A thermal field temperature equalization control method and system based on aerodynamic layout optimization

Abstract

This invention discloses a thermal field temperature equalization control method and system based on aerodynamic layout optimization, belonging to the field of thermal field temperature control technology. The method is applied to a gas delivery system in semiconductor thin film deposition processes, including: constructing an aerodynamic-thermal field coupling model to quantitatively analyze the influence of pipeline geometry on the temperature field; physically optimizing the pipeline structure based on the analysis results to reduce airflow disturbance; deploying a sensor network and implementing dual-domain collaborative dynamic control, combining a fuzzy PID algorithm to stabilize airflow, and using a model predictive control algorithm to pre-adjust heating power to compensate for thermal hysteresis; embedding microchannel heat exchangers filled with phase change materials and nanofluids in areas with the risk of instantaneous thermal disturbance to suppress temperature spikes; and performing self-optimization through a digital twin. This invention can systematically improve gas temperature uniformity, response speed, and long-term stability from source to end, meeting the stringent requirements of semiconductor processes.

Description

Technical Field

[0001] This invention relates to the field of thermal field temperature control technology, and in particular to a thermal field temperature equalization control method and system based on aerodynamic layout optimization. Background Technology

[0002] In semiconductor manufacturing processes, such as atomic layer deposition (ALD), chemical vapor deposition (CVD), and extreme ultraviolet (EUV) lithography gas preparation, the uniformity, stability, and response speed of the process gas during delivery to the reaction chamber are key factors determining the film thickness uniformity, compositional consistency, and device performance. Especially for 3nm and below technology nodes, the process window is extremely narrow, requiring the precursor gas to have its axial and radial temperature gradients controlled within ±0.5℃ or even smaller when entering the reaction chamber, and to possess sub-second response capabilities to rapid changes in gas flow and thermal field (such as the rapid cycling of ALD).

[0003] Currently, the industry mainly relies on the following technical approaches to improve the uniformity of the heat field in gas transportation, but all of them have significant limitations:

[0004] Independent control and hysteresis issues: Traditional temperature control often employs proportional-integral-derivative (PID) algorithms to adjust heater power through feedback loops. However, in long pipelines with complex bends and branching structures, the thermal inertia of the gas and pipe walls leads to significant thermal hysteresis. The inherent delay of PID feedback control makes it difficult to cope with rapid process switching, resulting in temperature overshoot and slow recovery, which fails to meet the rapid timing requirements of processes such as ALD.

[0005] Decoupled aerodynamic and thermal design: Existing systems typically treat airflow (pressure, flow rate) control and temperature control as relatively independent components. Aerodynamic control (e.g., via mass flow controllers) aims to stabilize flow rate, while thermal control is used for heating. This decoupled design ignores the decisive influence of airflow distribution on the temperature field. Airflow separation, vortices, and stagnation zones generated by pipe structures (e.g., right-angle bends, asymmetrical distributors) can cause significant differences in local heat transfer efficiency, forming inherent temperature gradients (often exceeding ±3°C) that are difficult to eliminate through end-point heating, becoming a fundamental bottleneck for improving uniformity.

[0006] Empirical layout and lack of quantitative guidance: The aerodynamic layout design of pipeline systems often relies on experience or rough fluid dynamics principles, lacking in-depth analysis and simulation guidance on the quantitative relationship between specific geometric structures (such as radius of curvature and flow splitting angle) and the final temperature field. This "trial and error" design makes it difficult to optimize the flow field from the source and cannot systematically eliminate hot and cold spots that cause temperature unevenness.

[0007] Limitations and lack of coordination in control strategies: While advanced control algorithms such as model predictive control (MPC) have been introduced to improve dynamic performance, their predictive models are often oversimplified and fail to fully consider the impact of complex flow fields on heat transfer. Furthermore, there is a lack of deep feedforward and coordination mechanisms between stable flow / pressure and controlled thermal field temperature. When process requirements necessitate flow rate switching, changes in airflow conditions act as severe disturbances impacting the thermal field, forcing a passive response and resulting in temperature fluctuations.

