Air separation unit health degree evaluation method and system based on data driving

Abstract

The invention relates to the technical field of mechanical equipment monitoring, in particular to an air separation unit health degree evaluation method and system based on data driving, and the method comprises the steps: obtaining a first data set of key parts of an air separation unit in an operation process; an intelligent edge measuring and calculating unit is arranged on a key part of the air separation device to carry out data preprocessing on the first data set, and operation deviation data and abnormal data are recognized and marked; constructing a health degree evaluation model to analyze the first standard data set, and generating a first health degree evaluation index; constructing a hybrid mechanism simulation model to perform analogue simulation on the operation deviation data and the abnormal data, analyzing differences between operation results, correcting the health degree evaluation model and the physical air separation device, correcting the first health degree evaluation index, and generating a second health degree evaluation index of the air separation device; and when the second health degree evaluation index is lower than the health degree threshold value of the air separation device, a self-adaptive operation and maintenance strategy is simulated and generated through a hybrid mechanism simulation model.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical equipment monitoring, and particularly to a data-driven health assessment method and system for an air separation unit. Background Art

[0002] With the rapid advancement of industrial automation and informatization, the air separation unit, as a core equipment in industrial production, plays a crucial role in multiple key fields such as chemical industry, metallurgy, and energy. However, due to its complex structure, variable operating conditions, and high-load operation characteristics, the health status of the air separation unit not only directly affects production efficiency but is also closely related to the safety and stability of the system.

[0003] Traditional health assessment methods mainly rely on offline analysis of operation data. Although this method can reflect the equipment status to a certain extent, it shows obvious lag when facing sudden failures, and it is often difficult to take effective maintenance measures in a timely manner. In addition, existing technical means, such as regular inspections, performance tests, and predictive maintenance based on historical data, although they can ensure the normal operation of the equipment to a certain extent, still have significant limitations. First of all, these methods can only detect problems after a failure has occurred or the performance has significantly declined, lacking the ability to actively identify early fault hidden dangers, resulting in delayed maintenance. Secondly, in the face of the complex and variable operating conditions of the air separation unit, a single data analysis method is difficult to comprehensively and accurately evaluate the overall health status of the equipment, especially in the case of alternating operation of multiple working conditions, it is more likely to have misjudgments or missed judgments. Finally, when a failure occurs, existing methods often lack efficient intelligent response capabilities, are difficult to quickly locate the root cause of the problem, and are even less able to provide dynamic optimization of maintenance and adjustment plans.

[0004] Therefore, a data-driven health assessment method and system for an air separation unit are proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a data-driven health assessment method and system for an air separation unit, which obtains a first data set of key components of the air separation unit during operation; performs data preprocessing on the first data set through intelligent edge measurement units arranged on the key components of the air separation unit, and identifies and marks operation deviation data and abnormal data; constructs a health assessment model to analyze the first standard data set to generate a first health assessment index; constructs a hybrid mechanism simulation model to simulate the operation deviation data and abnormal data, analyzes the differences between the operation results, corrects the health assessment model and the physical air separation unit, and corrects the first health assessment index to generate a second health assessment index of the air separation unit; when the second health assessment index is lower than the health threshold of the air separation unit, simulates through the hybrid mechanism simulation model and generates an adaptive operation and maintenance strategy.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for evaluating the health of an air separation unit based on data driving, comprising: Obtaining a first data set of key components of the air separation unit during operation; the first data set includes a historical first data set and a real-time first data set; arranging intelligent edge measurement units on the key components of the air separation unit; performing data preprocessing on the first data set through the intelligent edge measurement units to obtain a first standard data set, operation deviation data, and abnormal data; identifying and marking the obtained operation deviation data and abnormal data; Based on the historical first data set, constructing a health evaluation model for the air separation unit, analyzing the real-time first standard data set, and generating a first health evaluation index for the air separation unit; Coupling the health evaluation model of the air separation unit with the physical mechanism model of the air separation unit to construct a hybrid mechanism simulation model; Using the hybrid mechanism simulation model to perform simulation on the operation deviation data and abnormal data, analyzing the difference between the simulation result and the actual operation result, and correcting the health evaluation model and the physical mechanism model of the air separation unit; Based on the corrected hybrid mechanism simulation model, correcting the first health evaluation index to generate a second health evaluation index for the air separation unit; When the second health evaluation index is lower than the health threshold of the air separation unit, simulating and generating an adaptive operation and maintenance strategy through the hybrid mechanism simulation model to obtain a correction strategy for the air separation unit; the adaptive operation and maintenance strategy includes simulating and optimizing the operation parameters of the air separation unit, triggering preventive maintenance operations, and triggering device shutdown protection in a linked manner.

