A method for predicting phase shifter oil performance degradation based on real-time monitoring data

Abstract

The present application provides a method for predicting the performance degradation of a phase shifter oil product based on real-time monitoring data, including: using a constructed wear particle prediction model to realize online prediction of the corresponding oil wear particle parameter distribution for the input operation data of a specified part; comparing the prediction result with the actual wear particle detection data, analyzing the error between the two, and evaluating the wear prediction effect of the model at different parts; extracting lubricating oil samples and detecting elemental composition at various parts of the phase shifter based on the wear prediction effect of the wear particle prediction model at different parts, and determining the source of worn parts; analyzing the use environment and mode of vulnerable parts, proposing phase shifter status improvement measures, and realizing phase shifter operation optimization; synchronously monitoring the dynamic operation information of the phase shifter and the wear condition of the lubrication system oil product, and verifying the effects of phase shifter and oil circuit status monitoring and component maintenance optimization.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for predicting phase shifter oil performance degradation based on real-time monitoring data. Background Art

[0002] In modern industrial applications, especially in the use of precision equipment such as phase regulators, the durability and performance maintenance of equipment components is an important and complex issue. These equipment usually operate in a changing and complex working environment and face multiple challenges. In actual operation, the loads borne by various parts of the equipment often change. These changes may be caused by production requirements, changes in operating modes or external environmental factors. This constant change in load poses a challenge to the durability and performance stability of the equipment. During intermittent high-speed operation, there are significant differences in the speed and power experienced by various parts and parts. This difference affects the wear pattern of the components, which in turn affects the overall performance and safety of the equipment. During the drastic cold and hot cycle conversion, various parts of the equipment are subjected to different degrees of thermal stress. These thermal stresses may cause changes in material properties, thereby affecting the wear rate and pattern of the components. Due to the above factors, there are significant differences in the wear pattern and degree between equipment components. This difference not only affects the function of individual components, but also affects the operating efficiency and maintenance cycle of the entire equipment. During the operation of the equipment, the generation of wear particles and the degradation of oil composition are key indicators for monitoring the health of the equipment. Changes in these particles and chemical composition can reflect the degree of wear and potential failure risks of the equipment. Therefore, existing technologies face challenges in handling equipment maintenance and performance monitoring under these complex and changing conditions. These challenges include, but are not limited to, different wear modes caused by load, speed, and temperature changes, as well as the monitoring and analysis of wear particles and oil degradation. In order to effectively address these challenges, a method that can comprehensively monitor and analyze these complex factors is needed to optimize equipment operating efficiency and maintenance strategies. Summary of the invention

[0003] The present invention provides a method for predicting phase shifter oil performance degradation based on real-time monitoring data, which mainly includes:

[0004] Collect online monitoring information of various oil circuit components including operating data of load parameters, speed and temperature, and obtain data sample sets as input of condenser state through sampling and filtering methods; regularly extract lubricating oil samples from multiple positions of condenser oil circuit, use electron microscope and energy spectrum analysis technology to detect the quality characteristic data of wear particles in each sample, and construct wear particle data set corresponding to the position; perform data statistical feature calculation and feature mapping based on real-time dynamic monitoring data of condenser parts and wear particle detection results of various parts of oil circuit, and construct wear particle prediction model; use the constructed wear particle prediction model to realize online prediction of corresponding oil wear particle parameter distribution for input operating data of specified parts; compare the prediction results with actual wear particle detection data, analyze the errors between the two, and evaluate the wear prediction effect of the model at different parts; extract lubricating oil samples and detect elemental composition of various parts of condenser according to the wear prediction effect of wear particle prediction model at different parts, and judge the source of worn parts; analyze the use environment and mode of vulnerable parts, propose condenser state improvement measures, and realize condenser operation optimization; synchronously monitor condenser dynamic operation information and lubrication system oil wear condition, and verify the effect of condenser and oil circuit state monitoring and component maintenance optimization.

[0005] As a preferred solution, the online monitoring information of each component of the oil circuit containing the operating data of load parameters, speed and temperature is collected, and a data sample set is obtained as the phase shifter state input by sampling and filtering methods, including:

[0006] Continuously monitor each component of the oil circuit online and record load parameters, speed and temperature data; perform preliminary sampling of the monitoring data, perform initial sampling according to the first preset number, and obtain a data sample set; use a low-pass filter to process the data sample set to remove noise and interference; if the load parameter fluctuation is greater than the preset threshold, increase the sampling frequency and obtain a second preset number of data samples; if the temperature is detected to be higher than the preset threshold, expand the filter processing range to eliminate abnormal data caused by temperature fluctuations; based on the number of data samples, determine whether the preset sample number requirement is met. If the number of data samples is insufficient, further increase the number of samplings; based on the cleaned data set, determine the current state of the phase regulator. If the number of data samples is insufficient, increase the number of samplings; for the case where the speed is greater than the preset speed threshold or the load is greater than the preset load threshold, adjust the sampling strategy and filtering parameters to obtain the processed real-time data sample of the phase regulator in the current stable state; extract key features from the real-time monitoring data and the stored historical behavior data, including load, speed, and temperature; use the cosine similarity method to compare the similarity between the real-time data and the historical data, and extract the historical behavior data whose similarity to the current state is greater than the preset threshold.

[0007] As a preferred solution, the lubricating oil samples are regularly extracted from multiple positions of the phase shifter oil circuit, and the quality characteristic data of the wear particles in each sample are detected by using electron microscopy and energy spectrum analysis technology to construct a wear particle data set corresponding to the position, including:

[0008] According to the phase shifter structure diagram, the positions of various parts of the oil circuit are determined; a fixed cycle is set to extract lubricating oil samples from various designated positions of the oil circuit within a predetermined time interval; the extracted lubricating oil samples are imaged and observed through an electron microscope to obtain the morphological characteristic images of the wear particles; the size characteristics, shape characteristics, and surface property characteristics of the wear particles are determined based on the morphological characteristic images of the wear particles, the size characteristics include length, width and depth, the shape characteristics include round, angular, irregular, and the surface property characteristics include smooth, rough, and cracked; X-ray energy spectrum analysis technology is used to determine the elemental composition of lubricating oil wear particles; the chemical properties and chemical stability of the particles are evaluated based on the elemental composition, and the chemical properties of the particles, including Including metals, non-metals, and oxides; construct a wear particle data set based on the measured size, shape, surface quality and chemical property data; judge the difference in parameter values ​​of different parts based on the wear particle data set; if the difference is less than the preset threshold, adjust the sampling distance between the parts, increase the sample types, and obtain a wear particle data set with a difference greater than the preset threshold; use the decision tree algorithm to train the model based on the wear particle data set with a difference greater than the preset threshold to determine the wear type of different parts. If the parts cannot be distinguished, increase the number of test samples; estimate the theoretical wear type based on the phase shifter structure and operating parameters; compare the estimated results with the actual test results, analyze the attribution of the deviation, and optimize the data set construction plan based on the comparison results.

[0009] As a preferred solution, the method of calculating and mapping data statistics characteristics and constructing a wear particle prediction model based on the real-time dynamic monitoring data of the phase shifter parts and the wear particle detection results of various parts of the oil circuit includes:

[0010] Obtain real-time dynamic monitoring data of each part of the phase shifter, including load, speed, and temperature; calculate statistical characteristics of the dynamic data of each part, including mean, standard deviation, and skewness; obtain the wear particle detection results of the oil products in each part of the oil circuit, and extract the key characteristics of the wear particles, including length distribution and quantity ratio; map the real-time dynamic monitoring data with the corresponding wear particle data for each part, and the real-time monitoring data includes load, speed, and temperature; integrate the statistical characteristics of the dynamic data and the wear particle characteristics into a matrix, use the support vector machine algorithm to train the integrated feature matrix, obtain the wear particle prediction model, and determine the parameter relationship and feature weight; test whether the correlation relationship generated by the model is consistent with the operating conditions of the actual parts; if the model effect does not meet expectations, adjust the feature selection, update the model structure, and redefine the parameters; evaluate the prediction effect of different parts according to the model error value, if the error is greater than the preset threshold, increase the sample size of the corresponding part, and retrain the model; obtain verification data to evaluate whether the prediction effect of the model meets the standard; if the prediction effect meets the standard, apply the wear particle prediction model to the actual phase shifter state monitoring and wear prediction.

