Disilane mixed gas dissolving and mixing system and control method

Abstract

The invention discloses a disilane mixed gas dissolving and mixing system and a control method, and relates to the technical field of gas mixing preparation, the disilane mixed gas dissolving and mixing system comprises seven modules, a storage module realizes accurate pressure control through a honeycomb gas storage tank matrix, and the tank wall adopts a nano composite structure for corrosion prevention; the premixing module forms a chaotic flow field by using a double-helix micro-channel and pulse injection; the reaction module is combined with supercritical and electromagnetic resonance to accelerate the reaction; the monitoring module fuses multispectral detection; the control system is based on deep reinforcement learning, in addition, a gradient condensation-membrane distillation separation module, a bionic safety interlocking module and other modules are further arranged, and all the modules cooperate to achieve efficient mixing and production. According to the device and the method, high-precision preparation of disilane mixed gas, precise pressure-controlled material storage, molecular-level efficient mixed reaction, multispectral precise monitoring, intelligent regulation and control, stability guarantee, high safety interlocking and intelligent maintenance are realized, faults can be reduced, the product quality and the production efficiency are improved, and the cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of gas mixing preparation, and particularly to a dissolution and mixing system and a control method for disilane mixed gas. Background Art

[0002] Disilane mixed gas is widely used in fields such as semiconductor manufacturing, solar cell production, and special material coating. Its preparation quality directly affects the performance of downstream products. At present, the dissolution and mixing technology of disilane mixed gas faces many challenges. The traditional storage method of raw material gas uses a single gas storage tank or a simple gas storage tank group, which is difficult to accurately control the storage pressure and temperature of different gases. During storage, the gas is prone to component ratio deviation due to pressure fluctuations and temperature changes, affecting the subsequent mixing accuracy. At the same time, after the gas storage tank material is in long-term contact with corrosive gases such as disilane, problems such as corrosion and hydrogen embrittlement are likely to occur, posing safety hazards, and the frequent replacement of gas storage tanks increases production costs.

[0003] In the mixing process, existing equipment mostly uses static mixers or simple stirring devices, which cannot achieve sufficient mixing of disilane and other gases at the molecular level. Due to the active chemical properties of disilane, conventional mixing methods are prone to uneven mixing. Excessive local concentration may cause side reactions, reducing the quality of the mixed gas; while the low mixing efficiency prolongs the production cycle and is difficult to meet the requirements of large-scale industrial production. In addition, traditional reaction kettles lack effective reaction intensification means, with slow reaction rates and high energy consumption, and cannot fully exert the reaction activity of disilane.

[0004] In terms of monitoring and control, traditional systems rely on a single type of sensor for data acquisition, with incomplete information acquisition, poor data accuracy and real-time performance, and it is difficult to precisely control the complex dissolution and mixing process. The control strategy mostly uses PID control with fixed parameters, with a lag in adjustment when facing working condition changes, and cannot respond in a timely manner to fluctuations in parameters such as temperature, pressure, and concentration, resulting in unstable product quality. Moreover, the safety protection and maintenance functions of existing equipment are weak, unable to effectively prevent and handle sudden failures, with long equipment failure shutdown times, affecting production continuity and the economic benefits of enterprises. Therefore, there is an urgent need to develop a new type of dissolution and mixing system and control method for disilane mixed gas to solve the above technical problems. Summary of the Invention

[0005] The dissolution and mixing system and control method for disilane mixed gas proposed by the present invention are used to solve the problems mentioned in the above prior art.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A dissolution and mixing system for disilane mixed gas, comprising the following modules: Dynamic gradient partial pressure raw material storage module: It is composed of a gas storage tank matrix. The gas storage tanks are connected by intelligent variable-diameter pressure balance pipes. According to the formula Adjust the pipe diameter, D, is the real-time and initial pipe diameter, is the pipe diameter adjustment coefficient, is the pressure difference. The tank wall adopts a nano-layered composite structure. Through the formula Ultra-low heat conduction, is the composite thermal conductivity, , are the thermal conductivities of each layer, , are the volume ratios of each layer; Chaotic flow field molecular premixing module: Adopts a double helix nested microchannel structure. Through the formula Regulate the fluid state, is the critical Reynolds number, is the gas density, v is the flow velocity, D is the channel diameter, is the dynamic viscosity, is the pulse influence coefficient, f is the pulse frequency. A laser-induced fluorescence concentration monitor is set at the module inlet to feedback the mixing uniformity; Supercritical electromagnetic resonance reaction module: The horizontal reaction kettle adopts a double-layer jacket structure. The stirring system adopts a planetary stirring paddle. The stirring speed is adjusted through n is the real-time speed, is the initial speed, P is the reaction pressure, is the standard pressure; Multi-spectral fusion real-time monitoring module: It is composed of a distributed optical fiber spectral sensor network, a terahertz wave detector and a mass spectrometer. The terahertz wave detector analyzes the rotational-vibrational energy levels of gas molecules through the formula is the terahertz wave transmission time, l is the detection path length, v g is the group velocity; Deep reinforcement learning adaptive control system: Build a model with an edge computing unit as the core. Update the strategy through the formula Q is the action value function, r is the immediate reward, is the discount factor, s, a are the current state and action, s', a' are the next state and action; Gradient condensation-membrane distillation coupling separation module: Adopts a three-stage gradient condensation structure. After the third-stage condensation, a membrane distillation unit is connected. Separate through the formula J is the membrane flux, K is the membrane mass transfer coefficient, is the vapor pressure difference across the membrane; Bionic intelligent safety interlock module: A risk assessment model is constructed based on a biological neural network, with hierarchical responses in case of anomalies, and a cut-off valve driven by bionic muscles is equipped.