[0007] Preferably, the intelligent edge measurement unit includes a data preprocessing layer, an anomaly detection and marking layer, a data transmission layer, and a data update layer; The data preprocessing layer obtains a first standard data set through filtering, denoising, standardization, and feature extraction; The anomaly detection and marking layer classifies the feature vectors through a pre-trained random forest model to identify abnormal data; analyzes the unmarked data through a clustering model to obtain operation deviation data; marks and stores the abnormal data and operation deviation data; The data transmission layer stores the first standard data set, abnormal data, and operation deviation data; The data update layer applies the adaptive operation and maintenance strategy data generated by the hybrid mechanism simulation model to the key components of the air separation unit.

[0008] Preferably, the health assessment model of the air separation unit includes a feature extraction layer, a feature selection layer, a feature fusion unit, and a health assessment acquisition layer; The feature extraction layer extracts features from the first standard data set through a deep learning model to obtain a first feature vector of key components; the first feature vector includes time-domain features, frequency-domain features, and data transmission features; The feature selection layer analyzes the first feature vector through the principal component analysis method to obtain a second feature vector related to the health; The feature fusion unit fuses the second feature vectors to obtain the fusion features of the key components in the air separation unit; The health assessment acquisition layer analyzes the fusion features of the key components in the air separation unit to obtain the first health assessment data of the air separation unit, and optimizes the first health assessment data of the air separation unit through multiple regression analysis. By synthesizing the first optimized health assessment data of the air separation unit, a first health assessment index is obtained.

[0009] Preferably, the process of obtaining the first health assessment index is as follows: S10. Analyze the fusion features of the key components in the air separation unit to obtain the first health assessment data of the air separation unit; the formula for the first health assessment data is: ; Wherein, is the first health assessment data, is the number of key components in the air separation unit, is the th health of the key component, is the th health weight of the key component, is the health calculation function, is the th th operating parameter of the key component; S20. Optimize the first health assessment data of the air separation unit through multiple regression analysis to obtain the first optimized health assessment data. The specific formula is: ; Wherein, is the standardized first optimized health assessment data, and are the minimum and maximum values of, is the natural constant, is the regression coefficient, is the error term; S30. Obtain the first health evaluation index based on the first health evaluation data and the first optimized health evaluation data. The specific calculation formula is: ; where, is the first health evaluation index, is the fusion weight of the first health evaluation index.

[0010] Preferably, the process of correcting the health evaluation model of the air separation unit and the physical air separation unit by the hybrid mechanism simulation model is as follows: S100. Input the operation deviation data and abnormal data into the hybrid mechanism simulation model; S200. Analyze the input data based on the health evaluation model of the air separation unit to generate the health of the simulation key components; simulate the operation state of the device under different working conditions based on the physical characteristics of the physical air separation unit, and simulate the dynamic behavior of the air separation unit through finite element analysis; S300. Output the health of the simulated air separation unit based on the health evaluation model of the air separation unit; obtain the health of the actual air separation unit based on the physical air separation unit; S400. Obtain the difference in health score and the difference in operation parameters, and calculate the error; the error includes the health score error and the operation parameter error; S500. Adjust the health evaluation model of the air separation unit and the physical air separation unit through the error; S501. Add the error to the training set of the health evaluation model of the air separation unit, perform incremental learning on the health evaluation model of the air separation unit, and increase the penalty for abnormal data points; S502. Adjust the parameters in the physical mechanism model to minimize the error between the simulation result and the actual result; S600. Through iterative steps of S100 to S500, when the error is lower than the preset threshold, stop the correction process.