[0011] As a preferred solution, the wear particle prediction model constructed is used to realize online prediction of the corresponding oil wear particle parameter distribution for the input operation data of the designated part, including:

[0012] The real-time operating data of the specified part of the phase regulator is obtained to determine whether the data quality meets the input requirements of the wear particle prediction model; if the data is abnormal or missing, the median filtering method is used to compensate and correct it; based on the processed real-time dynamic data, the wear particle prediction model is used to predict the length range and quantity ratio distribution of the wear particles in the specified part; the predicted wear parameter distribution is compared with the preset normal wear range; if the predicted result exceeds the normal wear range, the auxiliary diagnosis module is started to determine the cause of the abnormality. The auxiliary diagnosis module matches the operating parameters with the historical operating data to determine whether the predicted result exceeds the normal wear range due to abnormal operation, and provides corresponding processing suggestions; if no abnormal operation is detected, the quality of the dynamic data will be rechecked, and correction and recalculation will be performed in the event of errors; through the wear particle prediction model, the verified wear particle parameter distribution of the specified part is obtained, and the wear condition under the current state is evaluated.

[0013] As a preferred solution, the prediction result is compared with the actual detection data of wear particles, the error between the two is analyzed, and the wear prediction effect of the model at different parts is evaluated, including:

[0014] According to the wear particle prediction model, the predicted wear parameter distribution results are obtained, including the quantitative proportion distribution of particle sizes; lubricating oil samples are collected from the corresponding phase shifter parts predicted by the wear particle prediction model, and the particles are measured to obtain the actual wear particle data; the number of particles in the same size segment is compared between the predicted results of the wear particle prediction model and the measured actual wear particle data; the error between the predicted number and the actual number of each size segment is calculated, and the size segment with the largest error is identified; the model prediction results of different phase shifter parts are evaluated, and the average prediction error and maximum error of each part are calculated; according to the error value of each part, the model prediction effects of different parts are arranged, and the parts with the worst prediction effect are identified; for parts with errors greater than the preset threshold, the cause of the prediction deviation is determined, and the model is adjusted. The attribution of the deviation includes insufficient data, improper feature selection, and limitations of the model itself; the adjusted model is retrained, and the error fluctuations of each part are continuously monitored, and the model is optimized in a timely manner.

[0015] As a preferred solution, according to the wear prediction effect of the wear particle prediction model at different parts, the lubricating oil samples of various parts of the condenser are extracted and the element composition is tested to determine the source of the worn parts, including:

[0016] According to the prediction results of the wear particle prediction model of each part of the phase shifter, the parts where the prediction effect error is greater than the preset threshold are determined; from the parts where the effect error is greater than the preset threshold, a predetermined number of batches of lubricating oil samples are extracted to obtain data at different time points; wear particles are separated from the collected lubricating oil samples, and the main element components are detected using X-ray energy spectrum analysis technology; based on the element detection results, a time series change curve of each element in the particles is constructed, and it is determined whether there is an element component increase greater than the preset threshold or a decrease greater than the preset threshold; for elements with an element component increase greater than the preset threshold, the material composition and structure of the component are associated, and possible wear parts are preliminarily matched to determine the candidates for the wear source component; the historical wear and replacement records of the wear parts are checked to verify whether the source of the judged wear parts is correct; the source of the wear parts where the prediction effect error of the wear particle prediction model is greater than the preset threshold is confirmed, and the specific location code of the component is obtained; it also includes: according to the historical experimental data of the phase shifter, the connection between the lubricating oil wear particle generation rate and the operating conditions is identified, the wear trend is predicted in real time, and the state optimization and preventive maintenance of the lubrication system are guided.

[0017] The method of identifying the relationship between the generation rate of lubricating oil wear particles and operating conditions based on the historical experimental data of the phase shifter, making real-time predictions of wear trends, and guiding the state optimization and preventive maintenance of the lubrication system specifically includes:

[0018] Based on historical experimental data, the potential relationship between the wear particle generation rate and temperature T, pressure P, and mechanical stress S is identified, and a relationship model between the wear particle generation rate and temperature, pressure, and stress is constructed. , predict the wear particle generation rate in real time, where a, b, c, d, and n are model parameters determined by historical experimental data; evaluate the model's goodness of fit and prediction accuracy. Integrate the relationship model between the wear particle generation rate and temperature, pressure, and stress into the existing lubricant monitoring system of the phase shifter to generate a real-time prediction of the wear particle generation rate; optimize the state of the phase shifter lubrication system and perform preventive maintenance based on the wear particle generation rate. Test the performance of the model in an actual working environment to verify its prediction accuracy under actual operating conditions. Collect user feedback and fine-tune the model based on user feedback.

[0019] As a preferred solution, the analysis of the use environment and mode of vulnerable parts, the proposal of phase shifter status improvement measures, and the optimization of phase shifter operation include:

[0020] According to the historical wear and replacement records of the parts, the vulnerable parts that need to be optimized are determined; the use environment data of the vulnerable parts, including temperature and load, are obtained; the historical working mode data of the vulnerable parts are extracted to determine the use characteristics and patterns of the phase shifter, the historical working mode data includes load frequency and peak distribution, and the use characteristics and patterns include repeated loads or frequent temperature changes; the obtained environment and use mode data are matched with the design constraints of the parts to determine whether there is over-limit use; the number and proportion of the parts running under over-limit conditions are calculated to evaluate the risk of excessive wear of the parts; if the risk of excessive wear of the parts caused by over-limit use is greater than the preset threshold, the environmental factors causing the over-limit are identified, the critical environmental conditions that are most likely to cause wear are calculated, and improvement measures for adjusting the equipment control parameters are proposed; if there are problems with the material or processing of the parts themselves, improvement measures such as material replacement or processing enhancement are proposed; the proposed improvement measures are implemented, the changes in the component status and the working conditions of the phase shifter are continuously monitored, and the improvement measures are further adjusted and optimized; it also includes: according to the chemical decay characteristics of the lubricating oil under different environmental conditions and the corresponding wear data of the vulnerable parts, a correlation analysis between the decay rate and the wear rate is established.

[0021] The above-mentioned method establishes a correlation analysis between decay rate and wear rate based on the chemical decay characteristics of lubricating oil under different environmental conditions and the corresponding wear data of vulnerable parts, specifically including:

[0022] Obtain the chemical decay characteristics of different types of lubricants under varying environmental conditions, including different temperatures, pressures, and contamination levels. Also record changes in the concentration of specific chemicals in the lubricant samples, including oxides and acid values; obtain wear data for the condenser parts corresponding to the lubricant samples, including wear rate, replacement frequency, and wear pattern. Establish a decay rate model Describes the chemical decay rate of lubricating oil, where D represents the chemical decay rate, k, a, b, c, n, m, p are coefficients and exponents, C represents the pollutant concentration, T is the temperature, t is the time, and k is the proportionality coefficient. Based on the relationship between the chemical decay rate D and the component wear rate, a component wear rate model is established. , where W represents the wear rate of the component, α, β, γ and q are model parameters; use component wear data to adjust and optimize the component wear rate model; integrate the decay rate model and component wear rate model into the phase-shifting lubricating oil monitoring system.