[0008] Lack of buffering mechanisms to cope with transient disturbances: Even with optimized control and layout, extreme disturbances such as instantaneous valve opening and closing, and sudden changes in airflow can still trigger transient temperature spikes that exceed the regulation bandwidth of closed-loop control. Existing technologies lack a passive or semi-passive buffering mechanism capable of rapidly absorbing / releasing large amounts of heat and actively smoothing out such spikes.

[0009] Insufficient system adaptability: As equipment operates, factors such as heater aging and slight deposits on the inner walls of pipes can cause the system's heat transfer characteristics to drift. Traditional fixed-parameter control models and strategies cannot adapt to this slow time-varying behavior, leading to a gradual deterioration in temperature uniformity after long-term operation. The system lacks online self-sensing, self-evaluation, and self-optimization capabilities, relying on regular manual maintenance and calibration, which affects production efficiency and consistency.

[0010] In summary, existing technical solutions are partial and isolated, failing to address the problem from a holistic perspective encompassing "system modeling - source design - collaborative control - passive buffering - self-evolution." Therefore, there is an urgent need to develop a systematic gas thermal field equalization control method capable of fundamentally analyzing and optimizing airflow distribution, achieving dynamic coordination between aerodynamics and thermal fields, suppressing extreme disturbances, and adapting to long-term equipment state changes to meet the stringent requirements of next-generation semiconductor manufacturing for extreme uniformity of process gas temperature, rapid response, and long-term stability. Summary of the Invention

[0011] The purpose of this invention is to overcome the shortcomings of the prior art and provide a thermal field temperature equalization control method and system based on aerodynamic layout optimization.

[0012] The objective of this invention is achieved through the following technical solution: The first aspect of this invention provides: a thermal field temperature equalization control method based on aerodynamic layout optimization, used in a gas delivery system for semiconductor thin film deposition processes, comprising the following steps:

[0013] Aerodynamic-thermal coupling modeling and quantitative analysis stage: Based on the pipeline structure of the gas delivery system, computational fluid dynamics and finite element analysis are used for joint simulation to construct an aerodynamic-thermal coupling model; the aerodynamic-thermal coupling model is used to quantitatively analyze the influence of pipeline geometric characteristics on airflow distribution and thermal field uniformity, and identify airflow vortices, stagnation zones and temperature gradient exceeding the standard areas to obtain pipeline analysis results;

[0014] Aerodynamic layout optimization design stage: Based on the pipeline analysis results, the pipeline structure is physically optimized to reduce airflow disturbance and thermal unevenness;

[0015] Dual-domain collaborative dynamic control stage: In the optimized pipeline structure, a sensor network is deployed, and a pneumatic control strategy and a thermal field control strategy are executed; the pneumatic control strategy is used to dynamically adjust the pneumatic components to stabilize the airflow; the thermal field control strategy is used to pre-adjust the heating power in a feedforward manner to compensate for thermal hysteresis;

[0016] Local phase change energy storage buffer stage: For the disturbance region that still has the risk of instantaneous thermal disturbance after dual-domain collaborative dynamic control, a microchannel heat exchanger filled with phase change material and built with nanofluid is embedded to absorb or release instantaneous thermal shock by utilizing the latent heat of phase change, thereby suppressing the temperature peak caused by airflow disturbance.

[0017] Real-time monitoring and self-optimization of thermal field temperature: Real-time monitoring of the operating status of the gas delivery system, evaluation of thermal field equalization performance, and automatic triggering of control parameter optimization process or model optimization process when performance deviation or equipment status change is detected, so as to keep the gas delivery system in the current optimal working state.

[0018] Preferably, the construction of the aerodynamic-thermal field coupling model includes the following steps:

[0019] Parametric geometric modeling steps: Based on the pipeline structure of the gas delivery system, the geometric features that affect airflow and heat transfer are parametrically defined, including at least the radius of curvature of the elbow, the bending angle of the elbow, the branching angle of the splitter, and the interface transition radius of the splitter.

