Intelligent fault prediction method and system for air separation equipment

Abstract

This invention relates to the field of data prediction technology, and in particular to an intelligent prediction method and system for air separation equipment faults. The method includes the following steps: acquiring operating data including the outlet temperature of the current cycle's cold blowing stage; comparing the outlet temperature of the current cycle's cold blowing stage with a reference cold blowing temperature to obtain a temperature deviation; obtaining a thermal hysteresis characteristic index for the current cycle based on a weighted average of the temperature deviation, the rate of change of the outlet temperature during the cold blowing stage, and a time exponent; obtaining a comprehensive aging index based on a weighted sum of the thermal hysteresis characteristic index and the bed resistance pressure drop at the end of the cold blowing stage; correcting the standard mass transfer coefficient in the adsorption prediction model to obtain a target mass transfer coefficient; using the target mass transfer coefficient in the adsorption prediction model to predict the particle breakthrough time; and achieving intelligent prediction of air separation equipment faults based on a comparison between the breakthrough time and the adsorption running time, effectively improving the accuracy of intelligent prediction of air separation equipment faults.

Description

Technical Field

[0001] This invention relates to the field of data prediction technology, and in particular to an intelligent method and system for predicting faults in air separation equipment. Background Technology

[0002] Air separation units (ASUs), also known as air separation equipment, include molecular sieve purification systems used to remove impurities such as moisture, carbon dioxide, and hydrocarbons from the raw air. This prevents impurities from freezing and clogging the plate heat exchanger channels under subsequent cryogenic conditions, leading to operational malfunctions. Therefore, accurately predicting the breakthrough curve of the molecular sieve adsorber is crucial for avoiding safety accidents caused by impurity breakthrough.

[0003] For predicting the breakthrough curve of a molecular sieve adsorber, existing technologies often employ the particle filter (PF) algorithm, which constructs a state transition equation based on the adsorption mechanism and uses observational data to track and predict the system state in real time.

[0004] However, traditional particle filtering methods generally assume that key physicochemical parameters of molecular sieves, such as the mass transfer coefficient, remain constant throughout their entire lifespan, or decay according to a simple linear law. In reality, during long-term operation, molecular sieves undergo complex nonlinear and non-monotonic evolution of their internal mass transfer resistance and effective adsorption sites due to factors such as pulverization, compaction, micropore blockage, and chemical poisoning. This contradiction between the real time-varying nature of these physical parameters and the fixed nature of existing model parameters causes the state transition equations in the algorithm to gradually deviate from the actual physical process, failing to capture the significant decrease in the mass transfer coefficient with aging, resulting in predicted breakthrough times often lagging behind the actual occurrence time.

[0005] Therefore, how to accurately obtain intelligent prediction results of air separation equipment faults is an urgent problem to be solved. Summary of the Invention

[0006] To address the technical problem of how to accurately obtain intelligent prediction results of air separation equipment faults, this invention provides an intelligent prediction method and system for air separation equipment faults.

[0007] In a first aspect, the present invention provides an intelligent prediction method for air separation equipment faults, employing the following technical solution: A method for intelligent prediction of faults in air separation equipment, comprising the following steps: The system acquires operational data including the outlet temperature of the current cycle's cold blowing stage; compares the outlet temperature of the current cycle's cold blowing stage with the reference cold blowing temperature to obtain the temperature deviation; weights the temperature deviation, the rate of change of the outlet temperature during the cold blowing stage, and the time index to obtain the thermal hysteresis characteristic index of the current cycle; calculates the comprehensive aging index based on the weighted sum of the thermal hysteresis characteristic index and the bed resistance pressure drop at the end of the cold blowing stage; corrects the standard mass transfer coefficient in the adsorption prediction model to obtain the target mass transfer coefficient, which is positively correlated with the standard mass transfer coefficient and negatively correlated with the comprehensive aging index; uses the target mass transfer coefficient in the adsorption prediction model to predict the particle breakthrough time; and achieves intelligent prediction of air separation equipment failures based on the comparison between the breakthrough time and the adsorption running time.

