A helium recovery and purification method and purification device used in the process

Through intelligent pretreatment, multi-stage thermal energy recovery and multi-stage catalytic reactor systems, combined with gas separation and distillation technology, combined with intelligent monitoring and maintenance systems, the problems of low efficiency and high energy consumption in helium recovery and purification technology are solved, and the problem of low efficiency and high energy consumption in helium recovery and purification technology is achieved. Efficient and energy-saving helium recovery and purification.

CN118702074BActive Publication Date: 2024-11-12SHENZHEN WALSHENG INTELLIGENT CONTROL TECH CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202411180126.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-12
Estimated Expiration
2044-08-27

AI Technical Summary

Technical Problem

The existing helium recycling and purification technology has low efficiency, high energy consumption, and lacks intelligent supervision, making it difficult to effectively deal with impurities and low-concentration helium in the helium-containing industrial waste gas.

Method used

Adopt intelligent pretreatment, multi-stage thermal energy recovery, multi-stage catalytic reactor system, gas separation technology, distillation module and deep purification technology, combined with intelligent monitoring and maintenance system, dynamically adjust the chemical adsorption device and catalyst regeneration cycle, and use multi-stage membranes Separation, vortex tube separation, supercritical fluid extraction and acoustic wave separation technologies achieve efficient enrichment and purification of helium.

Benefits of technology

It improves helium recovery and purification efficiency, reduces energy consumption, realizes the production of high-purity helium, and improves the reliability and life of the system through intelligent monitoring and predictive maintenance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118702074B_ABST
    Figure CN118702074B_ABST
Patent Text Reader

Abstract

The present application discloses a helium recovery and purification method and a purification device used in the process, the method comprising: intelligently pre-treating helium-containing industrial waste gas; recovering waste heat of the helium-containing industrial waste gas by a multi-stage heat recovery system; removing impurities from the helium-containing industrial waste gas by a multi-stage catalytic reactor system; enriching and preliminarily purifying helium by gas separation technology; rectifying and purifying helium by a rectification module; achieving fine purification of helium by purification technology; and performing predictive maintenance by an intelligent monitoring and maintenance system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the field of gas recovery technology, and specifically, relates to a helium recovery and purification method and a purification device used in the process. Background Art

[0002] As a rare gas, helium is widely used in industry, medicine, scientific research and other fields. However, the content of helium in nature is extremely low, and it mainly comes from the extraction of natural gas. With the continuous increase in global helium demand, the helium resources in natural gas are becoming increasingly scarce, and helium recovery and purification technology is becoming more and more important.

[0003] At present, industrial waste gas often contains a certain amount of helium, but due to the complex impurity composition and low concentration, the traditional helium recovery and purification method has the problems of low efficiency, low thermal energy utilization and lack of intelligent supervision. Summary of the invention

[0004] In view of this, the present application provides a helium recovery and purification method and device, which solves the problems of low helium recovery and purification efficiency, high energy consumption and insufficient intelligence in the prior art.

[0005] An embodiment of the present application provides a helium recovery and purification method, including: intelligent pretreatment of helium-containing industrial waste gas; using a multi-stage heat recovery system to recover waste heat from the helium-containing industrial waste gas; using a multi-stage catalytic reactor system to remove impurities from the helium-containing industrial waste gas; using gas separation technology to enrich and preliminarily purify helium; distilling and purifying helium through a distillation module; using purification technology to achieve fine purification of helium; and using an intelligent monitoring and maintenance system to perform predictive maintenance.

[0006] Intelligent pretreatment of helium-containing industrial waste gas includes: removing micron-sized particles in the helium-containing industrial waste gas through a filter; using a multi-stage chemical adsorption device to remove corrosive impurities, including acidic gases, alkaline gases and organic pollutants; and dynamically adjusting the regeneration cycle of the multi-stage chemical adsorption device based on a stochastic predictive control (SPC) system.

[0007] Based on the stochastic predictive control (SPC) system, the regeneration period of the multi-stage chemical adsorption device is dynamically adjusted, including: establishing an SPC framework, including: defining system state variables, the system state variables including temperature, pressure and flow; determining control input parameters, the input parameters including valve opening and compressor power; establishing a system dynamic model including unbounded process noise; establishing conceptual contraction time domain and rolling time domain formulas; collecting various parameters of the helium-containing industrial waste gas; constructing and training a deep neural network; inputting various parameters of the helium-containing industrial waste gas into the deep neural network, so that the deep neural network is combined with the SPC framework to realize the state prediction of the helium-containing industrial waste gas; based on the state prediction result, dynamically adjusting the adsorbent dosage and replacement frequency in the multi-stage chemical adsorption device.

[0008] The waste heat of the helium-containing industrial waste gas is recovered by using a multi-stage heat recovery system, including: obtaining medium-temperature waste gas by heat exchange with inlet cold gas; obtaining low-temperature waste gas by exchanging heat with intermediate process gas; using the low-temperature waste gas to preheat supplementary gas; and using a cogeneration system to convert heat energy into electrical energy during the heating and cooling process to supply control systems and auxiliary equipment.

[0009] A multi-stage catalytic reactor system is used to remove impurities from the helium-containing industrial waste gas, including: a fixed bed reactor is used in the first stage to remove the first impurity; a pulse reactor is used in the second stage to remove the second impurity by periodically changing the reaction conditions; a plasma-assisted catalysis is used in the third stage to remove the third impurity at low temperature; the gas separation technology is used to enrich and preliminarily purify helium, including: a rotating bed adsorption separation device is used to enrich helium by using centrifugal force to enhance the mass transfer process; the enriched helium enters a multi-stage membrane separation system, and membrane materials with different pore sizes and selectivities are used to gradually improve the purity of the helium; an ionized gas separation technology is introduced to separate gases by using the difference in ionization energy of different gases; wherein the gas flow and pressure between the separation units are coordinated and controlled by the intelligent monitoring and maintenance system.

[0010] Before the gas separation technology is used for helium enrichment and preliminary purification, the method also includes: establishing an adaptive scalar field mapping system, including: establishing a gas component concentration scalar field; using a high-intensity region recognition algorithm to detect and identify high-intensity regions; constructing an adaptive avoidance strategy; based on the adaptive scalar field mapping system, structurally optimizing and centrifugal force field regulation of the rotating bed adsorption separation device; based on the adaptive scalar field mapping system, dynamically adjusting the spatial arrangement and cascade structure of the membrane components, and adaptively adjusting the trans-membrane pressure difference; based on the adaptive scalar field mapping system, dynamically adjusting plasma gas parameters to avoid high-intensity regions during ion flow movement.

[0011] The fine purification of helium is achieved by using purification technology, including: introducing vortex tube separation technology to further purify helium by using the gas temperature separation effect; using a supercritical fluid extraction device to selectively remove trace organic impurities; using acoustic wave separation technology to perform fine purification using the acoustic characteristics of different gases, wherein the fine purification process is monitored by the intelligent monitoring and maintenance system through the model predicted path integral MPPI technology to ensure that the purity of helium is greater than 99.9999%.

[0012] Predictive maintenance is performed using an intelligent monitoring and maintenance system, including: In the intelligent monitoring and maintenance system, a distributed monitoring network is constructed through an enhanced Push-DIging algorithm to achieve full-process intelligent monitoring and maintenance; a diffusion model is used to generate general time series energy data for health status assessment and fault prediction of different equipment.

[0013] A distributed monitoring network is built through an enhanced Push-DIging algorithm to achieve full-process intelligent monitoring and maintenance, including: building a Push-DIging algorithm based on gradient boosting; using the Push-DIging algorithm to perform distributed average consistency estimation and build an adaptive step size strategy; realizing distributed fusion of multi-source heterogeneous data; checking data consistency and correcting abnormal data; using a diffusion model to generate universal time series energy data for health status assessment and fault prediction of different equipment, including: building a diffusion model architecture suitable for energy data; training and fine-tuning the diffusion model; using the trained diffusion model to generate time series energy data and perform quality assessment; using the generated time series energy data to simulate abnormal scenarios and inject faults.

[0014] An embodiment of the present application also provides a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the above-mentioned helium recovery and purification method.

[0015] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned helium recovery and purification method.

[0016] The embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the above-mentioned helium recovery and purification method.

[0017] The present application also provides a helium recovery, purification and purification device, including: an intelligent monitoring and maintenance system, including a central processing unit, a sensor network, a touch screen display and a communication unit; a multi-stage heat recovery system, including a high-temperature heat exchanger, a medium-temperature heat exchanger, a low-temperature heat exchanger and a thermoelectric conversion unit; a multi-stage catalytic reactor system, including a fixed bed reactor, a pulse reactor, a plasma-assisted catalytic unit and a catalyst regeneration system; a gas separation system, including a rotating bed adsorber, a multi-stage membrane separation unit and an ionized gas separation device; a distillation module, including an internal heat-integrated distillation tower, a distributed feeding system, a centrifugal distillation unit and a condenser and a reboiler; a deep purification module, including a vortex tube separator, a supercritical fluid extraction device and an acoustic wave separation unit.

[0018] This application has the following technical effects:

[0019] Through intelligent pretreatment and multi-stage heat recovery system, the treatment efficiency and energy utilization rate of helium-containing industrial waste gas are improved;

[0020] The multi-stage catalytic reactor system and various gas separation technologies are used to achieve efficient helium enrichment and preliminary purification;

[0021] Combining distillation and deep purification technology, the high purity of helium is guaranteed;

[0022] By utilizing the intelligent monitoring and maintenance system, intelligent control and predictive maintenance of the entire process are achieved, thus improving the reliability and life of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A flow chart of a helium recovery and purification method in one embodiment;

[0025] Figure 2 This is a schematic diagram of the structure of a helium recovery and purification device;

[0026] Figure 3 A schematic diagram of the structure of an intelligent control module in one embodiment;

[0027] Figure 4 A schematic diagram of the structure of a multi-stage heat recovery system in one embodiment;

[0028] Figure 5A schematic diagram of the structure of a multi-stage catalytic reactor system in one embodiment;

[0029] Figure 6 A schematic diagram of the structure of a gas separation system in one embodiment;

[0030] Figure 7 is a schematic structural diagram of a distillation module in one embodiment;

[0031] Figure 8 A diagram showing the structure of a deep purification module in one embodiment. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0033] Figure 1 : is a schematic diagram of a helium recovery and purification method according to an embodiment of the present invention. Figure 1 As shown, the embodiment of the present invention includes the following steps:

[0034] S1: Intelligent pretreatment of helium-containing industrial waste gas;

[0035] In this step, micron-sized particles in the helium-containing industrial waste gas are first removed by high-efficiency filters. These particles include dust, metal particles, etc., which will affect the efficiency of subsequent treatment processes. The choice of filter can be customized according to the specific characteristics of the waste gas, such as using electrostatic precipitators, bag filters or high-efficiency air filters (HEPA).

[0036] Next, a multi-stage chemical adsorption unit is used to remove corrosive impurities. These impurities typically include acidic gases (such as hydrogen sulfide, sulfur dioxide), alkaline gases (such as ammonia) and organic pollutants (such as benzene, toluene, etc.). The multi-stage chemical adsorption unit can use different adsorbents, such as activated carbon, molecular sieves, alumina, etc., and each stage is for the adsorption of specific types of impurities.

[0037] In order to improve the efficiency of pretreatment and reduce costs, this application introduces a dynamic adjustment mechanism based on a stochastic predictive control (SPC) system. The core of the SPC system is to establish a dynamic model that includes system state variables (such as temperature, pressure, flow) and control input parameters (such as valve opening, compressor power). This model also takes into account unbounded process noise to better simulate actual working conditions.

[0038] The SPC system predicts the state of helium-containing industrial waste gas through a deep neural network. Specifically, the system first collects various parameters of the waste gas, such as temperature, pressure, flow rate, concentration of various impurities, etc. These data are input into a pre-trained deep neural network and combined with the SPC framework to predict the state of the waste gas.

[0039] Based on the prediction results, the system can dynamically adjust the amount of adsorbent and the replacement frequency in the multi-stage chemical adsorption device. For example, if the concentration of a certain impurity is predicted to increase, the system can increase the amount of the corresponding adsorbent in advance or regenerate it in advance. This dynamic adjustment mechanism can significantly improve adsorption efficiency, extend the service life of the adsorbent, and reduce operating costs.

[0040] S1.1: removing micron-sized particles from the helium-containing industrial waste gas through a filter;

[0041] In S1.1, the embodiment of the present invention uses a multi-stage filtration system to effectively remove micron-sized particles in helium-containing industrial waste gas. This system includes the following key components:

[0042] 1. Primary mechanical filter:

[0043] This is a coarse filter, used to remove large particulate matter. It is usually made of metal mesh or fiber material with a pore size between 10-100 microns. This step can effectively remove most visible particulate matter such as dust, fibers, etc.

[0044] 2. Electrostatic precipitator:

[0045] Electrostatic precipitators use a high voltage electric field to charge particles in the gas, and then adsorb the charged particles to the collector. This method is particularly suitable for removing particles in the range of 1-10 microns. The efficiency of electrostatic precipitators can reach more than 99% and the energy consumption is low.

[0046] 3. High efficiency filter (HEPA filter):

[0047] A HEPA filter is a high-efficiency particulate air filter that can remove particles 0.3 microns and larger with an efficiency of up to 99.97%. HEPA filters are usually made of glass fiber or polypropylene fiber and have extremely high filtration efficiency.

[0048] 4. Ultra-high efficiency filter (ULPA filter):

[0049] For more demanding situations, the present invention can also use ULPA filters, which can remove particles of 0.12 microns and above with an efficiency of 99.9995%.

[0050] 5. Dynamic monitoring system:

[0051] The embodiment of the present invention integrates a dynamic monitoring system in the filtration system, including a differential pressure sensor and a particle concentration detector. This system can monitor the efficiency and blockage degree of the filter in real time, and automatically adjust the fan speed or trigger the cleaning program as needed.

[0052] Through this multi-stage filtration system, the embodiment of the present invention can effectively remove particles of various sizes in helium-containing industrial waste gas, creating good conditions for subsequent processing steps.

[0053] S1.2: Use multi-stage chemical adsorption equipment to remove corrosive impurities, including acid gases, alkaline gases and organic pollutants;

[0054] Different adsorbents and treatment methods are used for different types of corrosive impurities. This device includes the following key parts:

[0055] 1. Acid gas removal unit: mainly for acid gases such as hydrogen sulfide (H2S), sulfur dioxide (SO2), and hydrogen chloride (HCl). Use alkaline adsorbents such as activated alumina, zinc oxide, and sodium carbonate. Use a fixed bed reactor, and the gas flows from bottom to top to fully contact the adsorbent. Set up multiple parallel adsorption towers, which can be switched and regenerated without stopping the machine.