[0011] Preferably, the process of obtaining the second health evaluation index is as follows: Calculate the error based on the difference between the hybrid mechanism simulation model and the actual operation data; the error includes the health score error and the operation parameter error; Design a non-linear correction function according to the error characteristics. The calculation formula is: ; where, is the health score error, is the th key component th operation parameter error, , , and are correction coefficients, is the number of key components in the air separation unit, is the Sigmoid activation function; By comprehensively considering the first health index and the correction amount, and fusing them through the weight coefficient, the second health evaluation index is obtained; ; Among them, is the second health evaluation index, is the fusion weight of the second health evaluation index.

[0012] An air separation unit health evaluation system based on data-driven includes: A data acquisition unit that acquires the first data set of the key components of the air separation unit during operation; the first data set includes a historical first data set and a real-time first data set; an intelligent edge measurement unit is arranged on the key components of the air separation unit; the first data set is preprocessed by the intelligent edge measurement unit to obtain a first standard data set, operation deviation data, and abnormal data; the obtained operation deviation data and abnormal data are identified and marked; A first model construction unit that constructs an air separation unit health evaluation model based on the historical first data set; uses the model to analyze the real-time first standard data set to generate the first health evaluation index of the air separation unit; A second model construction unit that couples the air separation unit health evaluation model with the physical mechanism model of the air separation unit to construct a hybrid mechanism simulation model; A second model correction unit that uses the hybrid mechanism simulation model to simulate the operation deviation data and abnormal data, analyzes the difference between the simulation result and the actual operation result, and corrects the air separation unit health evaluation model and the physical mechanism model; A health index acquisition unit that corrects the first health evaluation index based on the corrected hybrid mechanism simulation model to generate the second health evaluation index of the air separation unit; An adaptive operation and maintenance unit that, when the second health evaluation index is lower than the air separation unit health threshold, simulates and generates an adaptive operation and maintenance strategy through the hybrid mechanism simulation model to obtain the correction strategy of the air separation unit; the adaptive operation and maintenance strategy includes simulating and optimizing the operation parameters of the air separation unit, triggering preventive maintenance operations, and triggering device shutdown protection in a linked manner.

[0013] Preferably, the process of obtaining the second health evaluation index is as follows: Based on the difference between the hybrid mechanism simulation model and the actual operation data, the error is calculated; the error includes the health score error and the operation parameter error; Design a non - linear correction function according to the error characteristics, and the calculation formula is: ; where, is the error of the health score, is the th error of the th th and are correction coefficients respectively, is the number of key components in the air separation unit, is the Sigmoid activation function; By comprehensively integrating the first health index and the correction amount, and through the weight coefficient, the second health evaluation index is obtained; ; where, is the second health evaluation index, is the fusion weight of the second health evaluation index.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By combining the health evaluation model of the air separation unit with the physical mechanism model, the present invention forms a hybrid mechanism simulation model to achieve accurate prediction of the device operation state. By using the machine learning model to mine and analyze multi - dimensional sensor data, historical data and real - time data, it also makes a realistic simulation of the dynamic behavior of the device under different working conditions through physical simulation, enhances the ability to identify abnormal working conditions and potential faults, and effectively makes up for the deficiencies of traditional technologies in modeling accuracy, fault prediction and anomaly detection.

[0015] 2. The present invention uses the intelligent edge measurement unit to perform real - time filtering, denoising and anomaly detection on the first data set, obtains the first standard data set, and timely marks and stores the operation deviation data and anomaly data. The present invention reduces data lag and network burden, and through the collaborative operation with the physical mechanism model, generates the corrected health evaluation index in real - time, enabling the operation and maintenance of the air separation unit to transform from post - event response to predictive and adaptive adjustment, significantly improving the efficiency and accuracy of operation and maintenance decisions.

[0016] 3. The adaptive operation and maintenance strategy proposed by the present invention dynamically corrects the deviation between the simulation result and the actual data. When the second health evaluation index is lower than the set threshold, it automatically triggers the correction strategy for the device. It includes simulating and optimizing the operation parameters of the air separation unit and linkage shutdown protection, and also supports preventive maintenance operations, forming an active defense ability against the signs of faults, thereby reducing the risks of fault shutdown and safety hazards, and comprehensively improving the overall operation efficiency and economy of the air separation unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG. is a schematic flow chart of a method for evaluating the health of an air separation unit based on data driving provided by the present invention; Figure 2 FIG. is a schematic structural diagram of a system for evaluating the health of an air separation unit based on data driving provided by the present invention; Figure 3 FIG. is a schematic diagram for obtaining the evaluation of health provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without any creative work shall fall within the protection scope of the present invention.

