A real-time monitoring method for oil status of condenser based on deep learning
Through deep learning and a variety of analysis technologies, real-time monitoring and prediction of the status of the camera oil product is achieved, which solves the problem of insufficient monitoring in the existing technology and improves the reliability and efficiency of equipment operation.
Patent Information
- Application Number
- CN202311853425.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-29
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2043-12-29
AI Technical Summary
The existing camera condition monitoring methods have shortcomings in real-time, accuracy of mixed oil monitoring and data processing efficiency, resulting in increased risk of equipment failure and downtime. The traditional methods rely on laboratory equipment and professional technology and are costly. And the analysis is complicated.
Using a deep learning-based method, combined with near-infrared spectrometer, principal component analysis method, multi-layer perceptron neural network and dynamic model, the oil additive components and impurity content are detected in real time, and the oil quality database and degradation prediction model are established to achieve the Real-time monitoring and prediction of mixed oil status.
It significantly improves the real-time and accuracy of oil condition monitoring, prevents equipment failures, optimizes maintenance plans, reduces equipment operation risks, and improves monitoring efficiency and scientificity of data analysis.
Smart Images

Figure CN117805055B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of information technology, and in particular to a real-time monitoring method for phase shifting oil product status based on deep learning. Background Art
[0002] The quality and condition of condenser oil are crucial to the stable operation of industrial equipment. Traditional oil monitoring methods are mainly based on regular sample testing. Although these methods can provide basic physical and chemical property information of oil, they have obvious limitations. Traditional methods cannot achieve continuous and real-time monitoring of oil status. This leads to fewer opportunities for timely detection and response when the oil status changes significantly, which increases the risk of equipment failure and downtime. During the operation of the equipment, new oil is often needed to replenish or replace part of the oil. However, the addition of new oil will change the overall chemical composition of the oil, especially in terms of additive composition and impurity content. Different batches of oil may vary in these parameters, which will affect the overall performance and degradation law of the oil. When new and old oils are mixed, their chemical properties and degradation laws may be significantly different from those of a single oil. This difference makes it more complicated and inaccurate to predict the degradation degree of mixed oils based on traditional methods. Traditional oil analysis methods usually rely on professional laboratory equipment and technicians. This not only increases the cost of monitoring, but also makes the data analysis process relatively cumbersome and time-consuming. Due to the intermittent nature of traditional monitoring methods and the need for post-data analysis, the response to oil status is often not timely enough. This may lead to equipment damage or reduced operating efficiency in some cases. Existing methods for monitoring the status of phase-shifting oil have significant deficiencies in terms of real-time performance, accuracy of mixed oil monitoring, and data processing efficiency, which are particularly prominent in industrial applications. Therefore, seeking a method that can monitor the status of mixed oils in real time and accurately in order to better maintain the stable operation of equipment has become an urgent need in this field. Summary of the invention
[0003] The present invention provides a method for real-time monitoring of the oil status of a condenser based on deep learning, which mainly includes:
[0004] Use a near-infrared spectrometer to detect the additive composition and impurity content data of newly added oil samples in the fuel tank in real time, and establish an oil quality database; use the principal component analysis method to compare the differences between the new oil quality data and the original oil quality database, and evaluate the oil quality based on the extracted principal components and oil performance indicators; establish a mixed oil degradation prediction model based on the historical performance data of the mixed oil to predict the degree of oil degradation; divide the operation stage according to the service life of the mixed oil and implement regular sampling, record the operation parameters and conduct quality inspections, establish an oil quality prediction model and an oil quality indicator threshold warning model, regularly feedback the operation diagnosis results, and optimize the use conditions of the mixed oil; compare the data detected during the use of the mixed oil with the prediction results of the oil quality prediction model to determine the actual degree of degradation of the mixed oil; determine the phase shifter oil status parameters that need to be monitored based on the structural design and use characteristics of the phase shifter; determine whether the oil status is abnormal during the sensor monitoring process, and take corresponding treatment measures based on the status judgment results if abnormal.
[0005] As a preferred solution, the near infrared spectrometer is used to detect the additive composition and impurity content data of the oil sample newly added to the fuel tank in real time to establish an oil quality database, including:
[0006] According to the user manual of the near-infrared spectrometer, set different wavelengths of light to illuminate the oil sample newly added to the tank; detect the spectrum of the sample, record the intensity of reflected or transmitted light, and identify the characteristic absorption peaks of different additive components and impurities; determine the additive component or impurity content based on the intensity of the characteristic peak; judge the content of each additive component and impurity by repeated scanning and testing and comparing with the standard; if the content of a certain component of the test sample exceeds the threshold, it is marked as an unqualified sample, otherwise, it is marked as a qualified sample; establish a one-to-one correspondence between the sample addition time, additive component and impurity content data according to the oil batch number and the test time, obtain the oil quality data of each batch at each time point, and store it in the oil quality database; when new oil is added subsequently, set the near-infrared spectrometer to monitor the new oil added to the tank in real time; obtain the additive component and impurity content data of the new test sample through the scanning results, and update the oil quality database in real time according to the addition time and batch number; if the real-time monitoring results show that the additive component and impurity content of the new sample meet the standard, confirm the addition; if it does not meet the standard, stop adding and record the abnormal data.
[0007] As a preferred solution, the principal component analysis method is used to compare the differences between the new oil quality data and the original oil quality database, and the oil quality is evaluated based on the extracted principal components and oil performance indicators, including:
[0008] A multidimensional quality data matrix is established based on the additive composition and impurity content data of each batch of samples in the original oil quality database; the principal components of the matrix are extracted through the singular value decomposition algorithm in the principal component analysis method, and the contribution rate of the extracted principal components to the sample differences is calculated; if the contribution rate of the first principal component is higher than the preset threshold, the first principal component is used as the evaluation index to judge the values of the new and old oil samples on the first principal component item to reflect the quality difference between the two; if the contribution rate of the first principal component is lower than the preset threshold, the contribution rate of the second principal component is continued to be judged until the cumulative contribution rate is higher than the preset threshold; the oil performance indicators are obtained, and the oil performance indicators include viscosity, flash point, acid value, alkali value, anti-wear ability, antioxidant performance, and impurity content; according to the extracted main components and oil performance indicators, the decision tree algorithm is used for model training to establish an oil quality evaluation model to obtain the quality evaluation of the oil; if the difference between the evaluation results of new oil and old oil is greater than the preset threshold, the new oil data is recorded and the quality database is updated, otherwise, the new oil data is directly saved without processing; according to the latest extracted main components and oil quality evaluation model, the quality data of each batch of new oil is analyzed to determine whether there is a quality difference with the old oil; if there is a quality difference with the old oil, an early warning is immediately fed back, and the oil is added after the oil test, and the database is updated in real time.
[0009] As a preferred solution, the mixed oil degradation prediction model is established based on the historical performance data of the mixed oil to predict the degradation degree of the oil, including:
[0010] Obtain historical data on the performance indicators of mixed oils, including viscosity, flash point, acid value, alkalinity value, anti-wear ability, antioxidant performance, and impurity content, and use the principal component analysis method to determine the indicators that have an impact on the degradation of mixed oils greater than a preset threshold; use a multi-layer perceptron neural network to train the model based on the indicators that have an impact on the degradation of mixed oils greater than a preset threshold and the consumption rate of additives, build a mixed oil degradation prediction model to predict the service life of the oil; compare the simulation results of the mixed oil degradation prediction model with the test data of the actual oil samples to determine the optimal consumption rate; during the use of the mixed oil, continuously perform status testing and analysis on the obtained oil samples; if there is a difference between the test results and the prediction results of the mixed oil degradation prediction model, fine-tune the parameters of the mixed oil degradation prediction model based on the obtained test data, and provide real-time feedback on the difference judgment results between the test and the prediction; also include: identifying key chemical reactions and determining kinetic parameters based on the chemical composition of the mixed oil, building a mixed oil degradation model and implementing real-time monitoring, and optimizing maintenance and replacement plans.
[0011] The method of identifying key chemical reactions and determining kinetic parameters based on the chemical composition of the mixed oil, building a mixed oil degradation model and implementing real-time monitoring, and optimizing maintenance and replacement plans specifically includes:
[0012] Use high-resolution mass spectrometry and nuclear magnetic resonance technology to analyze the various chemical components in the mixed oil and determine the chemical structure of various additives and possible reaction products. Use historical experimental records to identify key chemical reactions occurring in the mixed oil, including oxidation reactions and polymerization reactions. Experimentally determine the kinetic parameters of each key reaction, including the reaction rate constant k. Establish kinetic equations for each key reaction d[A] represents the concentration change of reactant A. Using experimental data, the reaction rate constant k is determined by nonlinear regression analysis. The kinetic equations of individual reactions are combined to construct a mixed oil degradation model. Where d[R] is the concentration change of the reaction product R, k i and k d are the rate constants for reaction and decomposition, respectively, and A i is the concentration of reactant A, B i is the concentration of reactant B, D i is the concentration of decomposition product D. During the use of mixed oil products, real-time monitoring is implemented to obtain the temperature, pressure and running time data of the oil products. The monitoring data is combined with the mixed oil degradation model to quantitatively analyze and simulate the rate and mode of the oil degradation process. The prediction accuracy of the model is verified by comparing with the data of laboratory tests and field sampling. The model parameters are adjusted according to the comparison results to ensure that the model can accurately reflect the degradation behavior of the mixed oil products under actual use conditions. The model prediction results are used to guide the maintenance and replacement plan of the oil products and optimize the operation strategy of the equipment.