[0023] As a preferred solution, the synchronous monitoring of the dynamic operation information of the condenser and the wear condition of the oil in the lubrication system to verify the effects of the condenser and oil circuit status monitoring and component maintenance optimization includes:

[0024] Through the dynamic data interface of the phase shifter, the real-time operation data of each component, including speed and temperature, is obtained; the oil status of different parts of the lubrication system is monitored synchronously, the changes of wear particles in the oil are regularly detected, and the lubricant status data is recorded and stored; based on the dynamic operation data and lubricant status data, the wear particle prediction model is used for verification to verify the accuracy and stability of the wear particle prediction model; if the model prediction effect decreases, a data quality check is performed or the model is retrained; the replaced or maintained wearing parts are continuously monitored, the wear status is continuously monitored, and the effectiveness of the improvement measures is confirmed; if the risk of excessive wear after replacing the parts is still greater than the preset threshold, the scope of measures is expanded and the overall optimization of the phase shifter components is performed; based on the monitoring results, the phase shifter monitoring plan is optimized to achieve coordinated monitoring and control among multiple systems of the phase shifter; a dynamic tracking and monitoring network for the entire life cycle of the phase shifter operation is established to form a continuous closed-loop control of component status assessment, usage optimization and system reliability improvement.

[0025] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0026] The present invention provides a method for predicting the performance degradation of phase-shifting oil products based on real-time monitoring data, which is used to optimize the maintenance and operation efficiency of phase-shifting equipment and similar precision equipment. By collecting key operating data including load parameters, speed, and temperature, this method can effectively monitor the real-time status of the equipment and predict the wear trend of components. The analysis of lubricating oil samples extracted regularly provides a detailed basis for identifying and constructing a wear particle data set, thereby enhancing the understanding and prediction ability of component wear conditions. By comparing the prediction results with the actual wear data, the method accurately evaluates the prediction effect of the model in different parts, especially for parts with poor prediction performance, it can deeply analyze the changes in the elemental composition of particles and accurately determine the source of worn parts. Furthermore, the method analyzes the performance of vulnerable parts in different use environments and working modes, and proposes targeted state improvement measures, thereby optimizing the operation of the phase-shifting equipment. Finally, the implementation of continuous tracking and monitoring ensures the long-term stable operation of the equipment and the sustainability of the maintenance effect. Overall, the patented method effectively improves the maintenance efficiency of the equipment, reduces the failure rate, extends the service life of the equipment, and improves the overall operation safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The present invention is a flow chart of a method for predicting phase shifter oil performance degradation based on real-time monitoring data.

[0028] Figure 2 It is a schematic diagram of a method for predicting phase shifter oil performance degradation based on real-time monitoring data of the present invention.

[0029] Figure 3 It is another schematic diagram of a method for predicting phase shifter oil performance degradation based on real-time monitoring data of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] In this embodiment, a method for predicting phase condenser oil performance degradation based on real-time monitoring data may specifically include:

[0032] S101, collect online monitoring information of various oil circuit components including operating data of load parameters, rotation speed and temperature, and obtain data sample sets as phase regulator state input through sampling and filtering methods.

[0033] Continuously monitor each component of the oil circuit online, and record the load parameters, speed and temperature data. Perform preliminary sampling of the monitoring data, perform initial sampling according to the first preset number, and obtain a data sample set. Use a low-pass filter to process the data sample set to remove noise and interference. If the load parameter fluctuation is greater than the preset threshold, increase the sampling frequency and obtain a second preset number of data samples. If the temperature is higher than the preset threshold, expand the filter processing range to eliminate abnormal data caused by temperature fluctuations. According to the number of data samples, determine whether the preset sample number requirements are met. If the number of data samples is insufficient, further increase the number of samplings. According to the cleaned data set, determine the current state of the phase regulator. If the number of data samples is insufficient, increase the number of samplings. For the case where the speed is greater than the preset speed threshold or the load is greater than the preset load threshold, adjust the sampling strategy and filtering parameters to obtain the real-time data sample of the processed phase regulator in the current stable state. Extract key features from the real-time monitoring data and the stored historical behavior data, including load, speed, and temperature; use the cosine similarity method to compare the similarity between the real-time data and the historical data, and extract the historical behavior data whose similarity to the current state is greater than the preset threshold.

[0034] For example, the preset threshold of the load parameter of a certain condenser is 100, the preset threshold of the speed is 2000, and the preset threshold of the temperature is 80. Initial sampling obtains 100 data samples, which are processed by a low-pass filter to remove noise and interference. When further evaluating the data sample set, it is found that the load parameter fluctuates greatly and exceeds the preset threshold. In order to obtain more accurate data, it is necessary to increase the sampling frequency and obtain a second preset number of data samples, which is 200. At the same time, it is monitored that the temperature is higher than the preset threshold, and the filtering processing range needs to be expanded to eliminate abnormal data caused by temperature fluctuations. According to the number of data samples, it is judged whether the preset sample number requirement is met. If the number of data samples is insufficient, it is necessary to further increase the number of sampling times to 300 times. The current state of the condenser can be determined by the cleaned data set. If the number of data samples is insufficient, it is necessary to increase the number of sampling times. For the case where the speed is greater than the preset speed threshold or the load is greater than the preset load threshold, the sampling strategy and filtering parameters can be adjusted to obtain the processed real-time data samples of the condenser in the current stable state. Key features such as load, speed, and temperature can be extracted from real-time monitoring data and stored historical behavior data. Using the cosine similarity method, we can compare the similarity between real-time data and historical data, and extract historical behavior data with a similarity greater than the preset threshold to the current state. If the current phase regulator has a load parameter of 120, a speed of 2100, and a temperature of 90, and in the real-time data sample set processed according to the above steps, there is a historical data with a load parameter of 115, a speed of 2000, and a temperature of 85. By calculating the cosine similarity, it is found that the similarity with the current state is 95, which is greater than the preset threshold. Therefore, this historical behavior data can be used as a reference to determine the behavior of the current phase regulator in a stable state.

[0035] S102. Regularly extract lubricating oil samples from multiple locations of the phase shifter oil circuit, use electron microscopy and energy spectrum analysis technology to detect the quality characteristic data of wear particles in each sample, and construct a wear particle data set corresponding to the location.

[0036] According to the phase shifter structure diagram, the positions of various parts of the oil circuit are determined. A fixed cycle is set to extract lubricating oil samples from various designated positions of the oil circuit within a predetermined time interval. The extracted lubricating oil samples are imaged and observed through an electron microscope to obtain the morphological characteristic images of the wear particles. The size characteristics, shape characteristics, and surface property characteristics of the wear particles are determined based on the morphological characteristic images of the wear particles. The size characteristics include length, width, and depth. The shape characteristics include round, angular, and irregular shapes. The surface property characteristics include smooth, rough, and cracked. X-ray energy spectrum analysis technology is used to determine the elemental composition of lubricating oil wear particles. The chemical properties and chemical stability of the particles are evaluated based on the elemental composition. The chemical properties of the particles include metals, non-metals, and oxides. A wear particle data set is constructed based on the measured size, shape, surface quality, and chemical property data. Based on the wear particle data set, the differences in parameter values ​​of different parts are determined. If the difference is less than the preset threshold, the sampling distance between the parts is adjusted, and the sample types are increased to obtain a wear particle data set with a difference greater than the preset threshold. Based on the wear particle data set with a difference greater than the preset threshold, the decision tree algorithm is used to train the model and determine the wear type of different parts. If the parts cannot be distinguished, the number of test samples is increased; the theoretical wear type is estimated based on the phase shifter structure and operating parameters. The estimated results are compared with the actual test results to analyze the attribution of the deviation, and the data set construction plan is optimized based on the comparison results.