[0007] Furthermore, it also includes a nano-bubble enhanced mass transfer module: A nano-bubble generator is set in front of the inlet of the supercritical electromagnetic resonance reaction module. Through shearing and in cooperation with a Venturi tube, a pressure drop of 0.5 - 2 MPa is achieved to generate nano-bubbles with an average particle size of 50 - 200 nm. The bubble generation rate is regulated according to the formula , where N is the bubble generation rate, Q is the gas flow rate, is the pressure drop, and K v is the volumetric mass transfer coefficient.

[0008] Furthermore, it also includes a self-repairing intelligent maintenance module: The key components of the equipment are coated with a self-repairing nano-coating. The coating consists of micro-capsules containing a repair agent and a polymer matrix. When the coating is damaged, the micro-capsules rupture to release the repair agent, and the repair is carried out through the formula , where t is the repair time, l is the diffusion distance, D is the diffusion coefficient of the repair agent, C0 and C are the initial concentration and the current concentration of the repair agent respectively. The module is also equipped with a micro 3D printing device for in-situ repair of worn mechanical components.

[0009] Furthermore, a magneto-rheological fluid damper is set in the pressure balance pipeline of the dynamic gradient partial pressure raw material storage module, and the gas flow rate in the pipeline is adjusted through the formula , where is the shear stress, K is the magneto-rheological fluid constant, and H is the magnetic field strength.

[0010] Furthermore, an acoustic standing wave field is set at the micro-channel outlet of the chaotic flow field molecular premixing module, and the standing wave is generated through the formula , where f is the acoustic wave frequency, n is the harmonic number, c is the speed of sound, and L is the length of the standing wave cavity.

[0011] Furthermore, a microwave resonance cavity is set in the reaction kettle of the supercritical electromagnetic resonance reaction module, and the vibration frequency of disilane molecules is matched through the formula , where c is the speed of light, L is the length of the cavity, is the relative dielectric constant; the stirring paddle of this module adopts a topology optimization design, and through finite element analysis combined with a genetic algorithm, the mass of the paddle blade is reduced while meeting the strength requirements.

[0012] Furthermore, the deep reinforcement learning adaptive control system introduces a transfer learning mechanism to transfer the optimization strategy under similar working conditions to the new working conditions, so as to improve the learning efficiency of the system in the new scenario.

[0013] Furthermore, a method for the dissolution and mixing system of disilane mixed gas includes the following steps: Multi-dimensional parameter collaborative planning steps: Input the target mixed gas composition and production parameters. The system calculates the optimal combination of 18 process parameters through a multi-objective optimization algorithm combined with the process database, and uses the formula to balance the product target and generate the initial control parameter set F as the comprehensive objective function value, where w i is the weight coefficient, and f i is each sub-objective function; Chaotic premixing dynamic regulation steps: After the raw material gas enters the premixing module, the laser-induced fluorescence monitor real-time feedbacks the mixing uniformity data. The control system dynamically adjusts the pulsed gas injection parameters according to the formula where is the pulse frequency adjustment amount, k is the adjustment coefficient, is the real-time concentration mean value, C set is the set concentration; Supercritical resonance reaction intelligent regulation steps: During the reaction process, the multi-spectral fusion monitoring module real-time collects temperature, pressure, and composition data. The deep reinforcement learning control system selects the regulation strategy according to the formula where a is the optimal action. By adjusting the electromagnetic coil frequency, microwave power, and stirring speed parameters, the reaction temperature fluctuation is controlled within ±1°C, the pressure fluctuation is controlled within ±0.05 MPa, and the concentration error of the key components is less than 0.5%; Gradient condensation-membrane distillation combined separation steps: After the mixed gas enters the separation module, according to the condensation temperature of each stage and the membrane distillation parameters, through the formula and cooperatively control the condensation and membrane distillation processes to improve the recovery rate of the target product and reduce the impurity content to less than 1 ppm. Among them, is the condensation heat transfer amount, U is the total heat transfer coefficient, A is the heat transfer area, is the temperature difference; Full life cycle intelligent maintenance steps: The self-repair intelligent maintenance module continuously monitors the status of the key components of the equipment. When coating damage or component wear is detected, the self-repair or 3D printing repair program is started. At the same time, according to the equipment operation duration and working condition data, the life prediction model is used to plan the maintenance plan in advance; Emergency response hierarchical processing steps: The bionic intelligent safety interlock module real-time evaluates the system risk. When the risk level reaches level one, it issues an audible and visual alarm and pushes a warning message to the operator's terminal; when it is a level two risk, it automatically adjusts the valve and equipment parameters to make the system tend to a safe state; when it is a level three risk, it immediately triggers the emergency cut-off and safety protection devices, and at the same time starts the accident simulation analysis program to provide the operator with the optimal disposal plan.