[0019] Please refer to Figure 1 and Figure 2 The present invention provides a method for evaluating the health of an air separation unit based on data driving, which is applied to a system for evaluating the health of an air separation unit based on data driving. The technical solution is as follows: Obtain a first data set of key components of the air separation unit during operation; the first data set includes a historical first data set and a real-time first data set; arrange an intelligent edge measurement unit on the key components of the air separation unit; perform data preprocessing on the first data set through the intelligent edge measurement unit to obtain a first standard data set, operation deviation data, and abnormal data; identify and mark the obtained operation deviation data and abnormal data; apply to the data acquisition unit of a system for evaluating the health of an air separation unit based on data driving; the preprocessing includes filtering, denoising, standardization, and generating feature vectors; the first data set is composed of the operation parameters of the key components; among them, the key components mentioned in the present invention and the corresponding first data are shown in Table 1; Table 1 Key components and corresponding parameters

[0020] Furthermore, the intelligent edge measurement unit includes a data preprocessing layer, an anomaly detection and marking layer, a data transmission layer, and a data update layer; The data preprocessing layer obtains a first standard data set through filtering, denoising, standardization, and feature extraction; The anomaly detection and marking layer classifies the feature vectors through a pre-trained random forest model to identify abnormal data; analyzes the unlabeled data through a clustering model to obtain operation deviation data; marks and stores the abnormal data and operation deviation data; The data transmission layer stores the first standard data set, abnormal data, and operation deviation data; The data update layer applies the adaptive operation and maintenance strategy data generated by the hybrid mechanism simulation model to the key components of the air separation unit.

[0021] The intelligent edge measurement unit realizes real-time data collection, preprocessing, and anomaly detection in the health evaluation method of the present invention, greatly improving the efficiency and accuracy of data processing. By deploying edge units at the key components of the device, data can be filtered and preliminarily analyzed at the source, reducing the bandwidth requirements and latency of data transmission.

[0022] The present invention constructs an intelligent edge measurement unit on the key components to preprocess the acquired data and obtain abnormal data, ensuring the timely identification of abnormal data and providing a reliable data basis for the subsequent health evaluation model and operation and maintenance strategy. For specific data, refer to Table 2; Table 2 Data Processing Efficiency Table

[0023] Based on the historical first data set, a health evaluation model for the air separation unit is constructed; the model is used to analyze the real-time first standard data set to generate the first health evaluation index of the air separation unit; it is applied to the first model construction unit of a data-driven health evaluation system for the air separation unit; Furthermore, the health evaluation model of the air separation unit includes a feature extraction layer, a feature selection layer, a feature fusion unit, and a health evaluation acquisition layer; The feature extraction layer extracts features from the first standard data set through a deep learning model to obtain the first feature vector of the key components; the first feature vector includes time-domain features, frequency-domain features, and data transmission features; The feature selection layer analyzes the first feature vector through the principal component analysis method to obtain the second feature vector related to the health degree; the second feature vector is obtained through the principal component analysis method; The feature fusion unit fuses the second feature vectors to obtain the fusion features of the key components in the air separation unit; The health assessment acquisition layer analyzes the fusion features of key components in the air separation unit to obtain the first health assessment data of the air separation unit, and optimizes the first health assessment data of the air separation unit through multiple regression analysis. By synthesizing the first optimized health assessment data of the air separation unit, the first health assessment index is obtained.

[0024] The health assessment model of the air separation unit comprehensively applies machine learning and statistical methods, combines real-time collected data and physical mechanism simulation to achieve accurate assessment of the device health status and fault prediction. The model construction and optimization process cover data preprocessing, feature selection, algorithm application, model training and verification, parameter calculation, and continuous correction to ensure its efficiency and reliability in complex industrial environments.

[0025] The health assessment model provided by the present invention not only improves the accuracy and reliability of health assessment, but also can respond to potential faults in a timely manner through an adaptive operation and maintenance strategy, solving the problem of lag in detection and maintenance.