[0037] For example, in the phase shifter structure, there are three locations for extracting lubricating oil samples, namely, location A, location B, and location C. Every 10 minutes, lubricating oil samples are extracted from these three locations for imaging observation. In the lubricating oil sample extracted from location A, a wear particle was observed, whose morphological characteristics are as follows: length 20μm, width 10μm, depth 5μm, round shape, and smooth surface properties. Through X-ray energy spectrum analysis, the element composition of the wear particle is determined to be iron, oxygen, and carbon. According to the element composition, the chemical properties and chemical stability of the particles are evaluated, and it can be determined that the wear particle is a metal particle. Based on the measured size, shape, surface quality, and chemical property data, a wear particle data set is constructed. In the lubricating oil samples at locations B and C, the size, shape, and surface property characteristics of the wear particles observed are the same as those of the particles at location A. Based on the wear particle data set, the differences in parameter values ​​at different locations are judged. If the differences in particle size, shape, and surface properties at locations A, B, and C are less than the preset threshold. Therefore, it is necessary to adjust the sampling distance between locations and increase the sample types to obtain a wear particle data set with a difference greater than the preset threshold. Based on the wear particle dataset with a difference greater than the preset threshold, the decision tree algorithm is used to train the model and determine the wear type of different parts. If the parts cannot be distinguished, the number of test samples is increased to improve the accuracy of the model. The theoretical wear type estimated based on the phase shifter structure and operating parameters is metal particles. The estimated results are compared with the actual test results, the cause of the deviation is analyzed, and the dataset construction plan is optimized based on the comparison results. If the estimated results are consistent with the actual test results, it means that the accuracy of the estimated model and the dataset construction plan is high. If there is a deviation, the cause of the deviation can be further analyzed, such as the inaccuracy of the operating parameters or the lack of model training, and then corresponding optimization can be performed.

[0038] S103, based on the real-time dynamic monitoring data of the phase regulator parts and the wear particle detection results of various parts of the oil circuit, perform data statistical feature calculation and feature mapping, and build a wear particle prediction model.

[0039] Obtain real-time dynamic monitoring data of each part of the condenser, including load, speed, and temperature. Calculate statistical characteristics of the dynamic data of each part, including mean, standard deviation, and skewness; obtain wear particle detection results of oil products in each part of the oil circuit, and extract key features of wear particles, including length distribution and quantity ratio. Map the real-time dynamic monitoring data with the corresponding wear particle data for each part, and the real-time monitoring data includes load, speed, and temperature. Integrate the statistical characteristics of the dynamic data and the wear particle characteristics into a matrix, use the support vector machine algorithm to train the integrated feature matrix, obtain the wear particle prediction model, and determine the parameter relationship and feature weight. Check whether the correlation relationship generated by the model is consistent with the operating conditions of the actual part. If the model effect does not meet expectations, adjust the feature selection, update the model structure, and re-determine the parameters. Evaluate the prediction effect of different parts according to the model error value. If the error is greater than the preset threshold, increase the sample size of the corresponding part and retrain the model. Obtain verification data to evaluate whether the prediction effect of the model meets the standard. If the prediction effect meets the standard, the wear particle prediction model is applied to the actual condenser state monitoring and wear prediction.

[0040] For example, a bearing, a key component in a condenser, is being monitored. The collected data for the bearing is an average load of 500N per hour, a speed of 1500 rpm, and an average temperature of 70°C. The mean load of the bearing is 500N, the standard deviation is 50N, and the skewness is -0.5. Through oil analysis, it is found that the length distribution of wear particles is mainly between 0.1mm and 0.5mm, and the number ratio is about 1000 particles per milliliter of oil. Associate the real-time monitoring data of the bearing with the wear particle data. Create a feature matrix containing load, speed, temperature, particle length distribution, and number ratio. Use the support vector machine algorithm to train the feature matrix and establish a prediction model. For example, the model predicts that the number of wear particles of the bearing under the current load and speed should be 1200 per milliliter of oil, but the actual test result is 1500 per milliliter of oil. Collect additional real-time monitoring data and wear particle data to re-verify the prediction effect of the model. If the prediction error of the model drops to an acceptable range, such as a prediction error of less than 5%, the model is considered to meet the standard. The optimized model is applied to actual phase regulator condition monitoring and bearing wear prediction to predict the bearing wear status in real time and perform maintenance or replacement in time.

[0041] S104, using the constructed wear particle prediction model, online prediction of corresponding oil wear particle parameter distribution is achieved for input operation data of a designated part.

[0042] Obtain the real-time operating data of the specified part of the condenser to determine whether the data quality meets the input requirements of the wear particle prediction model. If the data is abnormal or missing, the median filter method is used to compensate and correct it. According to the processed real-time dynamic data, the wear particle prediction model is used to predict the length range and quantity ratio distribution of the wear particles in the specified part. The predicted wear parameter distribution is compared with the preset normal wear range. If the predicted result exceeds the normal wear range, the auxiliary diagnosis module is started to determine the cause of the abnormality. The auxiliary diagnosis module matches the operating parameters with the historical operating data to determine whether the predicted result exceeds the normal wear range due to abnormal operation, and provides corresponding processing suggestions. If no abnormal operation is detected, the quality of the dynamic data will be rechecked, and correction and recalculation will be performed if there is an error. Through the wear particle prediction model, the verified wear particle parameter distribution of the specified part is obtained, and the wear condition under the current state is evaluated.

[0043] For example, it is necessary to obtain the real-time operating data of the phase-shifting machine bearing and judge the health of the bearing based on this data. The bearing temperature, vibration and noise data can be obtained every minute. First, it is necessary to determine whether the data quality meets the input requirements of the wear particle prediction model. For example, the sampling frequency of the bearing temperature data is required to be 1Hz. If the actual sampling frequency is lower than 1Hz, it means that the data is abnormal or missing. In this case, the median filter method can be used for compensation and correction to obtain smooth temperature data. Next, the wear particle prediction model is used to predict the length range and number ratio distribution of the bearing wear particles based on the processed real-time dynamic data. The model may predict that the length of the bearing wear particles is between 10-50 microns, the number ratio distribution is 70% 10-30 microns, and 30% 30-50 microns. Then, the predicted wear parameter distribution is compared with the preset normal wear range. The preset normal wear range of the bearing is between 0-20 microns in length and 100% 0-20 microns in number ratio distribution. If the predicted result exceeds the normal wear range, it means that the bearing may be abnormal. In this case, the auxiliary diagnosis module is started to determine the cause of the abnormality. The auxiliary diagnosis module can match the operating parameters with the historical operation data, determine whether the predicted results are beyond the normal wear range due to abnormal operation, and provide corresponding processing suggestions. If the bearing temperature is found to be abnormally high, it may be caused by insufficient lubricating oil or a lubrication system failure. It is recommended to add lubricating oil or repair the lubrication system in time. If the auxiliary diagnosis module does not detect abnormal operation, it will recheck the quality of the dynamic data and perform corrections and recalculations if there are errors. If noise interference is found in the vibration data, the filtering algorithm can be used to remove the noise and then rerun the prediction model. Finally, through the wear particle prediction model, the verified bearing wear particle parameter distribution can be obtained, and the wear condition of the bearing in the current state can be evaluated. According to the prediction results of the model, the average wear particle length of the bearing is 40 microns, and the number ratio distribution is 60% for 30-50 microns and 40% for 50-70 microns, so the wear condition of the bearing is evaluated as moderate wear.

[0044] S105. Compare the prediction results with the actual detection data of wear particles, analyze the errors between the two, and evaluate the wear prediction effect of the model in different parts.

