Intelligent control system of air separation device

Abstract

The invention relates to the technical field of air separation and purification, and discloses an intelligent control system of an air separation device, which comprises a sensor module used for collecting real-time operation parameters of the device; the data processing module is used for calculating a health degree index representing the reliability of the equipment and a risk probability vector of a future fault; the dynamic decision and optimization module is used for dynamically adjusting and controlling the safety margin according to the health degree index and the risk probability and constructing a comprehensive optimization objective function; the control execution module is used for calculating and outputting a control instruction to an execution mechanism according to the adjusted safety margin and optimization target; and the alarm module is used for generating an instant or predictive alarm according to a preset condition. According to the invention, through the arrangement of the data processing module, mathematical operation is carried out on a statistical baseline established by real-time operation parameters collected by the sensor and historical data, a quantitative health degree index and a risk probability are generated, adaptive optimization of a control strategy is realized, and operation benefits are improved while safety is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of air separation and purification technology, and in particular to an intelligent control system for an air separation device. Background Technology

[0002] Currently, the control of air separation units generally relies on distributed control systems (DCS) or programmable logic controllers (PLCs). The core logic of these systems revolves around a set of static process parameter setpoints. For example, a PID (proportional-integral-derivative) controller stabilizes critical parameters such as temperature and pressure around preset fixed values. Safety is ensured by a separate interlocking protection system with several rigid, immutable alarm thresholds. When an operating parameter exceeds these thresholds, the system triggers an alarm or even an emergency shutdown. Operators are responsible for monitoring the entire process and manually adjusting the setpoints based on experience or production plans.

[0003] The existing control paradigm is inherently passive. Whether it is control or alarm, the system only takes action when the operating parameters deviate significantly, that is, when a problem has occurred or is occurring. This mechanism lacks the ability to predict failures and cannot identify signs of equipment degradation or early failures from small, coordinated changes in multiple related parameters.

[0004] The safety margins set are rigid; these fixed upper and lower operating limits are determined at the initial design stage of the equipment and usually remain unchanged throughout the entire operating cycle. This setting fails to reflect the equipment's true, time-varying health condition. When the equipment is new or in good condition, overly conservative margins may limit the realization of production potential; while when the equipment experiences wear and tear or performance degradation due to long-term operation, these fixed margins may not provide sufficient additional protection, harboring potential risks. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent control system for an air separation device, which solves the problem that existing control systems, due to their static control strategies and safety boundaries, cannot optimize operation based on the actual and dynamically evolving health status and failure risks of the equipment, making it difficult to achieve an optimal balance between safety and economy.

[0006] To address the aforementioned technical problems, this invention provides an intelligent control system for an air separation device.

[0007] This system is applied to the cryogenic distillation production process of liquid oxygen, liquid nitrogen, and liquid argon. It includes: a sensor module, a data processing module, a dynamic decision-making and optimization module, a control execution module, an alarm module, and a remote monitoring module.

[0008] The sensor module is deployed in key process areas of the air separation unit to acquire critical operating parameters that comprehensively characterize the thermodynamic and hydrodynamic states of the cryogenic distillation process. The module includes: a PT100 platinum resistance temperature sensor or thermocouple temperature sensor for temperature monitoring; a piezoresistive pressure sensor or capacitive pressure sensor for pressure monitoring; and a vortex flow meter for flow monitoring.

[0009] The data processing module is connected to the sensor module. This module receives real-time operating parameters collected by the sensor module and performs two core calculations based on these parameters and pre-stored historical data:

[0010] First, a health index is generated to quantify the current health status of the equipment. Specifically, the instantaneous value of any operating parameter collected in real time is compared with a statistical baseline of the parameter's ideal operating status established based on historical data. The degree of deviation is quantified using a Gaussian function to obtain the parameter's health sub-index, calculated as follows:

[0011] ;

[0012] In the formula, For the first Each running parameter is Health sub-indices at any given time; For this parameter in Real-time collected values ​​at any given moment; This is the statistical mean of the parameter's long-term stable operation. This represents the statistical standard deviation of this parameter under long-term stable operation. Subsequently, different weights are assigned to each operating parameter based on its impact on the overall reliability of the equipment. We sum all the health sub-indices by weight to obtain a result. For the first Each running parameter is Health sub-indices at any given time; For this parameter in Real-time collected values ​​at any given moment; This is the statistical mean of the parameter's long-term stable operation. This represents the statistical standard deviation of this parameter under long-term stable operation. Subsequently, different weights are assigned to each operating parameter based on its impact on the overall reliability of the equipment. By weighting and summing all the health sub-indices, a normalized health index representing the current overall reliability level of the equipment is obtained. .