[0013] As a preferred solution, the operation stages are divided according to the service life of the mixed oil and regular sampling is implemented, the operation parameters are recorded and quality inspection is performed, an oil quality prediction model and an oil quality index threshold warning model are established, and the operation diagnosis results are regularly fed back to optimize the use conditions of the mixed oil, including:
[0014] According to the expected service life of the mixed oil, the entire use process is divided into three stages: initialization period, normal operation period, and exhaustion period; a fixed sampling time is set in each stage for regular sampling; if an abnormality is detected at the sampling point, including a sudden change in the indicator value, the sampling detection frequency is increased; during the sampling process, the operating parameters of the oil are recorded, including operating temperature, flow rate, and load; the collected oil samples are quality tested to obtain oil quality test data, including the additive content, oxidation product content, viscosity, acid value, and moisture content in the oil; through repeated multiple tests, the quality test data of the mixed oil at each stage of the use process is obtained; based on the quality test data of the mixed oil at each stage obtained, ARIMA is used for model training The system also includes: training, establishing an oil quality prediction model, determining the changing trend of oil quality indicators at different use stages; based on the oil quality monitoring data, using the logistic regression algorithm for model training, building an oil quality indicator threshold warning model, obtaining the probability of oil degradation, and determining the oil status based on the oil degradation rate. The oil degradation levels include normal, concern, critical, and expired; if the oil degradation probability is higher than the preset threshold, a warning signal is triggered, the status results are fed back and used for operation diagnosis and system optimization; it also includes: identifying and monitoring the chemical composition and reaction of additives in mixed oils, establishing a dynamic chemical interaction model and combining it with conventional performance tests to monitor oil stability, predicting degradation trends in long-term use, and adjusting the additive formula to optimize oil performance.
[0015] The above mentioned identification and monitoring of the chemical composition and reaction of additives in mixed oil products, establishment of dynamic chemical interaction model and combined with conventional performance testing to monitor oil stability, predict degradation trend in long-term use, and adjust additive formulation to optimize oil performance include:
[0016] Using high performance liquid chromatography, the chemical composition of various additives in the mixed oil is identified and quantitatively analyzed, and the type, concentration and expected chemical activity of each additive are recorded. Using a gas chromatography-mass spectrometry system, the chemical reaction products between additives are monitored in real time to determine the reaction paths between additives, including any side reactions that may cause performance degradation or oil degradation. A dynamic chemical interaction model based on historical data of the chemical composition of mixed oils is established, and the Schrödinger equation or wave function method is used to predict the interactions between additives. The historical data of the chemical composition of mixed oils includes the mixing ratio and operating conditions. The operating conditions include temperature, pressure, and running time. Combined with the conventional performance tests of oils and the results of chemical analysis of additives, the performance of oils is continuously monitored. Conventional performance tests include viscosity and acid value tests. Using MATLAB, the monitoring data is analyzed to identify abnormal trends or signs of performance degradation. The degradation trend and stability of oils during long-term use are predicted using the established chemical interaction model and historical monitoring data. Based on the type, concentration and expected chemical activity of additives, the ARIMA algorithm is used for model training to predict future performance changes of oils. According to the model prediction results and historical data of the chemical composition of the mixed oil, the additive formula is adjusted to reduce adverse chemical reactions and optimize oil performance.
[0017] As a preferred solution, the method of comparing the data detected during the use of the mixed oil product with the prediction results of the oil product quality prediction model to determine the actual degradation degree of the mixed oil product includes:
[0018] The quality inspection data obtained at different sampling stages of the mixed oil are compared with the oil quality data of the corresponding stages predicted by the oil quality prediction model, and the error between the actual value and the predicted value is calculated; the value with the largest error between the data of each state point in the oil quality inspection data during use and the predicted result is extracted; the quality problem of the mixed oil is determined based on the maximum error value, and whether the actual degree of degradation is consistent with the predicted result is judged. If the oil quality index in the obtained oil quality inspection data is higher than the predicted value of the oil quality prediction model, it is judged that the degree of oil degradation is serious; if the actual oil state is consistent with the oil state predicted by the oil quality index threshold warning model If there is a deviation, the oil quality index threshold warning model is readjusted for the sampling state point with the largest error, including correcting the degradation rate of relevant quality indicators, correcting the degradation path and degradation trend; and the adjusted oil quality index threshold warning model is iterated using the state detection data until the prediction result is consistent with the actual degradation degree; based on the detection data of the mixed oil, the prediction effect of the adjusted oil quality index threshold warning model is evaluated; it also includes: according to the amount of mixed oil added and the proportion of components, a measurement model based on key degradation indicators is established, the degradation prediction is adjusted through real-time monitoring and data comparison, and the mixed oil proportion and usage conditions are optimized.
[0019] According to the amount of mixed oil added and the proportion of components, a measurement model based on key degradation indicators is established, degradation prediction is adjusted through real-time monitoring and data comparison, and the mixed oil proportion and use conditions are optimized, specifically including:
[0020] Before each mixing, record the amount of each oil added V i . Calculate the proportion of each component in the mixed oil, Where ∑V is the total volume or weight of the mixed oil. Define the key degradation indicators of the oil, including viscosity V d 、Acid value A v and additive concentration C a For each indicator, establish the measurement model D x =a0+a1C a +a2V d +a3A v +a4T+a5L, where D x represents the degradation index of the oil, T represents temperature, L represents load, a0, a1, ..., a5 are coefficients determined by data analysis, reflecting the influence of each factor on the degradation index. Develop a degradation analysis model based on multivariate regression Y = b0 + b1P1 + b2P2 + Λ + b n P n +∈, where Y represents the degree of oil degradation, P n is the ratio of each oil component, b n is the regression coefficient, ∈ is the error term. Use high-precision sensors to monitor the key degradation indicators of oil products in real time D x , and apply data analysis methods to update the regression coefficient b of the degradation analysis model n Using the degradation analysis model, calculate the expected degradation degree Y of the mixed oil predicted =b0+b1P1+b2P2+Λ+b n P n Using the actual oil degradation data Y actual and the predicted value Y predicted Compare and verify the accuracy of the model. According to the difference between the actual data and the predicted value, adjust the regression coefficient b n . Regularly optimize the proportions of blended oils and usage conditions based on the latest monitoring data and degradation trends.
[0021] As a preferred solution, the phase shifter oil product status parameters to be monitored are determined based on the phase shifter structure design and usage characteristics, including:
[0022] According to the mechanical parts of the transmission gears and bearings in the phase shifter, the main mechanical failure modes are determined, including tooth surface wear and journal wear; through the use characteristics of the phase shifter, the viscosity and oil film strength of the oil are determined as the key monitoring status parameters. The use characteristics include the role of the oil in lubrication and isolation, and the use characteristics include frequent start and stop and high load; an online viscosity sensor is installed at the oil pool of the phase shifter to detect the dynamic viscosity change of the lubricating oil and monitor the lubrication status; an ultrasonic sensor is used to measure the thickness and strength of the oil film at the lubrication point to evaluate the lubrication performance of the oil; an online gas content detection sensor is installed in the hydraulic system pipeline to detect the bubble content and identify common fluid forces. Mechanical faults, including cavitation and emulsification; use temperature array sensors to measure the hot spot distribution and temperature rise at the bearing, and evaluate the lubrication and cooling status; monitor the generation and accumulation of bubbles in the oil, and evaluate the fluid state of the hydraulic system; set up an acceleration sensor on the bearing housing to collect bearing noise and vibration signals, perform spectrum analysis, and determine the degree of bearing wear; set up a torque sensor at the input and output ends of the phase regulator to detect the input torque and output torque, and calculate the power transfer efficiency; use a speed sensor to obtain the rotation speed data of the phase regulator to evaluate the impact of the lubricant on the speed and power transmission; evaluate the lubricant and hydraulic working conditions based on the speed signal and torque data.