[0008] To address the problem that existing particle filtering algorithms typically assume a constant mass transfer coefficient of the molecular sieve or decay according to a simple linear law, leading to prediction results lagging behind the actual breakthrough time after molecular sieve aging, this invention introduces a coupling mechanism between cold-blowing thermal hysteresis based on thermodynamic characteristics and bed resistance pressure drop based on structural characteristics. By calculating the thermal hysteresis characteristic index of the current cycle and the bed resistance pressure drop, a comprehensive aging index is obtained, and the standard mass transfer coefficient in the adsorption prediction model is dynamically corrected accordingly. This allows the model parameters to reflect in real time the actual physical state of increased mass transfer resistance and reduced effective adsorption sites within the molecular sieve, thereby accurately capturing the accelerated trend of the adsorption front after aging. This effectively improves the accuracy of air separation equipment fault prediction and thus effectively avoids downstream equipment freezing accidents caused by prediction lag.

[0009] According to the intelligent fault prediction method for air separation equipment provided by the present invention, the acquisition of operating data including the outlet temperature of the current cycle cold blowing stage includes: the current cycle includes an adsorption stage and a regeneration stage, the regeneration stage includes a cold blowing stage and a heating stage; in the adsorption stage, the inlet flow rate, inlet temperature, inlet pressure and outlet carbon dioxide concentration are collected; in the regeneration stage, the outlet temperature of the cold blowing stage and the bed resistance pressure drop at the end of the cold blowing are collected; the method also includes preprocessing the collected operating data, including using moving average filtering to remove noise, performing timestamp alignment and removing outliers.

[0010] This invention effectively solves the problems of high-frequency noise interference from sensors and inconsistent timing of multi-source data in the operating data of air separation equipment by acquiring multi-source data and combining preprocessing methods such as moving average filtering for noise reduction, timestamp alignment, and outlier removal, thus laying a solid foundation for subsequent data processing.

[0011] According to the intelligent prediction method for air separation equipment faults provided by the present invention, the method for obtaining the reference cold blowing temperature includes: continuously collecting outlet temperature data of the cold blowing stage for multiple cycles in the initial stage of new molecular sieve use; and obtaining the reference cold blowing temperature after preprocessing the outlet temperature data of the cold blowing stage.

[0012] According to the intelligent prediction method for air separation equipment faults provided by the present invention, obtaining the thermal hysteresis characteristic index of the current cycle includes: ; This is a characteristic indicator of thermal hysteresis in the current cycle. , These represent the start and end times of the current cycle's cold blowing phase, respectively. , These are the outlet temperature and the reference cold blowing temperature for the current cycle's cold blowing phase, respectively. The outlet temperature during the cold blowing stage of the current cycle. right The derivative at time t, For time coefficient, It is a natural constant. During the cold blowing stage time, For normalization function, It is the absolute value symbol.

[0013] This invention takes into account that the aging characteristics and structural damage of the deep molecular sieve bed are usually reflected in the outlet temperature curve in the later stage of the cold blowing process. Therefore, a specific integral formula including time exponent weighting is used to calculate the thermal hysteresis characteristic index. The time coefficient is used to amplify the weight of the time point in the later stage of the cold blowing stage, which improves the sensitivity of the diagnostic algorithm to the decay of heat transfer efficiency in the deep part of the bed, thereby accurately capturing potential aging signals.

[0014] According to the present invention, an intelligent prediction method for air separation equipment faults is provided, wherein the adsorption prediction model is a particle filter algorithm model.

[0015] This invention uses a particle filter algorithm model for prediction, which can simulate the complex physicochemical evolution during molecular sieve adsorption, thereby flexibly adapting to changes in operating conditions throughout the entire life cycle of the equipment and ensuring the robustness of the breakthrough time prediction under different aging levels.

[0016] According to the intelligent prediction method for air separation equipment failure provided by the present invention, the step of correcting the standard mass transfer coefficient in the adsorption prediction model to obtain the target mass transfer coefficient includes: ; This is the correction factor for the current cycle. This represents the effective surface area of ​​the molecular sieve in its initial state. It is a natural constant. The thermo-aging coefficient, The comprehensive aging index for the current cycle is used; the product of the correction factor for the current cycle and the standard mass transfer coefficient is used as the target mass transfer coefficient.

[0017] According to the present invention, an intelligent prediction method for air separation equipment faults is provided, wherein predicting the breakthrough time of particles using a target mass transfer coefficient in an adsorption prediction model includes: applying the target mass transfer coefficient of the current period to the adsorption prediction model, including embedding the target mass transfer coefficient into the state transition equation of the particle filtering algorithm to obtain the adsorption front position of the particles at each time in the next period; and obtaining the breakthrough time of the particles in response to the adsorption front position of the particles reaching the end of the molecular sieve bed at a certain time.