[0056] 2. Alkaline gas removal unit: mainly for alkaline gases such as ammonia (NH3) and amines. Use acidic adsorbents, such as activated carbon impregnated with phosphoric acid or sulfuric acid. Using a moving bed reactor, the adsorbent slowly descends and contacts the rising gas in reverse, improving the adsorption efficiency.

[0057] 3. Organic pollutant removal unit: mainly for volatile organic compounds (VOCs) such as benzene, toluene, and xylene. Use specially treated activated carbon, such as modified activated carbon or molecular sieve activated carbon. Use a rotating adsorption bed to enhance gas-solid contact and improve adsorption efficiency through rotation. The integrated pyrolysis regeneration system can regenerate activated carbon online to extend its service life.

[0058] 4. Composite adsorption unit: Using composite adsorbents, such as activated carbon / alumina composite materials, can remove multiple impurities at the same time. The honeycomb structure is used to increase the contact area and reduce pressure loss.

[0059] 5. Catalytic oxidation unit: For organic matter that is difficult to adsorb directly, catalytic oxidation technology is used. Precious metal catalysts (such as Pt, Pd) or transition metal oxide catalysts are used to achieve complete oxidation of organic matter at a relatively low temperature (200-300°C).

[0060] 6. Online monitoring system: Install gas component analyzers such as gas chromatograph, mass spectrometer, etc. at the inlet and outlet of each unit. Monitor the concentration of various impurities in real time to provide data support for the SPC system.

[0061] Through this multi-stage chemical adsorption device, the embodiment of the present invention can effectively remove various corrosive impurities in helium-containing industrial waste gas, creating favorable conditions for subsequent helium enrichment and purification.

[0062] S1.3: Based on the stochastic predictive control SPC system, dynamically adjust the regeneration period of the multi-stage chemical adsorption device.

[0063] In this step, the embodiment of the present invention develops an intelligent system based on stochastic predictive control (SPC) to dynamically adjust the regeneration cycle of the multi-stage chemical adsorption device. The main components and working principles of this system are as follows:

[0064] 1. Data acquisition module: collects the inlet and outlet gas component data of each adsorption unit in real time. Monitors the temperature, pressure, pressure drop and other parameters of the adsorbent bed. Records the operating time, adsorption amount and other historical data of the adsorption device.

[0065] 2. Data preprocessing module: clean, normalize and standardize the collected data. Use methods such as wavelet transform or Fourier transform to process time series data and extract useful features.

[0066] 3. State estimation module: Use algorithms such as extended Kalman filter (EKF) or particle filter (PF) to estimate the current saturation of the adsorbent. Considering the randomness and uncertainty of the system, the probability distribution of saturation is given.

[0067] 4. Prediction model module: Establish a prediction model based on deep learning, such as long short-term memory network (LSTM) or gated recurrent unit (GRU). Input historical data and current status to predict the change of adsorbent performance in the future. Consider random factors such as changes in inlet gas composition and temperature fluctuations to generate multiple future scenarios.

[0068] 5. Optimization decision module: Based on the model predictive control (MPC) framework, design multi-objective optimization problems. The objective functions include maximizing adsorption efficiency, minimizing energy consumption, minimizing regeneration frequency, etc. Genetic algorithm or particle swarm optimization algorithm is used to solve the optimal regeneration time.

[0069] 6. Adaptive learning module: Continuously compares the prediction results with the actual operation data and updates the parameters of the prediction model. Use reinforcement learning algorithms, such as the Deep Q Network (DQN), to optimize the control strategy.

[0070] 7. Execution control module: According to the optimization decision results, automatically adjust the switching time of each adsorption unit. Control the temperature, airflow and other parameters of the regeneration process to ensure the regeneration effect.

[0071] 8. Human-computer interaction interface: Provides a visual monitoring interface to display the operating status and prediction results of each unit. Allows operators to manually intervene or adjust control parameters.

[0072] Workflow:

[0073] 1. The system continuously collects and processes real-time data.

[0074] 2. The state estimation module gives the probability distribution of the current adsorbent saturation.

[0075] 3. The prediction model module generates multiple future scenarios.

[0076] 4. The optimization decision module solves the optimal regeneration time under these scenarios.

[0077] 5. The system automatically adjusts the regeneration cycle of each adsorption unit based on the decision results.

[0078] 6. The adaptive learning module continuously updates models and strategies to improve the accuracy of prediction and control.

[0079] Through this SPC-based intelligent system, the embodiment of the present invention can dynamically adjust the regeneration cycle of the multi-stage chemical adsorption device according to the real-time working conditions and future predictions. This can not only maximize the adsorption efficiency, but also reduce energy consumption and extend the service life of the adsorbent, thereby significantly improving the performance and economy of the entire pretreatment system.

[0080] S1.3.1: Establish an SPC framework, including:

[0081] defining system state variables, wherein the system state variables include temperature, pressure and flow;

[0082] Determining control input parameters, wherein the input parameters include valve opening and compressor power;

[0083] Build dynamic models of systems that include unbounded process noise;

[0084] Establish conceptual shrinking horizon and rolling horizon formulas;

[0085] S1.3.1: Establish an SPC framework;

[0086] 1. Define system state variables:

[0087] Temperature (T): Measure the temperature of each adsorption bed and gas stream in Kelvin.

[0088] Pressure (P): The pressure at each point in the measurement system, in Pa.

[0089] Flow (Q): measures the volume flow of gas in m³ / s.

[0090] Adsorbent saturation (S): indicates the degree of adsorbent usage, ranging from 0-1.

[0091] Impurity concentration (C): Measures the concentration of various impurities in ppm.

[0092] The state vector can be expressed as:

[0093] X = [T1, T2, ..., Tn, P1, P2, ..., Pn, Q1, Q2, ..., Qn, S1, S2, ..., Sn, C1, C2, ..., Cm].

[0094] 2. Determine the control input parameters:

[0095] Valve opening (V): controls gas flow direction and flow rate, ranging from 0-100%.

[0096] Compressor power (W): controls system pressure, in kW.

[0097] Regeneration heater power (H): controls the temperature of the regeneration process, in kW.

[0098] Switching time (t): Controls the switching time of the adsorption bed, in seconds.

[0099] The control input vector can be expressed as:

[0100] U = [V1, V2, ..., Vn, W1, W2, ..., Wn, H1, H2, ..., Hn, t1, t2, ...,tn].

[0101] 3. Establish a system dynamic model including unbounded process noise:

[0102] Use stochastic differential equations (SDEs) to describe the system dynamics:

[0103] dX = f(X, U, t)dt + G(X, U, t)dW;

[0104] where f(X, U, t) is the deterministic dynamic function, G(X, U, t) is the noise intensity matrix, and W is the Wiener process (representing unbounded process noise).

[0105] 4. Establish conceptual shrinking time domain and rolling time domain formulas:

[0106] Shrinking Horizon:

[0107] At time t, the time domain of the optimization problem is [t, T], where T is a fixed termination time.

[0108] Objective function: J = E[∫[t,T] L(X(τ), U(τ))dτ + Φ(X(T))];

[0109] Where L is the stage cost function and Φ is the terminal cost function.

[0110] Receding Horizon:

[0111] At each time instant t, an optimization problem is solved for a finite horizon [t, t+H], where H is the prediction horizon.

[0112] Only the first element of the optimal control sequence is executed, and then the time domain is rolled forward.

[0113] S1.3.2: Collect various parameters of the helium-containing industrial waste gas:

[0114] 1. Sensor Networks:

[0115] Temperature sensor: Use either a thermocouple or RTD (resistance temperature detector).

[0116] Pressure sensor: Use piezoelectric or capacitive pressure transmitter.

[0117] Flow meter: Use turbine flow meter or Coriolis mass flow meter.

[0118] Gas Analyzer: Use a gas chromatograph (GC) or a mass spectrometer (MS).

[0119] 2. Data acquisition system:

[0120] Use a high-speed A / D converter with a sampling rate of at least 10 Hz.

[0121] A distributed control system (DCS) or programmable logic controller (PLC) is used for data collection and preliminary processing.

[0122] 3. Data preprocessing:

[0123] Remove outliers: Use the 3σ criterion or machine learning-based anomaly detection algorithms.

[0124] Smoothing: Use moving average or Savitzky-Golay filter.

[0125] Data standardization: unify data of different dimensions to the same scale.

[0126] 4. Data Storage:

[0127] Use a time series database such as InfluxDB to store high-frequency sampling data.

[0128] Establish a data warehouse to support rapid retrieval and analysis of historical data.

[0129] S1.3.3: Build and train a deep neural network:

[0130] 1. Network architecture:

[0131] Use recurrent neural networks (RNNs), especially long short-term memory networks (LSTMs) or gated recurrent units (GRUs) to process time series data.

[0132] Network structure: input layer -> LSTM / GRU layer (multi-layer) -> fully connected layer -> output layer;

[0133] 2. Input features:

[0134] Current state variable X(t);

[0135] Historical state sequence X(t-1), X(t-2), ..., X(tn);

[0136] Current and planned control inputs U(t), U(t+1), ..., U(t+m);

[0137] 3. Output:

[0138] The predicted future state sequence X'(t+1), X'(t+2), ..., X'(t+k);

[0139] 4. Loss function:

[0140] Use mean squared error (MSE) or mean absolute error (MAE);

[0141] Add regularization terms to prevent overfitting;

[0142] 5. Training process:

[0143] Use historical running data for offline training; use mini-batch stochastic gradient descent (mini-batch SGD) or Adam optimizer; use early stopping to prevent overfitting; use learning rate decay strategy;

[0144] 6. Model Validation: Use cross-validation to evaluate model performance and perform final evaluation on an independent test set.

[0145] S1.3.4: Input various parameters of the helium-containing industrial waste gas into the deep neural network, so that the deep neural network is combined with the SPC framework to realize the state prediction of the helium-containing industrial waste gas.

[0146] 1. Data preparation: The real-time collected data is preprocessed and organized into a format suitable for neural network input. The input data is normalized.

[0147] 2. State prediction: Input the current state X(t) and historical data into the neural network.

[0148] The neural network outputs the state predictions X'(t+1), X'(t+2), ..., X'(t+k) for the next k time steps.

[0149] 3. Uncertainty quantification: Generate multiple sets of predictions using Monte Carlo methods to obtain the probability distribution of state predictions. Calculate the confidence interval of the prediction results.

[0150] 4. Combined with SPC framework: Input the prediction results of the neural network into the SPC optimization problem. Use the predicted state trajectory to evaluate the performance of different control strategies.

[0151] 5. Feedback correction: Compare the predicted results with the actual observed values ​​and calculate the prediction error. Use Kalman filter or particle filter to correct the state estimate.

[0152] 6. Online learning: Regularly update the neural network model with newly acquired data to adapt to dynamic changes in the system.

[0153] S1.3.5: Based on the state prediction result, dynamically adjust the amount and replacement frequency of the adsorbent in the multi-stage chemical adsorption device;

[0154] 1. Optimization problem construction: Objective function: maximize adsorption efficiency, minimize energy consumption and adsorbent replacement cost. Constraints: equipment capacity limit, adsorbent saturation threshold, product purity requirements, etc.

[0155] 2. Solve the optimization problem: Use the model predictive control (MPC) framework to solve nonlinear optimization problems in the receding horizon.

[0156] The problem is solved using sequential quadratic programming (SQP) or interior point method.

[0157] 3. Adjustment of adsorbent dosage: Dynamically adjust the adsorbent loading of each adsorption bed according to the optimization results. Consider the combined use of different types of adsorbents to deal with different types of impurities.

[0158] 4. Replacement frequency optimization: Based on the adsorbent saturation prediction, dynamically adjust the switching and regeneration frequency of the adsorption bed. Realize predictive maintenance and replace before the adsorption efficiency drops significantly.

[0159] 5. Adaptive control: Monitor system response in real time and compare actual results with expected results. Adaptively adjust control parameters and prediction models based on deviations.

[0160] 6. Decision support: Generate visual reports to display prediction results and optimization suggestions. Provide a human-computer interaction interface to allow operators to make fine adjustments based on experience. In this way, the SPC system can intelligently adjust the operating parameters of the multi-stage chemical adsorption device based on real-time data and future predictions to achieve optimal control of the adsorption process. This not only improves the efficiency and reliability of the system, but also reduces operating costs and energy consumption.

[0161] S2: Recovering waste heat from the helium-containing industrial waste gas using a multi-stage heat recovery system;

[0162] This step aims to maximize the use of heat energy in helium-containing industrial waste gas and improve the energy efficiency of the entire system. The multi-stage heat recovery system includes a high-temperature heat exchanger, a medium-temperature heat exchanger and a low-temperature heat exchanger.

[0163] First, the high-temperature exhaust gas exchanges heat with the inlet cold gas. This process not only reduces the exhaust gas temperature, but also preheats the inlet gas, reducing the energy required for subsequent heating. The medium-temperature exhaust gas obtained after the heat exchange continues to exchange heat with the intermediate process gas, further reducing the exhaust gas temperature and providing heat energy for other processes. Finally, the low-temperature exhaust gas is used to preheat the supplementary gas to achieve the cascade utilization of heat energy.

[0164] In the whole process of heat recovery, this application also introduces a cogeneration system. This system can convert part of the heat energy into electrical energy to supply control systems and auxiliary equipment. The cogeneration system can use technologies such as thermoelectric generators (TEGs) or organic Rankine cycle (ORC), which can be selected according to the specific temperature range and energy requirements.

[0165] Through this combination of multi-stage heat recovery and cogeneration, the overall energy utilization efficiency of the system can be significantly improved, external energy input can be reduced, and operating costs can be reduced.

[0166] S2.1: Exchanging heat between the high-temperature exhaust gas and the inlet cold gas to obtain medium-temperature exhaust gas;

[0167] In this step, the embodiment of the present invention uses a high-efficiency heat exchanger to recover the heat energy in the high-temperature exhaust gas.

[0168] 1. Heat exchanger selection: Use plate heat exchangers or spiral plate heat exchangers. These types of heat exchangers have high heat transfer efficiency and compact structure. The material should be high temperature resistant and corrosion resistant alloys such as Hastelloy or Inconel to cope with the presence of corrosive gases.

[0169] 2. Heat exchange process: High-temperature exhaust gas (usually in the range of 300-500°C) enters from one side of the heat exchanger. Inlet cold gas (usually ambient temperature or pre-treated temperature, about 20-50°C) enters from the other side. The two air flows exchange heat indirectly inside the heat exchanger without material exchange.