[0020] Multiphysics mesh generation steps: Perform multi-domain mesh generation on the parameterized geometric model, where the fluid domain uses an unstructured tetrahedral mesh with boundary layer refinement to capture near-wall flow and heat transfer, and the solid domain uses a hexahedral mesh;

[0021] Steps for setting boundary conditions and physical property parameters: Define simulation boundary conditions, including the mass of the gas inlet, the flow rate of the gas inlet, the temperature of the gas inlet, the pressure of the outlet, and the thermal boundary conditions between the pipe wall and the heater; and set the physical property parameters of the process gas as functions of temperature.

[0022] Two-way fluid-structure interaction solution: The computational fluid dynamics solver and the finite element analysis solver are used for co-simulation. In each time step or iteration step, the computational fluid dynamics solver transmits the calculated convective heat transfer coefficient at the interface between the fluid and the pipe wall to the finite element analysis solver, and the finite element analysis solver transmits the calculated pipe wall temperature field back to the computational fluid dynamics solver as a new thermal boundary. This process is repeated until convergence to obtain the aerodynamic-thermal field coupled model.

[0023] Preferably, the process of quantitative analysis using aerodynamic-thermal field coupling model includes the following steps:

[0024] The steps for quantitative analysis of airflow uniformity are as follows: extract the velocity cloud map and streamline map of the pipeline section, identify and quantify the location, size and intensity of the vortex region; calculate the local airflow residence time caused by the structure, which is the difference between the average time of fluid particles passing through the target local area and the time of passage through the mainstream area;

[0025] Steps for quantitative analysis of thermal field uniformity: Draw temperature distribution cloud maps of the pipeline along the axial and radial directions, and calculate the axial temperature gradient and radial temperature non-uniformity of the target section from the gas heating start point to the inlet of the reaction chamber.

[0026] Thermal dynamic response characteristic analysis steps: In the simulation, set a step change in heating power or a step change in inlet flow rate to simulate transient conditions, calculate the time required for the system to reach the preset percentage of the new steady-state value from the start of the step change, and define it as the system thermal hysteresis time.

[0027] Parameter impact analysis steps: Continuously change the geometric parameters of the pipeline structure geometric model and re-analyze to establish a quantitative relationship graph between geometric parameters and airflow uniformity index, temperature gradient, and system thermal lag time, providing optimization objectives and design basis for the aerodynamic layout optimization design stage.

[0028] Preferably, the solid domain includes the tube wall and the heating element; the thermal boundary conditions between the tube wall and the heater include constant heat flux density and convective heat transfer coefficient; the physical properties of the process gas include density, viscosity, specific heat capacity and thermal conductivity.

[0029] Preferably, the pipeline cross-section includes the cross-section downstream of each bend, before and after the distributor, and at the inlet of the reaction chamber.

[0030] Preferably, the aerodynamic layout optimization design stage further includes the following steps:

[0031] Elbow structure optimization steps: Replace right-angle elbows in the gas delivery pipeline with elbows with gradually decreasing curvature to eliminate flow separation and Dean vortices caused by sudden changes in airflow direction, and reduce local drag coefficient and energy loss.

[0032] Optimization steps for split / merge structure: Optimize the asymmetric split or merge nodes in the gas delivery pipeline into a symmetrical structure with a streamlined transition; for splitters that supply gas to multiple process chambers or heating zones, adopt a Y-type or improved T-type symmetrical design, and set a DC stabilization section upstream of the splitter to ensure that the airflow is fully developed and uniform, and reduce the flow distribution deviation of each branch.

[0033] Pipeline inner surface and connection optimization steps: Electropolishing is applied to the inner wall of the pipeline, and the interface at the pipeline connection is set to a structure with a flush inner wall and no step abrupt change, in order to minimize flow friction resistance and prevent secondary flow and particle deposition caused by abrupt change in inner diameter.