[0026] Further, the process of obtaining the first health assessment index is as follows: S10. Analyze the fusion features of key components in the air separation unit to obtain the first health assessment data of the air separation unit; the formula for the first health assessment data is: ; Where, is the first health assessment data, is the number of key components in the air separation unit, is the th health of the key component, is the th health weight of the key component, is the health calculation function, is the th th operating parameter of the th key component; The specific calculation formula of the health calculation function ; Where, is the th th weight of the th th operating parameter of the th key component, is the th th optimal operating parameter of the

[0027] When the operating parameter is equal to the optimal operating parameter, it indicates that the parameter is in the healthiest state; When the operating parameter deviates from the optimal operating parameter, the function value smoothly approaches 0 in a non-linear manner.

[0028] S20. Optimize the first health evaluation data of the air separation unit through multiple regression analysis to obtain the first optimized health evaluation data. The specific formula is: ; Among them, is the first optimized health evaluation data after standardization, and are the minimum and maximum values of, is the natural constant, is the regression coefficient, is the error term; S30. Obtain the first health evaluation index based on the first health evaluation data and the first optimized health evaluation data. The specific calculation formula is: ; Among them, is the first health evaluation index, is the fusion weight.

[0029] The first health evaluation index provided by the present invention provides a comprehensive quantitative evaluation of the operating state of the entire device by integrating the health scores of each key component in the air separation unit and using the methods of weighted average and multiple regression analysis. This formula can not only reflect the operating conditions of each key component, but also ensure the accuracy and reliability of the comprehensive index through the optimization of the weight coefficient and the regression model.

[0030] Couple the health evaluation model of the air separation unit with the physical mechanism model of the air separation unit to construct a hybrid mechanism simulation model; apply it to the second model construction unit of a data-driven air separation unit health evaluation system; Use the hybrid mechanism simulation model to simulate the operation deviation data and abnormal data, analyze the differences between the simulation results and the actual operation results, and correct the health evaluation model and the physical mechanism model of the air separation unit; apply it to the second model correction unit of a data-driven air separation unit health evaluation system; Further, the process of the hybrid mechanism simulation model correcting the health evaluation model and the physical air separation unit of the air separation unit is as follows: S100. Input the operation deviation data and abnormal data into the hybrid mechanism simulation model; S200. Analyze the input data based on the health assessment model of the air separation unit to generate the health of key components in the simulation; simulate the operating states of the unit under different working conditions based on the physical characteristics of the physical air separation unit, and through finite element analysis, simulate the dynamic behavior of the air separation unit; S300. Output the health of the simulated air separation unit based on the health assessment model of the air separation unit; obtain the health of the actual air separation unit based on the physical air separation unit; S400. Obtain the differences in health scores and operating parameters, and calculate the errors; the errors include health score errors and operating parameter errors; S500. Adjust the health assessment model of the air separation unit and the physical air separation unit based on the errors; S501. Add the errors to the training set of the health assessment model of the air separation unit, perform incremental learning on the health assessment model of the air separation unit, and increase the penalty for abnormal data points; S502. Adjust the parameters in the physical mechanism model to minimize the error between the simulation results and the actual results; S600. Through iterative steps of S100 to S500, when the error is lower than the preset threshold, stop the correction process.

[0031] By combining the data-driven model with the physical mechanism model, the health model can not only capture the potential patterns and anomalies in the data, but also provide a deeper understanding of the operating mechanism through physical simulation. This method effectively solves the lag problem of traditional air separation unit detection and maintenance, significantly improves the operating efficiency and reliability of the unit, and provides strong technical support for industrial automation and intelligent maintenance.

[0032] Based on the corrected hybrid mechanism simulation model, correct the first health assessment index to generate the second health assessment index of the air separation unit; applied to the health acquisition unit of a data-driven air separation unit health assessment system; Furthermore, the process of obtaining the second health assessment index is as follows: Calculate the errors based on the differences between the hybrid mechanism simulation model and the actual operating data; the errors include health score errors and operating parameter errors; Design a non-linear correction function according to the error characteristics, and the calculation formula is: ; where, is the health score error, is the th operating parameter error of the th key component, , , and are correction coefficients respectively, is the number of key components in the air separation unit, is the Sigmoid activation function; By comprehensively considering the first health index and the correction amount, and fusing them through weight coefficients, the second health evaluation index is obtained; ; wherein, is the second health evaluation index, is the fusion weight of the second health evaluation index.