[0170] 3. Temperature control: Use temperature sensors to monitor inlet and outlet temperatures in real time. Control heat exchange efficiency by adjusting gas flow or using bypass valves to ensure stable outlet temperature.

[0171] 4. Pressure loss control: Consider minimizing pressure loss when designing the heat exchanger. Use a pressure sensor to monitor the pressure drop and ensure that it does not exceed the set threshold.

[0172] 5. Anti-fouling measures: Install filters at the inlet of the heat exchanger to reduce the contamination of the heat exchange surface by particulate matter. Perform online cleaning regularly, such as using ultrasonic cleaning technology, to reduce downtime.

[0173] 6. Efficiency monitoring: Calculates and displays heat exchange efficiency in real time. When the efficiency drops to a certain threshold, a maintenance alarm is triggered.

[0174] Through this step, the high-temperature exhaust gas is cooled to medium temperature (usually in the range of 150-250°C), and the inlet cold gas is preheated, which greatly improves the energy utilization efficiency of the system.

[0175] S2.2: exchanging heat between the medium-temperature exhaust gas and the intermediate process gas to obtain low-temperature exhaust gas;

[0176] This step further utilizes the thermal energy in the medium-temperature exhaust gas and provides the required heat for other processes.

[0177] 1. Heat exchanger configuration: Use a multi-stream heat exchanger, such as a plate-fin heat exchanger or a compact heat exchanger. This configuration allows multiple air streams at different temperatures to exchange heat simultaneously.

[0178] 2. Intermediate process gas selection: Identify the intermediate processes in the system that need to be heated, such as the desorption gas flow during regeneration, the preheating gas required for certain catalytic reactions, etc. Arrange the appropriate heat exchange sequence according to the temperature requirements of each process.

[0179] 3. Temperature matching optimization: Use the Pinch Analysis method to optimize the heat exchange network. Determine the minimum temperature difference (ΔTmin) to maximize heat recovery while ensuring sufficient heat transfer driving force.

[0180] 4. Flow control: Use variable frequency fans or regulating valves to control the flow of each airflow. Dynamic flow distribution is achieved to adapt to changes in heat demand under different working conditions.

[0181] 5. Heat measurement and control: Install temperature and flow sensors at the inlet and outlet of each heat exchange unit. Calculate the heat exchange amount in real time and dynamically adjust the heat exchange parameters according to demand.

[0182] 6. Intelligent control system: Develop intelligent control algorithms based on model predictive control (MPC). Predict the heat demand of each process in the future and adjust the heat exchange network parameters in advance.

[0183] Through this step, the medium-temperature exhaust gas is further cooled to a low temperature (usually in the range of 50-100°C), while providing the required heat for multiple intermediate processes, realizing the cascade utilization of thermal energy.

[0184] S2.3: using the low temperature exhaust gas to preheat the supplementary gas;

[0185] This step utilizes the remaining heat in the low-temperature exhaust gas to further improve the energy efficiency of the system.

[0186] 1. Make-up gas identification: Identify the make-up gas in the system that needs to be preheated, such as fresh feed, purge gas for regeneration, etc.

[0187] Analyze the temperature requirements and flow characteristics of these make-up gases.

[0188] 2. Heat exchanger selection: For small flow rates of make-up gas, a shell and tube heat exchanger can be used.

[0189] For large flow rates of make-up gas, a heat pipe heat exchanger may be considered, which has the advantages of efficient heat transfer and no powered components.

[0190] 3. Temperature control strategy: Implement a graded preheating strategy to heat exchange the supplementary gas with different temperature requirements with the low-temperature exhaust gas in different temperature ranges. Use a three-way regulating valve to control the degree of heat exchange to ensure that the supplementary gas is heated to the required temperature.

[0191] 4. Condensate recovery: Condensate is generated during the cooling of low-temperature exhaust gas, and a condensate collection system is installed. The collected condensate can be used for other process water or reused after treatment.

[0192] 5. Anti-condensation design: Considering that low-temperature exhaust gas contains moisture, attention should be paid to anti-condensation when designing the heat exchanger.

[0193] Use hydrophobic design or heating device to prevent condensation on the heat exchange surface and affect heat transfer efficiency.

[0194] 6. Waste heat utilization evaluation: Real-time calculation and display of waste heat utilization rate.

[0195] When the waste heat utilization rate is lower than the preset threshold, the system automatically prompts optimization suggestions.

[0196] Through this step, the residual heat in the low-temperature exhaust gas is fully utilized, the make-up gas is preheated, and the energy efficiency of the entire system is further improved.

[0197] S2.4: During heating and cooling, use a cogeneration system to convert heat into electrical energy for use by control systems and auxiliary equipment;

[0198] This step converts part of the thermal energy into electrical energy through cogeneration technology, thus achieving efficient use of energy.

[0199] 1. CHP technology selection: For the high temperature section (>400°C), the Organic Rankine Cycle (ORC) system can be used.

[0200] For the medium and low temperature range (100-400°C), thermoelectric generation modules (TEG) can be used.

[0201] 2. Organic Rankine Cycle (ORC) System: Select a suitable working fluid, such as cyclopentane or R245fa, and optimize it according to the heat source temperature. Design a closed-loop system including evaporator, turbine generator, condenser and working fluid pump. Use a variable speed drive to control the turbine speed and optimize the power generation efficiency under different heat loads.

[0202] 3. Thermoelectric generator (TEG): Select high-performance thermoelectric materials such as bismuth telluride or silicon germanium alloy. Design the thermoelectric module array to maximize the temperature difference and power generation efficiency. Use MPPT (maximum power point tracking) technology to optimize output power.

[0203] 4. Power management system: Design intelligent power distribution system to prioritize supply to key control equipment. Use battery energy storage system to balance fluctuations in power generation and consumption. Develop load forecasting algorithms to optimize energy distribution.

[0204] 5. Waste heat management: Design a waste heat recovery device for the cogeneration system to further improve energy utilization. Use waste heat for preheating or other low-temperature heat needs. Through this cogeneration system, the embodiment of the present invention not only improves the energy utilization efficiency of the entire helium recovery and purification system, but also enhances the energy self-sufficiency and reliability of the system. The generated electrical energy can be directly supplied to auxiliary equipment such as control systems, sensor networks, and communication equipment, reducing external power demand and improving the economy and environmental protection of the entire system.

[0205] S3: removing impurities from the helium-containing industrial waste gas using a multi-stage catalytic reactor system;

[0206] The multi-stage catalytic reactor system is designed to effectively remove various impurities from helium-containing industrial waste gas, taking into account the chemical properties of different impurities and the reaction conditions required. The system includes three reactor stages, each stage treating a specific type of impurity.

[0207] The first stage uses a fixed bed reactor. This reactor is suitable for treating impurities that require a long residence time or have slow reaction kinetics. For example, it can be used to remove some stable organic compounds or some metal vapors. The advantages of a fixed bed reactor are simple structure, stable operation, and suitable for continuous operation.

[0208] The second stage uses a pulse reactor. This reactor improves the reaction efficiency by periodically changing the reaction conditions (such as temperature, pressure or reactant concentration). Pulse reactors are particularly suitable for treating impurities that are difficult to react under steady-state conditions or when frequent catalyst regeneration is required. For example, it can be used to remove some organic matter that is difficult to oxidize or some non-metallic gas impurities.

[0209] The third stage uses plasma-assisted catalysis. This is a new type of catalytic technology that activates reactant molecules by generating plasma, lowers the reaction temperature, and increases the reaction rate. Plasma-assisted catalysis is particularly suitable for treating some impurities that are difficult to remove under normal conditions, such as some inert gases or organic matter that is difficult to decompose. The advantage of this technology is that it can achieve efficient catalysis at a lower temperature, reduce energy consumption, and avoid side reactions caused by high temperature.

[0210] Each reactor stage is equipped with a corresponding catalyst regeneration system to ensure that the activity of the catalyst can be maintained at a high level for a long time. The entire multi-stage catalytic reactor system is coordinated by an intelligent control system, which dynamically adjusts the operating parameters of each reactor stage, such as temperature, pressure, gas flow rate, etc., according to the composition and concentration of the inlet gas to achieve the best impurity removal effect.

[0211] S3.1: The first stage uses a fixed bed reactor to remove the first impurity;

[0212] Fixed bed reactor is a commonly used gas-solid phase catalytic reaction device, which is suitable for treating impurities that require a long residence time or have slow reaction kinetics.

[0213] 1. Reactor design: A vertical cylindrical reactor is used, which is filled with catalyst particles. The reactor material is made of high temperature and corrosion resistant alloys, such as 316L stainless steel or Hastelloy. A multi-layer catalyst support grid is designed inside the reactor to ensure uniform distribution of the catalyst.

[0214] 2. Catalyst selection: Select a suitable catalyst according to the properties of the first impurity, for example: For sulfides: use zinc oxide (ZnO) or iron oxide (Fe2O3) catalysts. For halogen compounds: use activated carbon-supported noble metal catalysts (such as Pd / C). The catalyst shape is designed to be spherical or cylindrical to reduce pressure drop.

[0215] 3. Temperature control: Install a heating jacket or electric heating belt on the outside of the reactor. Set multiple temperature measurement points along the reactor axis to achieve precise temperature control. Use PID control algorithm to adjust the optimal temperature curve according to reaction kinetics.

[0216] 4. Pressure control: Install inlet and outlet pressure sensors to monitor pressure drop. Use back pressure regulator to control reactor pressure and optimize reaction conditions.

[0217] 5. Flow distribution: A gas distributor is designed at the top of the reactor to ensure uniform distribution of the inlet gas. A gas collector is set at the bottom to reduce short-circuit flow.

[0218] 6. Regeneration system: The parallel reactor system is designed to allow online regeneration. The regeneration process uses temperature programmed oxidation (TPO) technology to effectively remove carbon deposits.

[0219] 7. Monitoring and control: Use online gas chromatograph (GC) or mass spectrometer (MS) to monitor the outlet gas composition. Develop a machine learning-based prediction model to predict catalyst deactivation trends.

[0220] 8. Safety measures: Install emergency pressure relief devices and inert gas purge systems. Set safety interlocks for temperature, pressure and other parameters.

[0221] Through this fixed bed reactor, the embodiment of the present invention can effectively remove the first major impurity and create good conditions for subsequent processing.

[0222] S3.2: The second stage uses a pulse reactor to remove the second impurity by periodically changing the reaction conditions;

[0223] Pulse reactors can improve reaction efficiency by periodically changing reaction conditions (such as temperature, pressure or reactant concentration) and are particularly suitable for processing impurities that are difficult to react under steady-state conditions.

[0224] 1. Reactor design: adopts a multi-tube bundle structure, and each tube is filled with catalyst. A heating / cooling jacket is set outside the tube bundle to achieve rapid temperature change. The material is selected from alloys with high thermal conductivity and low heat capacity, such as aluminum alloy or copper-nickel alloy.

[0225] 2. Pulse system: Design a high-precision solenoid valve group to achieve millisecond-level gas pulse control. Use a mass flow controller (MFC) to accurately control the amount of gas in each pulse.

[0226] 3. Temperature control: A fast-response electric heating system and liquid nitrogen cooling system are used. A thermocouple array is used to monitor the temperature distribution in real time.

[0227] Develop a model predictive control (MPC) algorithm to achieve precise temperature trajectory control.

[0228] 4. Pressure control: Use a back pressure regulating valve with high response speed. Install a pressure buffer tank to reduce the impact of pressure fluctuations on upstream and downstream.

[0229] 5. Catalyst selection: Select catalysts with high specific surface area and good thermal stability, such as precious metal catalysts loaded with molecular sieves. The catalyst shape is designed to be a honeycomb or foam metal structure to improve mass transfer efficiency.

[0230] 6. Pulse optimization: Develop an adaptive pulse optimization algorithm to automatically adjust pulse parameters based on reaction kinetics. Parameters include pulse frequency, duration, amplitude, etc.

[0231] 7. Online monitoring: Use fast-response gas analyzers, such as Fourier transform infrared spectrometers (FTIR). Develop real-time data processing algorithms to evaluate the response effect of each pulse cycle.

[0232] 8. Regeneration strategy: Design in-situ regeneration function, use temperature pulse and gas component switching to achieve catalyst regeneration. Develop intelligent regeneration scheduling algorithm to minimize the impact of regeneration on production.

[0233] 9. Energy recovery: Design a heat recovery system to utilize the waste heat during the temperature pulse process. Use a thermoelectric converter to convert temperature fluctuations into electrical energy. Through the pulse reactor, the embodiments of the present invention can efficiently remove the second impurities, especially those components that are difficult to handle in conventional fixed bed reactors.

[0234] S3.3: The third stage uses plasma-assisted catalysis to remove the third impurity at low temperature;

[0235] Plasma-assisted catalytic technology combines the high activity of plasma and the selectivity of catalysts, which can achieve the removal of difficult-to-treat impurities at lower temperatures while reducing energy consumption.

[0236] 1. Reactor design: It adopts a coaxial cylindrical structure, with the internal electrode being a high-voltage electrode and the external electrode being a ground electrode. The catalyst is filled in the discharge area to form a dense catalyst bed. The dielectric barrier discharge (DBD) technology is used to cover the electrode surface with a dielectric layer.

[0237] 2. Power supply system: Design high-frequency and high-voltage power supply with a frequency range of 1-100 kHz and a voltage range of 5-30 kV.

[0238] The use of a tunable power supply can dynamically adjust the discharge parameters according to the gas composition.

[0239] 3. Catalyst selection: Use nanostructured catalysts, such as metal oxides such as TiO2 and CeO2. The catalyst is loaded on a porous carrier, such as γ-Al2O3 or molecular sieves. Consider the plasma-catalyst synergistic effect and optimize the catalyst composition.

[0240] 4. Temperature control: Design a precise temperature control system to maintain a low reaction temperature (usually <200°C).

[0241] A liquid cooling system is used to remove the heat generated by the discharge.

[0242] 5. Gas distribution: Design a special gas distributor to ensure that the gas flows evenly through the discharge area. Optimize the gas flow path to maximize the contact between plasma and catalyst.

[0243] 6. Energy optimization: Develop intelligent discharge control algorithms to dynamically adjust discharge power according to gas composition.

[0244] Use pulse discharge technology to reduce energy loss.

[0245] 7. Online monitoring: Integrated optical emission spectroscopy (OES) system to monitor plasma characteristics in real time.

[0246] Monitor the outlet gas composition using a mass spectrometer (MS) or gas chromatography-mass spectrometry (GC-MS).

[0247] 8. By-product treatment: Design post-treatment units to remove by-products such as ozone produced by plasma reactions.

[0248] These by-products are treated using catalytic decomposition or thermal decomposition techniques.