[0018] This invention directly embeds the dynamically corrected target mass transfer coefficient into the state transition equation of the particle filtering algorithm, updating the adsorption front movement velocity of simulated particles at the mechanistic level. This ensures that the model's evolution logic remains highly consistent with the actual physical mass transfer process, enabling precise calculation of the particle's position in the adsorption bed in the next cycle, thus providing an accurate prediction of the breakthrough time before the adsorption front reaches the end of the bed.

[0019] According to the present invention, an intelligent prediction method for air separation equipment faults is provided, wherein the intelligent prediction of air separation equipment faults is achieved based on the comparison result between the breakthrough time and the adsorption running time, including: taking the difference between the breakthrough time and the adsorption running time as the remaining safe time; and triggering an air separation equipment fault warning in response to the remaining safe time being lower than a safety threshold.

[0020] According to the intelligent prediction method for air separation equipment faults provided by the present invention, the intelligent prediction of air separation equipment faults further includes: pushing task work orders to operation and maintenance personnel, and simultaneously linking historical fault data of air separation equipment to assist in fault location.

[0021] Secondly, this invention provides an intelligent fault prediction system for air separation equipment, employing the following technical solution: An intelligent fault prediction system for air separation equipment includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned intelligent fault prediction method for air separation equipment is implemented.

[0022] By adopting the above technical solution, a computer program is generated from the above-mentioned intelligent fault prediction method for air separation equipment and stored in a memory so that it can be loaded and executed by a processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0023] The present invention has the following technical effects: Based on the above technical solutions, this invention provides an intelligent prediction method and system for air separation equipment faults. Addressing the problem that existing particle filter algorithms typically assume a constant molecular sieve mass transfer coefficient or decay according to a simple linear law, leading to prediction results lagging behind the actual breakthrough time after molecular sieve aging, this invention introduces a coupling mechanism between cold-blowing thermal hysteresis based on thermodynamic characteristics and bed resistance pressure drop based on structural characteristics. By calculating the thermal hysteresis characteristic index of the current cycle and the bed resistance pressure drop, a comprehensive aging index is obtained, and the standard mass transfer coefficient in the adsorption prediction model is dynamically corrected accordingly. This allows the model parameters to reflect in real time the actual physical state of increased internal mass transfer resistance and reduced effective adsorption sites within the molecular sieve, thereby accurately capturing the accelerated trend of the adsorption front after aging. This effectively improves the accuracy of air separation equipment fault prediction and thus effectively avoids downstream equipment freezing accidents caused by prediction lag. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an intelligent fault prediction method for air separation equipment provided in an embodiment of the present invention. Figure 2 This is a comparative schematic diagram of the cold blowing characteristics of molecular sieves provided in an embodiment of the present invention; Figure 3 This is a schematic diagram comparing the prediction effects of dynamic mass transfer coefficient and fixed mass transfer coefficient, provided as an embodiment of the present invention. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0026] To accurately obtain intelligent prediction results for air separation equipment faults, this invention discloses an intelligent prediction method for air separation equipment faults. For details, please refer to [link to details]. Figure 1 As shown, Figure 1 This is a flowchart illustrating an intelligent fault prediction method for air separation equipment provided in an embodiment of the present invention. The method specifically includes the following steps: S1: Obtain operating data including the outlet temperature of the current cycle's cold blowing phase.

[0027] For example, operational data can be collected from the molecular sieve purification system of the Air Separation Unit (ASU).

[0028] For example, in an embodiment of the present invention, acquiring operational data including the outlet temperature of the current cycle's cold blowing stage includes: the current cycle includes an adsorption stage and a regeneration stage, the regeneration stage includes a cold blowing stage and a heating stage; during the adsorption stage, the inlet air flow rate, inlet air temperature, inlet air pressure, and outlet carbon dioxide concentration are collected; during the regeneration stage, the outlet temperature of the cold blowing stage and the bed resistance pressure drop at the end of the cold blowing are collected.

[0029] Specifically, a flow meter can be used at the inlet pipeline of the adsorption tower to collect the inlet gas flow rate, a resistance temperature detector (RTD) or thermocouple can be used to collect the inlet gas temperature, a pressure transmitter can be used to collect the inlet gas pressure, and an online carbon dioxide analyzer can be used to collect the outlet carbon dioxide concentration. The sampling frequency can be set according to actual needs, and this embodiment of the invention does not impose too many restrictions.