[0045] According to the wear particle prediction model, the predicted wear parameter distribution results are obtained, including the distribution of the number ratio of particle size. Lubricating oil samples are collected from the corresponding condenser parts predicted by the wear particle prediction model, and the particles are measured to obtain the actual wear particle data; the prediction results of the wear particle prediction model and the measured actual wear particle data are compared for the number of particles in the same size segment. The error between the predicted number and the actual number of each size segment is calculated, and the size segment with the largest error is identified. The model prediction results of different condenser parts are evaluated, and the average prediction error and maximum error of each part are calculated. According to the error value of each part, the model prediction effects of different parts are arranged, and the parts with the worst prediction effect are identified. For parts with errors greater than the preset threshold, the cause of the prediction deviation is determined, and the model is adjusted. The attribution of the deviation includes insufficient data, improper feature selection, and limitations of the model itself. The adjusted model is retrained, and the error fluctuations of each part are continuously monitored, and the model is optimized in a timely manner.

[0046] For example, according to the wear particle prediction model, the wear parameter distribution results are obtained through prediction, and the number ratio distribution of particle size is obtained. In a certain phase shifter part, the predicted wear particle size distribution is that the number ratio of particles less than 1μm is 10%, the number ratio of particles between 1μm and 5μm is 50%, and the number ratio of particles greater than 5μm is 40%; then, lubricating oil samples are collected from the phase shifter part for particle measurement, and the actual wear particle data is obtained. The number of particles less than 1μm is 12, the number of particles between 1μm and 5μm is 45, and the number of particles greater than 5μm is 43; next, the predicted results and actual data are compared for the number of particles in the same size segment. Taking the particle size less than 1μm as an example, the predicted number is 10%, the actual number is 12, and the error is 2%. The errors of other size segments are calculated in the same way, and the error of particle size less than 1μm is 2%, the error of particle size between 1μm and 5μm is 10%, and the error of particle size greater than 5μm is 7%. According to the error value, the model prediction effect of different condenser parts can be evaluated. There are three condenser parts, namely A, B, and C. If their average prediction error and maximum error are, the average prediction error of condenser part A is 5%, the maximum error of condenser part A is 10%, the average prediction error of condenser part B is 8%, the maximum error of condenser part B is 12%, the average prediction error of condenser part C is 6%, and the maximum error of condenser part C is 9%; according to the value of the maximum error, the part with the worst prediction effect can be identified, which is condenser part B. For parts with errors greater than the preset threshold, it is necessary to determine the attribution of the prediction deviation and adjust the model. If the error of condenser part B is large, it may be due to insufficient data, improper feature selection, or limitations of the model itself. Further analysis can be carried out to find out the problem and make targeted model adjustments. After the adjusted model is retrained, it is necessary to continuously monitor the error fluctuations in each part and optimize the model in a timely manner to ensure the accuracy and stability of the prediction results.

[0047] S106. Based on the wear prediction effect of the wear particle prediction model at different parts, extract lubricating oil samples from various parts of the camera and perform element composition testing to determine the source of the worn parts.

[0048] According to the prediction results of the wear particle prediction model of each part of the condenser, the parts where the prediction effect error is greater than the preset threshold are determined; from the parts where the effect error is greater than the preset threshold, a predetermined number of batches of lubricant samples are extracted to obtain data at different time points. Wear particles are separated from the collected lubricant samples, and the main element components are detected using X-ray energy spectrum analysis technology. Based on the element detection results, a time series change curve of each element in the particles is constructed, and it is determined whether there is an element component increase greater than the preset threshold or a decrease greater than the preset threshold. For elements with an element component increase greater than the preset threshold, the material composition and structure of the component are associated, and a preliminary match is performed on the possible wear parts to determine the candidate wear source parts; the historical wear and replacement records of the wear parts are checked to verify whether the source of the judged wear parts is correct. Confirm the source of the wear parts where the prediction effect error of the wear particle prediction model is greater than the preset threshold, and obtain the specific location code of the part.

[0049] For example, according to the prediction results of the wear particle prediction model of each part of the phase shifter, the parts with prediction effect errors greater than the preset threshold are parts A and B. Two batches of lubricant samples were extracted from parts A and B, and data at different time points were obtained. At time point 1, 100 wear particles were separated from the lubricant sample of part A, and at time point 2, 150 wear particles were separated from the lubricant sample of part A. At the same time, 200 wear particles were separated from the lubricant sample of part B. The main element components of these wear particles were detected using X-ray energy spectrum analysis technology. According to the element detection results, it was found that the content of iron in the wear particles of part A was 50% at time point 1 and 60% at time point 2, with an increase of 10%. The content of iron in the wear particles of part B was 30% at time point 1 and 40% at time point 2, with an increase of 10%. According to the preset threshold, it can be judged that the increase in iron is greater than the preset threshold, that is, it is judged that both parts A and B are worn. Next, the material composition and structure of Part A and Part B are correlated to make a preliminary match of possible wear parts. The material composition and structure of Part A correspond to gears, while the material composition and structure of Part B correspond to bearings. The historical wear and replacement records of gears and bearings are further checked to verify whether the source of the judged wear parts is correct. If the historical records show that the gears and bearings are worn and replaced within the corresponding time period, it can be confirmed that the wear source parts of Part A are gears, and the wear source parts of Part B are bearings. Finally, the specific location codes of gears and bearings are obtained, and the sources of wear parts whose prediction effect errors of the wear particle prediction model are greater than the preset threshold are determined.

[0050] Based on the historical experimental data of the phase regulator, the relationship between the generation rate of lubricating oil wear particles and operating conditions is identified, the wear trend is predicted in real time, and the state optimization and preventive maintenance of the lubrication system are guided.

[0051] Based on historical experimental data, the potential relationship between the wear particle generation rate and temperature T, pressure P, and mechanical stress S is identified, and a relationship model between the wear particle generation rate and temperature, pressure, and stress is constructed. , predict the wear particle generation rate in real time, where a, b, c, d, and n are model parameters determined by historical experimental data; evaluate the model's goodness of fit and prediction accuracy. Integrate the relationship model between the wear particle generation rate and temperature, pressure, and stress into the existing lubricant monitoring system of the phase shifter to generate a real-time prediction of the wear particle generation rate; optimize the state of the phase shifter lubrication system and perform preventive maintenance based on the wear particle generation rate. Test the performance of the model in an actual working environment to verify its prediction accuracy under actual operating conditions. Collect user feedback and fine-tune the model based on user feedback.

[0052] For example, according to historical experimental data, the temperature T is [1, 2, 3, 4, 5], the pressure P is [2, 4, 6, 8, 10], the mechanical stress S is [3, 2, 3, 4, 5], and the wear particle generation rate R is [5, 8, 12, 15, 18]. Now, a relationship model between the wear particle generation rate and temperature, pressure and stress is constructed, and real-time prediction is performed. First, the model parameters a, b, c, d and n can be determined by fitting the data. Using the least squares method for fitting, the following relationship model can be obtained, R=5*T^2+6*e^(3*P)+5*log(S), where a=5, n=2, b=6, c=3, d=5. Now, the model can be used to predict the wear particle generation rate in real time. For example, at a certain moment, the temperature is 1°C, the pressure is 2MPa, and the mechanical stress is 3N, then R=5*1^2+6*e^(3*2)+5*log(3)=2427.97. The goodness of fit and prediction accuracy of the model can be evaluated using indicators such as root mean square error and coefficient of determination R^2. The goodness of fit and prediction accuracy of the model can be evaluated by comparing with actual data. The relationship model between the wear particle generation rate and temperature, pressure and stress can be integrated into the existing lubricating oil monitoring system of the phase shifter to generate the prediction results of the wear particle generation rate in real time. Based on the prediction results of the wear particle generation rate, the state of the phase shifter lubrication system can be optimized and preventive maintenance can be performed. Test the performance of the model in the actual working environment. For example, the lubricating oil monitoring system can be used to collect real-time temperature, pressure and mechanical stress data, and the wear particle generation rate can be predicted by the model. Then, it is compared with the actual measured wear particle generation rate to verify the prediction accuracy of the model under actual operating conditions. By collecting user feedback, potential problems of the model and directions for improvement can be understood. Based on user feedback, the model can be fine-tuned, such as adjusting model parameters or improving model structure, to improve prediction accuracy and adaptability.