[0014] Furthermore, it also includes the step of autonomous evolution of process knowledge: during the operation of the system, the optimized process parameters, control strategies and fault handling cases are stored in the knowledge graph database, and the knowledge reasoning algorithm is used to mine the potential relationship between the data to automatically generate new process optimization solutions and control strategies.

[0015] Furthermore, it also includes the step of multi-system collaborative optimization: when the system is running in conjunction with upstream and downstream production equipment, relevant equipment data is obtained through the industrial Internet platform, and the collaborative optimization algorithm is used to jointly optimize the operating parameters of multiple systems. Achieve full-process production efficiency improvement, n is the time step, is the weight coefficient, is the sub-objective function of each system.

[0016] Compared with the prior art, the present invention has the following beneficial effects: In the raw material storage link, the honeycomb gas tank matrix and intelligent pressure balance system of the dynamic gradient pressure raw material storage module realize precise control of gas storage pressure and temperature, effectively avoiding the deviation of component ratio. At the same time, the nano-layered composite structure of the tank wall greatly improves the corrosion resistance and hydrogen embrittlement resistance of the gas tank, prolongs the service life of the equipment, and reduces safety risks and maintenance costs.

[0017] During the mixing and reaction process, the innovative design of the chaotic flow field molecular premixing module and the supercritical electromagnetic resonance reaction module allows disilane to be fully mixed with other gases at the molecular level, and significantly increases the reaction rate with the help of electromagnetic resonance and other effects, shortens the production cycle, and improves production efficiency. At the same time, the reaction uniformity and product quality are greatly improved, the occurrence of side reactions is reduced, and the output mixed gas is of higher quality, which can better meet the stringent requirements of high-end industries such as semiconductors and solar energy.

[0018] In terms of monitoring and control, the multi-spectral fusion real-time monitoring module and deep reinforcement learning adaptive control system realize comprehensive and accurate monitoring and intelligent control of the production process. The system can quickly respond to changes in working conditions and adjust parameters in time to ensure a stable production process with minimal fluctuations in product quality. The intelligent safety interlock and maintenance module provides dual protection for system operation, which can not only effectively prevent and handle safety accidents, but also reduce equipment downtime through self-repair and intelligent maintenance functions, improve production continuity, reduce enterprise operating costs, and enhance enterprise market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic block diagram of a dissolving and mixing system for disilane mixed gas proposed by the present invention; Figure 2 A schematic block diagram of the dissolution and mixing control method of the disilane mixed gas proposed by the present invention; Figure 3It is a broken line graph for comparing the mixing uniformity of different technologies; Figure 4 It is a bar graph for comparing the product recovery rate and equipment failure rate of traditional technology and this technology. Specific implementation mode

[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0021] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present invention.

[0022] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more unless otherwise specifically defined. In addition, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations. The present invention will be further described in detail below with reference to the accompanying drawings.

[0023] Refer to Figures 1 to 4 : A dissolution and mixing system for disilane mixed gas, including the following modules: Dynamic Gradient Pressure Dividing Raw Material Storage Module: The honeycomb gas storage tank matrix of the raw material storage module consists of 9 gas storage tanks, which are distributed in a 3×3 array. Each group of gas storage tanks contains 3 tanks with different volumes, namely 200L, 500L, and 1000L, for storing different raw material gases such as disilane, hydrogen, and nitrogen. The gas storage tanks are made of high-strength Inconel625 alloy, which has excellent corrosion resistance and hydrogen embrittlement resistance, and can effectively cope with the erosion of gases such as disilane.

[0024] At the top of each gas storage tank, a Honeywell HMC1001 type MEMS pressure sensor (accuracy ±0.01MPa) and a Sensirion SHT45 type temperature and humidity sensor array are installed. The sensors transmit the collected data to the central control system at a frequency of 2 seconds / time through the RS-485 communication protocol. The intelligent variable-diameter pressure balance pipeline is made of polyvinylidene fluoride (PVDF), and the initial pipe diameter is set to 50mm, and the pipe diameter adjustment coefficient takes a value of 0.1. An electric control valve is installed on the pipeline, which is driven by a servo motor. When the pressure difference between adjacent gas storage tanks reaches 0.1MPa, the central control system calculates the required pipe diameter according to the formula and controls the valve to complete the pipe diameter adjustment within 1 - 2 seconds to achieve pressure balance.

[0025] The nano-layered composite structure of the tank wall is prepared by magnetron sputtering technology. First, a 20nm thick nickel-chromium alloy nanocrystalline layer is deposited on the surface of the tank body, and then a 50nm thick graphene aerogel layer is deposited, and this is alternately deposited 10 layers. After testing, the thermal conductivity of this composite structure is calculated according to the formula (where is the thermal conductivity of the nickel-chromium alloy, is its volume fraction; is the thermal conductivity of the graphene aerogel, is its volume fraction), which is 85% lower than that of single stainless steel, and the temperature fluctuation of gas storage is stably controlled within ±0.3℃, effectively ensuring the storage stability of raw material gases.