[0013] Second, it generates risk probabilities representing potential future failures. Specifically, this is achieved by using a pre-trained prediction model that combines real-time operating parameters with a calculated equipment health index. As input to the model, the model output is associated with the probability of occurrence of a set of pre-defined specific failure modes, forming a risk probability vector. This establishes a quantitative correlation between the current operating status and specific future failure modes, providing input for proactive maintenance decisions. The dynamic decision-making and optimization module is connected to the data processing module. This module receives the health index generated by the data processing module. and risk probability vector And based on this, the control strategy is dynamically adjusted.

[0014] On the one hand, this module is based on the health index. Dynamically adjust the control safety margin. Specifically, this is done by adjusting the health index... When the pressure is reduced, the upper or lower limit of the critical control parameters in the air separator is automatically tightened, thereby proactively avoiding operating conditions that may induce failures under equipment deterioration conditions within a controllable range. Taking pressure parameters as an example, the dynamically adjusted upper limit of the operation... The formula for calculation is:

[0015] ;

[0016] In the formula:

[0017] This represents the dynamically adjusted upper limit of operations;

[0018] This represents the upper limit of the static operation of this parameter under standard operating conditions;

[0019] This represents a preset risk adjustment factor, which takes values ​​in the range of (0,1).

[0020] The normalized health index is input to the data processing module 200.

[0021] On the other hand, this module is based on the risk probability vector Dynamically optimize operational control objectives. Specifically, this involves balancing multiple control objectives characterizing output, energy consumption, and equipment reliability based on a risk probability vector. The optimization weights of each objective are automatically adjusted to construct a comprehensive objective function that dynamically evolves with risk. This is provided for the control execution module to solve. The function is:

[0022] ;

[0023] In the formula, , , Sub-objective functions representing output, energy efficiency, and equipment lifespan, respectively; weighting coefficients. , , It is a risk probability vector The function. When the risk probability associated with a specific high-load or high-energy-consumption operation mode increases, the system automatically reduces the weight of the output target corresponding to that operation mode. Or energy efficiency target weight And accordingly increase the weight of equipment reliability targets. This allows the system to automatically transition to a more conservative and stable operating range without interrupting production. The control execution module is connected to the dynamic decision-making and optimization module. This module operates based on the adjusted control safety margin and optimized comprehensive objective function output by the dynamic decision-making and optimization module. The system performs automatic control operations on the actuators of the air separator. Specific operations include automatically adjusting the valve opening, compressor guide vane angle, and heater power, so that the actual operating state of the air separator continuously tracks the dynamic optimum determined by the comprehensive objective function.

[0024] The alarm module is connected to the data processing module. The alarm signals generated by this module are divided into two types: the first type is an immediate alarm generated when the operating parameters exceed the preset static threshold, which is used to prompt technicians to intervene or confirm immediately; the second type is a predictive warning generated when the equipment health index is lower than the preset threshold or any risk probability is higher than the preset threshold. This warning contains potential fault types and location information, aiming to transform maintenance work from reactive repair or periodic maintenance to condition-based predictive maintenance.

[0025] The remote monitoring module is used to send the operational data and alarm signals collected and processed by the system to the remote monitoring center. The sent data includes real-time operating parameters, equipment health index, risk probability, and graded alarm information, providing data support for remote assessment of the device's operational status and maintenance decision-making.

[0026] In summary, the present invention has at least one of the following beneficial technical effects:

[0027] 1. This invention, by setting up a data processing module, performs mathematical operations on the real-time operating parameters collected by sensors and the statistical baseline established by historical data to generate a quantified health index and risk probability. This transforms the judgment of equipment status from qualitative monitoring that relies on static thresholds to quantitative assessment based on dynamic data analysis, providing an objective basis for subsequent precise control and decision-making.