[0053] S107. Analyze the use environment and mode of vulnerable parts, propose measures to improve the state of the phase regulator, and optimize the operation of the phase regulator.

[0054] According to the historical wear and replacement records of the components, determine the vulnerable parts that need to be optimized. Obtain the usage environment data of the vulnerable parts, including temperature and load. Extract the historical working mode data of the vulnerable parts to determine the usage characteristics and patterns of the phase regulator. The historical working mode data includes load frequency and peak distribution. The usage characteristics and patterns include repeated loads or frequent temperature changes. Match the obtained environment and usage mode data with the design constraints of the components to determine whether there is over-limit use. Calculate the number and proportion of times the components operate under over-limit conditions to evaluate the risk of excessive wear of the components. If the risk of excessive wear of the components caused by over-limit use is greater than the preset threshold, identify the environmental factors that cause the over-limit, calculate the critical environmental conditions that are most likely to cause wear, and propose improvement measures to adjust the equipment control parameters. If there are problems with the material or processing of the component itself, propose improvement measures such as material replacement or processing enhancement. Implement the proposed improvement measures, continuously monitor changes in component status and phase regulator operating conditions, and further adjust and optimize the improvement measures.

[0055] For example, the vulnerable parts of the intermediate three-stage gear system are determined through preliminary analysis. The long-term operating environment temperature curve of the gear is obtained, and it is found that 15% of the operating time exceeds the maximum design limit of 100°C. At the same time, the load mode data of the three-stage gear is extracted, and it is found that the peak load exceeds the nominal maximum value many times, and the peak load usage time accounts for 18%. Referring to the design manual, the proportion of the maximum allowable temperature and load exceeding the limit at the same time does not exceed 5%. Obviously, the current three-stage gear exceeds the limit for too long, reaching 33%, which is an important cause of wear failure. It is judged to be the dual improper influence of the use environment and mode. The critical environmental condition that is most likely to accelerate wear under the current working condition is calculated as a composite state of temperature greater than 95°C and load exceeding 8.5 times the nominal value. This composite state corresponds to the main controller of the intermediate transmission system. Send optimization instructions to the main controller, set temperature control parameters, control the operating temperature of the three-stage gear box at 90°C, and set the load limiter to control the peak load below 8 times. Ensure that the environmental conditions meet the design requirements and reduce the probability of component loss. Continuously track and monitor the status of the third-stage gears, realize health assessment and service life prediction, verify the specific effects of the proposed measures, and improve system reliability.

[0056] According to the chemical decay characteristics of lubricating oil under different environmental conditions and the corresponding wear data of vulnerable parts, the correlation analysis between decay rate and wear rate is established.

[0057] Obtain the chemical decay characteristics of different types of lubricants under varying environmental conditions, including different temperatures, pressures, and contamination levels. Also record changes in the concentration of specific chemicals in the lubricant samples, including oxides and acid values; obtain wear data for the condenser parts corresponding to the lubricant samples, including wear rate, replacement frequency, and wear pattern. Establish a decay rate model Describes the chemical decay rate of lubricating oil, where D represents the chemical decay rate, k, a, b, c, n, m, p are coefficients and exponents, C represents the pollutant concentration, T is the temperature, t is the time, and k is the proportionality coefficient. Based on the relationship between the chemical decay rate D and the component wear rate, a component wear rate model is established. , where W represents the wear rate of the component, α, β, γ and q are model parameters; use component wear data to adjust and optimize the component wear rate model; integrate the decay rate model and component wear rate model into the phase-shifting lubricating oil monitoring system.

[0058] For example, for a lubricant, it is necessary to understand its chemical decay characteristics under different environmental conditions and the relationship between it and the wear of the consumable parts of the phase shifter. First, obtain the historical experimental data of lubricant samples at three different temperatures, temperature 1 is 25 degrees Celsius, temperature 2 is 50 degrees Celsius, and temperature 3 is 75 degrees Celsius. At each temperature, record the concentration changes of oxides and acid values ​​in the lubricant samples. At temperature 1, after a certain period of time, the concentration of oxides in the lubricant sample changes to 0.02mg / L, and the acid value changes to 0.05mgKOH / g. At temperature 2, the concentration of oxides changes to 0.05mg / L, and the acid value changes to 0.1mgKOH / g. At temperature 3, the concentration of oxides changes to 0.1mg / L, and the acid value changes to 0.15mgKOH / g. Next, consider the effect of contamination level on the decay characteristics of lubricants. Select two lubricant samples with different contamination levels for experiments: sample A is a low contamination level, and sample B is a high contamination level. Under the same temperature conditions, record the concentration changes of oxides and acid values ​​in the two samples. At low pollution levels, after a certain period of time, the oxide concentration in lubricating oil sample A changes to 0.03 mg / L, and the acid value changes to 0.08 mgKOH / g. At high pollution levels, the oxide concentration in lubricating oil sample B changes to 0.05 mg / L, and the acid value changes to 0.12 mgKOH / g. Based on the above experimental data, a decay rate model can be established: . If a lubricating oil sample A with a temperature of 50 degrees Celsius and a low contamination level is selected for modeling. By fitting the experimental data, the model parameters are obtained as follows: k=1, a=5, b=3, c=2, n=2, m=1, and p=5. Next, consider the relationship between component wear rate and chemical decay rate. The consumable parts of the phase shifter were used, and their wear data were recorded. It was found that there was a linear relationship between the component wear rate and the chemical decay rate D, and a component wear rate model was established: . By fitting the component wear data, the model parameters are obtained as α=2, β=3, γ=1, and q=8. Finally, the decay rate model and the component wear rate model are integrated into the phase-shifting lubricant monitoring system. By real-time monitoring of the chemical decay rate D in the lubricant sample and predicting the component wear rate W based on the component wear rate model, the performance of the lubricant and the wear of the components can be evaluated in a timely manner so as to make corresponding maintenance and replacement decisions.

[0059] S108. Synchronously monitor the dynamic operation information of the phase regulator and the wear condition of the oil in the lubrication system to verify the effects of phase regulator and oil circuit status monitoring and component maintenance optimization.

[0060] Through the dynamic data interface of the phase regulator, the real-time operation data of each component, including speed and temperature, is obtained. The oil status of different parts of the lubrication system is monitored synchronously, the changes of wear particles in the oil are detected regularly, and the lubricant status data is recorded and stored. According to the dynamic operation data and lubricant status data, the wear particle prediction model is used for verification to verify the accuracy and stability of the wear particle prediction model. If the model prediction effect decreases, the data quality check or retraining of the model is performed. Continuously monitor the replaced or maintained vulnerable parts, continuously monitor the wear status, and confirm the effect of the improvement measures. If the risk of excessive wear after replacing the parts is still greater than the preset threshold, the scope of measures is expanded to optimize the overall phase regulator components. According to the monitoring results, the phase regulator monitoring plan is optimized to realize the coordinated monitoring and control of multiple phase regulator systems. A dynamic tracking and monitoring network for the entire life cycle of the phase regulator operation is established to form a continuous closed-loop control of component status evaluation, usage optimization and system reliability improvement.