[0034] Preferably, the aerodynamic layout optimization design stage also includes local flow guidance and rectification steps: for areas where there is still a risk of uneven airflow, fixed guide vanes or rectification grids are added inside the pipe. The installation angle and number of the guide vanes or rectification grids are determined based on the streamline analysis results of the aerodynamic-thermal field coupling model, which are used to further guide the airflow, break up large-scale vortices, and achieve uniformity of the velocity field.

[0035] Preferably, the pneumatic control strategy is based on a fuzzy PID control algorithm; the thermal field control strategy is based on a model predictive control algorithm.

[0036] Preferably, the dual-domain collaborative dynamic control stage further includes the following steps:

[0037] Sensor network deployment steps: Deploy a multi-type sensor network at target locations in the optimized piping structure. The target locations in the optimized piping structure include at least the downstream of each gradually curvature bend, the outlets of each branch of the symmetrical splitter, and the inlet of the process reaction chamber. The multi-type sensor network includes at least a pressure sensor and a mass flow controller for monitoring airflow status, and a temperature sensor for monitoring thermal field temperature.

[0038] The pneumatic control strategy implementation steps are as follows: real-time pressure and flow data collected by pressure sensors and mass flow controllers are used as inputs. The opening degree of the upstream pneumatic proportional valve or mass flow controller is dynamically adjusted by a fuzzy PID control algorithm. The fuzzy PID controller takes the deviation between the pressure setpoint and the measured value and its rate of change as two-dimensional inputs. It infers online through a preset fuzzy rule base and outputs the correction amount of the proportional, integral and derivative parameters to suppress pipeline pressure fluctuations.

[0039] The thermal field control strategy implementation steps are as follows: using the stabilized flow and pressure data as feedforward input and temperature sensor data as feedback, a model predictive control algorithm is used for rolling optimization. The model predictive control algorithm is based on the simplified state-space equation of the aerodynamic-thermal field coupling model as the internal prediction model. In each control cycle, it predicts the temperature response of each zone of the wound heater in multiple future steps. By solving the constrained optimization problem, it calculates the optimal heating power sequence that makes the predicted temperature trajectory smoothest and closest to the target temperature. The control quantity is output to the actuator to compensate for the system thermal hysteresis, thereby shortening the overall thermal response time.

[0040] Dual-domain collaboration and signal interaction steps: The aerodynamic control strategy and the thermal field control strategy run in parallel on the same industrial control platform and interact with each other through shared memory or real-time communication bus; the model predictive control algorithm incorporates the stabilized flow and pressure data as feedforward disturbances into the internal prediction model, thereby adjusting the heating power in advance before the expected changes in airflow state, and realizing the collaborative control of aerodynamics and thermal field.

[0041] Preferably, the disturbance region includes at least the final confluence pipe section before multiple gases enter the process reaction chamber. When the airflow disturbance causes the pipe wall temperature in the disturbance region to rise instantaneously, the heat is conducted to the phase change material through the pipe wall, causing it to undergo a solid-liquid phase change and absorb a large amount of latent heat, thereby suppressing the gas temperature spike. At the same time, the nanofluid flowing through the microchannel continuously carries away the heat stored in the phase change material through forced convection, allowing it to periodically solidify and regenerate, thus providing sustainable buffering.

[0042] Preferably, the phase change material is a paraffin-based composite material, and the phase change material is encapsulated in a microchannel heat exchanger in the form of microencapsulation or in combination with a high thermal conductivity metal foam; the nanofluid is an Al2O3-water-based fluid.

[0043] Preferably, the thermal field equalization performance is evaluated using a digital twin, and the real-time monitoring and self-optimization stage of the thermal field temperature further includes the following steps:

[0044] Multi-source data real-time acquisition steps: Real-time acquisition of multi-dimensional operational data including gas temperature, pressure, mass flow rate and heating element power of pipeline cross-section; at the same time, acquisition of the currently executed process recipe information, including target temperature, gas flow rate setpoint and process step sequence;

[0045] Digital twin construction steps: Based on the aerodynamic-thermal field coupling model, a digital twin of the gas delivery system is constructed; the digital twin is synchronized with the physical system in real time through a data interface, and receives real-time data collected in the multi-source data real-time acquisition step as boundary conditions and initial conditions to drive simulation calculations, and virtually reproduces the airflow state and three-dimensional temperature field distribution of the physical system in real time.