[0028] 2. This invention achieves a dynamic balance between ensuring safety and pursuing efficiency. By setting up a dynamic decision-making and optimization module, it tightens or loosens the control safety margin of key parameters in real time based on the health index, and automatically adjusts the optimization weights of the three control objectives—output, energy efficiency, and reliability—based on the probability of risk. This allows the control system to automatically transition to a safer or more efficient operating range based on the equipment's own health condition, rather than always using a fixed control mode.

[0029] 3. This invention provides a decision-making basis for implementing condition-based predictive maintenance, which helps improve equipment reliability. The alarm module distinguishes between immediate alarms and predictive warnings. Predictive warnings are triggered based on a decrease in the health index or an increase in the probability of risk, and include information on potential fault types. This allows maintenance to shift from traditional passive repairs or fixed periodic maintenance to targeted, predictive maintenance that can intervene in the early stages of performance degradation, thereby effectively avoiding unplanned downtime. Attached Figure Description

[0030] Figure 1 This is a system framework diagram of the present invention;

[0031] Figure 2 This is a schematic diagram of the sensor module deployment according to the present invention;

[0032] Figure 3 This is a functional logic block diagram of the data processing module of the present invention;

[0033] Figure 4 This is a functional logic block diagram of the dynamic decision-making and optimization module of the present invention;

[0034] Figure 5 This is a functional logic block diagram of the control execution module of the present invention;

[0035] Figure 6 This is a functional logic block diagram of the alarm module of the present invention;

[0036] Figure 7 This is a functional logic block diagram of the remote monitoring module of the present invention.

[0037] Among them, 100 is the sensor module; 200 is the data processing module; 300 is the dynamic decision-making and optimization module; 400 is the control execution module; 500 is the alarm module; and 600 is the remote monitoring module. Detailed Implementation

[0038] The following is in conjunction with the appendix Figure 1 - Appendix Figure 7 The present invention will be further described in detail below.

[0039] See attached document Figure 1 , Figure 1 This is a functional block diagram of an intelligent control system according to an embodiment of the present invention. The present invention provides an intelligent control system for an air separation device, which is applied to the cryogenic distillation production process of liquid oxygen, liquid nitrogen, and liquid argon. The system may include the following modules:

[0040] The system includes a sensor module 100, a data processing module 200, a dynamic decision-making and optimization module 300, a control execution module 400, an alarm module 500, and a remote monitoring module 600.

[0041] The system's workflow during a complete operation cycle is as follows:

[0042] Sensor module 100 is deployed in key process areas of the air separation unit to continuously collect real-time operating parameters of the unit during the distillation process, such as temperature, pressure, and flow rate. The collected parameters are transmitted as raw data to data processing module 200.

[0043] The data processing module 200 is connected to the sensor module 100, and it receives raw data and performs quantitative analysis. This module first calculates the normalized health index, which characterizes the current reliability level of the equipment. The calculation method involves comparing the real-time value of any operating parameter with the statistical baseline of the parameter's ideal operating state in historical data to obtain the parameter's health sub-index. Then, a weighted summation of the health sub-indices of all parameters is performed. The specific calculation formula is:

[0044] ;

[0045] ;

[0046] In the formula:

[0047] Representing the Each running parameter is Health sub-indices at any given time;

[0048] Representing the Each running parameter is Real-time collected values ​​at any given moment;

[0049] Representing the Statistical mean of ideal operating state of each operating parameter;

[0050] Representing the The statistical standard deviation of the ideal operating state of each operating parameter;

[0051] Representing the Preset weighting coefficients for the impact of each operating parameter on the overall reliability of the equipment;

[0052] Representative equipment in Normalized health index at any given time.