[0249] 9. Safety measures: Install high voltage interlock system to prevent accidental contact. Design electromagnetic shielding to reduce electromagnetic interference.

[0250] Install ozone detectors and emergency shut-off systems.

[0251] 10. Regeneration strategy: Utilize the characteristics of plasma to design in-situ catalyst regeneration function.

[0252] By adjusting the discharge parameters and gas composition, the catalyst can be activated periodically.

[0253] 11. Intelligent control: Develop control algorithms based on machine learning to optimize plasma parameters and catalytic conditions.

[0254] Realize rapid response and adjustment to gas composition and flow fluctuations.

[0255] Through plasma-assisted catalytic technology, the embodiment of the present invention can efficiently remove the third type of difficult-to-treat impurities at low temperature while significantly reducing energy consumption. This technology is particularly suitable for treating low-concentration, difficult-to-degrade organic pollutants or inert gas impurities.

[0256] The combined use of this three-stage catalytic reactor system can effectively treat a variety of impurities in helium-containing industrial waste gas, creating favorable conditions for subsequent helium enrichment and purification. Each stage of the reactor is optimized for a specific type of impurity, ensuring the efficiency and reliability of the entire system.

[0257] S4: Helium enrichment and preliminary purification using gas separation technology;

[0258] This step uses three advanced gas separation technologies to efficiently enrich and initially purify helium.

[0259] First, a rotating bed adsorption separation device is used. This device uses centrifugal force to enhance the mass transfer process, significantly improving the helium enrichment efficiency. In a rotating bed, the adsorbent is placed in a high-speed rotating cylinder. When the helium-containing gas flows through the rotating bed, the centrifugal force makes the contact between the gas molecules and the adsorbent more complete, thereby enhancing the adsorption effect. This method is particularly suitable for processing large flow and low concentration helium mixtures.

[0260] Secondly, the enriched helium enters a multi-stage membrane separation system. This system uses membrane materials with different pore sizes and selectivities to gradually improve the purity of the helium. The advantages of membrane separation technology are low energy consumption, small footprint, and simple operation. In this application, a multi-stage cascade membrane separation configuration is adopted, and each stage uses membrane materials with different characteristics to specifically separate specific impurities. For example, the first stage uses a membrane with high selectivity for carbon dioxide and water molecules, the second stage uses a membrane with high selectivity for nitrogen, and so on.

[0261] Finally, we introduce ionized gas separation technology. This technology uses the difference in ionization energy of different gases for separation. Under the action of the electric field, different gas molecules are ionized into charged particles. Due to the difference in mass and charge, these ions move in different trajectories in the electric field, thus achieving separation. Ionized gas separation technology is particularly suitable for separating helium and other inert gases because helium has a higher ionization energy and its behavior in the electric field is significantly different from that of other gases.

[0262] The combination of these three separation technologies can effectively improve the purity of helium while maintaining a high recovery rate. The gas flow and pressure between the separation units are coordinated and controlled by an intelligent monitoring and maintenance system to ensure the optimal operation of the entire separation process.

[0263] In addition, before adopting gas separation technology for helium enrichment and preliminary purification, the method also introduces an adaptive scalar field mapping system to further optimize the separation process.

[0264] First, a scalar field of gas component concentration is established. This scalar field is a multidimensional data structure that describes the concentration distribution of each component in the gas mixture in space and time. Through real-time monitoring and data analysis, the system can build a dynamic, high-precision concentration distribution model.

[0265] Next, a high intensity region recognition algorithm is used to detect and identify high intensity regions. These regions typically represent locations with high helium concentrations or high concentrations of certain impurities. Identifying these regions is critical to optimizing the subsequent separation process.

[0266] Based on the identified high-intensity areas, the system builds an adaptive avoidance strategy. The purpose of this strategy is to avoid areas with high impurity concentrations (e.g., greater than 500 mol / m 3 ), while giving priority to areas with high helium concentration, thereby improving separation efficiency.

[0267] With the adaptive scalar field mapping system, the embodiment of the present invention can optimize the structure and control the centrifugal force field of the rotating bed adsorption separation device. For example, the rotation speed and gas inlet position of the rotating bed can be dynamically adjusted according to the distribution of the scalar field to maximize the helium enrichment effect.

[0268] For membrane separation systems, embodiments of the present invention can dynamically adjust the spatial arrangement and cascade structure of membrane components based on scalar fields. For example, in areas with high helium concentrations, the number of membrane components can be increased or the type of membrane can be adjusted to improve separation efficiency. At the same time, the system can also adaptively adjust the transmembrane pressure difference to optimize the separation effect at different concentrations.

[0269] In the process of ionized gas separation, the embodiments of the present invention can dynamically adjust the plasma gas parameters according to the scalar field. By adjusting the electric field strength, ionization degree and other parameters, the ion flow can avoid the high impurity concentration area as much as possible during the movement, thereby improving the separation efficiency.

[0270] S4.1: Use a rotating bed adsorption separation device to enrich helium by utilizing centrifugal force to enhance the mass transfer process.

[0271] The rotating bed adsorption separation device is an innovative gas separation technology that enhances the gas-solid mass transfer process and improves separation efficiency through high-speed rotation.

[0272] 1. Device structure design: A cylindrical rotating bed with a vertical axial design is used. The interior is filled with specially designed adsorbents, such as molecular sieves or activated carbon. High-strength, lightweight materials are used to make rotating parts, such as carbon fiber composite materials.

[0273] 2. Drive system: It is driven by a high-speed motor with a speed of 1000-3000 rpm. The frequency converter is used to accurately control the speed and achieve dynamic adjustment. The magnetic bearing system is designed to reduce friction and vibration.

[0274] 3. Adsorbent selection and optimization: Choose an adsorbent with low selectivity for helium and high selectivity for other gases. Optimize the shape and size of adsorbent particles, such as using spherical or cylindrical particles. Consider using composite adsorbents, such as molecular sieve / activated carbon composites.

[0275] 4. Gas distribution system: A special gas inlet distributor is designed to ensure that the gas enters the rotating bed evenly. A static gas collection ring is set outside the rotating bed.

[0276] 5. Pressure control: Use high-precision back pressure regulating valve to control system pressure. Design dynamic sealing system to ensure air tightness under high-speed rotation.

[0277] 6. Temperature control: A cooling jacket is set outside the rotating bed to control the temperature of the adsorption process. A non-contact infrared temperature sensor is used to monitor the bed temperature.

[0278] 7. Regeneration system: Design in-situ regeneration function, use centrifugal force to enhance thermal regeneration or pressure fluctuation regeneration. Develop intelligent regeneration scheduling algorithm to optimize regeneration time and frequency.

[0279] 8. Monitoring and control: Use high-speed data acquisition systems to monitor speed, pressure, temperature and other parameters in real time. Develop intelligent control algorithms based on model predictive control (MPC) to dynamically optimize operating parameters.

[0280] 9. Safety measures: Install a vibration monitoring system to detect abnormal vibration and automatically shut down the machine. Design an emergency braking system and overspeed protection device.

[0281] 10. Energy recovery: Use power generation braking during deceleration to recover part of the kinetic energy. Through the rotating bed adsorption separation device, the embodiment of the present invention can efficiently enrich helium, laying the foundation for the subsequent purification process.

[0282] S4.2: The enriched helium enters a multi-stage membrane separation system, which uses membrane materials with different pore sizes and selectivities to gradually improve the purity of the helium. The multi-stage membrane separation system uses the differences in size and permeability of different gas molecules to gradually improve the purity of the helium.

[0283] 1. Membrane material selection: First stage: Use polyimide (PI) or polysulfone (PSF) membrane to mainly separate large molecular impurities.

[0284] Second stage: further purification using polyetheretherketone (PEEK) or polytetrafluoroethylene (PTFE) membrane.

[0285] The third stage: Use carbon molecular sieve membrane (CMS) or zeolite membrane to achieve high selective separation.

[0286] 2. Membrane module design: Use hollow fiber membrane or spiral membrane structure to maximize membrane area. Design modular membrane modules for easy replacement and maintenance.

[0287] 3. Multi-stage cascade system: Design a 3-5-stage membrane separation system with compressors and heat exchangers between each stage. Use cross-flow or counter-flow configuration to optimize separation efficiency.

[0288] 4. Pressure control: Use high-precision pressure sensors and regulating valves to accurately control the pressure difference on both sides of the membrane. Develop a dynamic pressure optimization algorithm to adjust the operating pressure in real time according to the feed composition.

[0289] 5. Temperature management: Set up precise temperature control unit before each membrane separation stage. Optimize operating temperature and balance permeability and selectivity.

[0290] 6. Flow control: Use mass flow controller (MFC) to accurately control the feed flow of each stage. Design intelligent flow distribution algorithm to optimize the load of each membrane stage.

[0291] 7. Online monitoring: Install gas analyzers at the inlet and outlet of each membrane separation unit. Use a helium mass spectrometer leak detector to monitor membrane integrity.

[0292] 8. Membrane cleaning system: Design an online backwash system to regularly remove deposits on the membrane surface. Develop a chemical cleaning program for deep cleaning and regeneration.

[0293] 9. Intelligent control: Develop an adaptive control system based on machine learning to optimize the multi-stage membrane separation process. Predict membrane performance degradation in real time and arrange preventive maintenance.

[0294] 10. Energy optimization: Design membrane cascade heat integration system to reduce energy consumption. Use pressure energy recovery devices such as turbine expansion machines.

[0295] Through a multi-stage membrane separation system, the embodiments of the present invention can gradually improve the purity of helium while minimizing energy consumption and operating costs.

[0296] S4.3: Introducing ionized gas separation technology to separate gases using the difference in ionization energy of different gases; wherein the gas flow and pressure between the separation units are coordinated and controlled by the intelligent monitoring and maintenance system;

[0297] Ionized gas separation technology utilizes the difference in ionization energy of different gases to achieve high-precision separation through the action of electric field. It is particularly suitable for processing difficult-to-separate gas mixtures.

[0298] 1. Ionization chamber design: A multi-stage ionization chamber design is used, with each stage targeting specific impurities. The ionization chamber is made of high-purity quartz or ceramic materials to minimize the introduction of impurities.

[0299] 2. Ionization source: Design an efficient electron bombardment source to produce a stable electron beam. Use a radio frequency (RF) or microwave source as an auxiliary ionization method.

[0300] 3. Electric field system: Design a complex multi-level electric field system, including acceleration field, deflection field and focusing field. Use high-precision high-voltage power supply to ensure the stability of the electric field.

[0301] 4. Ion detector: Use a Faraday cup or secondary electron multiplier (SEM) as an ion detector. Design a multi-channel detection system to monitor multiple ions simultaneously.

[0302] 5. Vacuum system: Use a multi-stage vacuum pump system, including molecular pumps and turbomolecular pumps. Design a differential pumping system to maintain the best vacuum level in each area.

[0303] 6. Neutralization device: An electronic neutralization device is set after separation to reduce the separated ions to neutral gas.

[0304] 7. Cooling system: Design an efficient cooling system to remove the heat generated by the ionization process. Use liquid nitrogen cold trap to capture impurities.

[0305] 8. Online monitoring: Integrated mass spectrometer to monitor separation effect in real time. Use optical emission spectrometer (OES) to monitor the ionization process.

[0306] 9. Intelligent control system: Develop control algorithms based on artificial intelligence to dynamically optimize ionization and separation parameters. Achieve coordinated control of parameters such as electric field intensity, ion energy, and gas flow.

[0307] 10. Safety measures: Design multiple interlock systems to prevent high pressure and radiation hazards. Install emergency shut-off and inert gas injection systems.

[0308] 11. Energy management: Develop intelligent energy management systems to optimize power efficiency. Design energy recovery devices to utilize ion kinetic energy.

[0309] Coordinated control of intelligent monitoring and maintenance systems:

[0310] 1. Central Control Unit: Use high-performance industrial computers as the central control unit. Develop a distributed control architecture to ensure system reliability and response speed.

[0311] 2. Data collection and processing: Use high-speed data collection systems to collect the operating data of each unit in real time. Develop a big data analysis platform to process and analyze massive operating data.

[0312] 3. Model predictive control: Build a dynamic model of the entire separation system and use the model predictive control (MPC) algorithm to optimize the coordinated operation of each unit.

[0313] 4. Adaptive control strategy: Develop an adaptive control algorithm based on machine learning to adapt to changes in feed composition and operating conditions. Achieve dynamic load balancing between separation units.

[0314] 5. Fault diagnosis and predictive maintenance: Use artificial intelligence algorithms to monitor equipment health and predict faults. Develop a predictive maintenance scheduling system to optimize maintenance plans.

[0315] 6. Virtual sensing technology: Develop soft sensing models to estimate process parameters that are difficult to measure directly. Use data fusion technology to improve measurement accuracy and reliability.

[0316] 7. Human-computer interaction interface: Design an intuitive 3D visualization interface to display the operating status of the entire system. Develop a mobile monitoring app to achieve remote monitoring and control.

[0317] 8. Safety Management System: Develop a safety management system based on risk assessment. Design emergency response plans to automatically handle abnormal situations.

[0318] 9. Energy optimization: Develop energy management and optimization systems to minimize overall energy consumption. Realize comprehensive utilization of waste heat and pressure energy between units.

[0319] 10. Quality control: Establish an online quality control system to ensure the stability of the purity of the helium product. Develop an automated sampling and analysis system to achieve full-process quality traceability.

[0320] Through this intelligent coordinated control, the embodiment of the present invention can achieve efficient and stable operation of the entire helium separation system, maximize separation efficiency, and minimize energy consumption and cost. This integrated separation technology and intelligent control system represents the latest technology in the field of gas separation and can produce ultra-high purity helium products.

[0321] A1: Before the helium is enriched and initially purified by using gas separation technology, the method further comprises:

[0322] A1.1: Establish an adaptive scalar field mapping system, including:

[0323] Establish a scalar field of gas component concentrations; detect and identify high-intensity areas using a high-intensity area recognition algorithm; construct an adaptive avoidance strategy;

[0324] A1.2: Based on the adaptive scalar field mapping system, the rotating bed adsorption separation device is structurally optimized and the centrifugal force field is regulated;

[0325] A1.3: Based on the adaptive scalar field mapping system, the spatial arrangement and cascade structure of the membrane components are dynamically adjusted to adaptively adjust the transmembrane pressure difference;

[0326] A1.4: Based on the adaptive scalar field mapping system, the plasma gas parameters are dynamically adjusted to avoid high-intensity areas during the movement of the ion flow.

[0327] S5: distilling and purifying the helium through a distillation module;

[0328] Distillation is a separation technology based on boiling point differences, and is particularly suitable for processing low-boiling-point helium. This application uses advanced distillation technology, including an internal heat-integrated distillation tower, a distributed feed system, a centrifugal distillation unit, and an efficient condenser and reboiler.