[0023] As a preferred solution, the method of judging whether the oil product status is abnormal during the sensor monitoring process and taking corresponding treatment measures according to the status judgment result when it is abnormal includes:
[0024] For various parameters of oil status, including viscosity, oil film strength, bubble content, temperature, vibration, noise, speed, and torque data, difference thresholds are set as standards for normal operation and potential abnormalities; based on the change value of the oil status parameter and the preset difference threshold, identify whether there is a trend of accelerated decrease or accelerated increase, and judge whether the trend continues; if a parameter value continues to be abnormal or the change trend is abnormal, the situation is judged as a status warning; after the warning occurs, retrieve historical operation data, determine the cause of the warning, and formulate corresponding treatment measures for the determined attribution; if various parameters of the oil status are monitored to be abnormal at the same time, use the decision tree algorithm to train the model based on the various parameter data of the oil status, build a mechanical failure prediction model for the phase shifter, and identify the potential mechanical failure mode of the phase shifter; according to the potential mechanical failure mode of the phase shifter, based on the historical repair record of the phase shifter, correct the phase shifter operating parameters; if high temperature and high pressure occur at the same time, reduce the load and increase the oil supply to restore the equipment to normal working condition; through closed-loop feedback adjustment with the monitoring system, continuously optimize and improve the health of the lubrication system to achieve early prediction and avoidance of failures.
[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 real-time monitoring of the oil status of a phase shifter based on deep learning. The real-time and accuracy of the phase shifter oil status monitoring are significantly improved, and the risk of equipment failure and downtime is effectively prevented. By real-time monitoring of the quality of the newly added oil and comparing it with the existing database, the method can promptly detect any significant changes in the oil quality, so as to take preventive measures in advance. In addition, the quality evaluation and degradation prediction of the mixed oil provide an accurate prediction of the service life of the oil, making the maintenance and replacement plan more scientific and efficient. By sampling and testing the oil samples at different stages during the use of the mixed oil, the monitoring and analysis of the oil status are further optimized, ensuring the reliability of the equipment operation. In addition, the monitoring parameters determined according to the design and use characteristics of the phase shifter provide strong data support for the stable operation of the equipment. By timely identifying the abnormal oil status and taking corresponding treatment measures, the operating efficiency and safety of the entire system are effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 This is a flow chart of a method for real-time monitoring of phase-shifting oil status based on deep learning of the present invention.
[0028] Figure 2 It is a schematic diagram of a real-time monitoring method of phase-shifting oil status based on deep learning of the present invention.
[0029] Figure 3 This is another schematic diagram of a method for real-time monitoring of phase shifter oil status based on deep learning of the present invention. DETAILED DESCRIPTION
[0030] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.
[0031] In this embodiment, a method for real-time monitoring of oil status of a condenser based on deep learning may specifically include:
[0032] Step S101, using a near infrared spectrometer to detect in real time the additive composition and impurity content data of a newly added oil sample in the fuel tank, and to establish an oil quality database.
[0033] According to the user manual of the near-infrared spectrometer, set different wavelengths of light to illuminate the oil sample newly added to the tank; detect the spectrum of the sample, record the intensity of reflected or transmitted light, and identify the characteristic absorption peaks of different additive components and impurities. Determine the additive component or impurity content based on the intensity of the characteristic peak. Determine the additive component and impurity content of each additive component by repeated scanning and comparison with the standard. If the content of a certain component of the test sample exceeds the threshold, it is marked as an unqualified sample, otherwise, it is marked as a qualified sample. Establish a one-to-one correspondence between the sample addition time, additive component and impurity content data according to the oil batch number and the test time, obtain the oil quality data of each batch at each time point, and store it in the oil quality database. When new oil products are added later, set the near-infrared spectrometer to monitor the new oil added to the tank in real time. Obtain the additive component and impurity content data of the newly tested sample through the scanning results, and update the oil quality database in real time according to the addition time and batch number. If the real-time monitoring results show that the additive component and impurity content of the newly added sample meet the standard, confirm the addition. If it does not meet the standard, stop adding and record abnormal data.
[0034] For example, a near-infrared spectrometer is used to detect the additive components and impurity content of a diesel sample. First, the spectrometer is set to irradiate the diesel sample newly added to the tank at different wavelengths, the wavelength range is selected to be 900-1700 nanometers, and the light intensity of each wavelength is set to 1000. Next, the spectrum of the sample is scanned and recorded, and the reflected or transmitted light intensity is recorded. A characteristic absorption peak is detected at 1300 nanometers, and its light intensity is 800. According to the previous comparison with the standard, it is determined that when the characteristic peak intensity is lower than 600, the additive component or impurity content is low, when the characteristic peak intensity is between 600-900, the additive component or impurity content is medium, and when the characteristic peak intensity is higher than 900, the additive component or impurity content is high. By repeating the scanning test and comparing with the standard, the content of each additive component and impurity in the diesel sample can be determined. The detected characteristic peak intensity is 750. According to the previously set standard, it can be determined that the content of the additive component or impurity is medium. Then, a one-to-one correspondence is established between the sample addition time, additive composition and impurity content data, the oil batch number and the test time, and stored in the oil quality database. The sample addition time is set to January 1, 2021, the additive composition and impurity content are medium, the batch number is A001, and the test time is January 2, 2021. These data are stored in the oil quality database. When new oil products are added later, a near-infrared spectrometer is set to monitor the new oil added to the tank in real time. Through the scanning results, the additive composition and impurity content data of the newly tested sample can be obtained, and the oil quality database is updated in real time according to the addition time and batch number. If the real-time monitoring results show that the additive composition and impurity content of the newly added sample meet the standards, the addition is confirmed. If it does not meet the standards, the addition is terminated and the abnormal data is recorded. The additive composition and impurity content monitored in real time are low, the addition time is January 3, 2021, and the batch number is A002. According to the real-time monitoring results, the addition of the sample is confirmed.
[0035] Step S102, using principal component analysis to compare the differences between the new oil quality data and the original oil quality database, and evaluating the oil quality based on the extracted principal components and oil performance indicators.
[0036] According to the data of additive composition and impurity content of each batch of samples in the original oil quality database, a multidimensional quality data matrix is established. The principal components of the matrix are extracted by the singular value decomposition algorithm in the principal component analysis method, and the contribution rate of the extracted principal components to the sample difference is calculated. If the contribution rate of the first principal component is higher than the preset threshold, the first principal component is used as the evaluation index to judge the value of the new and old oil samples on the first principal component item to reflect the quality difference between the two. If the contribution rate of the first principal component is lower than the preset threshold, the contribution rate of the second principal component is continued to be judged until the cumulative contribution rate is higher than the preset threshold. Obtain oil performance indicators, which include viscosity, flash point, acid value, alkalinity value, anti-wear ability, antioxidant performance, and impurity content; according to the extracted principal components and oil performance indicators, the decision tree algorithm is used for model training, and an oil quality evaluation model is established to obtain the quality evaluation of the oil. If the difference between the evaluation results of the new oil and the old oil is greater than the preset threshold, the new oil data is recorded and the quality database is updated. Otherwise, the new oil data is directly saved without processing. According to the latest extracted main components and oil quality evaluation model, the quality data of each batch of new oil is analyzed to determine whether there is a quality difference with the old oil. If there is a quality difference with the old oil, an early warning will be immediately fed back, and the oil will be added after the oil test, and the database will be updated in real time.
[0037] For example, there is a batch of new oil samples and a batch of old oil samples, and we want to evaluate their quality differences through principal component analysis and decision tree algorithm. First, there is an oil quality database, which contains the additive composition and impurity content data of multiple batches of samples. These data are constructed into a multidimensional quality data matrix. Viscosity, flash point, acid value, base value, anti-wear ability, antioxidant performance and impurity content are selected as oil performance indicators. First, the singular value decomposition algorithm in the principal component analysis method is used to extract the principal components of the matrix, and 3 principal components are extracted. Then, the contribution rate of each principal component to the sample difference is calculated. The contribution rate of the first principal component is 70%, the contribution rate of the second principal component is 20%, and the contribution rate of the third principal component is 10%. Since the contribution rate of the first principal component is higher than the preset threshold of 50%, the first principal component will be used as an evaluation indicator to judge the values of the new and old oil samples on the first principal component item. The value of the new oil sample on the first principal component is 8, while the value of the old oil sample on the first principal component is 5. According to this value, it can be judged that there is a difference in quality between new and old oil products. If the contribution rate of the first principal component is lower than the preset threshold, the contribution rate of the second principal component will continue to be judged until the cumulative contribution rate is higher than the preset threshold. Next, the decision tree algorithm is used for model training to establish an oil quality evaluation model. The principal components and oil performance indicators are used as features, and the quality difference is used as a label. The trained model shows that if the difference between the evaluation results of new oil and old oil is greater than the preset threshold of 0.2, the new oil data is recorded and the quality database is updated. Otherwise, the new oil data is directly saved without processing. Finally, according to the latest extracted principal components and oil quality evaluation model, the quality data of each batch of new oil is analyzed to determine whether there is a quality difference with the old oil. The analysis results show that the new oils with quality differences from the old oils are batches A, B and D. Immediate feedback warnings are given, and the oil is added after the oil is tested, and the database is updated in real time. However, there is no quality difference between the new oil of batch C and the old oil, so the new oil data is directly saved.