[0053] Meanwhile, the data processing module 200 uses real-time operating parameters and the calculated health index as the basis for its operation. By using a pre-trained prediction model, the output is the probability of occurrence associated with a set of preset specific failure modes, forming a risk probability vector. The prediction model preferably employs a Long Short-Term Memory (LSTM) neural network because it can effectively capture time-series dependencies in industrial processes. The model's input layer receives time-series data consisting of multiple key process parameters (such as temperature, pressure, and flow rate) within a specific past time window (e.g., 10 minutes) and a real-time calculated health index H. All input data is normalized to eliminate the influence of dimensions. This input data is then fed into a core hidden layer containing two or more stacked LSTM layers (e.g., each layer has 128 neurons, followed by a dropout layer to prevent overfitting) for deep feature extraction. Finally, a fully connected output layer with N neurons uses the Softmax activation function to output an N-dimensional risk probability vector R. Where N is the total number of preset specific fault modes, Representing the This model calculates the probability of a certain failure mode occurring within a predetermined future timeframe. Its high performance relies on a comprehensive training dataset, constructed by fusing historical operational data from the entire equipment lifecycle, simulated failure data generated using a high-fidelity digital twin model, and accelerated aging test data for key components. Before training, domain experts pre-define key failure modes (e.g., compressor near-surge, main heat exchanger blockage), and supervise the annotation of historical data using equipment maintenance records and expert knowledge. Simulation and test data are then automatically and accurately labeled. The model training employs a standard supervised learning process, dividing the labeled dataset into training, validation, and test sets. Classification cross-entropy is used as the loss function, and adaptive optimizers such as Adam are employed for iterative optimization. An early stopping mechanism is introduced to monitor validation set loss and prevent overfitting, thereby achieving optimal generalization performance. This module calculates the health index. and risk probability vector The data is transmitted to the dynamic decision-making and optimization module 300 and the alarm module 500.

[0054] The dynamic decision-making and optimization module 300 is connected to the data processing module 200, and it processes the received health index. and risk probability vector Develop dynamic control strategies. This module first determines the control strategy based on the health index. Adjust the safety margin of key control parameters, such as their dynamically adjusted operating upper limit. The formula for calculation is:

[0055] ;

[0056] In the formula:

[0057] This represents the dynamically adjusted upper limit of operations;

[0058] This represents the standard static operating limit of this parameter;

[0059] This represents a preset risk adjustment coefficient.

[0060] Next, the module uses the risk probability vector Construct a comprehensive objective function This function is used to calculate output, energy consumption, and where:

[0061] The system dynamically optimizes among the three control objectives: equipment reliability, etc. Its function structure is as follows:

[0062] ;

[0063] In the formula:

[0064] Represents a comprehensive objective function;

[0065] , , These are sub-objective functions representing output, energy efficiency, and equipment lifespan, respectively.

[0066] These are respectively related to the risk probability vector The associated dynamic weighting coefficients used to characterize output, energy efficiency, and reliability targets, as described above. These are respectively related to the risk probability vector The associated dynamic weighting function, its specific function form and parameter setting method are explained as follows: The core idea of ​​this design is to dynamically adjust the weights among the three competing optimization objectives of output, energy efficiency, and reliability based on the predicted risk level, achieving a smooth transition from pursuing economic benefits to ensuring system safety. First, from the risk probability vector... The most direct and effective way to extract a scalar as a measure of overall risk is to take its largest component, i.e. This value represents the most pressing risk of failure. Subsequently, the weighted coefficients are defined as this risk measure. The functions. In a specific, non-limiting embodiment, these functions may take the form of: representing the weights of the output target. and the weights representing energy efficiency targets Designed as a monotonically decreasing function of risk, such as an exponentially decaying function, its form is: as well as ;

[0067] In the formula, and They represent the ideal safety state (i.e.) The basic weights of (time) output and energy efficiency, while and This is a parameter that adjusts sensitivity, determining how quickly the weight decreases as risk increases. Conversely, the weight representing the reliability objective... Designed as a monotonically increasing function of risk, such as a simple linear function. :

[0068] In the formula, It is the basic reliability weight. This ensures that when risks increase, the system will decisively increase its focus on reliability objectives, and all the aforementioned basic weights ( ) and sensitivity parameters ( All parameters are pre-set and adjustable, based on the characteristics, economic value, and safety redundancy strategies of the specific process object. Through the clear definition of the specific form of the function and the meaning of the parameters, the comprehensive objective function can be reliably constructed. The dynamic decision-making and optimization module 300 will then combine the adjusted safety margin with the determined comprehensive objective function. Transmitted to control execution module 400.