[0046] Performance evaluation and anomaly detection steps: Use a digital twin to calculate and output target performance indicators, including at least the temperature non-uniformity of the reaction chamber inlet section, the maximum axial temperature gradient of the pipeline, and the system thermal response time; set a safety threshold for the target performance indicators. When any performance indicator calculated in real time deviates from its threshold range for more than a preset time, or when abnormal sensor data is detected, it is determined that the system performance has deviated or the equipment status has changed, triggering the self-optimization process.

[0047] Self-optimization steps: Execute control parameter optimization procedures and / or model optimization procedures to ensure that the gas delivery system always maintains its current optimal operating state for the current equipment status and process requirements.

[0048] Preferably, the semiconductor thin film deposition process is an atomic layer deposition process.

[0049] Preferably, the semiconductor thin film deposition process is a chemical vapor deposition process or a metal-organic chemical vapor deposition process.

[0050] A second aspect of the present invention provides: a thermal field temperature equalization control system based on aerodynamic layout optimization, used to implement any of the above-mentioned thermal field temperature equalization control methods based on aerodynamic layout optimization, comprising:

[0051] The aerodynamic-thermal field coupled modeling and quantitative analysis module is used to construct an aerodynamic-thermal field coupled model based on the pipeline structure of the gas delivery system by using computational fluid dynamics and finite element analysis for joint simulation. The aerodynamic-thermal field coupled model is used to quantitatively analyze the influence of pipeline geometric features on airflow distribution and thermal field uniformity, identify airflow vortices, stagnation zones and temperature gradient exceeding the standard areas to obtain pipeline analysis results.

[0052] The aerodynamic layout optimization design module is used to physically optimize the pipeline structure based on pipeline analysis results, thereby reducing airflow disturbance and thermal unevenness.

[0053] The dual-domain collaborative dynamic control module is used to deploy a sensor network in the optimized pipeline structure and execute pneumatic control strategy and thermal field control strategy; the pneumatic control strategy is used to dynamically adjust pneumatic components to stabilize airflow; the thermal field control strategy is used to pre-adjust heating power in a feedforward manner to compensate for thermal hysteresis.

[0054] The local phase change energy storage buffer module is used to embed a microchannel heat exchanger filled with phase change material and built with nanofluid in the disturbance area that still has the risk of instantaneous thermal disturbance after dual-domain collaborative dynamic control. It uses the latent heat of phase change to absorb or release instantaneous thermal shock and suppress temperature peaks caused by airflow disturbance.

[0055] The real-time monitoring and self-optimization module for thermal field temperature is used to monitor the operating status of the gas transport system in real time, evaluate the thermal field equalization performance, and automatically trigger the control parameter optimization process or model optimization process when performance deviation or equipment status change is detected, so as to keep the gas transport system in the current optimal working state.

[0056] The beneficial effects of this invention are:

[0057] 1) This invention, through aerodynamic-thermal field coupling modeling and quantitative analysis, achieves for the first time a precise prediction of the quantitative relationship between geometric features (elbow curvature, flow splitting angle) and the final temperature distribution during the pipeline design stage. Based on this, aerodynamic layout optimization (gradual curvature elbows, symmetrical flow splitting structure) eliminates airflow vortices and stagnation zones that cause temperature unevenness to the greatest extent possible from a physical structure perspective. Compared to traditional designs that rely on experience or simple criteria, this invention minimizes the systematic temperature gradient caused by unreasonable layout from the source. Combined with subsequent coordinated control and buffering, the axial temperature gradient of the gas from the heating point to the reaction chamber inlet can be stably controlled within ±0.3℃ from over ±3℃ in traditional schemes, fully meeting the requirements of extreme temperature uniformity of precursor gases for nanoscale processes such as EUV lithography and advanced ALD.