[0030] It should be noted that air separation equipment generates a large amount of multi-source time-series data during the adsorption and regeneration stages. During adsorption, it is necessary to monitor inlet flow rate, inlet temperature, inlet pressure, and outlet carbon dioxide concentration in real time. During regeneration (including heating and cold blowing stages), it is necessary to collect the outlet temperature curve during the cold blowing stage and the bed resistance drop at the end of the cold blowing. This raw data may contain issues such as sensor noise, high-frequency fluctuations, and inconsistent timestamps. The aging state and adsorption performance of the molecular sieve are derived from this multi-dimensional data. If the raw data is not standardized and denoised, it will severely interfere with the accuracy of subsequent aging feature extraction and fault prediction. For example, high-frequency noise from sensors may distort the derivative calculation of the temperature curve, and misaligned timestamps may cause misalignment of input parameters in the adsorption model.

[0031] Therefore, after collecting the raw data, this step requires preprocessing the raw data.

[0032] For example, in an embodiment of the present invention, acquiring operating data including the outlet temperature of the current cycle cold blowing stage further includes: preprocessing the collected operating data, including using a moving average filter to remove noise, performing timestamp alignment, and removing outliers.

[0033] Specifically, during preprocessing, a moving average filtering algorithm is used to denoise the time-series data, such as the outlet temperature during the cold blowing stage, to eliminate interference from high-frequency noise from the sensors. Heterogeneous data from different sensors are aligned according to a unified time reference. Abnormal spikes caused by sensor malfunctions or communication interruptions are identified and removed. The specific preprocessing steps can be implemented using existing technologies, and will not be elaborated upon here.

[0034] Thus, by preprocessing the raw data, the embodiments of the present invention can effectively reduce the impact of sensor noise and time alignment issues, obtain stable and reliable operating data for the current cycle, and lay a data foundation for subsequent calculation of molecular sieve aging characteristics.

[0035] S2: Compare the outlet temperature of the current cycle's cold blowing stage with the reference cold blowing temperature to obtain the temperature deviation; based on the temperature deviation, the rate of change of the outlet temperature in the cold blowing stage, and the time exponent, obtain the thermal hysteresis characteristic index of the current cycle.

[0036] It should be noted that the above steps can obtain various operational data for the current cycle. However, the degree of molecular sieve aging is difficult to assess directly from adsorption data, and the lack of aging assessment will result in a lack of basis for correcting the prediction model. Once the degree of aging cannot be accurately assessed, particle filter prediction will continuously lag behind the actual breakthrough time, thus failing to achieve the purpose of early warning.

[0037] Furthermore, molecular sieve aging is a process of microstructural degradation, such as pulverization, compaction, or micropore blockage. This microstructural damage directly leads to a reduction in the effective specific surface area of ​​the molecular sieve, thereby decreasing the gas-solid heat transfer efficiency when the airflow passes through the molecular sieve bed during the regeneration cold blowing stage. Therefore, the reduced heat transfer efficiency manifests externally as a slower cooling rate during cold blowing, and temporally as a temperature lag phenomenon.

[0038] Based on this, the embodiments of the present invention can obtain the thermodynamic characteristics of the regeneration stage, and by evaluating the cold blowing thermal hysteresis phenomenon, reverse the aging state of the molecular sieve's microstructure.

[0039] For example, in an embodiment of the present invention, the method for obtaining the reference cold blowing temperature includes: continuously collecting outlet temperature data of the cold blowing stage for multiple cycles in the initial stage of the molecular sieve being put into use; and obtaining the reference cold blowing temperature after preprocessing the outlet temperature data of the cold blowing stage.

[0040] The baseline curve, constructed from the reference cold-blowing temperature, is established by averaging and preprocessing the outlet temperature data collected continuously over multiple cycles of cold-blowing during the initial use of the molecular sieve under design conditions. It represents the cooling curve under optimal conditions for the molecular sieve. By comparing the current cycle's cold-blowing temperature curve with the baseline curve, the cold-blowing thermal hysteresis characteristic index can be calculated. Preprocessing can be configured according to actual needs, and will not be elaborated upon in this embodiment.

[0041] For details, please refer to Figure 2 As shown, Figure 2 This is a schematic diagram comparing the characteristics of cold blowing of molecular sieves provided in an embodiment of the present invention, wherein the horizontal axis represents time in seconds, and the vertical axis represents the temperature at the outlet of the cold blowing stage.

[0042] As shown in the figure, due to the aging of the molecular sieve, its effective heat transfer surface area decreases, leading to a reduction in heat transfer efficiency and a slower cooling rate as the airflow passes through the bed. Therefore, the outlet temperature curve lags behind the baseline cold-blowing temperature curve. The shaded area between the two curves in the figure represents the thermal hysteresis region.