[0026] Chaotic Flow Field Molecular Premixing Module: It adopts a double-helix nested microchannel structure. The inner wall of the main channel is etched with turbulence grooves with fractal geometry patterns (fractal dimension 1.6 - 1.8), and a pulsed gas injection device is introduced into the sub-channel. Through the formula ( is the critical Reynolds number, is the gas density, v is the flow velocity, D is the channel diameter, is the dynamic viscosity, The pulse impact coefficient is [pulse impact coefficient], and f is the pulse frequency), which is used to regulate the fluid state and form a chaotic flow field in the channel. A laser-induced fluorescence (LIF) concentration monitor is set at the module inlet to provide real-time feedback on the mixing uniformity. Pulse parameters are adjusted through closed-loop control to improve the premixing efficiency.

[0027] Supercritical electromagnetic resonance reaction module: The horizontal reactor of the supercritical electromagnetic resonance reaction module has an inner diameter of 1 m and a length of 2 m, and adopts a double-layer jacket structure. The inner reaction chamber is made of Hastelloy C-276, which has good high-temperature resistance and corrosion resistance; the outer jacket is made of carbon steel and is used to install high-frequency electromagnetic coils. The power of the high-frequency electromagnetic coil is 50 kW, and the frequency can be adjusted in the range of 50 - 200 kHz. The output is controlled by an IGBT power amplifier, and an adjustable alternating magnetic field can be generated in the reaction chamber. The stirring system uses a planetary stirrer paddle driven by magnetic coupling. The paddle blade has a diameter of 0.3 m and is powered by neodymium iron boron permanent magnets. The surface of the paddle blade is coated with a superhydrophobic nano-coating by chemical vapor deposition. After testing, its contact angle reaches 165°, effectively preventing the adhesion of reactants on the surface of the paddle blade. The stirring speed is adjusted adaptively according to the formula (n is the real-time rotation speed, is the initial rotation speed set to 100 rpm, P is the real-time reaction pressure, is the standard pressure of 1 MPa), which is calculated and controlled in real time by the central control system according to the data fed back by the pressure sensor. The microwave resonance cavity set in the reactor adopts a rectangular cavity structure with dimensions of 0.5 m × 0.3 m × 0.2 m. By adjusting the cavity length and according to the formula (where f is the microwave resonance frequency, c is the speed of light, L is the cavity length, is the relative dielectric constant), the vibration frequency of disilane molecules is precisely matched. In practical applications, when the reaction is under supercritical conditions (pressure 10 MPa, temperature 300 °C), combined with the electromagnetic resonance effect, the reaction rate is increased by 4 times compared with the traditional method, and the standard deviation of reaction uniformity is reduced to 0.03, greatly improving the reaction efficiency and product quality.

[0028] Multi-spectral fusion real-time monitoring module: The multi-spectral fusion real-time monitoring module consists of a distributed optical fiber spectral sensor network, a terahertz wave detector and a mass spectrometer. The distributed optical fiber spectral sensor network arranges 10 monitoring nodes at key positions such as the reactor and pipelines. It uses OceanInsightNIRQuest512 type sensors based on surface plasmon resonance (SPR) technology, and its detection accuracy can reach 0.001 ppm, capable of precisely detecting trace components in the mixed gas. The terahertz wave detector is based on the formula ( is the terahertz wave transmission time, l is the detection path length, v gAnalyze the rotational-vibrational energy levels of gas molecules by group velocity (the group velocity); the mass spectrometer uses time-of-flight (TOF) technology. The system is constructed based on the D-S evidence theory through a multi-source data fusion algorithm. The central control system fuses the data from different detection devices every 100 ms. After actual verification, the accuracy of the fused data reaches 99.6%, and the response time is less than 1 ms, providing reliable data support for the precise control of the system.

[0029] Deep reinforcement learning adaptive control system: With the edge computing unit as the core, a reinforcement learning model based on the deep Q-network (DQN) is constructed. The system takes 12 state parameters such as temperature deviation , pressure deviation , concentration deviation as inputs, and updates the policy through the formula (Q is the action value function, r is the immediate reward, is the discount factor, s is the current state, a is the current action, s' and a' are the next state and action). An experience replay mechanism and a double-network structure are introduced, and the genetic algorithm is combined to optimize the neural network weights, so that the adjustment response speed of the system under complex working conditions is improved, and the steady-state error is reduced to less than 0.1%.

[0030] Gradient condensation-membrane distillation coupling separation module: The three-stage gradient condensation structure of the gradient condensation-membrane distillation coupling separation module uses a finned heat exchanger. The temperatures of each stage of the condenser are set to -20°C, -40°C, and -60°C respectively, and an independent refrigeration system is equipped. The refrigeration system uses a Danfoss scroll compressor, which can accurately control the condensation temperature. The condensing pipe is made of copper, and a nanostructure with special wettability gradient is prepared on the surface through anodic oxidation technology. From the first stage to the third stage, the contact angle gradually changes from 50° to 150°. This structure is conducive to gas-liquid separation and improves the condensation efficiency. The membrane distillation unit selects a polyvinylidene fluoride (PVDF) hollow fiber membrane module with a membrane pore size of 0.15 μm, a porosity of 85%, and a membrane area of 10 m². During the membrane distillation process, separation is carried out according to the formula (J is the membrane flux, K is the membrane mass transfer coefficient, is the vapor pressure difference across the membrane). This module integrates a heat recovery system, which uses the Rankine cycle principle to recover the latent heat of condensation, and the energy recovery efficiency reaches more than 75%.