[0061] For example, after the condenser has been running for a period of time, the input shaft speed and load data are continuously obtained, as well as the changes in the proportion of 40-100 micron wear particles in the corresponding bearing area oil. After the new data is input into the model to evaluate the effect, it is found that the prediction error of the bearing particles remains at about 5%, which is consistent with the long-term level of the model, verifying the stable matching of the dynamic wear assessment. However, after the increase in samples, it is found that the prediction deviation of the three-stage gearbox is large, reaching 10%. If there is no abnormality in the input data analysis, the model parameters are retrained for this part. After multiple rounds of iterative optimization, the prediction error of the three-stage gearbox is controlled within 3%, and the model effect is improved. In addition, for the three-stage gear that was previously proposed to be replaced, its wear data set was tested, and it was found that the proportion of large particles above 100 microns continued to increase, and the replacement effect was not obvious. Therefore, the detection range was expanded, and the connected intermediate bearings were strengthened. At the same time, the lubricating oil was improved, and finally the wear condition of the three-stage gear transmission system was fully improved, verifying the effect of the proposed solution. In the subsequent operation, the model will continue to be optimized and the state evaluation system will be improved, so that the condenser can achieve continuous and stable full life cycle monitoring and comprehensively improve the system reliability level.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting phase shifter oil performance degradation based on real-time monitoring data, characterized in that: The method comprises: Collect online monitoring information of various oil circuit components including operating data of load parameters, speed and temperature, and obtain data sample sets as phase regulator status input through sampling and filtering methods; Lubricating oil samples are regularly extracted from multiple locations of the condenser oil circuit. The quality characteristic data of wear particles in each sample are detected using electron microscopy and energy spectrum analysis technology to construct a wear particle data set corresponding to the location. According to the real-time dynamic monitoring data of the condenser parts and the wear particle detection results of various parts of the oil circuit, the data statistical feature calculation and feature mapping are performed, and the wear particle prediction model is constructed; Using the constructed wear particle prediction model, the input operation data of the designated part is used to realize online prediction of the corresponding oil wear particle parameter distribution; Compare the predicted wear parameter distribution results with the actual wear particle detection data, the predicted wear parameter distribution results include the quantitative proportion distribution of particle size, analyze the error between the two, and evaluate the wear prediction effect of the model at different parts; According to the wear prediction effect of the wear particle prediction model in different parts, extract the lubricating oil samples from various parts of the camera and test the element composition to determine the source of the worn parts; Analyze the use environment and mode of vulnerable parts, propose measures to improve the state of the phase regulator, and optimize the operation of the phase regulator; Synchronously monitor the dynamic operation information of the phase regulator and the wear condition of the lubrication system oil to verify the effects of phase regulator and oil circuit status monitoring and component maintenance optimization.

2. The method according to claim 1, wherein: The online monitoring information of each component of the oil circuit containing the operating data of load parameters, speed and temperature is collected, and a data sample set is obtained as the phase shifter state input by sampling and filtering methods, including: Continuously monitor each component of the oil circuit online and record load parameters, speed and temperature data; perform preliminary sampling of the monitoring data, perform initial sampling according to a first preset number, and obtain a data sample set; Process the data sample set using a low-pass filter to remove noise and interference; If the load parameter fluctuation is greater than the preset threshold, the sampling frequency is increased to obtain a second preset number of data samples; if the monitored temperature is higher than the preset threshold, the filtering processing range is expanded to eliminate abnormal data caused by temperature fluctuations; based on the number of data samples, determine whether the preset sample number requirement is met. If the number of data samples is insufficient, the number of samplings is further increased; based on the cleaned data set, the current state of the phase regulator is determined. If the number of data samples is insufficient, the number of samplings is increased; for the case where the speed is greater than the preset speed threshold or the load is greater than the preset load threshold, the sampling strategy and filtering parameters are adjusted to obtain the processed real-time data samples of the phase regulator in the current stable state; key features, including load, speed and temperature, are extracted from the real-time monitoring data and the stored historical behavior data; the cosine similarity method is used to compare the similarity between the real-time data and the historical data, and the historical behavior data whose similarity to the current state is greater than the preset threshold is extracted.

3. The method according to claim 1, wherein: The lubricating oil samples are regularly extracted from multiple positions of the phase shifter oil circuit, and the quality characteristic data of the wear particles in each sample are detected by using electron microscope and energy spectrum analysis technology to construct the wear particle data set corresponding to the position, including: According to the phase shifter structure diagram, the positions of various parts of the oil circuit are determined; a fixed cycle is set to extract lubricating oil samples from various designated positions of the oil circuit within a predetermined time interval; the extracted lubricating oil samples are imaged and observed through an electron microscope to obtain morphological characteristic images of wear particles; the size characteristics, shape characteristics and surface property characteristics of the wear particles are determined according to the morphological characteristic images of the wear particles, wherein the size characteristics include length, width and depth, the shape characteristics include round, angular and irregular shapes, and the surface property characteristics include smooth, rough and cracked; the elemental composition of lubricating oil wear particles is determined by X-ray energy spectrum analysis technology; the chemical properties and chemical stability of the particles are evaluated according to the elemental composition, and the chemical properties of the particles include metals, non-metals and oxides; a wear particle data set is constructed according to the measured size, shape, surface quality and chemical property data; Determining differences in parameter values ​​at different locations based on the wear particle data set; If the difference is less than the preset threshold, the sampling distance between parts is adjusted, and the sample types are increased to obtain a wear particle data set with a difference greater than the preset threshold. Based on the wear particle data set with a difference greater than the preset threshold, a decision tree algorithm is used to train the model to determine the wear types of different parts. If the parts cannot be distinguished, the number of test samples is increased. Based on the phase-shifting machine structure and operating parameters, the theoretical wear type is estimated. The estimated results are compared with the actual test results to analyze the attribution of the deviations, and the data set construction plan is optimized based on the comparison results.

4. The method according to claim 1, wherein: According to the real-time dynamic monitoring data of the phase regulator parts and the wear particle detection results of various parts of the oil circuit, data statistical feature calculation and feature mapping are performed, and a wear particle prediction model is constructed, including: The real-time dynamic monitoring data of each part of the phase shifter are obtained, including load, speed and temperature; the statistical characteristics of the dynamic data of each part are calculated, including mean, standard deviation and skewness; the wear particle detection results of the oil products in each part of the oil circuit are obtained, and the key characteristics of the wear particles, including length distribution and quantity ratio, are extracted; the real-time dynamic monitoring data is mapped with the corresponding wear particle data for each part, and the real-time dynamic monitoring data includes load, speed and temperature; the statistical characteristics of the dynamic data and the wear particle characteristics are integrated into a matrix, and the integrated feature matrix is ​​trained by the support vector machine algorithm to obtain a wear particle prediction model, and the parameter relationship and feature weight are determined; it is tested whether the correlation relationship generated by the model is consistent with the operating conditions of the actual parts; if the model effect does not meet expectations, the feature selection is adjusted, the model structure is updated, and the parameters are re-determined; the prediction effect of different parts is evaluated according to the model error value, and if the error is greater than the preset threshold, the sample size of the corresponding part is increased and the model is retrained; verification data is obtained to evaluate whether the prediction effect of the model meets the standard; if the prediction effect meets the standard, the wear particle prediction model is applied to the actual phase shifter state monitoring and wear prediction.

5. The method according to claim 1, wherein: The wear particle prediction model constructed is used to predict the wear particle parameter distribution of the corresponding oil product based on the input operation data of the designated part online, including: Obtain the real-time operating data of the specified part of the condenser to determine whether the data quality meets the input requirements of the wear particle prediction model; if the data is abnormal or missing, use the median filtering method to compensate and correct it; based on the processed real-time dynamic data, use the wear particle prediction model to predict the length range and quantity ratio distribution of the wear particles in the specified part; compare the obtained predicted wear parameter distribution with the preset normal wear range; if the predicted result exceeds the normal wear range, start the auxiliary diagnosis module to determine the cause of the abnormality, and the auxiliary diagnosis module matches the operating parameters with the historical operating data to determine whether the predicted result exceeds the normal wear range due to abnormal operation, and provide corresponding processing suggestions; If no operational abnormality is detected, the quality of the dynamic data will be rechecked, and corrections and recalculations will be performed if there are errors. The wear particle prediction model will be used to obtain the verified wear particle parameter distribution of the specified parts, and the wear condition under the current state will be evaluated.