[0058] 2) This invention deeply couples pneumatic control and thermal field control through dual-domain collaborative dynamic control. Pneumatic control (fuzzy PID) rapidly stabilizes the airflow, laying the foundation for temperature control; thermal field control (MPC), using a feedforward approach, pre-adjusts the heating power based on real-time airflow data to actively compensate for thermal hysteresis. This "airflow state sensing + power pre-output" mode overcomes the hysteresis bottleneck of traditional PID feedback control. It reduces the system's thermal response time (from step change to stability) from the traditional 10 seconds to less than 2 seconds, making it particularly suitable for the rapid gas switching (pulsing) several times per second in ALD processes, ensuring the accuracy of process timing and the periodicity of thin film growth.

[0059] 3) This invention reduces unnecessary turbulence and pressure loss through layout optimization, physically improving airflow smoothness. Simultaneously, the fuzzy PID algorithm in pneumatic control effectively suppresses pressure pulsations caused by valve actuation and pump speed fluctuations, reducing the overall pressure fluctuation amplitude of the pipeline system by more than 40%. The more stable airflow significantly reduces the risk of particle generation on the pipeline inner wall due to airflow impact, which is crucial for advanced processes sensitive to defect density. Furthermore, the optimized flow field and stable pressure also reduce energy consumption for gas delivery, resulting in an overall reduction of system energy consumption of approximately 15%.

[0060] 4) Based on aerodynamic design and advanced control, this invention introduces a local phase change energy storage buffer in the critical region (multi-path convergence point) where airflow disturbance is most intense. The microchannel heat exchange unit utilizes the high latent heat of phase change materials to instantly absorb or release large amounts of heat, acting like a "thermal capacitor" to quickly smooth out instantaneous temperature spikes that any control loop cannot fully keep up with. Combined with the heat transfer capabilities enhanced by nanofluids, the unit's response time is less than 2 seconds. This adds a passive and reliable final barrier to the entire temperature control system, suppressing instantaneous temperature spikes within an extremely narrow range, greatly enhancing the robustness and anti-interference capability of the process.

[0061] 5) This invention achieves virtual mirroring and real-time performance evaluation of the physical system by constructing a high-fidelity digital twin. When performance drift caused by equipment aging (such as decreased heater efficiency or slight deposition in pipelines) is detected, the system can automatically trigger control parameter optimization or online calibration of the predictive model. This allows the control system to adapt to changes in equipment status and automatically adjust itself back to the optimal operating point without requiring manual calibration during downtime. This solves the problem of inevitable performance degradation after long-term operation of existing systems, ensuring the long-term consistency and stability of gas temperature performance throughout the equipment's lifecycle, and improving the overall availability of the equipment and the consistency of produced wafers. Attached Figure Description

[0062] Figure 1 This is a flowchart illustrating the overall method of the present invention;

[0063] Figure 2 This is a flowchart of the construction process for the aerodynamic-thermal coupling model;

[0064] Figure 3 Flowchart for quantitative analysis;

[0065] Figure 4 Flowchart for aerodynamic layout optimization design;

[0066] Figure 5 This is a flowchart of the dual-domain collaborative dynamic control process.

[0067] Figure 6 This is a flowchart for real-time monitoring and self-optimization of thermal field temperature. Detailed Implementation

[0068] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] See Figures 1-6The first aspect of this invention provides: a thermal field temperature equalization control method based on aerodynamic layout optimization, for a gas delivery system in a semiconductor thin film deposition process, comprising the following steps:

[0070] Aerodynamic-thermal coupling modeling and quantitative analysis stage: Based on the pipeline structure of the gas delivery system, computational fluid dynamics and finite element analysis are used for joint simulation to construct an aerodynamic-thermal coupling model; the aerodynamic-thermal coupling model is used to quantitatively analyze the influence of pipeline geometric characteristics on airflow distribution and thermal field uniformity, and identify airflow vortices, stagnation zones and temperature gradient exceeding the standard areas to obtain pipeline analysis results;