[0033] The second health evaluation index provided by the present invention realizes the accurate evaluation of the health of the air separation unit by combining the first health evaluation index and the correction amount calculated based on the difference between the simulation model and the actual operation data. The design of the non-linear correction function enables the correction process to flexibly handle different types and degrees of errors, and the fusion of weight coefficients ensures the stability and accuracy of the comprehensive index. This method improves the real-time performance and reliability of the health evaluation by dynamically adjusting the health score, providing strong support for the efficient operation and preventive maintenance of the air separation unit; for the specific process of obtaining the health evaluation, refer to Figure 3 .

[0034] When the second health evaluation index is lower than the health threshold of the air separation unit, an adaptive operation and maintenance strategy is simulated and generated through a hybrid mechanism simulation model to obtain a correction strategy for the air separation unit; the adaptive operation and maintenance strategy includes simulating and optimizing the operation parameters of the air separation unit, triggering preventive maintenance operations, and triggering device shutdown protection in a linked manner. It is applied to the adaptive operation and maintenance unit of a data-driven air separation unit health evaluation system; when it is not lower than the health threshold of the air separation unit, the health evaluation of the air separation unit is output.

[0035] The present invention first realizes the real-time collection and preprocessing of historical data and real-time data during operation by arranging intelligent edge measurement units on key components; at the same time, the pre-trained random forest model and clustering analysis method are used to identify and mark operation deviation data and abnormal data, improving the accuracy and response speed of anomaly detection. Secondly, a health evaluation model based on deep learning and multiple regression analysis, combined with a hybrid mechanism simulation model coupled with a physical mechanism model, analyzes operation data to generate the first health evaluation index. By analyzing the difference between the simulation results and the actual operation results, the health evaluation model and the physical model are dynamically corrected, improving the accuracy and reliability of the health evaluation. In addition, based on the corrected hybrid mechanism simulation model, the first health evaluation index is non-linearly corrected to generate a more accurate second health evaluation index. The automated operation and maintenance strategy effectively reduces the failure risk and maintenance cost, significantly improving the operation efficiency and reliability of the air separation unit, refer to Table 3.

[0036] Table 3 Optimization Table of Operation and Maintenance Strategy

[0037] Table 3 Based on the revised health evaluation index, the adaptive operation and maintenance strategy generated by the present invention effectively reduces the equipment downtime, increases the number of preventive maintenance times, reduces the maintenance cost, and improves the operation efficiency and economy of the air separation unit.

[0038] Example 2 A data-driven health evaluation method and system for air separation units provided by the present invention, on the basis of Example 1, further includes coupling the health evaluation model of the air separation unit with the physical mechanism model of the air separation unit to construct a hybrid mechanism simulation model; the coupling process includes: 1) Data synchronization and integration In this embodiment, through a unified data interface, real-time data exchange between the two models is realized. The key operating parameters (such as temperature, pressure, flow rate, etc.) output by the health evaluation model will be passed as inputs to the physical mechanism model; at the same time, the simulation results (such as system response, energy consumption, etc.) of the physical mechanism model are fed back to the health evaluation model.

[0039] 2) Model interface design To achieve seamless coupling of the two models, a standardized interface protocol is designed. JSON or XML format is used for data exchange to ensure the compatibility and integrity of data during transmission. In addition, the format, frequency, and error handling mechanism of data transmission are defined in the interface to ensure the stability and reliability of data exchange.

[0040] 3) Parameter coordination and calibration Establish the corresponding relationship between the parameters of the health evaluation model and the parameters of the physical mechanism model. Map the pressure coefficient in the health evaluation model to the pressure transfer parameter in the physical mechanism model. Based on historical operation data, the initial parameters of the coupled model are calibrated, and the least squares method is used to adjust the key parameters to minimize the error between the simulation results and the actual operation data.