[0038] Step S103, establishing a mixed oil degradation prediction model based on the historical performance data of the mixed oil to predict the degradation degree of the oil.
[0039] Obtain historical data on the performance indicators of mixed oils, including viscosity, flash point, acid value, alkalinity value, anti-wear ability, antioxidant performance, and impurity content, and use the principal component analysis method to determine the indicators that have an impact on the degradation of mixed oils greater than the preset threshold. Based on the indicators that have an impact on the degradation of mixed oils greater than the preset threshold and the consumption rate of additives, use a multi-layer perceptron neural network for model training to build a mixed oil degradation prediction model to predict the service life of the oil; compare the simulation results of the mixed oil degradation prediction model with the test data of the actual oil samples to determine the optimal consumption rate. During the use of mixed oils, continuously perform status testing and analysis on the obtained oil samples. If there is a difference between the test results and the prediction results of the mixed oil degradation prediction model, fine-tune the parameters of the mixed oil degradation prediction model based on the obtained test data, and provide real-time feedback on the difference judgment results between the test and the prediction.
[0040] For example, there is a new oil sample data, the indicators in the historical data are viscosity 50, flash point 200, acid value 5,
[0041] The base value is 2, the anti-wear ability is 80, the antioxidant performance is 90, and the impurity content is 0.1. If the preset threshold is 0.8, it is necessary to determine the indicators whose impact on the degradation of the mixed oil is greater than the threshold. First, the historical data is processed using principal component analysis to obtain the principal component score of each indicator. The principal component scores obtained are principal component 1: 0.8, principal component 2: 0.6, principal component 3: 0.9, principal component 4: 0.5, principal component 5: 0.7, principal component 6: 0.4, and principal component 7: 0.3. According to the principal component scores, the indicators whose impact on the degradation of the mixed oil is greater than the preset threshold can be determined. If the scores of principal component 1 and principal component 3 are greater than or equal to the threshold of 0.8, it can be determined that viscosity and antioxidant performance are important indicators that affect the degradation of the mixed oil. Next, a multi-layer perceptron neural network is used for model training, using the indicators and additive consumption rate in the historical data as input features and the service life of the oil as output. The model obtained by training has the following input layers: viscosity, flash point, acid value, alkalinity value, anti-wear ability, antioxidant performance, impurity content, additive consumption rate, hidden layer: 10 neurons, and output layer: oil service life. The model is used to predict the service life of new oil sample data, and the predicted result is 500 hours. The simulation results are compared with the test data of actual oil samples. The service life of the actual oil sample is 480 hours, which is different from the simulation results. According to the obtained test data, the mixed oil degradation prediction model can be fine-tuned, such as adjusting the number of neurons in the hidden layer or increasing the amount of training data. The adjusted model can be re-predicted to improve the prediction results. At the same time, according to the difference judgment results, the difference between detection and prediction can be fed back in real time.
[0042] Based on the chemical composition of the mixed oil, key chemical reactions are identified and kinetic parameters are determined. A mixed oil degradation model is constructed and real-time monitoring is implemented to optimize maintenance and replacement plans.
[0043] Use high-resolution mass spectrometry and nuclear magnetic resonance technology to analyze the various chemical components in the mixed oil and determine the chemical structure of various additives and possible reaction products. Use historical experimental records to identify key chemical reactions occurring in the mixed oil, including oxidation reactions and polymerization reactions. Experimentally determine the kinetic parameters of each key reaction, including the reaction rate constant k. Establish kinetic equations for each key reaction d[A] represents the concentration change of reactant A. Using experimental data, the reaction rate constant k is determined by nonlinear regression analysis. The kinetic equations of individual reactions are combined to construct a mixed oil degradation model. Where d[R] is the concentration change of the reaction product R, k i and k d are the rate constants for reaction and decomposition, respectively, and A i is the concentration of reactant A, B i is the concentration of reactant B, D i is the concentration of decomposition product D. During the use of mixed oil products, real-time monitoring is implemented to obtain the temperature, pressure and running time data of the oil products. The monitoring data is combined with the mixed oil degradation model to quantitatively analyze and simulate the rate and mode of the oil degradation process. The prediction accuracy of the model is verified by comparing with the data of laboratory tests and field sampling. The model parameters are adjusted according to the comparison results to ensure that the model can accurately reflect the degradation behavior of the mixed oil products under actual use conditions. The model prediction results are used to guide the maintenance and replacement plan of the oil products and optimize the operation strategy of the equipment.
[0044] For example, using high-resolution mass spectrometry and nuclear magnetic resonance technology to analyze a diesel fuel, it was found that there were two different additives, antioxidant A and synergist B. Using historical experimental records, it was found that the key chemical reactions in diesel fuel are oxidation and polymerization. First, the kinetic parameters of the oxidation reaction were experimentally determined. By experimentally measuring the oxidation rate of diesel fuel under different temperature conditions, the oxidation reaction rate constant k1 can be obtained. The initial concentration of antioxidant A in diesel fuel is A i According to the kinetic equation where represents the concentration of antioxidant A. The value of the reaction rate constant k1 can be obtained by nonlinear regression analysis. Next, the kinetic parameters of the polymerization reaction are experimentally determined. By experimentally measuring the polymerization rate of diesel fuel under different temperature conditions, the polymerization reaction rate constant k2 can be obtained. The initial concentration of a reactant B in diesel fuel is B i According to the kinetic equation The value of the reaction rate constant k2 can be obtained by nonlinear regression analysis. After the reaction is completed, the concentration of the decomposition product D is D i and the decomposition rate k d The kinetic equations of oxidation reaction and polymerization reaction are combined to construct a mixed oil degradation model. By real-time monitoring of the temperature, pressure and running time data of diesel fuel, relevant monitoring data can be obtained. The monitoring data is combined with the degradation model to quantitatively analyze and simulate the rate and mode of the oil degradation process. If at a certain moment, the diesel fuel temperature is 50°C, the pressure is 2MPa, and the running time is 1000 hours, the reactant concentrations and in the degradation model can be calculated based on the monitoring data. Then, according to the kinetic equation of the degradation model, the degradation rate of diesel fuel can be calculated. The prediction accuracy of the model can be verified by comparing with the data of laboratory tests and field sampling. If the degradation rate predicted by the model is consistent with the experimental or field data, it means that the model can accurately reflect the degradation behavior of the mixed oil under actual use conditions. If there is a difference, the model parameters need to be adjusted, such as adjusting the values of the reaction rate constants k1 and k2, to ensure the accuracy of the model. Finally, according to the results predicted by the model, corresponding oil maintenance and replacement plans can be formulated to optimize the operation strategy of the equipment. When the model predicts that the concentration of antioxidant A in diesel fuel drops to a certain level, antioxidant A can be added in time to slow down the degradation rate of diesel fuel and increase its service life.
[0045] Step S104, divide the operation stages according to the service life of the mixed oil and implement regular sampling, record the operation parameters and perform quality inspection, establish an oil quality prediction model and an oil quality index threshold warning model, regularly feedback the operation diagnosis results, and optimize the use conditions of the mixed oil.
[0046] According to the expected service life of the mixed oil, the entire use process is divided into three stages: initialization period, normal operation period, and exhaustion period. A fixed sampling time is set in each stage for regular sampling. If an abnormality is detected at the sampling point, including a sudden change in the indicator value, the sampling detection frequency is increased. During the sampling process, the operating parameters of the oil are recorded, including operating temperature, flow rate, and load. The collected oil samples are quality tested to obtain oil quality test data, including the additive content, oxidation product content, viscosity, acid value, and water content in the oil. By repeating multiple tests, the quality test data of the mixed oil at each stage of the use process is obtained. Based on the quality test data of the mixed oil at each stage obtained, ARIMA is used for model training to establish an oil quality prediction model to determine the changing trends of oil quality indicators at different use stages. Based on the oil quality monitoring data, the logistic regression algorithm is used for model training, and an oil quality index threshold warning model is constructed to obtain the probability of oil degradation. The oil status is determined based on the oil degradation rate. The degree of oil degradation includes normal, concern, critical, and expired. If the probability of oil degradation is higher than the preset threshold, a warning signal is triggered, the status results are fed back and the results are used for operation diagnosis and system optimization.