[0031] Bionic intelligent safety interlock module: The bionic intelligent safety interlock module constructs a risk assessment model based on the principle of biological neural network. This model is built using the TensorFlow framework of Python and contains 100 neuron nodes. The system analyzes the data collected by the sensors every 100 ms to evaluate the risk level of the system. The hierarchical response mechanism is as follows: In the first-level warning, a high-decibel audible and visual alarm (decibel: 85 dB, flash frequency: 2 Hz) is activated to give a warning to the operator; in the second-level warning, the electric control valve (regulation accuracy ±1%) automatically adjusts the valve opening according to the instructions of the control system to adjust the system parameters and make the system tend to a safe state; in the third-level warning, the quick cut-off valve driven by bionic muscles (artificial pneumatic muscle Festo DMSP-20-100-P-A-B) acts quickly. Its response time is only 48 ms, and it can cut off the gas path in an extremely short time to prevent the accident from expanding. At the same time, the system starts the accident simulation analysis program, uses SolidWorks Simulation software for 3D modeling and simulation, and provides detailed and accurate optimal disposal solutions for the operator to ensure the safety of personnel and equipment.

[0032] In the present invention, it also includes a nano-bubble enhanced mass transfer module: A nano-bubble generator is arranged in front of the inlet of the supercritical electromagnetic resonance reaction module. Through high-speed shearing and in cooperation with a Venturi tube, a pressure drop of 0.5 - 2 MPa is achieved to generate nano-bubbles with an average particle size of 50 - 200 nm. The bubble generation rate is regulated according to the formula (N is the bubble generation rate, Q is the gas flow rate, is the pressure drop, K v is the volume mass transfer coefficient). The surface of the nano-bubbles adsorbs catalyst nanoparticles (particle size 5 - 10 nm), which increases the gas-liquid mass transfer coefficient by 3 - 8 times and significantly promotes the dissolution and reaction of disilane.

[0033] In the present invention, it also includes a self-repairing intelligent maintenance module: The key components of the equipment (such as the inner wall of the reaction kettle and the surface of the microchannel) are coated with a self-repairing nano-coating. The coating is composed of micro-capsules containing a repair agent (particle size 1 - 5 μm) and a polymer matrix. When the coating is damaged, the micro-capsules rupture to release the repair agent, and rapid repair is achieved through the formula (t is the repair time, l is the diffusion distance, D is the diffusion coefficient of the repair agent, C0 and C are the initial concentration and the current concentration of the repair agent respectively). The module is also equipped with a micro 3D printing device, which can in-situ repair the worn mechanical components with a repair accuracy of ±20 μm.

[0034] In the present invention, a magnetorheological fluid damper is arranged in the pressure balance pipeline of the dynamic gradient partial pressure raw material storage module. Through the formula ( (where τ is the shear stress, K is the magnetorheological fluid constant, and H is the magnetic field strength) to adjust the gas flow rate in the pipeline, shortening the pressure balance response time to 1 - 2 seconds and controlling the pressure fluctuation within ±0.02 MPa.

[0035] In the present invention, an acoustic standing wave field is set at the microchannel outlet of the chaotic flow field molecular premixing module. Through the formula (where f is the acoustic wave frequency, n is the harmonic number, c is the speed of sound, and L is the length of the standing wave cavity) to generate standing waves, further enhancing molecular mixing and reducing the standard deviation of the mixing uniformity to 0.02.

[0036] In the present invention, the stirring paddle of the supercritical electromagnetic resonance reaction module adopts a topology optimization design. Through finite element analysis combined with genetic algorithms, the mass of the paddle blade is reduced while meeting the strength requirements, and at the same time, the stirring power consumption is reduced and the stirring efficiency is improved.

[0037] In the present invention, the deep reinforcement learning adaptive control system introduces a transfer learning mechanism to transfer the optimization strategy under similar working conditions to the new working conditions, improving the learning efficiency of the system in the new scenario and reducing the training data requirement by 90%.

[0038] The present invention includes the following steps: Multidimensional parameter collaborative planning step: The operator inputs the target mixed gas composition, such as silane concentration 10%, hydrogen concentration 80%, nitrogen concentration 10%, and production rate 100 kg / h, etc. into the human - machine interaction interface. After the system receives the input information, it calls the process database containing 5000 sets of historical data and uses the NSGA - II multi - objective optimization algorithm for calculation. During the optimization process, taking product quality, energy consumption, and efficiency as the optimization objectives, through the formula (where F is the comprehensive objective function value, w1 is the product quality weight with a value of 0.4; w2 is the energy consumption weight with a value of 0.3; w3 is the efficiency weight with a value of 0.3; f i is each sub - objective function) to calculate the comprehensive objective function value to seek the optimal combination of process parameters. After calculation, the optimal settings of 18 process parameters such as the raw material gas flow rate, reaction temperature (300 °C), and pressure (10 MPa) are finally determined and sent to each relevant device to complete parameter initialization.