[0043] For example, in an embodiment of the present invention, the thermal hysteresis characteristic index of the current cycle is calculated, and the specific relationship can be found in the following formula: ; This is a characteristic indicator of thermal hysteresis in the current cycle. This marks the start of the current cycle's cold-blowing phase and represents the lower limit of the integral. This marks the end of the current cycle's cold-blowing phase and represents the maximum score. This refers to the outlet temperature during the cold blowing phase of the current cycle. This is the reference cold blowing temperature for the current cycle. The outlet temperature during the cold blowing stage of the current cycle. right The derivative at time t, For time coefficient, It is a natural constant. During the cold blowing stage time, It is an integral variable. For normalization function, It is the absolute value symbol.

[0044] The time coefficient is used to amplify the weight of the data in the later stage of cold blowing, thereby improving the diagnostic sensitivity of the deep aging characteristics of the molecular sieve bed. Its value ranges from 0.05 to 0.2; optionally, in this embodiment, it can be set to 0.1, and can be set according to actual needs. The normalization function can be a linear normalization function, used to normalize the calculation results and convert them into dimensionless parameters that can be used for model correction.

[0045] In the above relation, This temperature deviation can be used to assess the difference between the current cycle's temperature curve and the baseline cold-blowing temperature curve, directly reflecting the thermal hysteresis caused by molecular sieve aging. The larger the temperature deviation, the more severe the aging, and the greater the corresponding thermal hysteresis characteristic index.

[0046] This represents the rate at which the outlet temperature decreases during the cold blowing stage, and is used to weight temperature differences.

[0047] Indicates time-indexed weighting. This index indicates the duration of the cold blowing process and is used to reflect aging characteristics at data points in the later, more prominent stages of the bed. Because heat transfer degradation accumulates during the cold blowing phase, aging characteristics and structural damage in the deeper bed near the outlet only become apparent on the outlet temperature curve in the later stages of the cold blowing process. Therefore, by using time-weighted indexing, the weight of data points near the end of the cold blowing phase can be amplified, thus more accurately capturing and assessing the heat transfer efficiency degradation caused by deep bed aging.

[0048] Based on the above steps, the thermal hysteresis characteristic index for the current cycle can be obtained. The larger the value, the lower the gas-solid heat transfer efficiency and the more accelerated the aging of the physical structure. Thus, by comparing the deviation between the cold blow outlet temperature of the current cycle and the baseline curve, and by combining the cooling rate and time exponent for weighted integration and normalization, the present invention can accurately quantify the degree of heat transfer efficiency reduction caused by the aging of the molecular sieve microstructure, and obtain the aging index for subsequent model correction.

[0049] S3: The comprehensive aging index is obtained by weighting the thermal hysteresis characteristic index and the pressure drop of the bed resistance at the end of cold blowing; the standard mass transfer coefficient in the adsorption prediction model is corrected to obtain the target mass transfer coefficient, which is positively correlated with the standard mass transfer coefficient and negatively correlated with the comprehensive aging index.

[0050] It should be noted that the aging of molecular sieves is a complex physical process, which is not only reflected in a decrease in heat transfer efficiency, but may also manifest in changes in the bed structure, such as pulverization and compaction, ultimately leading to an increase in bed resistance. Therefore, increased bed resistance is another important indicator for assessing the structural state of molecular sieves. Using only a single thermal hysteresis index to assess the degree of aging will not fully reflect the structural degradation caused by aging, potentially leading to biased estimates of the aging state. Coupled thermodynamic and structural characteristics, a more comprehensive assessment of the overall aging degree of molecular sieves can be achieved.

[0051] Based on this, embodiments of the present invention can couple the thermal hysteresis characteristic index obtained in the above steps with the bed resistance pressure drop to synthesize a comprehensive aging index, so as to accurately assess the aging degree of molecular sieve and reduce the error in breakthrough time prediction.

[0052] For example, in this embodiment of the invention, the adsorption prediction model is a particle filtering algorithm model.

[0053] It should be understood that, in order to adapt to the input of the adsorption prediction model, the bed resistance pressure drop at the end of the cold blowing stage is a normalized dimensionless value.

[0054] The ratio of the initial bed resistance pressure drop to the initial reference value can be used as the bed resistance pressure drop.