[0047] For example, the estimated service life of a mixed oil is 1000 hours, and the sampling interval is a scheduled sampling every 100 hours. During the initialization period, the first sampling is performed, and the initial operating temperature of the oil is recorded as 50°C, the flow rate is 10m / s, and the load is 500kg. The quality inspection data shows that the additive content is 10%, the oxidation product content is 5ppm, the viscosity is 50cSt, the acid value is 5mgKOH / g, and the water content is 1%. During the normal operation period, the second sampling is performed, and the operating temperature of the oil is recorded as 60°C, the flow rate is 12m / s, and the load is 600kg. The quality inspection data shows that the additive content is 9%, the oxidation product content is 8ppm, the viscosity is 55cSt, the acid value is 7mgKOH / g, and the water content is 2%. During the exhaustion period, the third sampling is performed, and the operating temperature of the oil is recorded as 70°C, the flow rate is 15m / s, and the load is 800kg. The quality inspection data showed that the additive content was 8%, the oxidation product content was 10ppm, the viscosity was 60cSt, the acid value was 9mgKOH / g, and the water content was 3%. By repeating the test multiple times, more stages of quality inspection data were obtained. Based on these data, the ARIMA model was used for training to establish an oil quality prediction model, which can predict the change trend of oil quality indicators over time, such as viscosity may gradually increase with the increase of usage time. At the same time, the logistic regression algorithm was used for model training to construct an oil quality indicator threshold warning model. If the preset threshold is 20% oil degradation probability, if the oil degradation probability in a certain sampling is 25%, the warning signal is triggered and the status result is fed back. The oil status is determined according to the oil degradation rate, and the 25% degradation probability can be classified as a state of concern.
[0048] By identifying and monitoring the chemical composition and reactions of additives in mixed oils, a dynamic chemical interaction model is established and combined with conventional performance tests to monitor oil stability, predict degradation trends in long-term use, and adjust additive formulations to optimize oil performance.
[0049] Using high performance liquid chromatography, the chemical composition of various additives in the mixed oil is identified and quantitatively analyzed, and the type, concentration and expected chemical activity of each additive are recorded. Using a gas chromatography-mass spectrometry system, the chemical reaction products between additives are monitored in real time to determine the reaction paths between additives, including any side reactions that may cause performance degradation or oil degradation. A dynamic chemical interaction model based on historical data of the chemical composition of mixed oils is established, and the Schrödinger equation or wave function method is used to predict the interactions between additives. The historical data of the chemical composition of mixed oils includes the mixing ratio and operating conditions. The operating conditions include temperature, pressure, and running time. Combined with the conventional performance tests of oils and the results of chemical analysis of additives, the performance of oils is continuously monitored. Conventional performance tests include viscosity and acid value tests. Using MATLAB, the monitoring data is analyzed to identify abnormal trends or signs of performance degradation. The degradation trend and stability of oils during long-term use are predicted using the established chemical interaction model and historical monitoring data. Based on the type, concentration and expected chemical activity of additives, the ARIMA algorithm is used for model training to predict future performance changes of oils. According to the model prediction results and historical data of the chemical composition of the mixed oil, the additive formula is adjusted to reduce adverse chemical reactions and optimize oil performance.
[0050] For example, the additives in a mixed oil are identified and quantitatively analyzed by high performance liquid chromatography. Now, the additives in a certain automobile engine oil, including anti-wear agents and antioxidants, are to be analyzed. The concentrations of anti-wear agents and antioxidants in the sample are determined by high performance liquid chromatography, and their types and concentrations are recorded. If the anti-wear agent in the oil is found to be dioctyl dithiocarbonate DODTC, with a concentration of 5wt%; the antioxidant is diphenyl diisopropylphenol DPPD, with a concentration of 3wt%. Next, the chemical reaction products between the additives are monitored in real time using a gas chromatography-mass spectrometry system. It is found that a reaction occurs between DODTC and DPPD, producing thiophenol compounds, which may cause the performance of the oil to deteriorate. In order to determine the reaction path between the additives, a dynamic chemical interaction model based on the historical data of the chemical composition of the mixed oil is established using the Schrödinger equation or wave function method. Based on the historical data of the chemical composition of the mixed oil, a model is established to predict the reaction path between DODTC and DPPD. The historical data of the chemical composition of the mixed oil includes the mixing ratio and the operating conditions, and the operating conditions include temperature, pressure, and running time. At the same time, routine performance tests of the oil were conducted, including viscosity and acid value tests. The viscosity of the oil was measured to be 10 cSt and the acid value was 5 mgKOH / g. MATLAB was used to analyze the monitoring data to identify abnormal trends or signs of performance degradation. It was found that the viscosity of the oil had a significant upward trend over the past period of time, while the acid value also increased slightly. Based on the established chemical interaction model and historical monitoring data, the degradation trend and stability of the oil during long-term use can be predicted. According to the model prediction, as the use time increases, the thiophenol compounds in the oil may further increase, resulting in a decrease in oil performance. Based on the type and concentration of the additive and the expected chemical activity, an ARIMA model was trained using historical data to predict future changes in viscosity. Based on the model prediction results and the historical data of the chemical composition of the mixed oil, the viscosity was less than 8 cSt after 3 weeks. The additive formula can be adjusted to reduce adverse chemical reactions and optimize oil performance. If the model predicts that the thiophenol content of the oil will continue to increase, it can be considered to reduce DODTC or replace the antioxidant to reduce the occurrence of adverse reactions.
[0051] Step S105, comparing the data detected during the use of the mixed oil with the prediction results of the oil quality prediction model to determine the actual degradation degree of the mixed oil.
[0052] The quality inspection data obtained at different sampling stages of the mixed oil products are compared with the oil quality data of the corresponding stages predicted by the oil quality prediction model, and the error between the actual value and the predicted value is calculated. The value with the largest error between the data of each state point in the oil quality inspection data during use and the predicted result is extracted. The quality problem of the mixed oil is determined according to the maximum error value, and whether the actual degradation degree is consistent with the predicted result is judged. If the oil quality index in the obtained oil quality inspection data is higher than the predicted value of the oil quality prediction model, it is judged that the oil degradation degree is serious. If there is a deviation between the actual oil state and the oil state predicted by the oil quality index threshold warning model, the oil quality index threshold warning model is readjusted for the sampling state point with the largest error, including correcting the degradation rate of the relevant quality index, correcting the degradation path and degradation trend. And for the adjusted oil quality index threshold warning model, the state inspection data is used for iteration until the prediction result is consistent with the actual degradation degree. According to the inspection data of the mixed oil, the prediction effect of the adjusted oil quality index threshold warning model is evaluated.
[0053] For example, the quality inspection data of mixed oil products at different sampling stages are collected, and the oil quality prediction model is used to predict the oil quality at each stage. Now let's calculate the error between the actual value and the predicted value, and find the sampling state point with the largest error to get the data. The actual value is [85,88,90,92,86], and the predicted value is [82,87,89,91,85]. The error value of each stage can be calculated, and the error value is [3,1,1,1,1]. From these error values, the maximum value is found, and the maximum error value is 3, which means that at a certain sampling state point, the actual value exceeds the predicted value by 3 units. According to the maximum error value, the quality problem of the mixed oil can be judged. If the actual oil quality index is higher than the predicted value, it means that the oil degradation is serious. However, if there is a deviation between the actual oil state and the oil state predicted by the oil quality index threshold warning model, the oil quality index threshold warning model needs to be readjusted. The sampling state point with the largest error can be corrected, including correcting the degradation rate of the relevant quality index, correcting the degradation path and degradation trend. The quality indicator threshold warning model was readjusted, and the prediction results were obtained. The adjusted prediction values were [83, 87, 89, 91, 85]. The error between the actual value and the adjusted prediction value can be calculated again. The adjusted error value is [2, 1, 1, 1, 1]. The maximum value is found from the adjusted error value, and the maximum error value is 2. This means that at a certain sampling state point, the actual value exceeds the adjusted prediction value by 2 units. By repeatedly iteratively adjusting the quality indicator threshold warning model, the prediction results can be gradually made consistent with the actual degree of degradation. Finally, the prediction effect of the adjusted oil quality indicator threshold warning model can be evaluated, and other statistical methods or evaluation indicators can be used for analysis.
[0054] According to the amount of mixed oil added and the proportion of components, a measurement model based on key degradation indicators is established. Through real-time monitoring and data comparison, the degradation prediction is adjusted, and the mixed oil proportion and usage conditions are optimized.