[0039] Chaotic premixing dynamic adjustment step: When the raw material gas enters the chaotic flow field molecular premixing module, the Hamamatsu Photonics C15076 type laser - induced fluorescence (LIF) concentration monitor installed at the outlet of the microchannel mixer starts to collect the mixing uniformity data in real - time, with a collection frequency of 50 ms / time. Once it detects that the mixing uniformity is lower than 98%, the central control system immediately according to the formula (where k takes a value of 0.5, is the real - time concentration mean, Calculate the pulse frequency adjustment amount (for the set concentration), and send a control signal to the proportional valve of the pulsed gas injection device to automatically adjust the pulse frequency, thereby changing the gas injection method and mixing state, ensuring that the premixing uniformity is always maintained above 98%, and providing a uniform raw material mixture gas for the subsequent reaction process.

[0040] Intelligent regulation steps for supercritical resonance reaction: During the supercritical electromagnetic resonance reaction process, the multi-spectral fusion real-time monitoring module collects data such as temperature, pressure, and composition in the reaction kettle at a frequency of 100 ms / time, and transmits this data to the deep reinforcement learning adaptive control system. The control system analyzes and makes decisions according to the formula (where a is the optimal action, a' is the candidate action, and Q(s,a') is the action value function for executing action a' in state s), and selects the optimal regulation strategy. Then, the control system precisely controls the reaction process by adjusting parameters such as the frequency of the high-frequency electromagnetic coil, the power of the microwave resonance cavity, and the rotation speed of the stirring paddle, controlling the reaction temperature fluctuation within ±1°C, the pressure fluctuation within ±0.05 MPa, and the concentration error of key components less than 0.5%, ensuring that the reaction proceeds under the best conditions and improving product quality and production efficiency.

[0041] Gradient condensation - membrane distillation combined separation steps: The mixed gas after the reaction first enters the three-stage gradient condenser of the gradient condensation - membrane distillation coupling separation module. Under the stepwise condensation effects at -20°C, -40°C, and -60°C, most of the condensable gases in the mixed gas gradually condense into liquids and are separated. Subsequently, the remaining gas enters the membrane distillation unit. According to the formula (where the membrane mass transfer coefficient K = 0.01 kg / (m²·h·Pa)), by adjusting the operating pressure, control the membrane flux to achieve further separation of the remaining impurities and moisture in the gas. During the entire separation process, the heat recovery system operates synchronously, using an organic Rankine cycle device to recover the latent heat generated during the condensation process, and improving the energy recovery efficiency. Finally, after being processed by this module, the recovery rate of the target product reaches 99.2%, and the impurity content is reduced to 0.8 ppm, obtaining a disilane mixed gas that meets the quality requirements.

[0042] Intelligent maintenance steps for the entire life cycle: The intelligent maintenance module for the entire life cycle uses an infrared thermal imager (FLIRT1040) and an eddy current flaw detector (GE Panametrics 38DLPLUS) to monitor the key components of the equipment in real time, with a monitoring frequency of once per hour. When detecting situations such as damage to the inner wall coating of the reaction kettle in the supercritical electromagnetic resonance reaction module or wear on the surface of the microchannels in the chaotic flow field molecular premixing module, the system immediately starts the self-repair or 3D printing repair program. For coating damage, the microcapsules in the self-repairing nano-coating rupture, releasing the repair agent, according to the formula (where t is the repair time, l is the diffusion distance, D is the diffusion coefficient of the repair agent, C0 is the initial concentration of the repair agent, and C is the current concentration) for rapid repair; for component wear, the micro 3D printing device uses the same material as the component for in-situ repair according to the preset model and parameters, and the repair accuracy can reach ±20μm. At the same time, the life prediction model based on the LSTM neural network predicts the remaining life of key components of the equipment based on the equipment's operating time, operating conditions, historical fault data and other information, plans maintenance plans 30 days in advance, reduces unplanned downtime of the equipment, and effectively ensures the normal operation of the equipment and the continuity of production.

[0043] Emergency response hierarchical processing steps: The bionic intelligent safety interlock module evaluates the system risk every 100ms. When the risk level reaches level 1, the high-decibel sound and light alarm (decibel: 85dB, flash frequency: 2Hz) is immediately activated, and the warning information is pushed to the operator's terminal device to remind the operator to pay attention to the abnormal situation of the system. When the risk level rises to level 2, the system automatically controls the electric control valve (adjustment accuracy ±1%) to adjust the opening of the relevant valves, and adjusts the gas flow, pressure and other parameters to gradually make the system tend to a safe state. If the risk level reaches level 3, the bionic muscle-driven quick shut-off valve will be quickly closed within 48ms, cutting off the raw gas supply, and starting the accident simulation analysis program at the same time. The program uses SolidWorksSimulation software to perform three-dimensional modeling and simulation of the accident scene, analyze the cause of the accident and possible development trends, and provide operators with detailed, scientific and optimal disposal plans to help operators quickly and effectively handle accidents, minimize the losses caused by accidents, and ensure the safety of personnel and the integrity of equipment.

[0044] The present invention also includes the step of autonomous evolution of process knowledge: during the operation of the system, the optimized process parameters, control strategies and fault handling cases are stored in the knowledge graph database. The knowledge reasoning algorithm (based on graph neural network) is used to mine the potential relationship between data, automatically generate new process optimization solutions and control strategies, and regularly update the process database to achieve autonomous evolution of process knowledge and continuous improvement of system performance.