[0329] The design of the internal heat integration distillation tower can significantly improve energy efficiency. By setting up a heat exchange device inside the tower, the heat can be used in a step-by-step manner, reducing the need for external heat sources. This design is particularly suitable for helium distillation, because helium has an extremely low boiling point, and traditional distillation towers often require a lot of energy to maintain low temperatures.

[0330] The distributed feeding system allows feeding at different heights of the distillation column. This design can select the optimal feeding position according to the concentration and properties of different components, thereby improving the separation efficiency. For helium distillation, this means that the feeding strategy can be flexibly adjusted according to the concentration of helium and the properties of other impurities.

[0331] The centrifugal distillation unit is an innovative design that incorporates the principles of centrifugal force and distillation. Through high-speed rotation, the contact and mass transfer process between the gas and liquid phases can be enhanced, thereby improving the separation efficiency. This technology is particularly suitable for processing difficult-to-separate mixtures, such as mixtures of helium and other low-boiling point gases.

[0332] Efficient condensers and reboilers are key components of the distillation system. Considering the low boiling point of helium, the condenser adopts a multi-stage refrigeration system to achieve ultra-low temperature condensation. The reboiler adopts an efficient heat transfer design to ensure that sufficient vaporization heat can be provided while minimizing energy loss.

[0333] The entire distillation process is monitored and regulated in real time by an intelligent control system. The system can dynamically adjust operating parameters such as reflux ratio and heat load according to parameters such as temperature, pressure, and component concentration in the tower to achieve the best separation effect.

[0334] 1. Distillation tower design: Ultra-low temperature distillation tower is used, and the material is low temperature resistant alloy, such as 316L stainless steel or aluminum alloy. The tower body adopts vacuum multi-layer insulation design to minimize heat loss. Structured packing, such as Sulzer Mellapak or Koch-Glitsch FLEXIPAC, is used in the tower to improve mass transfer efficiency. The tower diameter and height are optimized according to the processing capacity and required purity, and the height-to-diameter ratio is usually between 20-30.

[0335] 2. Cooling system: Use a multi-stage refrigeration cycle system, such as a nitrogen-helium cascade refrigeration system. The first stage uses liquid nitrogen pre-cooling, and the temperature can reach -196°C. The second stage uses helium expansion refrigeration, which can reduce the temperature to about -269°C. Design an efficient heat exchange network to maximize energy utilization.

[0336] 3. Feeding system: Design a multi-point feeding system and optimize the feeding position according to the feed composition. Use an accurate mass flow controller (MFC) to control the feeding rate. Design a feed pre-cooling system to reduce cooling loss.

[0337] 4. Reflux system: Design a top full condenser and a bottom reboiler. Use high-efficiency plate heat exchangers as condensers and reboilers. Design a variable ratio reflux control system to dynamically adjust the reflux ratio.

[0338] 5. Internal structure optimization: Design an efficient liquid distributor to ensure uniform distribution of liquid. Use a demister to reduce droplet entrainment.

[0339] Install temperature sensor arrays at key locations to accurately monitor temperature distribution.

[0340] 6. Pressure control: Use high-precision pressure sensors and control valves to accurately control the pressure at the top of the tower. Design a pressure buffer system to reduce the impact of external disturbances on the pressure inside the tower.

[0341] 7. Sampling system: Set up automatic sampling points at key locations of the tower. Use an online gas chromatograph or mass spectrometer to analyze components in real time.

[0342] 8. Intelligent control system: Develop advanced control strategies based on model predictive control (MPC). Use neural network models to predict product purity and implement feedforward control. Develop adaptive control algorithms to dynamically optimize operating parameters.

[0343] 9. Energy integration: Design an internal heat integration system to utilize the temperature difference at different locations in the tower. Use heat pump technology to improve energy efficiency.

[0344] 10. Safety measures: Install emergency exhaust system and safety valve. Design inert gas purge system for startup and shutdown. Install helium leak detection system to detect leaks in time.

[0345] 11. Product purification: A molecular sieve adsorption bed is set at the top of the distillation tower to further remove trace impurities. A hydrogen purifier (such as palladium membrane) is used to remove the existing hydrogen.

[0346] 12. Online analysis: Install high-precision helium purity analyzers, such as mass spectrometers or laser spectrometers. Develop real-time purity prediction models and combine soft sensing technology to estimate parameters that are difficult to measure directly.

[0347] 13. Automated operation: Develop automatic start-up and shutdown procedures for the distillation process. Design dynamic load adjustment strategies to adapt to changes in feed volume and composition.

[0348] 14. Maintenance strategy: Develop a predictive maintenance system to predict maintenance needs based on equipment status and operating data. Design an online cleaning system to reduce downtime for maintenance.

[0349] 15. Data management: Establish a digital twin model of the distillation process for optimization and training. Develop a big data analysis platform to explore potential optimization opportunities in long-term operation data.

[0350] 16. Environmental control: Design a temperature and humidity control system in the distillation area to prevent external environmental influences. Implement strict cleanliness management to prevent the introduction of external impurities.

[0351] Through this advanced distillation system, the embodiments of the present invention can improve the purity of helium to a very high level (usually up to 99.99% or more). The advantage of the distillation process is that it can run continuously, process a large amount of gas, and can effectively separate components with similar boiling points. Combined with intelligent control and optimization strategies, this distillation module can not only produce high-purity helium, but also achieve efficient, stable and economical operation.

[0352] This distillation technology is combined with the other separation technologies mentioned above (such as rotating bed adsorption, membrane separation, and ionized gas separation) to form a comprehensive helium purification system that can efficiently recover and purify helium from various helium-containing industrial waste gases.

[0353] S6: Use purification technology to achieve fine purification of helium;

[0354] In this step, the embodiment of the present invention adopts three advanced purification technologies to purify the helium to ultra-high purity (>99.9999%).

[0355] First, the vortex tube separation technology is introduced. The vortex tube is a gas separation device without moving parts, which uses the gas temperature separation effect for separation. When the compressed gas enters the vortex tube tangentially, a high-speed rotating airflow is formed. Due to the action of centrifugal force, the airflow is divided into two streams: one is the hot end airflow with a higher temperature, and the other is the cold end airflow with a lower temperature. For helium mixtures, due to the different thermodynamic properties of different components, their behaviors in the vortex tube are also different, thereby achieving further separation. The advantages of this technology are simple structure, no moving parts, low maintenance cost, and it is particularly suitable for processing small flow and high purity gases.

[0356] Secondly, a supercritical fluid extraction device is used. Supercritical fluid (usually supercritical carbon dioxide) has unique dissolving properties and can selectively dissolve and extract certain trace organic impurities. In this process, helium comes into contact with the supercritical fluid, and the impurities are dissolved and taken away by the supercritical fluid, while helium, due to its special physical and chemical properties, will not be dissolved by the supercritical fluid. This method is particularly effective in removing trace organic pollutants that are difficult to remove by conventional methods.

[0357] Finally, acoustic wave separation technology is used. This technology uses the acoustic properties of different gases for fine purification. When sound waves pass through a gas mixture, different components will exhibit different acoustic behaviors due to differences in physical properties such as mass and density. Through a carefully designed sound field, further separation of gas components can be achieved. The advantage of acoustic wave separation technology is that it is a non-contact separation method that does not introduce any new impurities.

[0358] The entire fine purification process is monitored by the intelligent monitoring and maintenance system through the model predictive path integral (MPPI) technology. MPPI technology is an advanced control algorithm that can achieve optimal control in complex nonlinear systems. In this application, MPPI technology can predict the future system status based on the real-time monitored gas composition data and generate the optimal control strategy to ensure that the helium purity is always maintained above 99.9999%.

[0359] Through the combined use of these three technologies, the embodiment of the present invention can effectively remove various trace impurities in helium, including other inert gases, organic matter, moisture, etc., and finally obtain an ultra-high purity helium product.

[0360] S6.1: Introduce vortex tube separation technology to further purify helium using the gas temperature separation effect;

[0361] A vortex tube is a gas separation device without moving parts that uses the temperature separation effect produced by high-speed rotating airflow to separate gas mixtures.

[0362] 1. Vortex tube design: Adopt bidirectional vortex tube design to maximize separation effect. Optimize vortex tube geometric parameters, such as aspect ratio, inlet nozzle angle, etc. Use high-strength, wear-resistant materials such as tungsten carbide to make key components.

[0363] 2. Air intake system: Design an efficient gas compression system to provide a high-pressure gas source (usually 4-10 bar). Use a precision filter to remove particles that damage the vortex tube.

[0364] 3. Temperature control: Install high-precision temperature sensors at the cold and hot ends. Develop adaptive control algorithms to dynamically adjust the ratio of cold and hot airflows.

[0365] 4. Multi-stage cascade: Design a multi-stage vortex tube system to further improve the separation effect. Optimize inter-stage heat exchange and improve energy efficiency.

[0366] 5. Pressure control: Use precision back pressure regulating valve to control the hot and cold end pressure. Develop dynamic pressure optimization algorithm to maximize separation efficiency.

[0367] 6. Online monitoring: Integrate a micro mass spectrometer to monitor the gas components at the hot and cold ends in real time. Use a high-precision flow meter to monitor the flow of each gas stream.

[0368] 7. Energy recovery: Design a pressure energy recovery system, such as using a turbo expander. Use the heat energy of the hot end gas for thermoelectric conversion.

[0369] 8. Intelligent control: Develop control algorithms based on machine learning to optimize operating parameters and achieve coordinated control of parameters such as intake pressure, temperature, and flow.

[0370] S6.2: Use a supercritical fluid extraction device to selectively remove trace organic impurities.

[0371] Supercritical fluid extraction uses the special properties of supercritical fluid (usually supercritical CO2) to selectively dissolve and remove trace organic impurities.

[0372] 1. Extractor design: Use a multi-stage countercurrent extraction tower design to improve extraction efficiency. Use high-pressure resistant materials such as Inconel alloy to make the extractor. Optimize the internal structure, such as using structured fillers to increase the contact area.

[0373] 2. Supercritical fluid system: Design an efficient CO2 compression and refrigeration system to maintain the supercritical state. Use a precision temperature control system to keep CO2 in the optimal supercritical state (usually 31.1°C, above 73.8 bar).

[0374] 3. Co-solvent addition: Design a trace co-solvent (such as ethanol) addition system to improve the solubility of specific impurities.

[0375] Develop an intelligent ratio algorithm to dynamically adjust the co-solvent ratio according to the impurity type.

[0376] 4. Separation system: Design a multi-stage separator to gradually reduce the pressure to separate CO2 from impurities. Use a cyclone separator and an electrostatic demister to improve the gas-liquid separation efficiency.

[0377] 5. CO2 circulation system: Design a CO2 purification and recycling system to reduce CO2 consumption. Use a molecular sieve adsorption bed to remove moisture and other impurities in CO2.

[0378] 6. Online monitoring: Integrated gas chromatography-mass spectrometry (GC-MS) to monitor the impurity removal effect in real time. Use infrared spectrometer to monitor CO2 purity.

[0379] 7. Safety measures: Install high-pressure safety valves and bursting discs to prevent overpressure. Design CO2 leak detection and emergency discharge systems.

[0380] 8. Intelligent control: Develop an operation optimization algorithm based on model predictive control to achieve coordinated control of parameters such as pressure, temperature, flow rate, and co-solvent ratio.

[0381] S6.3: Use acoustic separation technology to perform fine purification using the acoustic properties of different gases.

[0382] Acoustic separation technology uses the differences in the behavior of different gases in the acoustic field to separate them, and is particularly suitable for processing difficult-to-separate gas mixtures.

[0383] 1. Acoustic wave generator: Design a high-power, frequency-adjustable piezoelectric transducer array. The frequency range is adjustable (usually 20kHz-1MHz) to adapt to different gas components.

[0384] 2. Sound field optimization: Design complex standing wave sound fields to maximize separation effects. Use numerical simulation to optimize the sound field distribution and create multiple separation nodes.

[0385] 3. Separation cavity design: Use multi-cavity series design to achieve multi-stage separation. Optimize cavity geometry, such as using special focusing structures.

[0386] 4. Temperature control: Design a precise temperature control system to maintain the optimal separation temperature. Use thermoelectric cooling technology to achieve fast and accurate temperature regulation.

[0387] 5. Pressure control: Use a high-precision pressure control system to maintain the optimal separation pressure. Develop a pressure fluctuation compensation algorithm to ensure the stability of the sound field.

[0388] 6. Gas flow control: Design a laminar flow intake system to minimize the impact of turbulence on separation. Use precision flow control valves to achieve precise flow regulation.

[0389] 7. Online monitoring: Integrate a high-precision mass spectrometer to monitor the concentration of each component in real time. Use an acoustic sensor array to monitor the distribution of the acoustic field.

[0390] 8. Energy optimization: Develop intelligent power control algorithms to ensure efficient energy utilization. Design acoustic energy recovery systems to reduce energy losses.

[0391] Model Predictive Path Integral (MPPI) technology for intelligent monitoring and maintenance systems:

[0392] 1. MPPI algorithm implementation: Develop a GPU-based parallel computing framework to improve the computational efficiency of the MPPI algorithm. Implement an adaptive sampling strategy to optimize control trajectory generation.

[0393] 2. System modeling: Establish a high-precision dynamic model of the acoustic wave separation process. Use deep learning techniques, such as recurrent neural networks (RNN) or long short-term memory networks (LSTM), to improve the model's prediction accuracy.

[0394] 3. Real-time optimization: Achieve millisecond-level control decision updates. Develop multi-objective optimization algorithms to balance purity, yield, and energy consumption.

[0395] 4. Uncertainty handling: Integrate Bayesian inference techniques to quantify and handle model and measurement uncertainties. Implement robust control strategies to cope with system parameter fluctuations.

[0396] 5. Adaptive learning: Develop online learning algorithms to continuously update and improve system models. Implement anomaly detection and automatic correction mechanisms.

[0397] 6. Human-computer interaction: Design an intuitive 3D visualization interface to display the control process and decision-making basis. Implement expert knowledge integration and allow manual intervention and adjustment.

[0398] 7. Predictive maintenance: Predict equipment performance degradation trends based on MPPI. Implement intelligent maintenance scheduling to minimize downtime.

[0399] 8. Multi-system collaboration: Extend the MPPI algorithm to achieve collaborative optimization of vortex tube, supercritical extraction and acoustic separation systems. Develop a global optimization strategy to ensure optimal overall system performance.