[0055] Before each mixing, record the amount of each oil added V i . Calculate the proportion of each component in the mixed oil, Where ∑V is the total volume or weight of the mixed oil. Define the key degradation indicators of the oil, including viscosity V d 、Acid value A v and additive concentration C a For each indicator, establish the measurement model D x =a0+a1C a +a2V d +a3A v +a4T+a5L, where D x represents the degradation index of the oil, T represents temperature, L represents load, a0, a1, ..., a5 are coefficients determined by data analysis, reflecting the influence of each factor on the degradation index. Develop a degradation analysis model based on multivariate regression Y = b0 + b1P1 + b2P2 + Λ + b n P n +∈, where Y represents the degree of oil degradation, P n is the ratio of each oil component, b n is the regression coefficient, ∈ is the error term. Use high-precision sensors to monitor the key degradation indicators of oil products in real time D x , and apply data analysis methods to update the regression coefficient b of the degradation analysis model n Using the degradation analysis model, calculate the expected degradation degree Y of the mixed oil predicted =b0+b1P1+b2P2+Λ+b n P n Using the actual oil degradation data Y actua l and predicted value Y predicted Compare and verify the accuracy of the model. According to the difference between the actual data and the predicted value, adjust the regression coefficient b n . Regularly optimize the proportions of blended oils and usage conditions based on the latest monitoring data and degradation trends.
[0056] For example, there are two oils, A and B, which are used to manufacture lubricants for certain mechanical equipment. It is necessary to calculate the proportion of each component based on the amount of each oil added in the mixed oil, and establish a degradation analysis model to evaluate the degree of degradation of the oil. If 100 liters of oil are mixed, of which 60 liters of oil A and 40 liters of oil B are added. The proportion of oil A is 60 / 100 = 60%, and the proportion of oil B is 40 / 100 = 40%. Establish a measurement model for the key degradation indicators of oil products and select viscosity V d 、Acid value A v and additive concentration C a As the key degradation indicator. The measurement model can be expressed as D x =a0+a1C a +a2V d +a3A v +a4T+a5L, where a0, a1, a2, a3, a4, and a5 are coefficients determined through data analysis. Develop a degradation analysis model based on multivariate regression. There are three oil component ratios P1, P2, and P3. The degradation analysis model can be expressed as Y=b0+b1P1+b2P2+b3P3+∈, where Y represents the degree of oil degradation and ∈ is the error term. Use high-precision sensors to monitor key degradation indicators of oil products in real time D x , and apply data analysis methods to update the regression coefficient b of the degradation analysis model n The viscosity V of the oil measured by the sensor d 50, acid value A v is 2, the additive concentration C a is 5. The expected degradation degree Y of the mixed oil is calculated using the degradation analysis model predicted According to the actual measured oil component ratio and degradation index, substitute them into the regression model. If the regression coefficients are b0=1, b1=2, b2=5, and b3=3, the expected degradation degree is calculated as Y predicted =1+2*6+5*4+3*0+∈. Using the actual collected oil degradation data Y actual and the predicted value Y predicted Compare and verify the accuracy of the model. If the actual oil degradation data collected is Y actual =8, the regression coefficient b can be adjusted according to the difference between the actual data and the predicted value n , to improve the accuracy of the model. According to the latest monitoring data and degradation trends, the proportion of mixed oils and usage conditions are regularly optimized. Based on the new data analysis, the addition amount of oils A and B can be adjusted to obtain better degradation control effect.
[0057] Step S106, determining the phase shifter oil product status parameters that need to be monitored based on the phase shifter structure design and usage characteristics.
[0058] According to the mechanical components of the transmission gears and bearings in the phase regulator, the main mechanical failure modes are determined, including tooth surface wear and journal wear. Based on the use characteristics of the phase regulator, the viscosity and oil film strength of the oil are determined as the key monitoring state parameters. The use characteristics include the role of the oil in lubrication and isolation, and the use characteristics include frequent start and stop and high load. An online viscosity sensor is set at the oil pool of the phase regulator to detect the dynamic viscosity change of the lubricating oil and monitor the lubrication state. An ultrasonic sensor is used to measure the thickness and strength of the oil film at the lubrication point to evaluate the lubrication performance of the oil. An online gas content detection sensor is installed in the hydraulic system pipeline to detect the bubble content and identify common fluid mechanics failures, including cavitation and emulsification. A temperature array sensor is used to measure the hot spot distribution and temperature rise at the bearing to evaluate the lubrication and cooling status. The generation and accumulation of bubbles in the oil are monitored to evaluate the fluid state of the hydraulic system. An acceleration sensor is set at the bearing housing to collect bearing noise and vibration signals, perform spectrum analysis, and determine the degree of bearing wear. A torque sensor is set at the input and output ends of the phase regulator to detect the input torque and output torque and calculate the power transfer efficiency. Use a speed sensor to obtain the rotation speed data of the condenser to evaluate the impact of lubricating oil on speed and power transmission. Based on the speed signal and torque data, evaluate the working status of lubricating oil and hydraulics.
[0059] For example, according to the mechanical parts of the transmission gears and bearings in the phase shifter, the main mechanical failure modes, including tooth surface wear and journal wear, are determined. Through the use characteristics of the phase shifter, the key role of oil in lubrication and isolation is judged, and the viscosity and oil film strength of the oil are determined as key monitoring state parameters, such as viscosity of 50cSt and oil film strength of 5kPa. An online viscosity sensor is set at the oil pool of the phase shifter to detect the dynamic viscosity change of the lubricating oil and monitor the lubrication state. During operation, the viscosity sensor detected that the viscosity of the oil rose to 60cSt, indicating that the lubricating oil may produce metal particles due to tooth surface wear, affecting the lubrication effect. Ultrasonic sensors are used to measure the thickness and strength of the oil film at the lubrication point to evaluate the lubrication performance of the oil. The ultrasonic sensor measured that the oil film thickness at the lubrication point was 20 microns and the oil film strength was 4kPa, indicating that the lubrication performance of the lubricating oil is still within the normal range. An online gas content detection sensor is installed in the hydraulic system pipeline to detect the bubble content and identify common fluid mechanics failures such as cavitation and emulsification. When the gas content sensor detects that the bubble content in the hydraulic system exceeds the normal range, it can be judged that there may be cavitation or emulsification faults. Use the temperature array sensor to measure the hot spot distribution and temperature rise at the bearing to evaluate the lubrication and cooling status. During operation, the temperature array sensor detected a local hot spot at the bearing, and the temperature rise reached 60°C, indicating that there may be problems with the lubrication and cooling status. Monitor the generation and accumulation of bubbles in the oil and evaluate the fluid state of the hydraulic system. When the bubble sensor detects a sharp increase in the number of bubbles in the oil, it can be judged that the hydraulic system may have an abnormal fluid state and needs to be handled in time. Set up an acceleration sensor on the bearing housing to collect bearing noise and vibration signals, perform spectrum analysis, and determine the degree of bearing wear. The vibration signal collected by the acceleration sensor shows an obvious peak after spectrum analysis, indicating that the bearing may have a wear problem. Set a torque sensor at the input and output ends of the phase regulator to detect the input torque and output torque and calculate the power transfer efficiency. The torque sensor detects that the input torque is 100Nm and the output torque is 80Nm, and the power transfer efficiency is calculated to be 80%. Use the speed sensor to obtain the rotation speed data of the condenser to evaluate the effect of the lubricant on the speed and power transmission. The speed sensor records that the rotation speed of the condenser is 5000rpm, indicating that the lubricant has a good effect on the speed and power transmission. According to the speed signal and torque data, the lubricant and hydraulic working conditions are evaluated. According to the speed signal and torque data analysis, it is found that the high viscosity of the lubricant leads to poor hydraulic working conditions, and it is necessary to replace the appropriate lubricant.
[0060] Step S107, during the sensor monitoring process, determine whether the oil product status is abnormal, and if abnormal, take corresponding treatment measures according to the status determination result.
[0061] For various parameters of oil status, including viscosity, oil film strength, bubble content, temperature, vibration, noise, speed, and torque data, difference thresholds are set as the standard for normal operation and potential abnormalities. According to the change value of the oil status parameter and the preset difference threshold, identify whether there is a trend of accelerated decrease or accelerated increase, and judge whether the trend continues. If a parameter value is continuously abnormal or the change trend is abnormal, the situation is judged as a status warning. After the warning occurs, retrieve the historical operation data, determine the attribution of the warning, and formulate corresponding treatment measures for the determined attribution. If the various parameters of the oil status are monitored to be abnormal at the same time, the decision tree algorithm is used for model training based on the various parameter data of the oil status to build a mechanical failure prediction model for the phase shifter, identify the potential mechanical failure mode of the phase shifter; according to the potential mechanical failure mode of the phase shifter, based on the historical repair record of the phase shifter, the phase shifter operating parameters are corrected. If high temperature and high pressure occur at the same time, reduce the load and increase the oil supply to restore the equipment to normal working state. Through closed-loop feedback adjustment with the monitoring system, the health state of the lubrication system is continuously optimized and improved, and early prediction and avoidance of failures are achieved.