[0069] The control execution module 400 is connected to the dynamic decision-making and optimization module 300. Based on the received safety margin and comprehensive objective function, it calculates and outputs specific control commands to the actuators of the air separator, such as adjusting valve opening, compressor guide vane angle, or heater power, so that the actual operating status of the device can continuously track the changes caused by the control execution module. The determined dynamic optimal point is achieved while ensuring that the operating parameters are within the safety margin after dynamic adjustment.

[0070] The alarm module 500 is connected to the data processing module 200. When it receives operating parameters and health index... or risk probability vector When the values ​​in the vector meet preset alarm conditions, a corresponding immediate alarm or predictive warning signal is generated. This establishes a strict mapping relationship between the inherent structured definition of the risk probability vector R and a pre-established fault mode knowledge base. During the model design and training phases, this N-dimensional vector... Each component (i.e., probability value) These are not abstract numbers, but rather uniquely and permanently bound to a specific failure mode predefined by experts. Crucially, this predefined failure mode description itself already contains precise type and location information. For example, someone skilled in the art can pre-create a mapping list or lookup table, where the first index position of a vector (corresponding to a probability)... The bearings of the main compressor unit K-101 are clearly associated with a risk of excessive wear. (Second index position (corresponding probability)) The A channel associated with the main heat exchanger E-201 may be blocked or leaking, and the Nth index position corresponds to another specific fault containing a specific equipment tag number. Therefore, when the alarm module receives the risk probability vector R generated at the time of purchase, its built-in alarm judgment logic does not guess, but first finds the probability value with the largest value in the vector. and its corresponding index Next, determine the maximum probability value. Has the preset predictive alarm threshold been exceeded? If it is confirmed that it has been exceeded, the system will then use an index to... The descriptive text containing the specific fault type and location information that is uniquely matched to the pre-established mapping list is retrieved directly from the list and used as the core content of the warning signal to generate and output the warning signal.

[0071] The remote monitoring module 600 is responsible for sending the operating parameters collected by the sensor module 100, the health index and risk probability generated by the data processing module 200, and the alarm signals generated by the alarm module 500 to the remote monitoring center through the communication network, so as to realize remote status assessment and management of the device.

[0072] See attached document Figure 2 , Figure 2 This is a schematic diagram of the deployment of a sensor module in an air separation device according to an embodiment of the present invention.

[0073] This section details the implementation of the sensor module 100. This module forms the basis of the system's data input, and its function is to provide raw, real-time operating parameters for the subsequent data processing module 200.

[0074] The sensor module 100 is configured to be deployed at preset key process locations of the air separation unit, including but not limited to: the inlet and outlet of the air compressor, the fluid channels of the main heat exchanger, the top and bottom of the distillation column, and the process piping connecting the various equipment units.

[0075] In one specific embodiment, the sensor module 100 includes a temperature sensor, a pressure sensor, and a flow sensor.

[0076] The specific type of temperature sensor is either a PT100 platinum resistance temperature sensor or a K-type thermocouple temperature sensor. This type of sensor is installed at the location where temperature monitoring is required to obtain temperature parameters of the thermodynamic state of the reaction process.

[0077] The specific types of pressure sensors are piezoresistive pressure sensors or capacitive pressure sensors. These types of sensors are installed in pressure-bearing equipment or pipelines to obtain pressure parameters that reflect the fluid dynamics state.

[0078] The specific type of flow sensor is a vortex flow meter. This type of sensor is installed in the main pipeline of the process fluid to obtain the volumetric flow rate or mass flow rate parameter of the fluid.

[0079] By deploying appropriate types of sensors at the aforementioned locations, the sensor module 100 can acquire and output a set of key parameters in real time that can comprehensively characterize the thermodynamic and hydrodynamic states of the cryogenic distillation process. These parameters form the data basis for all subsequent analysis, decision-making, and control operations.

[0080] See attached document Figure 3 , Figure 3 This is a functional logic block diagram of a data processing module according to an embodiment of the present invention. This section describes in detail the implementation of the data processing module 200. This module is electrically connected to the sensor module 100, and its function is to receive real-time operating parameters from the sensor module 100 and convert them into quantitative indicators for subsequent decision-making and optimization. The output of this module is a device health index. and failure risk probability vector .