[0055] For example, when obtaining the comprehensive aging index based on the weighted sum of the thermal hysteresis characteristic index and the bed resistance pressure drop at the end of cold blowing, the sum of the weights of the thermal hysteresis characteristic index and the bed resistance pressure drop at the end of cold blowing can be set to 1; optionally, in this embodiment of the invention, the weight of the thermal hysteresis characteristic index can be set to 0.6 and the weight of the bed resistance pressure drop can be set to 0.4; the specific settings can be made according to actual needs.

[0056] It's important to note that in traditional adsorption prediction models, the mass transfer coefficient is a core parameter, directly determining the velocity of the adsorption front and the breakthrough time. However, due to the aging of molecular sieves leading to a reduction in the effective adsorption surface area, the mass transfer coefficient also decreases. If the model uses a constant standard mass transfer coefficient, the predicted breakthrough time will lag significantly behind the actual situation, failing to provide accurate early warnings.

[0057] Furthermore, the linear driving force (LDF) mass transfer coefficient of a molecular sieve is directly proportional to its effective adsorption surface area. Based on this, embodiments of the present invention can use the comprehensive aging index obtained in the previous step to assess the degree of decay of the effective surface area, and then dynamically correct the standard mass transfer coefficient.

[0058] For example, in an embodiment of the present invention, the standard mass transfer coefficient in the adsorption prediction model is corrected to obtain the target mass transfer coefficient, as shown in the following formula: ; This is the correction factor for the current cycle. This represents the effective surface area of ​​the molecular sieve in its initial state. It is a natural constant. The thermo-aging coefficient, This is the comprehensive aging index for the current cycle.

[0059] The thermo-mass aging coefficient is used to adjust the sensitivity of the effective adsorption surface area to the overall aging degradation, and can be obtained through engineering calibration. Optionally, in this embodiment of the invention, it can be set to 0.15, and can be set according to actual needs.

[0060] In the above relationship, the correction factor ranges from 0 to 1, and its formula can be simplified to: Therefore, the correction factor is negatively correlated with the comprehensive aging index and is used to describe the rate of decline in mass transfer performance.

[0061] The effective adsorption surface area of ​​the molecular sieve in the current cycle represents the sites available for mass transfer. This exponential decay model accurately simulates the physical process of nonlinearly accelerated loss of effective mass transfer sites during the aging of the molecular sieve. The ratio of the effective adsorption surface area to the effective surface area of ​​the molecular sieve in its initial state is the correction factor for the current cycle.

[0062] After obtaining the correction factor for the current cycle according to the above steps, the standard mass transfer coefficient of the molecular sieve can be dynamically corrected using the correction factor to obtain the target mass transfer coefficient for the current cycle.

[0063] The standard mass transfer coefficient is the coefficient of the molecular sieve in its initial state, which can be obtained through existing technology. The specific details of this invention will not be elaborated here.

[0064] It should be understood that the target mass transfer coefficient for the current cycle is used to characterize the actual mass transfer rate of the molecular sieve. The higher the degree of aging and the lower the mass transfer efficiency, the smaller the corresponding target mass transfer coefficient should be. Therefore, the target mass transfer coefficient is negatively correlated with the comprehensive aging index. If the target mass transfer coefficient does not decrease with the degree of aging, it will continue the drawback of the fixed parameters of the traditional particle filter model, failing to reflect the decline in mass transfer capacity caused by aging. This will lead to the model's predicted breakthrough time lagging behind the actual time, causing accidents such as downstream equipment freezing.

[0065] For example, the product of the correction factor for the current period and the standard mass transfer coefficient can be used as the target mass transfer coefficient for the current period.

[0066] In this way, the embodiments of the present invention calculate the target mass transfer coefficient of the current cycle, so that the target mass transfer coefficient decreases as the degree of aging increases. The state transition equation of the particle filter can fit the real mass transfer process, thereby accurately capturing the trend of accelerated movement of the adsorption front after aging, predicting the penetration risk in advance, and ensuring the accuracy of the prediction.

[0067] S4: In the adsorption prediction model, the target mass transfer coefficient is used to predict the breakthrough time of particles; based on the comparison between the breakthrough time and the adsorption running time, intelligent prediction of air separation equipment failure is realized.

[0068] It should be noted that the above steps yield a target mass transfer coefficient that reflects the current degree of aging. By applying this target mass transfer coefficient to the adsorption prediction model, accurate prediction of the molecular sieve breakthrough time can be achieved.