[0062] For example, viscosity is an important oil status parameter. The preset difference threshold is 20, which means that when the difference between the oil status parameter and the preset upper and lower thresholds is less than 20, it can be considered that there is an accelerated decreasing trend. If the viscosity value is 80 in the last detection, the current viscosity value is 90, and the viscosity change value is 10, which is less than the preset difference threshold of 20, it means that there is no accelerated increasing trend. If the viscosity value is continuously observed to decrease gradually, it can be judged that the viscosity change trend is abnormal. If the viscosity value is continuously observed to decrease continuously, it can be judged that the situation is a status warning, and the attribution of the warning can be determined based on historical operation data. According to historical operation data, when the viscosity continues to decrease, the temperature rises rapidly, and it can be judged that the viscosity decrease is caused by the increase in temperature. For this attribution, corresponding treatment measures can be formulated, such as adjusting the operating parameters of the cooling system and reducing the temperature. If the parameters of the oil status are monitored to be abnormal at the same time, the decision tree algorithm can be used for model training to build a phase-shifting machine failure prediction model. By monitoring the abnormalities of multiple parameters such as viscosity decrease, oil film strength decrease, and bubble content increase, the decision tree algorithm can be used to identify the mechanical failure mode as tooth surface wear. According to the tooth surface wear, the operating parameters of the condenser can be corrected according to the historical repair records. According to the historical repair records, under the condition of tooth surface wear, increasing the oil supply can reduce the incidence of mechanical failures, and the amount of oil supply can be adjusted according to this attribution. If high temperature and high pressure occur at the same time, the load can be reduced and the oil supply can be increased to restore the equipment to normal working condition. Through the closed-loop feedback adjustment with the monitoring system, when high temperature and high pressure are detected at the same time, the temperature and pressure can be reduced by reducing the load and increasing the oil supply to restore the equipment to normal working condition.
[0063] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. A real-time monitoring method for phase-shifting oil product status based on deep learning, characterized in that: The method comprises: Use near-infrared spectrometer to detect the additive composition and impurity content of new oil samples added to the fuel tank in real time and establish an oil quality database; The principal component analysis method is used to compare the differences between the new oil quality data and the original oil quality database, and the oil quality is evaluated based on the extracted principal components and oil performance indicators; Based on the historical performance data of mixed oil products, a mixed oil degradation prediction model is established to predict the degradation degree of the oil products; Divide the operation phases according to the service life of the mixed oil and implement regular sampling, record the operation parameters and conduct quality inspections, establish an oil quality prediction model and an oil quality index threshold warning model, regularly feedback the operation diagnosis results, and optimize the use conditions of the mixed oil; Compare the data detected during the use of the mixed oil with the prediction results of the oil quality prediction model to determine the actual degradation degree of the mixed oil; Determine the phase shifter oil status parameters that need to be monitored based on the phase shifter's structural design and usage characteristics; During the sensor monitoring process, determine whether the oil status is abnormal, and take corresponding treatment measures according to the status judgment results if it is abnormal; The mixed oil degradation prediction model is established based on the historical performance data of the mixed oil to predict the degradation degree of the oil, including: Obtain historical data of mixed oil performance indicators, including viscosity, flash point, acid value, alkalinity value, anti-wear ability, antioxidant performance and impurity content, and use the principal component analysis method to determine the indicators that have an impact on the degradation of the mixed oil that is greater than a preset threshold; use a multi-layer perceptron neural network to train the model based on the indicators that have an impact on the degradation of the mixed oil that is greater than the preset threshold and the consumption rate of additives, and build a mixed oil degradation prediction model to predict the service life of the oil; compare the simulation results using the mixed oil degradation prediction model with the test data of the actual oil sample to determine the optimal consumption rate; during the use of the mixed oil, continuously test and analyze the status of the obtained oil samples; if there is a difference between the test results and the prediction results of the mixed oil degradation prediction model, fine-tune the parameters of the mixed oil degradation prediction model based on the obtained test data, and provide real-time feedback on the difference judgment results between the test and the prediction; also includes: identifying key chemical reactions and determining kinetic parameters based on the chemical composition of the mixed oil, building a mixed oil degradation model and implementing real-time monitoring, and optimizing maintenance and replacement plans; The method of identifying key chemical reactions and determining kinetic parameters based on the chemical composition of the mixed oil, constructing a mixed oil degradation model and implementing real-time monitoring, and optimizing maintenance and replacement plans specifically includes: using high-resolution mass spectrometry and nuclear magnetic resonance technology to analyze various chemical components in the mixed oil and determine the chemical structures of various additives and possible reaction products; using historical experimental records to identify key chemical reactions occurring in the mixed oil, including oxidation reactions and polymerization reactions; experimentally determining the kinetic parameters of each key reaction, experimentally determining the oxidation rate of diesel fuel under different temperature conditions, and obtaining the oxidation reaction rate constant According to the kinetic equation ,in, represents the concentration of antioxidant A, and the reaction rate constant is obtained by nonlinear regression analysis. The polymerization rate constant can be obtained by experimentally measuring the polymerization rate of diesel fuel under different temperature conditions. , the initial concentration of reactant B in diesel fuel is According to the kinetic equation , the reaction rate constant was obtained by nonlinear regression analysis The kinetic equations of individual reactions are combined to construct a mixed oil degradation model ,in, is the concentration change of the reaction product R, and are the rate constants for reaction and decomposition, respectively, is the concentration of reactant A, is the concentration of reactant B, is the concentration of decomposition product D; implement real-time monitoring during the use of mixed oil products to obtain the temperature, pressure and running time data of the oil products; combine the monitoring data with the mixed oil degradation model to quantitatively analyze and simulate the rate and mode of the oil degradation process; verify the prediction accuracy of the model by comparing with the data of laboratory tests and field sampling; adjust the model parameters according to the comparison results to ensure that the model can accurately reflect the degradation behavior of the mixed oil products under actual use conditions; use the model prediction results to guide the maintenance and replacement plan of the oil products and optimize the operation strategy of the equipment; The method of comparing the data detected during the use of the mixed oil product with the prediction result of the oil product quality prediction model to determine the actual degradation degree of the mixed oil product includes: The quality inspection data obtained at different sampling stages of the mixed oil are compared with the oil quality data of the corresponding stages predicted by the oil quality prediction model, and the error between the actual value and the predicted value is calculated; the value with the largest error between the data of each state point in the oil quality inspection data during use and the predicted result is extracted; the quality problem of the mixed oil is determined based on the maximum error value, and whether the actual degree of degradation is consistent with the predicted result is judged. If the oil quality index in the obtained oil quality inspection data is higher than the predicted value of the oil quality prediction model, it is judged that the degree of oil degradation is serious; if the actual oil state is consistent with the oil state predicted by the oil quality index threshold warning model If there is a deviation, the oil quality index threshold warning model is readjusted for the sampling state point with the largest error, including correcting the degradation rate of the relevant quality index, and correcting the degradation path and degradation trend; and the adjusted oil quality index threshold warning model is iterated using the state detection data until the prediction result is consistent with the actual degradation degree; the prediction effect of the adjusted oil quality index threshold warning model is evaluated based on the detection data of the mixed oil; and it also includes: according to the amount of mixed oil added and the proportion of components, a measurement model based on key degradation indicators is established, the degradation prediction is adjusted through real-time monitoring and data comparison, and the proportion of mixed oil and the use conditions are optimized; According to the amount of mixed oil added and the proportion of components, a measurement model based on key degradation indicators is established, degradation prediction is adjusted through real-time monitoring and data comparison, and the proportion of mixed oil and usage conditions are optimized, specifically including: before each mixed use, the amount of each oil added is recorded ; Calculate the proportion of each component in the mixed oil, ,in Is the total volume or weight of the mixed oil; defines key oil degradation indicators, including viscosity , acid value and additive concentration ; For each indicator, establish a measurement model ,in Represents the degradation index of oil, T represents temperature, L represents load, , ..., It is a coefficient determined by data analysis, reflecting the influence of various factors on the degradation index; develop a degradation analysis model based on multivariate regression , where Y represents the degree of oil degradation, is the ratio of each oil component, is the regression coefficient, is an error term; use high-precision sensors to monitor key degradation indicators of oil products in real time , and apply data analysis methods to update the regression coefficients of the degradation analysis model ; Use the degradation analysis model to calculate the expected degradation degree of the mixed oil ; Use actual oil degradation data collected With the predicted value Compare and verify the accuracy of the model; adjust the regression coefficient according to the difference between the actual data and the predicted value ; Regularly optimize the proportion and usage conditions of mixed oils based on the latest monitoring data and degradation trends.
