, the target layer determines the importance assessment of the signal;

[0066] The criterion layer includes lubrication signal, temperature signal, and vibration signal;

[0067] If there are multiple plans to be compared in the plan layer, you can list the different plans at this layer;

[0068] Then compare the signals in the criterion layer in pairs and construct a judgment matrix; use the 1-9 notation method to determine the degree of comparison;

[0069] Among them, 1 means that both are equally important, 3 means that one is 60% important, 5 means that one is 70% important, 7 means that one is 80% important, and 9 means that one is 90% important; 2, 4, 6, and 8 are between the above two scores; 2 is between 1 and 3; 4 is between 3 and 5, 6 is between 5 and 7, and 8 is between 7 and 9;

[0070] The calculation method for constructing the judgment matrix is ​​as follows:

[0071] Among them, A represents the judgment matrix, a ij It is expressed as the importance score of signal i relative to signal j;

[0072] After constructing the judgment matrix, the sum of the i-th row in the judgment matrix is ​​calculated. The specific calculation method of the sum of the i-th row in the judgment matrix is:

[0073] Among them, B i It is represented as the sum of the i-th row in the judgment matrix, a ij It is represented as the importance score of signal i relative to signal j, and b is represented as the total number of signals;

[0074] Then calculate the normalized judgment matrix. The calculation method of the normalized judgment matrix is ​​as follows:

[0075] Among them, C ij Represented as the normalized judgment matrix, a ij It is expressed as the importance score of signal i relative to signal j, B i It is represented as the sum of the i-th row in the judgment matrix;

[0076] The average value of each row of the normalized judgment matrix is ​​the weight of each factor, and the specific method for calculating the weight is:

[0077] Among them, w i Represented as the weight value of the i-th signal, C ijIt is represented as the normalized judgment matrix, and b is represented as the total number of signals;

[0078] After the weights are calculated, the judgment matrix is ​​checked for consistency. The calculation method of the consistency ratio is as follows:

[0079] Among them, D represents the consistency ratio, E represents the consistency index, and F represents the random consistency index;

[0080] Among them, the random consistency index is obtained by looking up the matrix order table;

[0081] The calculation method of consistency index is as follows:

[0082] Among them, E represents the consistency index, α max It is represented as the maximum eigenvalue of the judgment matrix, and b is represented as the total number of signals;

[0083] If the calculated consistency ratio is less than 0.1, the judgment matrix is ​​of acceptable consistency; if the calculated consistency ratio is greater than 0.1, the judgment matrix is ​​of unacceptable consistency and needs to be readjusted;

[0084] Step A5: construct a comprehensive evaluation model based on the monitoring data of lubrication state, vibration state and temperature state;

[0085] In step A5, each state indicator is standardized; since the monitoring data units and dimensions of the lubrication state, vibration state and temperature state are different, directly performing weighted calculation may lead to distorted results; therefore, each state indicator needs to be Z-score standardized, and the Z-score standardization method converts the value of each state indicator into a standard normal distribution;

[0086] The standardized state index values ​​are multiplied by the corresponding weighting coefficients, and then weighted summed to obtain the overall state evaluation value of the equipment. The specific calculation method of the overall state evaluation value is:

[0087] Q=w1G+w2H+w3K, where Q represents the overall state evaluation value, G represents the standardized lubrication state, H represents the standardized vibration state, K represents the standardized temperature state, w1 represents the weight of the lubrication state, w2 represents the weight of the vibration state, and w3 represents the weight of the temperature state;

[0088] The calculated overall status evaluation value is compared with the preset overall status evaluation threshold. If the calculated overall status evaluation value is greater than the preset overall status evaluation threshold, it indicates that the operating status of the equipment is gradually deteriorating; if the calculated overall status evaluation value is less than the preset overall status evaluation threshold, it indicates that the operating status of the equipment is normal and monitoring continues;

[0089] The calculation method of the preset overall status evaluation threshold is specifically as follows:

[0090] Where Q0 represents the preset overall state evaluation threshold, n represents the calculated historical overall state evaluation value, and in order to reduce the error of the calculated value, the value of n is n≥5; G q It is expressed as the lubrication state after normalization of the historical data for the qth time, H q It is expressed as the standardized vibration state of the historical data for the qth time, K q It is represented by the standardized temperature state of the historical data for the qth time;

[0091] Verify and optimize the comprehensive evaluation model; verify the accuracy and reliability of the comprehensive evaluation model through actual application and comparison with historical data; if deviations are found in the model, further improve the performance of the model by adjusting the weighting coefficients or optimizing the model structure;

[0092] Among them, the lubrication state is evaluated by analyzing the viscosity, moisture, impurities, acid value, alkalinity and other parameters of the lubricating oil; the changes in these parameters can directly reflect the performance of the lubricating oil and the operating status of the equipment; for example, the change in viscosity can reflect the degree of oxidation and contamination of the lubricating oil, the increase in moisture and impurities can indicate that the lubricating oil is contaminated, and the changes in acid value and alkalinity can reflect the anti-oxidation and anti-corrosion properties of the lubricating oil;

[0093] By analyzing the state changes of the preset characteristic values ​​of mechanical vibration, the overall operating status level of the equipment and the possible fault types, severity and development trends can be evaluated. The definition of the characteristic value depends on the type of equipment being measured and the type of signal. For mechanical vibration, the amplitudes of different measurement methods reflect the different types of force, and force is the main cause of vibration. The amplitude changes of characteristic components are related to the type and degree of faults and are the main information source for diagnosing faults. The evaluation parameters include absolute value, rate of change and distribution.