[0069] For example, in an embodiment of the present invention, in the adsorption prediction model, using the target mass transfer coefficient to predict the breakthrough time of particles includes: applying the target mass transfer coefficient of the current period to the adsorption prediction model, including embedding the target mass transfer coefficient and the running data of the adsorption stage into the state transition equation of the particle filtering algorithm to obtain the adsorption front position of the particles at each time in the next period; in response to the adsorption front position of the particles reaching the end of the molecular sieve bed at a certain time, the breakthrough time of the particles is obtained.

[0070] The operational data for the adsorption phase includes the collection of inlet air flow rate, inlet air temperature, inlet air pressure, and outlet carbon dioxide concentration. This adsorption phase data serves as input to the state transition equation, allowing the predicted breakthrough time of particles to more closely approximate the actual operating conditions after aging.

[0071] It should be understood that the adsorption prediction model used in this embodiment of the invention is the particle filter (PF) algorithm model. The modified target mass transfer coefficient is embedded into the state transition equation of the particle filter to update the moving velocity of the adsorption front. The predicted breakthrough time is the moment when the particle set with a confidence level of 80% predicts that the adsorption front has reached the end of the bed. The specific steps can be implemented by existing technology, and will not be described in detail in this embodiment of the invention.

[0072] It should be noted that if the molecular sieve penetrates and is not switched on in time, it will cause a serious accident involving the freezing of downstream equipment by carbon dioxide. Therefore, it is necessary to convert the prediction results into an intuitive indicator of remaining safety time and set a threshold to trigger an early warning.

[0073] For example, in an embodiment of the present invention, intelligent prediction of air separation equipment failure is achieved based on the comparison result between the breakthrough time and the adsorption running time, including: taking the difference between the breakthrough time and the adsorption running time as the remaining safe time; and triggering an air separation equipment failure warning in response to the remaining safe time being lower than the safety threshold.

[0074] The remaining safety time is used to assess the remaining safe operating time of the molecular sieve adsorber before a breakthrough accident occurs within the current cycle. The safety threshold can be set to 30 minutes, but the specific setting can be adjusted according to actual needs.

[0075] For example, in this embodiment of the invention, intelligent prediction of air separation equipment faults is realized, and then the method further includes: pushing task work orders to operation and maintenance personnel, and linking historical fault data of air separation equipment to assist in fault location.

[0076] The task work order can be to switch the adsorber in advance, etc. The specific settings can be made according to actual needs. This embodiment of the invention does not impose too many restrictions here.

[0077] Thus, by embedding the dynamically corrected target mass transfer coefficient into the particle filter model to predict the breakthrough time, and calculating the remaining safe time based on the predicted breakthrough time and the running time, this embodiment of the invention can ensure that the prediction results can reflect the breakthrough advance trend caused by molecular sieve aging in real time, and achieve high-precision, adaptive, and intelligent early warning of air separation equipment failures.

[0078] For details, please refer to Figure 3 As shown, Figure 3 This is a schematic diagram comparing the prediction effects of dynamic mass transfer coefficient and fixed mass transfer coefficient provided in an embodiment of the present invention. The horizontal axis represents the comprehensive aging index of the molecular sieve, and the vertical axis represents the prediction delay time.

[0079] As shown in the figure, with accelerated aging, the fixed mass transfer coefficient in the traditional scheme increases rapidly and significantly with the increase of the molecular sieve aging index, causing the predicted value to lag behind the actual breakthrough, which can easily lead to safety accidents. The dynamic mass transfer coefficient provided by this scheme introduces thermal hysteresis characteristics for dynamic correction of the mass transfer coefficient, keeping the prediction error at a consistently low level and effectively reducing the prediction delay problem of the traditional method.

[0080] As can be seen, in this embodiment of the invention, when realizing intelligent prediction of air separation equipment faults, operating data including the outlet temperature of the current cycle's cold blowing stage can be obtained; the outlet temperature of the current cycle's cold blowing stage is compared with the reference cold blowing temperature to obtain the temperature deviation; the thermal hysteresis characteristic index of the current cycle is obtained by weighting the temperature deviation, the rate of change of the outlet temperature of the cold blowing stage, and the time index; the comprehensive aging index is obtained by weighting the thermal hysteresis characteristic index and the bed resistance pressure drop at the end of cold blowing; the standard mass transfer coefficient in the adsorption prediction model is corrected to obtain the target mass transfer coefficient, which is positively correlated with the standard mass transfer coefficient and negatively correlated with the comprehensive aging index; in the adsorption prediction model, the target mass transfer coefficient is used to predict the breakthrough time of particles; based on the comparison results between the breakthrough time and the adsorption running time, intelligent prediction of air separation equipment faults is realized, effectively improving the accuracy of intelligent prediction of air separation equipment faults.