6. The method according to claim 1, wherein: The predicted wear parameter distribution result is compared with the actual detection data of wear particles, the predicted wear parameter distribution result includes the quantitative proportion distribution of particle size, the error between the two is analyzed, and the wear prediction effect of the model at different parts is evaluated, including: According to the wear particle prediction model, the predicted wear parameter distribution results are obtained, including the quantitative proportion distribution of particle sizes; lubricating oil samples are collected from the corresponding phase shifter parts predicted by the wear particle prediction model, and the particles are measured to obtain the actual wear particle data; the number of particles in the same size segment is compared between the prediction results of the wear particle prediction model and the measured actual wear particle data; the error between the predicted number and the actual number of each size segment is calculated, and the size segment with the largest error is identified; the model prediction results of different phase shifter parts are evaluated, and the average prediction error and maximum error of each part are calculated; according to the error value of each part, the model prediction effects of different parts are arranged, and the parts with the worst prediction effect are identified; for parts with errors greater than the preset threshold, the cause of the prediction deviation is determined, and the model is adjusted. The attribution of the deviation includes insufficient data and improper feature selection, and the limitations of the model itself; the adjusted model is retrained, and the error fluctuations of each part are continuously monitored, and the model is optimized in a timely manner.

7. The method according to claim 1, wherein: According to the wear prediction effect of the wear particle prediction model at different parts, the lubricating oil samples of various parts of the condenser are extracted and the element composition is tested to determine the source of the worn parts, including: According to the prediction results of the wear particle prediction model of each part of the condenser, the parts where the prediction effect error is greater than a preset threshold are determined; Extract a predetermined number of batches of lubricant samples from the parts where the effect error is greater than a preset threshold, and obtain data at different time points; separate wear particles from the collected lubricant samples, and use X-ray energy spectrum analysis technology to detect the main elemental composition; based on the element detection results, construct a time series change curve of each element in the particles, and determine whether there is an elemental component increase greater than a preset threshold or a decrease greater than a preset threshold; for elements with an elemental component increase greater than a preset threshold, associate the material composition and structure of the component where they are located, perform a preliminary match on possible wear parts, and determine candidates for wear source parts; check and determine the historical wear and replacement records of the wear parts, and verify whether the source of the judged wear parts is correct; confirm the source of the wear parts whose effect error predicted by the wear particle prediction model is greater than the preset threshold, and obtain the specific location code of the part; It also includes: based on the historical experimental data of the phase shifter, identifying the relationship between the lubricating oil wear particle generation rate and the operating conditions, making real-time predictions of the wear trend, and guiding the state optimization and preventive maintenance of the lubrication system; The method of identifying the relationship between the lubricating oil wear particle generation rate and operating conditions based on the historical experimental data of the phase shifter, performing real-time prediction of the wear trend, and guiding the state optimization and preventive maintenance of the lubrication system specifically includes: identifying the potential relationship between the wear particle generation rate and temperature T, pressure P, and mechanical stress S based on the historical experimental data, and constructing a relationship model between the wear particle generation rate and temperature, pressure, and stress; , make real-time predictions of the wear particle generation rate, where a, b, c, d, and n are model parameters determined by historical experimental data; evaluate the model's goodness of fit and prediction accuracy; integrate the relationship model between the wear particle generation rate and temperature, pressure, and stress into the existing lubricant monitoring system of the phase shifter to generate a real-time prediction of the wear particle generation rate; optimize and perform preventive maintenance on the phase shifter's lubrication system based on the wear particle generation rate; test the model's performance in an actual working environment to verify its prediction accuracy under actual operating conditions; collect user feedback and fine-tune the model based on user feedback.

8. The method according to claim 1, wherein: The analysis of the use environment and mode of vulnerable parts, the proposal of phase shifter status improvement measures, and the optimization of phase shifter operation include: Determine the vulnerable parts that need to be optimized based on the historical wear and replacement records of the parts; obtain the usage environment data of the vulnerable parts, including temperature and load; extract the historical working mode data of the vulnerable parts to determine the usage characteristics and patterns of the phase regulator, the historical working mode data includes load frequency and peak distribution, and the usage characteristics and patterns include repeated loads or frequent temperature changes; match the obtained environment and usage mode data with the design constraints of the parts to determine whether there is over-limit use; calculate the number and proportion of parts running under over-limit conditions to evaluate the risk of excessive wear of the parts; If the risk of excessive wear of components caused by overuse is greater than the preset threshold, the environmental factors causing the overuse are identified, the critical environmental conditions that are most likely to cause wear are calculated, and improvement measures for adjusting equipment control parameters are proposed; If there are problems with the material or processing of the component itself, propose improvement measures such as material replacement or processing enhancement; implement the proposed improvement measures, continuously monitor the changes in the component status and the phase shifter working conditions, and further adjust and optimize the improvement measures; also include: establishing a correlation analysis between the decay rate and the wear rate based on the chemical decay characteristics of the lubricating oil under different environmental conditions and the corresponding wear data of the vulnerable parts; The method establishes a correlation analysis between decay rate and wear rate based on the chemical decay characteristics of lubricating oil under different environmental conditions and the corresponding wear data of vulnerable parts, specifically including: obtaining the chemical decay characteristics of different types of lubricating oil under changing environmental conditions, the changing environmental conditions include different temperatures, pressures and pollution levels; recording the concentration changes of specific chemical substances in the lubricating oil samples, including oxides and acid values; obtaining the wear data of the vulnerable parts of the phase shifter corresponding to the lubricating oil samples, including wear rate, replacement frequency and wear mode; establishing a decay rate model Describe the chemical decay rate of lubricating oil, where D represents the chemical decay rate, k, a, b, c, n, m, p are coefficients and exponents, C represents the pollutant concentration, T is the temperature, t is the time, and k is the proportionality coefficient; according to the relationship between the chemical decay rate D and the component wear rate, a component wear rate model is established , where W represents the wear rate of the component, α, β, γ and q are model parameters; use component wear data to adjust and optimize the component wear rate model; integrate the decay rate model and component wear rate model into the phase-shifting lubricating oil monitoring system.

9. The method according to claim 1, wherein: The synchronous monitoring of the dynamic operation information of the condenser and the wear condition of the oil in the lubrication system verifies the effects of the condenser and oil circuit status monitoring and component maintenance optimization, including: Through the dynamic data interface of the phase regulator, the real-time operation data of each component, including speed and temperature, is obtained; the oil status of different parts of the lubrication system is synchronously monitored, the changes of wear particles in the oil are regularly detected, and the lubricant status data is recorded and stored; based on the dynamic operation data and lubricant status data, the wear particle prediction model is used for verification to verify the accuracy and stability of the wear particle prediction model; if the model prediction effect decreases, the data quality is checked or the model is retrained; the wearing parts that have been replaced or maintained are continuously monitored, the wear status is continuously monitored, and the effect of improvement measures is confirmed; If the risk of excessive wear is still greater than the preset threshold after replacing parts, expand the scope of measures and conduct overall optimization of the phase modulator components; optimize the phase modulator monitoring plan based on the monitoring results to achieve coordinated monitoring and control among multiple systems of the phase modulator; establish a dynamic tracking and monitoring network for the entire life cycle of the phase modulator operation to form a continuous closed-loop control of component status assessment, usage optimization and system reliability improvement.

Images