[0400] In addition, in another embodiment, different from the above method, the MPPI technology implementation may further include the following steps:

[0401] 1.1 Establish system dynamic model: construct mathematical models of vortex tube, supercritical fluid extraction and acoustic separation; identify key state variables and control inputs; realize online estimation and update of model parameters;

[0402] 1.2 Implementation of the path integral method: Design a path generation algorithm based on Monte Carlo sampling; implement the path cost evaluation function;

[0403] Develop efficient path optimization strategies;

[0404] 1.3 Predictive control framework design: define the regressive horizon optimization problem; implement real-time control law calculation; design feedback mechanism to improve robustness;

[0405] 2. Reach-Avoid task integration;

[0406] 2.1 Target state definition: clarify the target purity and efficiency indicators of deep purification; design a multi-objective optimization framework to balance purity and energy consumption; implement a dynamic target adjustment mechanism to adapt to changes in working conditions;

[0407] 2.2 Obstacle avoidance strategy design: Identify dangerous conditions during deep purification (such as excessive pressure and abnormal temperature);

[0408] Establish safety constraint mapping and define prohibited areas; implement real-time obstacle avoidance path planning;

[0409] 2.3 Task execution optimization: design task completion evaluation indicators; implement multi-stage task decomposition and optimization; develop task switching strategies to ensure smooth transition;

[0410] 3. Control Barrier Function (CBF) implementation;

[0411] 3.1 Safety set definition: Based on the system physical limitations and operating specifications, define the safe operation area; design state-dependent dynamic safety boundaries; implement a real-time update mechanism for the safety set;

[0412] 3.2 Barrier function construction: Design a barrier function that satisfies Lipschitz continuity; implement barrier function calculation based on system state; develop an adaptive barrier function adjustment strategy;

[0413] 3.3 Control law synthesis: Integrate CBF constraints into the MPPI framework; implement control input solution based on quadratic programming; design control smoothing mechanism to avoid operation jitter;

[0414] 4. Enhanced vortex tube separation;

[0415] 4.1 Intelligent temperature field control: Use MPPI to predict the evolution of temperature distribution; realize safe temperature control based on CBF; optimize the flow ratio of hot and cold ends to improve separation efficiency;

[0416] 4.2 Pressure fluctuation suppression: Design a pressure prediction model based on MPPI; implement active pressure fluctuation compensation; use CBF to ensure that the pressure is always within a safe range;

[0417] 4.3 Flow field optimization: predict gas flow trajectory based on MPPI; realize dynamic nozzle geometry adjustment; optimize vortex intensity and improve separation effect;

[0418] 5. Supercritical fluid extraction optimization;

[0419] 5.1 Phase equilibrium prediction and control: Use MPPI to predict the phase behavior of supercritical fluid-solute systems; achieve accurate pressure-temperature trajectory planning; design safe phase transition strategies based on CBF;

[0420] 5.2 Solubility optimization: Design a co-solvent addition strategy based on MPPI; achieve dynamic regulation of solubility; optimize extraction efficiency while ensuring operational safety;

[0421] 5.3 Enhanced fractional separation: Implement MPPI-based multi-stage separation process optimization; design intelligent inter-stage flow distribution strategy; use CBF to ensure the safety of operating parameters at each level;

[0422] 6. Improvement of acoustic wave separation technology;

[0423] 6.1 Accurate control of the sound field: Use MPPI to predict the distribution and evolution of the sound field; realize adaptive sound frequency and intensity adjustment; design a safe sound energy input strategy based on CBF;

[0424] 6.2 Standing wave optimization: Design a standing wave position prediction model based on MPPI; achieve precise gas flow and pressure control; optimize separation efficiency while avoiding equipment resonance risks;

[0425] 6.3 Multi-frequency collaborative separation: realize multi-frequency sound field collaborative control based on MPPI; design intelligent frequency switching strategy; use CBF to ensure that the sound field intensity is always within a safe range;

[0426] 7. System integration and collaborative control;

[0427] 7.1 Multi-unit collaborative optimization: Design a global optimization algorithm based on MPPI; realize the collaborative control of vortex tube, supercritical extraction and acoustic separation; develop a safe collaboration mechanism based on CBF;

[0428] 7.2 Adaptive control strategy: realize online model update and parameter identification; design adaptive MPPI algorithm to improve control accuracy;

[0429] Develop dynamic CBF adjustment mechanisms to adapt to system changes;

[0430] 7.3 Intelligent fault diagnosis and self-healing: Use MPPI to predict potential faults; design CBF-based fail-safe operation strategies;

[0431] Achieve automatic fault isolation and system reconstruction;

[0432] This enhanced deep purification technology solution significantly improves the accuracy, efficiency and safety of the deep purification process by integrating MPPI technology. The main improvements include:

[0433] 1. The model predictive path integral (MPPI) system is introduced to achieve precise control and optimization of complex nonlinear systems.

[0434] 2. By integrating the Reach-Avoid task, the system can effectively avoid dangerous conditions while pursuing high purity goals, improving operational safety.

[0435] 3. The implementation of the control barrier function (CBF) ensures that the system always operates within the safe operating range, greatly reducing the risk of accidents.

[0436] 4. During the vortex tube separation process, the separation efficiency is significantly improved through intelligent temperature field control, pressure fluctuation suppression and flow field optimization.

[0437] 5. Supercritical fluid extraction technology has been optimized, including accurate phase equilibrium prediction and control, solubility optimization, and enhanced fractionation separation.

[0438] 6. Acoustic wave separation technology achieves a more efficient and safer separation process through precise control of the sound field, standing wave optimization and multi-frequency collaborative separation.

[0439] 7. The integration and collaborative control of the entire system have been optimized, including multi-unit collaborative optimization, adaptive control strategy, and intelligent fault diagnosis and self-healing capabilities.

[0440] These improvements enable the deep purification process to more accurately control the operating status of each unit, significantly enhancing the safety and reliability of the system while improving purification efficiency. The application of the MPPI method enables the system to better cope with complex nonlinear dynamics, while the introduction of CBF provides a strong safety guarantee for the system. This advanced control strategy not only improves the final purity of helium, but also optimizes energy utilization, reduces operating costs, and makes the entire deep purification process more intelligent and efficient.

[0441] Through the combined use of these three advanced technologies and the strong support of the intelligent monitoring and maintenance system, the embodiment of the present invention can achieve ultra-high purity refining of helium (>99.9999%). This multi-technology integrated method can not only effectively remove various types of trace impurities, but also achieve efficient and stable operation through the intelligent control system, maximizing product quality and production efficiency.

[0442] S7: Use intelligent monitoring and maintenance systems for predictive maintenance;

[0443] The intelligent monitoring and maintenance system is the "brain" of the entire helium recovery and purification device. It is not only responsible for real-time monitoring and control of the entire process, but also can perform predictive maintenance, greatly improving the reliability and life of the system.

[0444] First, the embodiment of the present invention constructs a distributed monitoring network through an enhanced Push-DIging algorithm (push-distributed inexact gradient method and gradient tracking algorithm). The Push-DIging algorithm is a distributed optimization algorithm based on gradient advancement, which can efficiently propagate information and reach consensus in a distributed network. In this system, the embodiment of the present invention enhances the Push-DIging algorithm so that it can better adapt to complex industrial environments.

[0445] Specifically, the embodiment of the present invention first constructs a Push-DIging algorithm based on gradient advancement. This algorithm allows each node (such as a sensor, controller, etc.) to exchange information only with its neighboring nodes without requiring global information. This greatly reduces communication overhead and improves the scalability of the system.

[0446] Next, the embodiment of the present invention uses the Push-DIging algorithm to perform distributed average consistency estimation and construct an adaptive step size strategy. This means that the system can quickly reach a consensus among all nodes, such as the estimated value of a key parameter. The adaptive step size strategy can dynamically adjust the step size of the algorithm according to the current convergence speed to accelerator the convergence process.

[0447] On this basis, the embodiments of the present invention realize the distributed fusion of multi-source heterogeneous data. This function allows the system to integrate data from different types of sensors, such as temperature, pressure, flow, component concentration, etc., to form a comprehensive system status assessment.

[0448] To ensure data reliability, the system will check data consistency and correct abnormal data. This step can effectively filter out abnormal data caused by sensor failure or communication errors, and improve the stability of the entire system.

[0449] In addition to real-time monitoring, this system also uses the diffusion model to generate general time series energy data for health status assessment and fault prediction of different equipment. The diffusion model is a powerful generative model that can learn and simulate complex time series data.

[0450] First, the embodiment of the present invention constructs a diffusion model architecture suitable for energy data. This architecture takes into account the particularities of industrial equipment energy data, such as periodicity, long-term trends, emergencies, etc.

[0451] Then, the embodiment of the present invention trains and fine-tunes the diffusion model. The training data includes energy data during normal operation and energy data under various known fault conditions. Through careful training and fine-tuning, the model can accurately capture subtle changes in the operating status of the equipment.

[0452] After the training is completed, the embodiment of the present invention uses this model to generate time series energy data and perform quality assessment. The generated data includes not only normal operating status, but also various fault states. These generated data provide rich samples for subsequent health status assessment and fault prediction.

[0453] Finally, the embodiment of the present invention uses the generated time series energy data to simulate abnormal scenarios and inject faults. This step allows the embodiment of the present invention to test the performance of the system under various extreme conditions and optimize the fault detection and prediction algorithms.

[0454] In this way, the intelligent monitoring and maintenance system can realize intelligent monitoring of the entire process and perform accurate predictive maintenance. It can predict equipment failures, schedule maintenance in advance, avoid unexpected downtime, optimize maintenance plans, and reduce unnecessary overhauls, thereby greatly improving the reliability and economy of the system.

[0455] In summary, the helium recovery and purification method and device provided in this application realizes efficient, energy-saving and intelligent helium recovery and purification through intelligent pretreatment, multi-stage heat recovery, multi-stage catalytic reaction, advanced gas separation technology, distillation technology, deep purification technology and intelligent monitoring and predictive maintenance system. This method not only improves the recovery rate and purity of helium, but also significantly reduces energy consumption and operating costs, providing an effective technical solution to solve the problem of helium resource shortage.

[0456] S7.1: In the intelligent monitoring and maintenance system, a distributed monitoring network is constructed through an enhanced Push-DIging algorithm to achieve full-process intelligent monitoring and maintenance;

[0457] The Push-DIging (Push-Distributed Informaion Gradient) algorithm is an efficient distributed optimization algorithm suitable for large-scale network systems. In this system, the embodiment of the present invention enhances it to better adapt to the complex helium recovery and purification process.

[0458] 1. Network topology design: Adopt a hierarchical network structure, including sensor layer, edge computing layer and central processing layer. Use dynamic topology adjustment algorithm to adapt to the increase and decrease of devices and network changes.

[0459] 2. Data collection: Deploy a high-precision sensor network to cover key points in the entire process. Use Industrial Internet of Things (IIoT) technology to achieve real-time data collection.

[0460] 3. Edge computing: Deploy edge computing nodes in each main processing unit to achieve local data preprocessing and preliminary analysis, reducing the burden on the central processor.

[0461] 4. Enhanced Push-DIging algorithm implementation: Improved gradient estimation method to increase convergence speed and accuracy. Introduced adaptive step size strategy to optimize algorithm performance. Integrated asynchronous communication mechanism to improve network robustness.

[0462] 5. Distributed consensus: Implement a fast distributed average consensus algorithm. Develop a dynamic weight distribution mechanism to adjust influence based on node reliability.

[0463] 6. Multi-objective optimization: Integrate a multi-objective optimization framework to balance multiple objectives such as efficiency, quality and energy consumption. Use the Pareto optimization method to generate the optimal solution set.

[0464] S7.1.1: Construct a gradient-based Push-DIging algorithm;

[0465] The Push-DIging algorithm is an efficient distributed optimization method, and the embodiment of the present invention enhances its performance through gradient boosting.

[0466] 1. Algorithm framework design:

[0467] Define the network topology G=(V,E), where V is the node set and E is the edge set.

[0468] Each node i maintains local variables xi and gradient estimates gi.

[0469] 2. Improved gradient estimation:

[0470] The Nesterov accelerated gradient method is used to improve the convergence speed.

[0471] Introduce momentum term: mi(k+1) = βmi(k) + (1-β)gi(k);

[0472] Update rule: xi(k+1) = xi(k) α(k)(gi(k) + γmi(k));

[0473] Where α(k) is the step size and γ is the momentum coefficient.

[0474] 3. Communication mechanism optimization: Implement asynchronous communication protocol to reduce waiting time. Design adaptive communication frequency strategy to adjust communication frequency according to gradient changes.

[0475] 4. Robustness enhancement: Integrate variance reduction techniques, such as the control variable method. Implement gradient compression and quantization to reduce communication overhead.

[0476] 5. Convergence analysis: Derive the theoretical convergence rate to ensure the theoretical guarantee of the algorithm. Design adaptive termination conditions to avoid excessive iterations.

[0477] 6. Parallel computing optimization: Use GPU to accelerate gradient calculation. Implement pipeline parallel strategy, overlap calculation and communication.

[0478] S7.1.2: Use the Push-DIging algorithm to perform distributed average consistency estimation and construct an adaptive step size strategy;

[0479] In this step, the embodiment of the present invention uses an improved Push-DIging algorithm to implement distributed average consistency estimation, and introduces an adaptive step size strategy to optimize performance.

[0480] 1. Consistency estimation framework:

[0481] Define the global objective function: f(x) = (1 / n)Σfi(x), where fi(x) is the local function of node i.

[0482] The goal is to make the xi of all nodes converge to the global optimal solution x*.

[0483] 2. Advance process design:

[0484] Each node i executes at each iteration k:

[0485] xi(k+1) = Σj∈Ni wij(xj(k) α(k)gj(k));

[0486] Where Ni is the neighbor set of node i, and wij is the weight.

[0487] 3. Adaptive weight calculation:

[0488] Design a dynamic weight adjustment mechanism: wij(k) = exp(-||xi(k)-xj(k)||² / σ²);

[0489] σ is an adaptive parameter, which is adjusted dynamically according to the network status.

[0490] 4. Adaptive step size strategy:

[0491] Implement the Barzilai-Borwein (BB) step size method:

[0492] α(k) = ||x(k)-x(k-1)||² / ⟨x(k)-x(k-1), g(k)-g(k-1)>;

[0493] Combined with line search technology, the stability of step length is ensured.

[0494] 5. Convergence acceleration: Integrate Nesterov acceleration technology. Implement periodic restart strategy to avoid oscillation.

[0495] 6. Robustness enhancement: Design an abnormal node detection mechanism. Implement an adaptive fault-tolerant strategy and dynamically adjust the impact of abnormal nodes.