[0081] This invention also discloses an intelligent fault prediction system for air separation equipment, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an intelligent fault prediction method for air separation equipment provided by this invention.

[0082] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0083] In this invention, the aforementioned memory can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0084] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An intelligent fault prediction method for air separation equipment, characterized in that, include: Obtain operational data including the outlet temperature of the current cycle's cold blowing phase; The temperature deviation is obtained by comparing the outlet temperature of the current cycle's cold blowing stage with the reference cold blowing temperature. The thermal hysteresis characteristic index of the current cycle is obtained by weighting the temperature deviation, the rate of change of outlet temperature during the cold blowing stage and the time index. The method for obtaining the thermal hysteresis characteristic index of the current period includes: ； This is a characteristic indicator of thermal hysteresis in the current cycle. , These represent the start and end times of the current cycle's cold blowing phase, respectively. , These are the outlet temperature and the reference cold blowing temperature for the current cycle's cold blowing phase, respectively. The outlet temperature during the cold blowing stage of the current cycle. right The derivative at time t, For time coefficient, It is a natural constant. During the cold blowing stage time, For normalization function, It is the absolute value symbol; A comprehensive aging index is obtained by weighting the thermal hysteresis characteristic index and the pressure drop of the bed resistance at the end of cold blowing. The target mass transfer coefficient is obtained by correcting the standard mass transfer coefficient in the adsorption prediction model, including: ； This is the correction factor for the current cycle. This represents the effective surface area of ​​the molecular sieve in its initial state. It is a natural constant. The thermo-aging coefficient, The comprehensive aging index for the current period is used; the product of the correction factor for the current period and the standard mass transfer coefficient is used as the target mass transfer coefficient; the target mass transfer coefficient is positively correlated with the standard mass transfer coefficient and negatively correlated with the comprehensive aging index. In the adsorption prediction model, the target mass transfer coefficient is used to predict the breakthrough time of particles; based on the comparison between the breakthrough time and the adsorption running time, intelligent prediction of air separation equipment failure is realized.

2. The method of claim 1, wherein, The acquisition of operational data including the outlet temperature of the current cycle's cold blowing phase includes: The current cycle includes an adsorption stage and a regeneration stage. The regeneration stage includes a cold blowing stage and a heating stage. During the adsorption stage, the inlet air flow rate, inlet air temperature, inlet air pressure, and outlet carbon dioxide concentration are collected. During the regeneration stage, the outlet temperature of the cold blowing stage and the bed resistance pressure drop at the end of the cold blowing stage are collected. The system also includes preprocessing the collected operating data, including using moving average filtering to remove noise, performing timestamp alignment, and removing outliers.

3. The method of claim 1, wherein, The method for obtaining the reference cold blowing temperature includes: In the initial stage of the new molecular sieve being put into use, outlet temperature data of the cold blowing stage were continuously collected for multiple cycles; after preprocessing the outlet temperature data of the cold blowing stage, the reference cold blowing temperature was obtained.

4. The method of claim 1, wherein, The adsorption prediction model is a particle filtering algorithm model.

5. The method of claim 1, wherein, The method of predicting the particle penetration time using the target mass transfer coefficient in the adsorption prediction model includes: The target mass transfer coefficient of the current cycle is applied to the adsorption prediction model, including embedding the target mass transfer coefficient into the state transition equation of the particle filtering algorithm to obtain the adsorption front position of the particle at each time in the next cycle; in response to the adsorption front position of the particle reaching the end of the molecular sieve bed at a certain time, the particle penetration time is obtained.

6. The method of claim 1, wherein, The method of intelligently predicting air separation equipment malfunctions based on the comparison between the breakthrough time and the adsorption running time includes: The difference between the penetration time and the adsorption time is used as the remaining safety time; if the remaining safety time is lower than the safety threshold, a fault warning for the air separation equipment is triggered.

7. The method of claim 1, wherein, The intelligent prediction of air separation equipment faults further includes: Task work orders are pushed to maintenance personnel, and historical fault data of air separation equipment is used to assist in fault location.

8. An intelligent fault prediction system for air handling equipment, the system comprising: include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a method for intelligent prediction of faults in an air separation unit according to any one of claims 1-7.

Images