[0094] Temperature monitoring is an important part of equipment status assessment, which can effectively reflect the operating status of the equipment; especially the temperature changes of key parts such as bearings, motors, gearboxes, hydraulic systems, etc., which can reflect the operating status and potential faults of the equipment; the evaluation parameters include absolute temperature, temperature rise, temperature change rate and temperature distribution; the monitoring of absolute temperature can reflect the thermal load of the equipment, and too high or too low absolute temperature may indicate that the equipment is faulty. The monitoring of temperature rise can reflect the heat dissipation of the equipment, and too high or too low temperature rise may indicate that the equipment is faulty; the monitoring of temperature change rate can reflect the thermal stability of the equipment, and too large or too small temperature change rate indicates that the equipment is faulty; the monitoring of temperature distribution can reflect the thermal balance of the equipment, and uneven temperature distribution may indicate that the equipment is faulty.

[0095] Step A6: Based on the comprehensive evaluation results, predict the future state of the equipment.

[0096] In step A6, a prediction model is constructed using a long short-term memory network method; parameters need to be continuously adjusted during model training to improve prediction accuracy; the performance of the model is evaluated by cross-validation and other methods, and indicators such as accuracy and recall are calculated; the model is optimized based on the evaluation results, the algorithm parameters are adjusted, and the model structure is replaced to improve the accuracy and stability of the prediction; the future state of the equipment is predicted based on the output results of the prediction model; and corresponding maintenance plans and operation plans are formulated in combination with the operation status and maintenance strategy of the equipment to ensure that the equipment operates in the best state;

[0097] Among them, the method of predictive model training is specifically as follows:

[0098] Step A1: Collect historical power grid equipment data, including lubrication status, vibration status, and temperature status of the equipment;

[0099] Step A2: Clean the collected data, remove outliers and missing values, and standardize or normalize the data; divide the data into training set, validation set, and test set;

[0100] Step A3: construct a prediction model using a long short-term memory network method, the model includes the settings of an input layer, a hidden layer, and an output layer;

[0101] Step A4: Perform forward propagation, calculate loss, back propagation and parameter update by iterating the training data;

[0102] Step A5: Select the loss function mean square error and the coefficient of determination to evaluate the prediction performance of the model, and select the optimizer to update the model weights;

[0103] Step A6: Calculate and predict the overall state evaluation value through the trained model, and judge the predicted overall state evaluation value and the overall state evaluation value;

[0104] Among them, if the calculated predicted overall state evaluation value is equal to the overall state evaluation value, there is no need to continue iterative training; if the calculated predicted overall state evaluation value is not equal to the overall state evaluation value, continue iterative training until the predicted overall state evaluation value is equal to the overall state evaluation value.

[0105] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A triple weighted equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature, characterized in that: include: Step A1: perform data fusion and unification processing on different systems of the existing system; For new systems, use sensors to collect lubrication signals, temperature signals, and vibration signals; Step A2: performing missing value and normalization processing on the collected lubrication signal, temperature signal and vibration signal; Step A3: extract key features from the processed lubrication signal, temperature signal and vibration signal using time domain analysis; Step A4: Determine the weights of the lubrication signal, the temperature signal, and the vibration signal by constructing a hierarchical model; Step A5: construct a comprehensive evaluation model based on the monitoring data of lubrication state, vibration state and temperature state; Step A6: Based on the comprehensive evaluation results, predict the future state of the equipment.

2. The triple weighted equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature according to claim 1 is characterized in that: In the step A1, data is obtained from different systems in the existing system; the extracted data is corrected and deduplicated, and the data cleaning process includes removing duplicate data, filling missing values, and standardizing data formats; the data formats from different sources are standardized, and the standardized data are analyzed on a unified platform; this includes format conversion of structured data and parsing and extraction of unstructured data; the processed data is stored in a database; A database is a centralized data storage system used to store and manage data extracted from multiple heterogeneous data sources; For new systems, the lubrication signal acquisition uses a capacitive lubricant real-time online monitoring sensor to obtain the lubrication signal by measuring the capacitance value of the lubricant; Use a temperature sensor to collect temperature signals through the circuit board; The vibration signal is collected using an eddy current sensor and a high-precision signal converter, and the vibration signal is collected in real time through the DH5928W data collector.