2. The method according to claim 1, wherein: The method uses a near infrared spectrometer to detect the additive composition and impurity content data of the oil sample newly added to the fuel tank in real time, and establishes an oil quality database, including: According to the user manual of the near-infrared spectrometer, set different wavelengths of light to illuminate the oil sample newly added to the tank; detect the spectrum of the sample, record the intensity of reflected or transmitted light, and identify the characteristic absorption peaks of different additive components and impurities; determine the additive component or impurity content based on the intensity of the characteristic peak; judge the content of each additive component and impurity by repeated scanning and testing and comparing with the standard; if the content of a certain component of the test sample exceeds the threshold, it is marked as an unqualified sample, otherwise, it is marked as a qualified sample; establish a one-to-one correspondence between the sample addition time and the additive component and impurity content data according to the oil batch number and the test time, obtain the oil quality data of each batch and each time point, and store it in the oil quality database; when new oil is added subsequently, set the near-infrared spectrometer to monitor the new oil added to the tank in real time; obtain the additive component and impurity content data of the new test sample through the scanning results, and update the oil quality database in real time according to the addition time and batch number; if the real-time monitoring results show that the additive component and impurity content of the new sample meet the standard, confirm the addition; if it does not meet the standard, stop adding and record the abnormal data.
3. The method according to claim 1, wherein: The principal component analysis method is used to compare the differences between the new oil quality data and the original oil quality database, and the oil quality is evaluated based on the extracted principal components and oil performance indicators, including: A multidimensional quality data matrix is established based on the additive composition and impurity content data of each batch of samples in the original oil quality database; the principal components of the matrix are extracted through the singular value decomposition algorithm in the principal component analysis method, and the contribution rate of the extracted principal components to the sample differences is calculated; if the contribution rate of the first principal component is higher than the preset threshold, the first principal component is used as an evaluation index to judge the values of the new and old oil samples on the first principal component item to reflect the quality difference between the two; if the contribution rate of the first principal component is lower than the preset threshold, the contribution rate of the second principal component is continued to be judged until the cumulative contribution rate is higher than the preset threshold; Obtaining oil performance indicators, including viscosity, flash point, acid value, alkalinity value, anti-wear ability, anti-oxidation performance and impurity content; using a decision tree algorithm to train a model based on the extracted main components and oil performance indicators, establishing an oil quality evaluation model, and obtaining an oil quality evaluation; if the difference between the evaluation results of the new oil and the old oil is greater than a preset threshold, then recording the new oil data and updating the quality database, otherwise, directly saving the new oil data without processing; Based on the most recently extracted principal components and oil quality evaluation model, the quality data of each batch of new oil is analyzed to determine whether there is a quality difference with the old oil. If there is a quality difference with the old oil, an early warning will be immediately fed back, and the oil will be added after testing, and the database will be updated in real time.
4. The method according to claim 1, wherein: The method divides the operation phases according to the service life of the mixed oil and implements regular sampling, records the operation parameters and performs quality inspection, establishes an oil quality prediction model and an oil quality index threshold warning model, regularly feeds back the operation diagnosis results, and optimizes the use conditions of the mixed oil, including: According to the expected service life of the mixed oil, the entire use process is divided into three stages: initialization period, normal operation period and exhaustion period; a fixed sampling time is set in each stage for regular sampling; If an abnormality is detected at the sampling point, and the abnormality includes a sudden change in the indicator value, the sampling detection frequency is increased; during the sampling process, the operating parameters of the oil are recorded, including the operating temperature, flow rate and load; the collected oil samples are quality tested to obtain oil quality test data, including the additive content, oxidation product content, viscosity, acid value and water content in the oil; through repeated multiple tests, the quality test data of the mixed oil at each stage of the use process is obtained; based on the quality test data of the mixed oil at each stage obtained, ARIMA is used for model training to establish an oil quality prediction model to determine the changing trend of oil quality indicators at different use stages; based on the oil quality monitoring data, a logistic regression algorithm is used for model training to construct an oil quality indicator threshold warning model to obtain the probability of oil degradation, and the oil status is determined based on the oil degradation rate. The degree of oil degradation includes normal, concern, critical and expired; If the probability of oil degradation is higher than the preset threshold, an early warning signal is triggered, the status results are fed back and the results are used for operation diagnosis and system optimization; it also includes: by identifying and monitoring the chemical composition and reaction of additives in mixed oils, establishing a dynamic chemical interaction model and combining it with conventional performance tests to monitor oil stability, predict degradation trends in long-term use, and adjust additive formulations to optimize oil performance; The method identifies and monitors the chemical composition and reactions of additives in mixed oils, establishes a dynamic chemical interaction model and combines it with conventional performance tests to monitor oil stability, predicts degradation trends during long-term use, and adjusts the additive formula to optimize oil performance. Specifically, it includes: using high performance liquid chromatography to identify and quantitatively analyze the chemical composition of various additives in mixed oils, and recording the type, concentration and expected chemical activity of each additive; using a gas chromatography-mass spectrometry system to monitor the chemical reaction products between additives in real time, and determine the reaction paths between additives, including any side reactions that may cause performance degradation or oil degradation; establishing a dynamic chemical interaction model based on historical data on the chemical composition of mixed oils, using the Schrödinger equation or wave function method to predict the interactions between additives. The historical data of the chemical composition of mixed oils include the mixing ratio and operating conditions, and the operating conditions include temperature, pressure, and running time; the conventional performance tests of oils and the chemical analysis results of additives are combined to continuously monitor the performance of oils. Conventional performance tests include viscosity and acid value tests; MATLAB is used to analyze the monitoring data to identify abnormal trends or signs of performance degradation; the established chemical interaction model and historical monitoring data are used to predict the degradation trend and stability of oils during long-term use; according to the type, concentration, and expected chemical activity of additives, the ARIMA algorithm is used for model training to predict future performance changes of oils; according to the model prediction results and the historical data of the chemical composition of mixed oils, the additive formula is adjusted to reduce adverse chemical reactions and optimize oil performance.
5. The method according to claim 1, wherein: According to the structural design and usage characteristics of the condenser, the condenser oil product status parameters that need to be monitored are determined, including: According to the mechanical components of the transmission gears and bearings in the phase shifter, the main mechanical failure modes are determined, including tooth surface wear and journal wear; through the use characteristics of the phase shifter, the viscosity and oil film strength of the oil are determined as the key monitoring status parameters. The use characteristics include the role of the oil in lubrication and isolation, and the use characteristics include frequent start and stop and high load; an online viscosity sensor is installed at the oil pool of the phase shifter to detect the dynamic viscosity change of the lubricating oil and monitor the lubrication status; an ultrasonic sensor is used to measure the thickness and strength of the oil film at the lubrication point to evaluate the lubrication performance of the oil; an online gas content detection sensor is installed in the hydraulic system pipeline to detect the bubble content and identify common fluid forces. Mechanical faults, including cavitation and emulsification; use temperature array sensors to measure the hot spot distribution and temperature rise at the bearing, and evaluate the lubrication and cooling status; monitor the generation and accumulation of bubbles in the oil, and evaluate the fluid state of the hydraulic system; set up an acceleration sensor on the bearing housing to collect bearing noise and vibration signals, perform spectrum analysis, and determine the degree of bearing wear; set up a torque sensor at the input and output ends of the phase regulator to detect the input torque and output torque, and calculate the power transfer efficiency; use a speed sensor to obtain the rotation speed data of the phase regulator to evaluate the impact of the lubricant on the speed and power transmission; evaluate the lubricant and hydraulic working conditions based on the speed signal and torque data.
6. The method according to claim 1, wherein: The method of determining whether the oil product status is abnormal during the sensor monitoring process and taking corresponding treatment measures according to the status determination result when the oil product status is abnormal includes: For various parameters of oil status, including viscosity, oil film strength, bubble content, temperature, vibration, noise, speed and torque data, difference thresholds are set as the standard for normal operation and potential abnormalities; according to the change value of the oil status parameter and the preset difference threshold, whether there is a trend of accelerated decrease or accelerated increase, and whether the trend continues; if a parameter value continues to be abnormal or the change trend is abnormal, it is judged as a status warning; After an early warning occurs, retrieve historical operating data to determine the cause of the early warning, and formulate corresponding treatment measures for the determined attribution; if all parameters of the oil status are monitored to be abnormal at the same time, use the decision tree algorithm to train the model based on the various parameter data of the oil status, build a phase-shifting machine mechanical failure prediction model, and identify the potential mechanical failure mode of the phase-shifting machine; according to the potential mechanical failure mode of the phase-shifting machine and based on the historical repair records of the phase-shifting machine, correct the phase-shifting machine operating parameters; if high temperature and high pressure occur at the same time, reduce the load and increase the oil supply to restore the equipment to normal working condition; through closed-loop feedback adjustment with the monitoring system, continuously optimize and improve the health of the lubrication system to achieve early prediction and avoidance of failures.