[0496] S7.1.3: Implement distributed fusion of multi-source heterogeneous data;

[0497] In a complex helium recovery and purification system, it is necessary to process heterogeneous data from different sources. The embodiment of the present invention designs an efficient distributed data fusion framework.

[0498] 1. Data preprocessing: Design an adaptive sampling strategy to dynamically adjust the sampling rate according to the importance of the data. Achieve multi-scale data alignment and process data with different temporal resolutions.

[0499] 2. Feature extraction and representation: Use autoencoders to extract features and capture the potential structure of the data. Implement multimodal feature fusion, such as early fusion, mid-term fusion, and late fusion.

[0500] 3. Distributed fusion algorithm: Design a distributed fusion model based on graph neural network (GNN). Implement the attention mechanism to automatically learn the importance of different data sources.

[0501] 4. Uncertainty modeling: Integrate the Bayesian inference framework to quantify the uncertainty of the fusion results. Implement multi-hypothesis tracking and handle the uncertainty of data association.

[0502] 5. Knowledge transfer: Design cross-domain adaptive learning algorithms to leverage knowledge from different data sources. Implement a meta-learning framework to quickly adapt to new data sources.

[0503] 6. Privacy protection: Implement a federated learning framework to protect sensitive data. Design a differential privacy mechanism to protect privacy during data sharing.

[0504] 7. Real-time processing: Develop a stream processing architecture, such as Apache Flink or Spark Streaming. Implement a sliding window algorithm to process continuously arriving data streams.

[0505] S7.1.4: Check data consistency and correct abnormal data;

[0506] Ensuring data consistency and accuracy is crucial for the reliable operation of the system. The embodiment of the present invention designs a comprehensive data consistency check and anomaly correction mechanism.

[0507] 1. Consistency check framework: Design a multi-level consistency check strategy: local check, neighborhood check, and global check. Implement a distributed consistency protocol based on blockchain to ensure that data cannot be tampered with.

[0508] 2. Anomaly detection algorithms: Develop deep learning-based anomaly detection models, such as variational autoencoders (VAE). Implement ensemble learning frameworks, such as isolation forests. Design time series anomaly detection algorithms to capture abnormal patterns in time series.

[0509] 3. Data correction strategy: Implement data correction based on physical models and use domain knowledge. Develop multiple interpolation algorithms to handle missing values. Design adaptive smoothing algorithms to remove noise.

[0510] 4. Causal Inference: Implement anomaly diagnosis based on causal graphs to identify the root causes of anomalies. Develop a counterfactual reasoning framework to evaluate the impact of data corrections.

[0511] 5. Adaptive threshold: Design a dynamic threshold adjustment algorithm to adapt to changes in system status. Implement multi-dimensional anomaly detection and consider the correlation between variables.

[0512] 6. Confidence Assessment: Develop a data fusion framework based on evidence theory to quantify the reliability of data. Implement a fuzzy logic system to handle uncertainty and ambiguity.

[0513] 7. Human-machine collaboration: Design an interactive exception verification interface, combine expert knowledge, implement an active learning framework, and optimize the efficiency of manual intervention.

[0514] 8. Real-time response: Develop incremental learning algorithms to update anomaly detection models in real time. Implement fast anomaly handling strategies to minimize the impact of abnormal data.

[0515] 9. Performance evaluation: Design comprehensive evaluation indicators, including accuracy, recall, F1 score, etc.

[0516] Implement an A / B testing framework to evaluate the effectiveness of different correction strategies.

[0517] Through these four detailed steps, the embodiment of the present invention builds a powerful and flexible distributed monitoring network. The Push-DIging algorithm based on gradient advancement provides an efficient optimization framework, and the adaptive step size strategy ensures the rapid convergence of the algorithm. The distributed fusion of multi-source heterogeneous data enables the system to fully understand and analyze the complex helium recovery and purification process. Finally, strict data consistency checks and anomaly correction mechanisms ensure the reliability of system operation and the accuracy of data.

[0518] This advanced monitoring network can not only monitor the operating status of the entire helium recovery and purification system in real time, but also quickly identify and respond to potential problems, thereby improving the overall efficiency and reliability of the system. It lays a solid foundation for realizing an intelligent and automated helium recovery and purification process.

[0519] S7.2: Use diffusion models to generate generic time series energy data for health status assessment and failure prediction of different equipment.

[0520] The diffusion model is a powerful generative model that can learn and generate complex time series data. In this system, the embodiment of the present invention uses the diffusion model to generate high-quality energy data for equipment health status assessment and fault prediction.

[0521] 1. Data preprocessing: Design multi-source heterogeneous data fusion algorithms. Implement automatic data cleaning and outlier detection. Develop time series feature extraction modules to capture key time series patterns.

[0522] 2. Diffusion model architecture: Deep neural network is used as the backbone network of the diffusion model. The attention mechanism is integrated to capture long-term dependencies. The conditional generation mechanism is implemented to generate specific data according to different devices and working conditions.

[0523] 3. Model training: Design a distributed training framework to support large-scale data sets. Implement a progressive training strategy to gradually increase model complexity. Use mixed precision training technology to improve training efficiency.

[0524] 4. Data generation: Develop MCMC-based sampling algorithms to generate high-quality energy data. Implement diversity control mechanisms to ensure the diversity and representativeness of generated data.

[0525] 5. Health status assessment: Design a multimodal health index calculation model. Develop anomaly detection algorithms based on generated data.

[0526] Realize equipment performance degradation trend analysis.

[0527] 6. Fault prediction: Build a predictive maintenance model based on generated data. Implement a multi-step prediction algorithm to provide long-term fault risk assessment. Develop a fault root cause analysis module to support precision maintenance.

[0528] 7. Uncertainty Quantification:

[0529] The Bayesian inference framework is integrated to quantify the uncertainty of the prediction. Confidence interval estimation is achieved to provide reliable decision support. Through these two steps, the embodiment of the present invention builds a powerful intelligent monitoring and maintenance system. The enhanced Push-DIging algorithm enables the embodiment of the present invention to efficiently process large-scale data in a distributed network and achieve real-time monitoring and optimization of the entire process. The high-quality energy data generated by the diffusion model provides a reliable basis for equipment health status assessment and fault prediction, making predictive maintenance a reality.

[0530] This system can not only monitor all aspects of the helium recovery and purification process in real time, but also predict potential equipment failures, optimize maintenance plans, maximize equipment utilization, and minimize downtime. By integrating advanced algorithms and technologies, the system provides intelligent and automated management and control for the entire helium recovery and purification process, greatly improving production efficiency and product quality.

[0531] S7.2.1: Construct diffusion model architecture suitable for energy data;

[0532] The diffusion model is a powerful generative model that is particularly suitable for processing complex time series data. The embodiment of the present invention designs a diffusion model architecture specifically for energy data.

[0533] 1. Model infrastructure:

[0534] U-Net is used as the backbone network to capture multi-scale features. A time embedding layer is introduced to encode the diffusion time step as the model input.

[0535] 2. Time series specialization: Integrate long short-term memory (LSTM) or gated recurrent unit (GRU) layers to capture long-term dependencies. Implement causal convolutional layers to maintain causality of time series.

[0536] 3. Attention mechanism: Introduce a self-attention layer to capture long-distance dependencies within a sequence. Design a temporal attention module to focus on key time points.

[0537] 4. Conditional generation: Design a conditional embedding layer that allows data to be generated based on specific conditions (e.g. device type, operating mode). Implement a feature control module that allows precise control over specific attributes of the generated data.

[0538] 5. Noise scheduling: Design an adaptive noise scheduling strategy to adjust the noise addition process according to the characteristics of the energy data. Implement variance preservation technology to ensure the stability of the generation process.

[0539] 6. Multi-resolution processing: Design a multi-resolution analysis module to process energy fluctuations at different time scales. Implement a wavelet transform layer to capture the time-frequency characteristics of energy data.

[0540] 7. Uncertainty modeling: Integrate the variational inference module to quantify the uncertainty of the generated results. Implement a probabilistic encoder to model the potential distribution of data.

[0541] 8. Computational efficiency optimization: Design lightweight network structures, such as MobileNet-style depthwise separable convolutions. Implement model quantization and pruning to reduce computational complexity.

[0542] S7.2.2: training and fine-tuning the diffusion model;

[0543] Training and fine-tuning are key steps to ensure model performance. The present invention will use advanced training strategies to optimize the diffusion model.

[0544] 1. Data preparation:

[0545] Implement data augmentation strategies such as time warping, amplitude scaling, etc. Design stratified sampling strategies to ensure balanced representation of different operating modes.

[0546] 2. Pre-training: Use large-scale unlabeled energy data for self-supervised pre-training. Implement pre-training tasks such as mask self-encoding and learn general feature representations.

[0547] 3. Training process: Use a progressive training strategy to gradually increase model complexity and data difficulty. Implement mixed precision training to balance computational efficiency and numerical stability.

[0548] 4. Loss function design: Use denoising score matching loss as the main training objective. Integrate auxiliary tasks such as reconstruction loss, contrastive learning loss, etc.

[0549] 5. Optimizer selection: Adopt Adam optimizer, and implement learning rate warm-up and cosine annealing strategy. Introduce gradient clipping and weight decay to prevent overfitting.

[0550] 6. Regularization technology: Implement Spectral Normalization to improve the stability of the model. Use Dropout and LayerNorm to enhance the generalization ability of the model.

[0551] 7. Fine-tuning strategies: Design transfer learning frameworks that allow for fast adaptation on specific device data. Implement few-shot learning techniques to handle data scarcity situations.

[0552] 8. Distributed training: Use a distributed training strategy that combines data parallelism and model parallelism. Implement gradient accumulation technology to handle large-scale training.

[0553] 9. Training monitoring: Design comprehensive evaluation indicators, including FID (Fréchet Inception Distance), IS (Inception Score), etc. Implement early stopping strategy to prevent overfitting.

[0554] 10. Model Ensemble: Train multiple model variants, such as different initializations, different architectures. Implement model ensemble techniques, such as Snapshot Ensemble.

[0555] S7.2.3: Generate time series energy data using the trained diffusion model and perform quality assessment:

[0556] Generating high-quality time series energy data is the key to model application. The embodiments of the present invention require the design of a rigorous generation process and evaluation method.

[0557] 1. Data generation process: Implement a reverse diffusion sampling algorithm, such as DDIM (Denoising Diffusion Implicit Models). Design a guided sampling strategy to control specific properties of the generated data.

[0558] 2. Diversity control: Implement temperature regulation techniques to balance the diversity and quality of generated data. Design clustering to guide generation and ensure coverage of different energy modes.

[0559] 3. Conditional generation: Implement category conditional generation for different device types or operating modes. Design attribute control modules to allow precise adjustment of specific characteristics of generated data.

[0560] 4. Statistical consistency assessment: Calculate the statistical moments (mean, variance, skewness, kurtosis) of the generated data and the real data. Implement KL divergence and Wasserstein distance calculations to assess distribution similarity.

[0561] 5. Time-frequency characteristics analysis: Perform power spectrum density analysis to compare the frequency characteristics of generated data and real data. Implement wavelet consistency evaluation and analyze multi-scale time-frequency characteristics.

[0562] 6. Physical constraint verification: Design a physical constraint checker based on domain knowledge. Implement energy conservation and thermodynamic law verification.

[0563] 7. Manual evaluation: Design an expert rating system to let domain experts evaluate the authenticity of the generated data. Implement an A / B testing framework to compare the discrimination between generated data and real data.

[0564] 8. Downstream task evaluation: Use the generated data to train the prediction model and evaluate its performance on real data. Implement anomaly detection tests to verify abnormal patterns in the generated data.

[0565] 9. Robustness testing: Conduct noise sensitivity analysis to evaluate the model’s robustness to input disturbances. Implement extreme scenario testing to verify the model’s performance under boundary conditions.

[0566] 10. Long-term consistency check: Design long sequence generation tasks to evaluate the consistency of the model over long time scales. Implement periodicity and trend analysis to verify the long-term patterns of generated data.

[0567] S7.2.4: Utilize the generated time series energy data to perform abnormal scenario simulation and fault injection;

[0568] The generated high-quality energy data can be used to simulate various abnormal scenarios and fault conditions to improve the robustness and reliability of the system.

[0569] 1. Construction of abnormal scenario library: Design a comprehensive abnormal type classification system, including equipment failure, system abnormality, external interference, etc.

[0570] Implement an abnormal scenario generator based on knowledge graph and integrate domain expert knowledge.

[0571] 2. Progressive fault injection: Design a progressive fault evolution model to simulate the development process of faults from minor to severe.

[0572] Implement multi-source fault superposition mechanism and simulate complex fault combinations.

[0573] 3. Conditional fault generation: Using the conditional diffusion model, data is generated based on specific fault types and severity.

[0574] Implement interpolation and extrapolation techniques to generate unseen failure modes.

[0575] 4. Rare event simulation: Design importance sampling strategies to increase the probability of generating rare faults.

[0576] Implement extreme value theory models to generate extreme anomalies.

[0577] 5. System response simulation: Integrate physical models to simulate the dynamic response of the system to faults.

[0578] Enable multi-scale simulation covering the effects of failures from component to system level.

[0579] 6. Fault propagation analysis: Design a fault tree analysis module to simulate the propagation path of faults in the system.

[0580] Implement a fault impact prediction model based on graph neural network.

[0581] Through these four detailed steps, the embodiment of the present invention builds a powerful diffusion model system for generating high-quality time series energy data and applying it to abnormal scenario simulation and fault injection. This system can not only generate synthetic data that is highly consistent with real data, but also simulate various complex fault scenarios. This provides a powerful tool for the optimization, fault diagnosis and predictive maintenance of the helium recovery and purification system, significantly improving the reliability and efficiency of the system.

[0582] By generating a large amount of simulation data and abnormal scenarios, the embodiments of the present invention can train a more robust monitoring and control system, identify potential problems in advance, and develop more effective response strategies. This data-driven approach, combined with domain expert knowledge, opens up new possibilities for the intelligent and automated operation of helium recovery and purification systems.