[0081] In one specific embodiment, the data processing module 200 performs the following calculation steps:

[0082] First, this module calculates the device health index. This calculation process uses the real-time operating parameter values ​​collected by sensor module 100. Compare with the statistical baseline of ideal operating conditions established based on historical data. For the first... Each operating parameter, its health sub-index The formula for calculation is:

[0083] ;

[0084] In the formula, For the first Each running parameter is Health sub-indices at any given time; For this parameter in Real-time collected values ​​at any given moment; This is the statistical mean of the parameter's long-term stable operation. This represents the statistical standard deviation of the parameter under long-term stable operation.

[0085] Subsequently, different preset weighting coefficients were assigned to each operating parameter based on its impact on the overall reliability of the equipment. By weighting and summing all the health sub-indices, a normalized health index representing the current overall reliability level of the equipment is obtained. The formula for its calculation is:

[0086] ;

[0087] in, .

[0088] Secondly, this module generates a failure risk probability vector. This process is achieved through a pre-trained prediction model. This model will run the parameters in real time. And the equipment health index calculated in the previous step As input.

[0089] The prediction model outputs a risk probability vector. Each element in the vector This corresponds to the probability of a specific, pre-defined fault mode (e.g., compressor bearing wear, heat exchanger leakage, etc.) occurring within a pre-defined time window in the future. The predictive model can be configured as a neural network model or a support vector machine model trained based on historical operating data and fault record data.

[0090] After completing the above calculations, the data processing module 200 will generate a health index. and risk probability vector The data is transmitted to the dynamic decision-making and optimization module 300 and the alarm module 500.

[0091] See attached document Figure 4 , Figure 4 This is a functional logic block diagram of a dynamic decision-making and optimization module according to an embodiment of the present invention.

[0092] This section details the implementation of the dynamic decision-making and optimization module 300. This module is electrically connected to the data processing module 200, and its function is to receive the device health index output by the data processing module 200. and failure risk probability vector Based on these two inputs, a dynamically adjusted control safety margin and a dynamically optimized operational control objective are generated.

[0093] In one specific embodiment, the dynamic decision-making and optimization module 300 performs two operations in parallel:

[0094] The first step is to check the device health index. Dynamically adjust and control the safety margin. When the health index... When the pressure is reduced, the module automatically tightens the upper or lower limit of one or more preset key control parameters in the air separator. Taking pressure as an example, its dynamically adjusted upper limit... The formula for calculation is:

[0095] ;

[0096] In the formula:

[0097] This represents the dynamically adjusted upper limit of operations;

[0098] This represents the upper limit of the static operation of this parameter under standard operating conditions;

[0099] This represents a preset risk adjustment factor, which takes values ​​in the range of (0,1).

[0100] The normalized health index is input to the data processing module 200.

[0101] The second operation is based on the failure risk probability vector. Dynamically optimize operational control objectives. This module constructs a comprehensive objective function that dynamically evolves with risk. This function is the result of a weighted sum of multiple sub-objective functions characterizing output, energy consumption, and equipment reliability. Its function expression is:

[0102] ;

[0103] In the formula:

[0104] Represents a comprehensive objective function;

[0105] , , These are sub-objective functions representing output, energy efficiency, and equipment lifespan, respectively.

[0106] , , These are respectively related to the risk probability vector Associated dynamic weighting coefficients. These weighting coefficients are configured as follows: The function. For example, when the risk probability vector When the risk probability value associated with a specific high-load or high-energy-consumption operating mode increases, the weighting function... or The output value of the weight function decreases accordingly. The output value will then be increased accordingly. After completing the above two operations, the dynamic decision-making and optimization module 300 will calculate the adjusted control safety margin (e.g., and the determined comprehensive objective function Transmitted to control execution module 400.

[0107] See attached document Figure 5 , Figure 5 This is a functional logic block diagram of a control execution module according to an embodiment of the present invention.