[0041] 4) Comparison and analysis of simulation results During the simulation process, the output results of the health evaluation model and the physical mechanism model are compared in real time. For example, monitor the temperature and pressure changes of the two models under the same operating conditions and calculate the differences. Through statistical indicators such as mean square error and coefficient of determination, the model fitting effect is evaluated, and systematic deviations and outliers are identified.

[0042] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A data-driven health assessment method for air separation units, characterized in that, Including: Obtain the first data set of the key components of the air separation unit during operation; The first data set includes a historical first data set and a real-time first data set; Arrange intelligent edge measurement units on the key components of the air separation unit; Preprocess the first data set through the intelligent edge measurement unit to obtain a first standard data set, operation deviation data, and abnormal data; Identify and mark the obtained operation deviation data and abnormal data; Based on the historical first data set, construct a health evaluation model for the air separation unit, analyze the real-time first standard data set, and generate the first health evaluation index of the air separation unit; Couple the health evaluation model of the air separation unit with the physical mechanism model of the air separation unit to construct a hybrid mechanism simulation model; Use the hybrid mechanism simulation model to simulate the operation deviation data and abnormal data, analyze the difference between the simulation result and the actual operation result, and correct the health evaluation model and the physical mechanism model of the air separation unit; Based on the corrected hybrid mechanism simulation model, correct the first health evaluation index to generate the second health evaluation index of the air separation unit; When the second health evaluation index is lower than the health threshold of the air separation unit, simulate and generate an adaptive operation and maintenance strategy through the hybrid mechanism simulation model to obtain the correction strategy of the air separation unit; The adaptive operation and maintenance strategy includes simulating and optimizing the operation parameters of the air separation unit, triggering preventive maintenance operations, and triggering device shutdown protection in a linked manner.

2. The method for evaluating the health of an air separation unit based on data driving according to claim 1, wherein: The intelligent edge measurement unit includes a data preprocessing layer, an anomaly detection and marking layer, a data transmission layer, and a data update layer; The data preprocessing layer obtains a first standard data set through filtering, denoising, standardization, and feature extraction; The anomaly detection and marking layer classifies the feature vectors through a pre-trained random forest model to identify abnormal data; analyzes the unmarked data through a clustering model to obtain operation deviation data; marks and stores the abnormal data and operation deviation data; The data transmission layer stores the first standard data set, abnormal data, and operation deviation data; The data update layer applies the adaptive operation and maintenance strategy data generated by the hybrid mechanism simulation model to the key components of the air separation unit.

3. A method for evaluating the health of an air separation unit based on data-driven according to claim 1, characterized in that: The health evaluation model of the air separation unit includes a feature extraction layer, a feature selection layer, a feature fusion unit, and a health evaluation acquisition layer; The feature extraction layer extracts features from the first standard data set through a deep learning model to obtain the first feature vector of the key component; the first feature vector includes time domain features, frequency domain features, and data transmission features; The feature selection layer analyzes the first feature vector through the principal component analysis method to obtain a second feature vector related to health; The feature fusion unit fuses the second feature vectors to obtain the fusion features of the key components in the air separation unit; The health assessment acquisition layer analyzes the fusion features of key components in the air separation unit to obtain the first health assessment data of the air separation unit, and optimizes the first health assessment data of the air separation unit through multiple regression analysis. By synthesizing the first optimized health assessment data of the air separation unit, the first health assessment index is obtained.

4. The method for evaluating the health of an air separation unit based on data driving according to claim 3, wherein: The process of obtaining the first health assessment index is as follows: S10. Analyze the fusion features of key components in the air separation unit to obtain the first health assessment data of the air separation unit; the formula for the first health assessment data is: ； Among them, is the first health evaluation data, is the number of key components in the air separation unit, is the health of the is the health weight of the is the health calculation function, is the operation parameter of the key component; S20. Optimize the first health assessment data of the air separation unit through multiple regression analysis to obtain the first optimized health assessment data. The specific formula is: ； Among them, is the first optimized health evaluation data after standardization, and is the minimum and maximum values of, is the natural constant, is the regression coefficient, is the error term; S30. Obtain the first health assessment index based on the first health assessment and the first optimized health assessment data. The specific calculation formula is: ； Among them, is the first health evaluation index, is the fusion weight of the first health evaluation index.