[0583] The present application also provides a helium recovery and purification device for implementing the above method, such as Figure 2 As shown, including:

[0584] The intelligent monitoring and maintenance system 10 includes a central processing unit (not shown in the figure), a sensor network (not shown in the figure), a touch screen display (not shown in the figure), and a communication unit (not shown in the figure);

[0585] The multi-stage heat recovery system 20 includes a multi-stage heat recovery device 21 and a thermoelectric conversion unit 22, wherein the multi-stage heat recovery device 21 includes a high-temperature heat exchanger 211, a medium-temperature heat exchanger 212 and a low-temperature heat exchanger 213;

[0586] A multi-stage catalytic reactor system 30, comprising a fixed bed reactor 31, a pulse reactor 32, a plasma-assisted catalytic unit 33 and a catalyst regeneration system 34;

[0587] A gas separation system 40, comprising a rotating bed adsorber 41, a multi-stage membrane separation unit 43 and an ionized gas separation device 42;

[0588] The distillation module 50 includes an internal heat-integrated distillation column 512, a distributed feed system 513, a centrifugal distillation unit 511, a condenser (not shown in the figure), and a reboiler (not shown in the figure);

[0589] A deep purification module 60, comprising a vortex tube separator 61, a supercritical fluid extraction device 63 and a sonic separation unit 62;

[0590] The storage tank 70 includes a coarse filtering device 71 , a pressurized temporary storage tank 72 and a pressurized helium collection tank 73 .

[0591] Among them, the intelligent control module 10, such as Figure 3 As shown, it includes:

[0592] Central processing unit: An industrial-grade computer is installed in a sealed control box on the top of the device.

[0593] Sensor network: temperature, pressure, flow, and gas composition sensors distributed in each module.

[0594] Touch screen display, mounted on the front of the unit, for easy operation.

[0595] Communication unit: includes wired and wireless communication interfaces, located on the side of the control box.

[0596] Structure: The control box is square, with dimensions of approximately 60cm×40cm×20cm, and is dustproof and waterproof.

[0597] Multi-stage heat recovery system 20, such as Figure 4 As shown:

[0598] High temperature heat exchanger (211): Located at the top of the system, usually used to handle the highest temperature heat exchange.

[0599] Medium temperature heat exchanger (212): located below the high temperature heat exchanger and used for heat exchange in a medium temperature range.

[0600] Low temperature heat exchanger (213): Located at the bottom of the system, used to handle heat exchange at lower temperatures.

[0601] Thermoelectric conversion unit (22): Located on the right side of the entire system, it is used to convert thermal energy into electrical energy.

[0602] Energy conversion medium (23): It is a fluid or other medium used to transfer energy between the various parts of the system. As can be seen from the figure, it connects the various heat exchangers and thermoelectric conversion devices through pipes.

[0603] Among them, three heat exchangers (high, medium and low temperature) are arranged vertically and connected to each other through pipes. The energy conversion medium connects all heat exchangers and thermoelectric converters through the pipeline system to form a closed-loop system. The thermoelectric converter is connected to the heat exchanger system through pipes to convert the recovered heat energy into electrical energy. This design allows the system to recover heat energy at different temperature levels in stages, improving the overall energy recovery efficiency. Heat energy can be transferred between various temperature levels, and ultimately the energy is effectively utilized through the thermoelectric converter.

[0604] Multi-stage catalytic reactor system 30, such as Figure 5 As shown:

[0605] Fixed bed reactor (31):

[0606] Located in the center upper part of the system is a cylindrical container with a feed port on the top. This reactor is used for helium catalytic reactions, where the catalyst is fixed inside the reactor.

[0607] Pulse Reactor (32):

[0608] Located on the left side of the system, it is a vertical multi-layer structure. This reactor is used for rapid, periodic reactions or treatments.

[0609] Plasma assisted catalyst (33):

[0610] Located in the lower middle of the system is a horizontal cylindrical device with a porous structure inside. This device uses plasma technology to enhance the efficiency of catalytic reactions.

[0611] Catalyst regeneration system (34):

[0612] Located on the right side of the system is a vertical cylindrical vessel. This system is used to treat and regenerate the spent catalyst to extend its service life.

[0613] The fixed bed reactor is connected to the pulse reactor and the plasma-assisted catalyst through pipelines to form a reaction network. The plasma-assisted catalyst is connected to the fixed bed reactor and the catalyst regeneration system through pipelines to process the products from the fixed bed or provide support for catalyst regeneration. The catalyst regeneration system is connected to other components through pipelines to circulate the catalyst in the entire system.

[0614] This design allows the system to perform multi-stage catalytic reactions under different reaction conditions while providing the ability to regenerate the catalyst. By combining different types of reactors and auxiliary equipment, the system can achieve complex chemical processes and improve reaction efficiency and catalyst utilization.

[0615] Gas separation system 40, such as Figure 6 As shown:

[0616] Rotating bed adsorber 41: disc type, 0.8 m in diameter, 0.2 m in thickness, located at the top.

[0617] Multi-stage membrane separation unit 43: hollow fiber membranes, arranged in multiple cylindrical bundles, located in the middle.

[0618] Temporary storage tank 42,

[0619] Ionized gas separation device 44: parallel electrode plate structure, located at the bottom.

[0620] Structure: The overall shape is a rectangular parallelepiped, with dimensions of approximately 1.5m×1m×2m, and each unit is connected in parallel through pipes and valves.

[0621] Distillation module 50, such as Figure 7 As shown:

[0622] Internal heat integration distillation tower 512: a cylinder with a diameter of 0.5m and a height of 2m, located in the center.

[0623] Distributed feeding system 513: multiple feeding ports are evenly distributed along the tower body.

[0624] Centrifugal distillation unit 511: located at the top of the tower, a disc-shaped structure with a diameter of 0.3m.

[0625] Condenser and reboiler: integrated inside the tower.

[0626] Structure: The overall shape is cylindrical, with a diameter of about 0.8m and a height of 2.5m, and all components are vertically integrated.

[0627] Deep purification module 60, such as Figure 8 As shown:

[0628] The first temporary storage tank 601 is located on the left side of the system and is a smaller cylindrical container used to temporarily store the substances to be processed or processed.

[0629] Vortex tube separator 61: Located in the center of the system, it is a large cylindrical container. This device uses the vortex principle to separate substances of different components.

[0630] Supercritical Fluid Extraction Device 63: Located on the right side of the system, this is a high pressure vessel. This device uses the properties of supercritical fluid (usually carbon dioxide) to perform efficient extraction.

[0631] Acoustic separation unit 62: Located on the far right side of the system, it is a smaller rectangular device that uses sound waves to separate substances.

[0632] The second temporary storage tank 602 is located at the bottom of the entire system and serves as a base for supporting and connecting various devices.

[0633] All devices are installed on a common base to form an integrated system.

[0634] The temporary storage tank is connected to the vortex tube separation technology device through a pipeline for conveying the material to be processed. The vortex tube separation technology device is connected to the supercritical fluid extraction device through a pipeline, and the two devices are processed in sequence.

[0635] There is also a pipeline connecting the supercritical fluid extraction device and the sonic separation unit.

[0636] This design integrates multiple advanced separation and purification technologies into one system, which can achieve deep purification of complex mixtures. By combining technologies such as eddy current separation, supercritical fluid extraction and acoustic separation, the module is able to process various types of substances and improve separation efficiency and purity.

[0637] Connection between modules: Each module is connected through standardized flanges and quick-connect interfaces for easy assembly and disassembly. Flexible pipes are used to connect modules to reduce vibration transmission. Waterproof plug-in connectors are used for electrical connections to ensure safety and reliability. Industrial-grade Ethernet is used for data transmission to achieve real-time communication between modules.

[0638] Overall structure: All modules are installed on a common base, for example, the base size is 3m×2m. The base is equipped with adjustable feet to adapt to different ground conditions. The overall height is about 3m, and the modular stacking design is adopted, and the modules can be increased or decreased according to the needs. The exterior adopts a lightweight alloy shell with good heat insulation and protection performance. The key components adopt explosion-proof design to meet the needs of various industrial environments.

[0639] This modular, miniaturized, multifunctional purification device is designed with full consideration of flexibility and adaptability. Each module can be combined or used separately according to the specific industrial scenario requirements, while the standardized interface design ensures the convenience of rapid deployment and maintenance. The overall structure is compact and efficient, suitable for use in various industrial environments.

[0640] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A helium recovery and purification method, characterized in that: The method comprises: Intelligent pretreatment of helium-containing industrial waste gas; Recovering waste heat from the helium-containing industrial waste gas using a multi-stage heat recovery system; Using a multi-stage catalytic reactor system to remove impurities from the helium-containing industrial waste gas; Gas separation technology is used for helium enrichment and preliminary purification; The helium is distilled and purified by a distillation module; Use purification technology to achieve fine purification of helium; Use smart monitoring and maintenance systems for predictive maintenance; Among them, the intelligent pretreatment of helium-containing industrial waste gas includes: Removing micron-sized particles from the helium-containing industrial waste gas through a filter; Use multi-stage chemical adsorption units to remove corrosive impurities, including acid gases, alkaline gases, and organic pollutants; Based on a stochastic predictive control (SPC) system, dynamically adjusting the regeneration period of the multi-stage chemical adsorption device; Wherein, based on the stochastic predictive control SPC system, the regeneration period of the multi-stage chemical adsorption device is dynamically adjusted, including: Establish an SPC framework, including: defining system state variables, wherein the system state variables include temperature, pressure and flow; Determining control input parameters, wherein the input parameters include valve opening and compressor power; A system dynamics model with unbounded process noise is established, where the stochastic differential equation (SDE) is used to describe the system dynamics: dX = f(X,U,t)dt+G(X,U,t)dW, where f(X,U,t) is the deterministic dynamic function, G(X,U,t) is the noise intensity matrix, W is the Wiener process, representing unbounded process noise, X is the state vector, U is the control input vector, and t is the switching time; The conceptual shrinking horizon and rolling horizon formulas are established, where the shrinking horizon is: At time t, the time domain of the optimization problem is [t,T], where T is a fixed termination time; Objective function: J = E[∫[t,T]L(X(τ),U(τ))dτ+Φ(X(T))]; where L is the stage cost function and Φ is the terminal cost function; The rolling horizon is: at each time t, solve the optimization problem of the finite horizon [t, t+H], where H is the prediction horizon; only execute the first element of the optimal control sequence, and then roll the horizon forward; Collecting various parameters of the helium-containing industrial waste gas; Build and train deep neural networks; Inputting various parameters of the helium-containing industrial waste gas into the deep neural network, so that the deep neural network is combined with the SPC framework to realize the state prediction of the helium-containing industrial waste gas; Based on the state prediction result, dynamically adjust the amount and replacement frequency of the adsorbent in the multi-stage chemical adsorption device; Among them, the use of intelligent monitoring and maintenance systems for predictive maintenance includes: In the intelligent monitoring and maintenance system, a distributed monitoring network is constructed through an enhanced Push-DIging algorithm to achieve full-process intelligent monitoring and maintenance; Generate universal time series energy data using diffusion model for health status assessment and fault prediction of different equipment; Among them, the enhanced Push-DIging algorithm is used to build a distributed monitoring network to achieve full-process intelligent monitoring and maintenance, including: Construct a Push-DIging algorithm based on gradient advancement; The Push-DIging algorithm is used to perform distributed average consistency estimation and construct an adaptive step size strategy, including: Define a global objective function: f(x) = (1 / n)Σfi(x), where fi(x) is a local function of node i, and the goal of the global objective function is to make the xi of all nodes converge to the global optimal solution x*; Each node i executes at each iteration k: xi(k+1)=Σj∈Ni wij(xj(k)α(k)gj(k)); where Ni is the neighbor set of node i and wij is the weight; Design a dynamic weight adjustment mechanism: wij(k)=exp(-||xi(k)-xj(k)|| 2 / σ 2 ), where σ is an adaptive parameter that is dynamically adjusted according to the network status; Implement the Barzilai-Borwein step method: α(k) = ||x(k)-x(k-1)|| 2 / <x(k)-x(k-1),g(k)-g(k-1)> ; Combined with line search technology to ensure the stability of step length; Integrate Nesterov acceleration technology to implement periodic restart strategy; Realize distributed fusion of multi-source heterogeneous data; Check data consistency and correct abnormal data; Generate generic time series energy data using diffusion models for health status assessment and failure prediction of different equipment, including: Build a diffusion model architecture suitable for energy data; Performing model training and fine-tuning on the diffusion model; Generate time series energy data using the trained diffusion model and perform quality assessment; The generated time series energy data is used to perform abnormal scenario simulation and fault injection.

2. The method according to claim 1, characterized in that The waste heat of the helium-containing industrial waste gas is recovered by a multi-stage heat recovery system, comprising: The high-temperature exhaust gas is heat exchanged with the inlet cold gas to obtain the medium-temperature exhaust gas; Exchanging heat between the medium-temperature exhaust gas and the intermediate process gas to obtain low-temperature exhaust gas; using the low temperature exhaust gas to preheat supplementary gas; During the heating and cooling process, a cogeneration system is used to convert thermal energy into electrical energy for use in control systems and auxiliary equipment.

3. The method according to claim 1, characterized in that The impurities in the helium-containing industrial waste gas are removed using a multi-stage catalytic reactor system, comprising: The first stage uses a fixed bed reactor to remove the first impurity; The second stage uses a pulse reactor to remove the second impurity by periodically changing the reaction conditions; The third stage uses plasma-assisted catalysis to remove the third impurity at low temperature; The gas separation technology is used to enrich and initially purify helium, including: A rotating bed adsorption separation device is used to enrich helium by utilizing centrifugal force to enhance the mass transfer process; The enriched helium enters a multi-stage membrane separation system, using membrane materials with different pore sizes and selectivities to gradually increase the purity of the helium. The ionized gas separation technology is introduced to separate gases by utilizing the difference in ionization energy of different gases; wherein the gas flow and pressure between the separation units are coordinated and controlled by the intelligent monitoring and maintenance system.

4. The method according to claim 3, characterized in that Before the helium is enriched and initially purified by using gas separation technology, the method further comprises: Establish an adaptive scalar field mapping system, including: Establish a scalar field of gas component concentration; Detect and identify high-intensity regions using a high-intensity region recognition algorithm; Building adaptive avoidance strategies; Based on the adaptive scalar field mapping system, the rotating bed adsorption separation device is structurally optimized and the centrifugal force field is regulated; Based on the adaptive scalar field mapping system, the spatial arrangement and cascade structure of the membrane components are dynamically adjusted, and the transmembrane pressure difference is adaptively adjusted; Based on the adaptive scalar field mapping system, plasma gas parameters are dynamically adjusted to avoid high-intensity areas during the movement of ion flow.

5. The method according to claim 1, characterized in that Fine purification of helium is achieved using purification technologies, including: Introducing vortex tube separation technology to further purify helium using the gas temperature separation effect; Use supercritical fluid extraction equipment to selectively remove trace organic impurities; The acoustic wave separation technology is adopted to utilize the acoustic characteristics of different gases for fine purification, wherein the fine purification process is monitored by the intelligent monitoring and maintenance system through the model prediction path integral MPPI technology to ensure that the purity of helium is greater than 99.9999%.

Citation Information

Patent Citations

  • Method and device for recycling and purifying helium gas

    CN104176717A