[0045] The present invention also includes a multi-system collaborative optimization step: when the system is linked with upstream and downstream production equipment, relevant equipment data is obtained through the industrial Internet platform, and a collaborative optimization algorithm (based on distributed model predictive control) is used to jointly optimize the operating parameters of multiple systems. (m is the number of systems, n is the time step, is the weight coefficient, is the sub-objective function of each system) to improve the production efficiency of the entire process and reduce energy consumption.

[0046] ReferenceFigure 3 , in the traditional technology, the mixing equipment uses a static mixer or a simple stirring device, without enhanced mixing structures such as double helix microchannels, pulse injection, and acoustic standing wave fields; when storing raw materials, the raw material gas is stored in a single gas storage tank or a simple gas storage tank group, and it is impossible to accurately control the storage pressure and temperature. The gas composition ratio deviates due to pressure fluctuations and temperature changes. After the gas storage tank material is in long-term contact with corrosive gases such as disilane, problems such as corrosion and hydrogen embrittlement are likely to occur, affecting the initial state of the raw materials; reaction conditions: supercritical electromagnetic resonance reaction is not adopted, the reaction pressure and temperature are conventional conditions, and there is no microwave resonance cavity matching the vibration frequency of disilane molecules, so it is impossible to accelerate the reaction and improve the uniformity; monitoring and regulation: rely on a single type of sensor for data acquisition, the information obtained is incomplete, the data accuracy and real-time performance are poor, it is impossible to precisely control the mixing process, and the mixing uniformity cannot be optimized in real-time closed-loop.

[0047] Refer to Figure 4 , in the traditional technology, for the reaction link: the traditional reaction kettle lacks effective reaction intensification means, without supercritical conditions, electromagnetic resonance, and topology-optimized stirring paddles, the reaction rate is slow, and there are many side reactions, resulting in a low product recovery rate; separation and maintenance: there is no gradient condensation - membrane distillation coupling separation module, and the impurity separation relies on single condensation, with poor separation effect; the key components of the equipment are not coated with a self-healing nano-coating, and there is no equipped micro 3D printing device, so in-situ repair cannot be carried out, and the equipment failure rate is high; full-process coordination: the deep reinforcement learning adaptive control system is not integrated, and the fixed-parameter PID control is adopted. When facing changes in working conditions, the adjustment lags, and it is impossible to respond in time to fluctuations in parameters such as temperature, pressure, and concentration, and the long-term operation stability is poor.

[0048] The above is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A dissolution and mixing system for disilane mixed gas, characterized in that It includes the following modules: Dynamic gradient partial pressure raw material storage module: It is composed of a gas storage tank matrix. The gas storage tanks are connected by intelligent variable-diameter pressure balance pipes. According to the formula Adjust the pipe diameter, D, is the real-time and initial pipe diameter, is the pipe diameter adjustment coefficient, is the pressure difference. The tank wall adopts a nano-layered composite structure. Through the formula Ultra-low heat conduction, is the composite thermal conductivity, , are the thermal conductivities of each layer, , are the volume ratios of each layer; Chaotic flow field molecular premixing module: Adopting a double-helix nested microchannel structure, through the formula to regulate the fluid state, is the critical Reynolds number, is the gas density, v is the flow velocity, D is the channel diameter, is the dynamic viscosity, is the pulse influence coefficient, f is the pulse frequency, and a laser-induced fluorescence concentration monitor is set at the module inlet to feedback the mixing uniformity; Supercritical electromagnetic resonance reaction module: The horizontal reactor adopts a double-layer jacket structure, and the stirring system uses planetary stirring paddles. The stirring speed is adjusted by , where n is the real-time speed, is the initial speed, P is the reaction pressure, is the standard pressure; Multi-spectral fusion real-time monitoring module: It consists of a distributed optical fiber spectral sensor network, a terahertz wave detector and a mass spectrometer. The terahertz wave detector analyzes the rotational-vibrational energy levels of gas molecules through the formula where is the terahertz wave transmission time, l is the detection path length, and v g is the group velocity; Deep Reinforcement Learning Adaptive Control System: Build a model with the edge computing unit as the core, and update the policy through the formula where Q is the action-value function, r is the immediate reward, is the discount factor, s and a are the current state and action, and s' and a' are the next state and action; Gradient condensation-membrane distillation coupling separation module: Adopting a three-stage gradient condensation structure, a membrane distillation unit is connected after the third-stage condensation. Through the formula for separation, J is the membrane flux, K is the membrane mass transfer coefficient, is the vapor pressure difference across the membrane; Bionic intelligent safety interlock module: Build a risk assessment model based on biological neural networks, respond hierarchically in case of anomalies, and be equipped with a cut-off valve driven by bionic muscles.

2. The dissolution and mixing system of disilane mixed gas according to claim 1, characterized in that It also includes a nano-bubble enhanced mass transfer module: a nano-bubble generator is set in front of the inlet of the supercritical electromagnetic resonance reaction module, and a pressure drop of 0.5 - 2 MPa is achieved through shearing and in cooperation with a Venturi tube to generate nano-bubbles with an average particle size of 50 - 200 nm. The bubble generation rate is regulated according to the formula , where N is the bubble generation rate, Q is the gas flow rate, is the pressure drop, and K v is the volumetric mass transfer coefficient.