5. The method for evaluating the health of an air separation unit based on data driving according to claim 1, wherein: The process of the hybrid mechanism simulation model correcting the health assessment model of the air separation unit and the physical air separation unit is as follows: S100. Input the operation deviation data and abnormal data into the hybrid mechanism simulation model; S200. Analyze the input data based on the health assessment model of the air separation unit to generate the simulation health of key components; simulate the operation state of the device under different working conditions based on the physical characteristics of the physical air separation unit, and simulate the dynamic behavior of the air separation unit through finite element analysis; S300. Output the simulated health of the air separation unit based on the health assessment model of the air separation unit; obtain the actual health of the air separation unit based on the physical air separation unit; S400. Obtain the difference in health score and the difference in operation parameters, and calculate the error; the error includes the health score error and the operation parameter error; S500. Adjust the health assessment model of the air separation unit and the physical air separation unit through the error; S501. Add the error to the training set of the health assessment model of the air separation unit, perform incremental learning on the health assessment model of the air separation unit, and increase the penalty for abnormal data points; S502. Adjust the parameters in the physical mechanism model to minimize the error between the simulation result and the actual result; S600. Through iterative steps S100 to S500, when the error is lower than the preset threshold, stop the correction process.

6. The method for evaluating the health of an air separation unit based on data driving according to claim 1, wherein: The process of obtaining the second health assessment index is as follows: Calculate the error based on the difference between the hybrid mechanism simulation model and the actual operation data; the error includes the health score error and the operation parameter error; Design a non-linear correction function according to the error characteristics. The calculation formula is: ； Among them, is the health score error, is the th operating parameter error of the th key component, , , and are correction coefficients respectively, is the number of key components in the air separation unit, is the Sigmoid activation function; Obtain the second health assessment index through integration by weight coefficient of the first health index and the correction amount; ； Among them, is the second health evaluation index, is the fusion weight of the second health evaluation index, is the first health evaluation index.

7. A data-driven health assessment system for air separation units, characterized in that, Including: A data acquisition unit that acquires the first data set of key components of the air separation unit during operation; The first data set includes a historical first data set and a real-time first data set; Arrange intelligent edge measurement units on the key components of the air separation unit; Perform data preprocessing on the first data set through the intelligent edge measurement unit to obtain a first standard data set, operation deviation data, and abnormal data; identify and mark the obtained operation deviation data and abnormal data; The first model construction unit constructs a health evaluation model for the air separation unit based on the historical first data set, analyzes the real-time first standard data set, and generates the first health evaluation index of the air separation unit; The second model construction unit couples the health evaluation model of the air separation unit with the physical mechanism model of the air separation unit to construct a hybrid mechanism simulation model; The second model correction unit uses the hybrid mechanism simulation model to perform simulation on the operation deviation data and abnormal data, analyzes the difference between the simulation result and the actual operation result, and corrects the health evaluation model and physical mechanism model of the air separation unit; The health index acquisition unit corrects the first health evaluation index based on the corrected hybrid mechanism simulation model to generate the second health evaluation index of the air separation unit; The adaptive operation and maintenance unit, when the second health evaluation index is lower than the health threshold of the air separation unit, simulates and generates an adaptive operation and maintenance strategy through the hybrid mechanism simulation model to obtain the correction strategy of the air separation unit; The adaptive operation and maintenance strategy includes simulating and optimizing the operation parameters of the air separation unit, triggering preventive maintenance operations, and triggering the shutdown protection of the linkage trigger device.

8. The health assessment system of an air separation unit based on data-driven according to claim 7, wherein: The process of obtaining the second health evaluation index is as follows: Calculate the error based on the difference between the hybrid mechanism simulation model and the actual operation data; the error includes the health score error and the operation parameter error; Design a nonlinear correction function according to the error characteristics, and the calculation formula is: ； Among them, is the health score error, is the th operating parameter error of the th key component, , , and are correction coefficients respectively, is the number of key components in the air separation unit, is the Sigmoid activation function; Obtain the second health evaluation index by integrating the first health index and the correction amount through the weight coefficient; ； Among them, is the second health evaluation index, is the fusion weight of the second health evaluation index, is the first health evaluation index.

Images