3. The method for evaluating and predicting the equipment's comprehensive state based on triple weighting of lubrication vibration and temperature according to claim 1 is characterized in that: In step A2, the vibration signal needs to be subjected to anti-aliasing filtering, narrow-band filtering, singular point elimination and missing value processing to improve the signal-to-noise ratio of the data; the vibration signal, lubrication signal and temperature signal are then normalized to convert data of different dimensions into the same dimension; Identify missing values ​​in the data and use interpolation to fill in the missing data.

4. The method for evaluating and predicting the equipment's comprehensive state based on triple weighting of lubrication vibration and temperature according to claim 1 is characterized in that: In the step A3, a signal to be analyzed is selected from the processed lubrication, temperature and vibration signals, and then basic statistical characteristics, absolute fluctuation amplitude and root mean square value are calculated, and key features of the lubrication signal, temperature signal and vibration signal are extracted by calculation through the basic statistical characteristics, absolute fluctuation amplitude and root mean square value; The key features extracted from lubrication signals include the viscosity, particle size and moisture content of the lubricating oil; the key features extracted from temperature signals include the temperature change rate and fluctuation range; the key features extracted from vibration signals include the spectrum, amplitude and phase of the vibration.

5. The triple weighted equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature according to claim 1 is characterized in that: In step A4, the problem is decomposed into three levels, namely, the target level, the criterion level and the solution level; The signals in the criterion layer are compared pairwise and a judgment matrix is ​​constructed; the degree of comparison is determined using the 1-9 notation method; Among them, 1 means that both are equally important, 3 means that one is 60% important, 5 means that one is 70% important, 7 means that one is 80% important, and 9 means that one is 90% important; 2, 4, 6, and 8 are between the above two scores; 2 is between 1 and 3; 4 is between 3 and 5, 6 is between 5 and 7, and 8 is between 7 and 9; The calculation method for constructing the judgment matrix is ​​as follows: Among them, A represents the judgment matrix, a ij It is expressed as the importance score of signal i relative to signal j; After constructing the judgment matrix, the sum of the i-th row in the judgment matrix is ​​calculated. The specific calculation method of the sum of the i-th row in the judgment matrix is: Among them, B i It is represented as the sum of the i-th row in the judgment matrix, a ij It is represented as the importance score of signal i relative to signal j, and b is represented as the total number of signals.

6. The method for evaluating and predicting the equipment's comprehensive state based on triple weighting of lubrication vibration and temperature according to claim 5 is characterized in that: Calculate the normalized judgment matrix. The calculation method of the normalized judgment matrix is ​​as follows: Among them, C ij Represented as the normalized judgment matrix, a ij It is expressed as the importance score of signal i relative to signal j, B i It is represented as the sum of the i-th row in the judgment matrix; The average value of each row of the normalized judgment matrix is ​​the weight of each factor, and the specific method for calculating the weight is: Among them, w i Represented as the weight value of the i-th signal, C ij It is represented as the normalized judgment matrix, and b is represented as the total number of signals; After the weights are calculated, the judgment matrix is ​​checked for consistency. The calculation method of the consistency ratio is as follows: Among them, D represents the consistency ratio, E represents the consistency index, and F represents the random consistency index.

7. The method for evaluating and predicting the equipment's comprehensive state based on triple weighting of lubrication vibration and temperature according to claim 1 is characterized by: In step A5, each state indicator is standardized; since the monitoring data units and dimensions of the lubrication state, vibration state and temperature state are different, directly performing weighted calculation may lead to distorted results; therefore, each state indicator needs to be Z-score standardized, and the Z-score standardization method converts the value of each state indicator into a standard normal distribution; The specific calculation method of the overall status evaluation value is as follows: Q=w1G+w2H+w3K, where Q represents the overall state evaluation value, G represents the standardized lubrication state, H represents the standardized vibration state, K represents the standardized temperature state, w1 represents the weight of the lubrication state, w2 represents the weight of the vibration state, and w3 represents the weight of the temperature state; The calculated overall status evaluation value is compared with the preset overall status evaluation threshold. If the calculated overall status evaluation value is greater than the preset overall status evaluation threshold, it indicates that the operating status of the equipment is gradually deteriorating; if the calculated overall status evaluation value is less than the preset overall status evaluation threshold, it indicates that the operating status of the equipment is normal and monitoring should continue.

8. The triple weighted equipment comprehensive state evaluation and prediction method based on lubrication vibration and temperature according to claim 1 is characterized by: In step A6, a prediction model is constructed using a long short-term memory network method; parameters need to be continuously adjusted during model training; the performance of the model is evaluated by cross-validation and other methods, and the accuracy and recall rate indicators are calculated; the model is optimized based on the evaluation results, the algorithm parameters are adjusted, and the model structure is replaced; based on the output results of the prediction model, the future state of the equipment is predicted; and corresponding maintenance plans and operation plans are formulated in combination with the operating conditions and maintenance strategies of the equipment.

Images