3. The disolving and mixing system of disilane mixed gas according to claim 1, wherein It also includes a self-healing intelligent maintenance module: a self-healing nano-coating is applied to the key components of the device. The coating consists of microcapsules containing a repair agent and a polymer matrix. When the coating is damaged, the microcapsules rupture and release the repair agent. Through the formula for repair, where t is the repair time, l is the diffusion distance, D is the diffusion coefficient of the repair agent, C0 and C are the initial concentration and the current concentration of the repair agent respectively. The module is also equipped with a micro 3D printing device for in-situ repair of worn mechanical components.

4. The dissolution and mixing system of disilane mixed gas according to claim 1, characterized in that, A magnetorheological fluid damper is arranged in the pressure balance pipeline of the dynamic gradient partial pressure raw material storage module. Through the formula the gas flow rate in the pipeline is adjusted, where τ is the shear stress, K is the magnetorheological fluid constant, and H is the magnetic field strength.

5. The dissolution and mixing system of disilane mixed gas according to claim 1, characterized in that, An acoustic standing wave field is set at the microchannel outlet of the chaotic flow field molecular premixing module, and standing waves are generated through the formula where \(f\) is the acoustic wave frequency, \(n\) is the harmonic number, \(c\) is the speed of sound, and \(L\) is the length of the standing wave cavity.

6. The dissolution and mixing system of disilane mixed gas according to claim 1, wherein A microwave resonance cavity is arranged inside the reactor of the supercritical electromagnetic resonance reaction module. Through the formula matches the vibration frequency of disilane molecules, c is the speed of light, and L is the cavity length. is the relative permittivity; the stirring paddle of this module adopts a topology optimization design. Through finite element analysis combined with a genetic algorithm, the mass of the paddle blade is reduced while meeting the strength requirements.

7. The dissolution and mixing system of disilane mixed gas according to claim 1, characterized in that, The deep reinforcement learning adaptive control system introduces a transfer learning mechanism to transfer the optimized strategies under similar working conditions to new working conditions, improving the learning efficiency of the system in new scenarios.

8. A method for applying the dissolution and mixing system of the disilane mixed gas according to any one of claims 1-7, characterized in that, It includes the following steps: Multi-dimensional parameter collaborative planning steps: Input the target mixed gas composition and production parameters. The system calculates the optimal combination of 18 process parameters through a multi-objective optimization algorithm combined with a process database, and uses the formula to balance the product objectives, generating an initial control parameter set F as the comprehensive objective function value, w i as the weight coefficient, f i as each sub-objective function; Chaotic premixing dynamic adjustment steps: After the raw material gas enters the premixing module, the laser-induced fluorescence monitor real-time feeds back the data of the mixing uniformity. The control system dynamically adjusts the pulsed gas injection parameters according to the formula where is the pulse frequency adjustment amount, k is the adjustment coefficient, is the real-time concentration mean value, and C set is the set concentration. Intelligent regulation steps for supercritical resonance reaction: During the reaction process, the multi-spectral fusion monitoring module collects temperature, pressure, and composition data in real time. The deep reinforcement learning control system selects the adjustment strategy according to the formula a is the optimal action. By adjusting the electromagnetic coil frequency, microwave power, and stirring speed parameters, the reaction temperature fluctuation is controlled within ±1°C, the pressure fluctuation is controlled within ±0.05 MPa, and the concentration error of key components is less than 0.5%. Gradient condensation - membrane distillation combined separation step: After the mixed gas enters the separation module, according to the condensation temperatures at all levels and the membrane distillation parameters, through the formula and cooperatively control the condensation and membrane distillation processes to improve the recovery rate of the target product and reduce the impurity content to less than 1 ppm, where is the condensation heat transfer, U is the total heat transfer coefficient, A is the heat transfer area, is the temperature difference; Full-life-cycle intelligent maintenance step: The self-repair intelligent maintenance module continuously monitors the states of key components of the equipment. When coating damage or component wear is detected, it starts the self-repair or 3D printing repair program. At the same time, according to the equipment operation duration and working condition data, it uses the life prediction model to plan the maintenance plan in advance. Emergency response hierarchical processing step: The bionic intelligent safety interlock module evaluates the system risk in real time. When the risk level reaches level one, it gives an audible and visual alarm and pushes a warning message to the operator's terminal; in case of level two risk, it automatically adjusts the valve and equipment parameters to make the system tend to a safe state; in case of level three risk, it immediately triggers the emergency cut-off and safety protection devices, and at the same time starts the accident simulation analysis program to provide the operator with the optimal disposal plan.

9. The method for the dissolution and mixing system of disilane mixed gas according to claim 8, characterized in that, It also includes the step of autonomous evolution of process knowledge: During the operation of the system, the optimized process parameters, control strategies, and fault handling cases are stored in the knowledge graph database. The knowledge reasoning algorithm is used to mine the potential relationships between the data, and new process optimization plans and control strategies are automatically generated.

10. The method of the dissolution and mixing system of disilane mixed gas according to claim 8, wherein, It also includes the multi-system collaborative optimization step: when the system operates in linkage with upstream and downstream production equipment, relevant equipment data is obtained through the industrial Internet platform, and the collaborative optimization algorithm is used to jointly optimize the operating parameters of multiple systems. Through the formula the overall production efficiency is improved. Here, n is the time step, is the weight coefficient, is the sub-objective function of each system.

Images