[0108] This section details the implementation of the control execution module 400. This module is electrically connected to the dynamic decision-making and optimization module 300, and its function is to receive the adjusted control safety margin and optimized comprehensive objective function output by the dynamic decision-making and optimization module 300. Based on these inputs, specific physical control operations for various actuators in the air separation device are generated and executed.

[0109] In one specific embodiment, the control execution module 400 is configured as a model predictive control (MPC) controller. This controller internally contains a preset process model capable of describing the dynamic behavior of the air separation device.

[0110] In each control cycle, the MPC controller will receive the comprehensive objective function. The target of its optimization solution is the dynamically adjusted control safety margin, which is received as a hard constraint in its optimization solution process.

[0111] The MPC controller uses an online numerical optimization algorithm to solve the aforementioned constrained optimization problem within a preset prediction time domain, in order to calculate a set of comprehensive objective functions.

[0112] The future control operation sequence that minimizes the predicted value.

[0113] The control execution module 400 sends the first control action in the calculated optimal control sequence as the output command for the current moment to one or more corresponding actuators of the air separation device. These actuators include valves for regulating the fluid passage, with the control command being the valve opening degree; compressors for regulating gas flow and pressure, with the control command being the guide vane angle; and heaters for regulating the process temperature, with the control command being the heating power.

[0114] By periodically repeating the above process of receiving, solving, and outputting instructions, the control execution module 400 enables the actual operating state of the air separation device to continuously track the dynamic optimal point determined by the comprehensive objective function, while ensuring that key operating parameters are always within the safety margin range set by the dynamic decision-making and optimization module 300.

[0115] See attached document Figure 6 , Figure 6 This is a functional logic block diagram of an alarm module according to an embodiment of the present invention.

[0116] This section details the implementation of the alarm module 500. This module is electrically connected to the data processing module 200, and its function is to generate and differentiate alarm signals of different levels based on the data received from the data processing module 200.

[0117] In one specific embodiment, the alarm module 500 is configured to perform two independent alarm logic judgments. The first is an immediate alarm logic. This logic compares the real-time operating parameter values ​​collected by the sensor module 100 and transmitted via the data processing module 200 with a set of preset static thresholds. When the real-time value of any operating parameter exceeds its corresponding static threshold range, the alarm module 500 generates an immediate alarm signal.

[0118] The second type is predictive early warning logic. This logic is based on the health index output by the data processing module 200. and risk probability vector The logic involves two independent triggering conditions. The first triggering condition is when the health index... An alert is triggered when the value of the risk probability vector is below a preset warning threshold and the duration of this state exceeds a preset duration threshold. The second triggering condition is that when the risk probability vector... any element in The value represents the probability of a specific fault mode occurring. When this value exceeds a preset probability threshold, an early warning is triggered.

[0119] When a predictive warning is triggered, the warning signal generated by the alarm module 500 also includes additional information. If the warning is triggered by a risk probability, the additional information includes the identification information of the potential fault mode corresponding to that high risk probability. All generated instant alarm signals and predictive warning signals are transmitted to the local human-machine interface and simultaneously sent to the remote monitoring module 600.

[0120] This section details the implementation of the remote monitoring module 600. This module has data connections with all other modules in the system, and its function is to aggregate, transmit, and visualize key information of the system.

[0121] In one specific embodiment, the module includes a data interface unit and a communication unit. The data interface unit is responsible for collecting the raw operating parameters of the sensor module 100, the health index and risk probability vector generated by the data processing module 200, and the alarm signal generated by the alarm module 500.

[0122] The communication unit will send the collected data to a remote monitoring center deployed in the central control room or cloud via industrial Ethernet using MQTT or OPC UA communication protocols.

[0123] The remote monitoring center is configured to have the following functions:

[0124] Data display and visualization: The air separation unit's various operating parameters and calculated indicators are displayed in real time in the form of dashboards, trend curves, and equipment topology diagrams.

[0125] Data storage and historical query: All received data is stored in a time-series database, supporting querying, playback, and analysis of historical data for any time period.

[0126] Report generation: Based on preset templates, automatically generate daily, weekly, or monthly reports that include key performance indicators (KPIs), alarm statistics, health trends, etc.

[0127] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent control system for an air separation device, characterized in that, include: The sensor module is used to collect the operating parameters of the air separation unit in real time during the distillation process; The data processing module, connected to the sensor module, generates a health index that quantifies the current health status of the equipment and a risk probability that characterizes potential future failures based on the operating parameters and historical data. The dynamic decision-making and optimization module, connected to the data processing module, is used to dynamically adjust the control safety margin of the air separation device according to the equipment health index, and to dynamically optimize the operation control target of the device according to the risk probability. The control execution module, connected to the dynamic decision-making and optimization module, is used to perform automatic control operations on the actuators of the air separation device based on the adjusted control safety margin and the optimized operation control target. An alarm module, connected to the data processing module, is used to generate an alarm signal based on the analysis results for confirmation by technicians. The remote monitoring module is used to send system operation data and alarm signals to the remote monitoring center.

2. The intelligent control system for an air separation device according to claim 1, characterized in that, The specific method by which the data processing module generates the device health index includes: comparing the real-time collected operating parameters with the ideal operating state statistical baseline established based on historical data to quantify the degree to which the current operating state of the device deviates from its ideal state, and converting it into a normalized index that characterizes the current reliability level of the device.

3. The intelligent control system for an air separation device according to claim 1, characterized in that, The data processing module generates risk probabilities by using a pre-trained prediction model, taking the operating parameters and equipment health index as input, and outputting the probability of occurrence associated with preset fault modes, thereby establishing the correlation between the current operating state and future fault modes, and providing quantitative input for forward-looking maintenance decisions.

4. The intelligent control system for an air separation device according to claim 1, characterized in that, The dynamic decision-making and optimization module is used to dynamically adjust the control safety margin in the following way: when the equipment health index decreases, the upper or lower limit of the operation of the key control parameters in the air separation device is automatically tightened, thereby sacrificing some extreme performance within a controllable range to actively avoid operating conditions that induce failures in the equipment deterioration state.

5. The intelligent control system for an air separation device according to claim 1, characterized in that, The dynamic decision-making and optimization module is used to dynamically optimize the operation control objectives in the following way: among multiple control objectives that characterize output, energy consumption and equipment reliability, the optimization weights of each objective are automatically adjusted according to the risk probability to construct a comprehensive objective function that evolves dynamically with the risk, which is then solved by the control execution module.

6. The intelligent control system for an air separation device according to claim 5, characterized in that, When the probability of risk associated with high-load or high-energy-consumption operation modes increases, the dynamic decision-making and optimization module will automatically reduce the optimization weight of the production or energy efficiency targets corresponding to the operation mode and correspondingly increase the optimization weight of the equipment reliability target, thereby guiding the system to automatically transition to a more conservative and stable operating range without interrupting production.

7. The intelligent control system for an air separation device according to claim 5, characterized in that, The automatic control operations performed by the control execution module include: automatically adjusting the valve opening, compressor guide vane angle, or heater power of the air separation device, so that the actual operating state of the air separation device can continuously track the dynamic optimal point determined by the comprehensive objective function.

8. The intelligent control system for an air separation device according to claim 1, characterized in that, The alarm signals generated by the alarm module are classified as follows: An immediate alarm is generated when the operating parameters exceed a preset static threshold, which is used to prompt technicians to intervene or confirm immediately; A predictive warning is generated when the equipment health index is lower than a preset threshold or the risk probability is higher than a preset threshold. This predictive warning includes potential fault types and location information, transforming maintenance work from reactive repair or periodic maintenance into condition-based predictive maintenance.

9. The intelligent control system for an air separation device according to claim 1, characterized in that, The sensor module is deployed in key process areas of the air separation device, including: PT100 platinum resistance temperature sensor or thermocouple temperature sensor; Piezoresistive pressure sensor or capacitive pressure sensor; Vortex flowmeter; used to obtain key parameters that can comprehensively characterize the thermodynamic and hydrodynamic states of cryogenic distillation processes.

10. The intelligent control system for an air separation device according to claim 1, characterized in that, The system operation data and alarm signals sent by the remote monitoring module include the operating parameters, equipment health index, risk probability, and graded alarm information, providing data support for remotely assessing the operating status of the device and making maintenance decisions.

Images