[0071] Aerodynamic layout optimization design stage: Based on the pipeline analysis results, the pipeline structure is physically optimized to reduce airflow disturbance and thermal unevenness;

[0072] Dual-domain collaborative dynamic control stage: In the optimized pipeline structure, a sensor network is deployed, and a pneumatic control strategy and a thermal field control strategy are executed; the pneumatic control strategy is used to dynamically adjust the pneumatic components to stabilize the airflow; the thermal field control strategy is used to pre-adjust the heating power in a feedforward manner to compensate for thermal hysteresis;

[0073] Local phase change energy storage buffer stage: For the disturbance region that still has the risk of instantaneous thermal disturbance after dual-domain collaborative dynamic control, a microchannel heat exchanger filled with phase change material and built with nanofluid is embedded to absorb or release instantaneous thermal shock by utilizing the latent heat of phase change, thereby suppressing the temperature peak caused by airflow disturbance.

[0074] Real-time monitoring and self-optimization of thermal field temperature: Real-time monitoring of the operating status of the gas delivery system, evaluation of thermal field equalization performance, and automatic triggering of control parameter optimization process or model optimization process when performance deviation or equipment status change is detected, so as to keep the gas delivery system in the current optimal working state.

[0075] In this embodiment, a precursor gas delivery system for atomic layer deposition (ALD) is used as an example for illustration. However, the invention is not limited to this and is also applicable to semiconductor thin film deposition processes requiring high-precision gas temperature control, such as chemical vapor deposition (CVD) and metal-organic chemical vapor deposition (MOCVD). For the gas delivery system of the ALD equipment, the system is responsible for heating two precursor gases (TMA and H2O) to 150°C and then delivering them to the reaction chamber for alternating deposition. The original system suffers from uneven reaction chamber inlet temperature (gradient of ±3°C) and slow thermal response (approximately 10 seconds) during process switching. The present invention employs a systematic optimization and control method, the process flow of which is as follows: Figure 1 As shown.

[0076] In some embodiments, the construction of the aerodynamic-thermal field coupling model includes the following steps:

[0077] Parametric geometric modeling steps: Based on the pipeline structure of the gas delivery system, the geometric features that affect airflow and heat transfer are parametrically defined, including at least the radius of curvature of the elbow, the bending angle of the elbow, the branching angle of the splitter, and the interface transition radius of the splitter.

[0078] Multiphysics mesh generation steps: Perform multi-domain mesh generation on the parameterized geometric model, where the fluid domain uses an unstructured tetrahedral mesh with boundary layer refinement to capture near-wall flow and heat transfer, and the solid domain uses a hexahedral mesh;

[0079] Steps for setting boundary conditions and physical property parameters: Define simulation boundary conditions, including the mass of the gas inlet, the flow rate of the gas inlet, the temperature of the gas inlet, the pressure of the outlet, and the thermal boundary conditions between the pipe wall and the heater; and set the physical property parameters of the process gas as functions of temperature.

[0080] Two-way fluid-structure interaction solution: The computational fluid dynamics solver and the finite element analysis solver are used for co-simulation. In each time step or iteration step, the computational fluid dynamics solver transmits the calculated convective heat transfer coefficient at the interface between the fluid and the pipe wall to the finite element analysis solver, and the finite element analysis solver transmits the calculated pipe wall temperature field back to the computational fluid dynamics solver as a new thermal boundary. This process is repeated until convergence to obtain the aerodynamic-thermal field coupled model.

[0081] In this embodiment, the aerodynamic-thermal field coupling model is constructed as follows: Figure 2 As shown. It mainly includes the following steps:

[0082] Parametric geometric modeling: Accurate modeling of the gas delivery pipeline is performed using 3D CAD software (such as SolidWorks). The focus is on parametrically defining the geometric features affecting airflow, including: the diameter of the main gas supply line (D = 1 / 4 inch), the radius of curvature of all bends (R, initially designed as right angles, i.e., R / D≈0.5), the branching angle of the Y-shaped splitter before gas convergence (θ, initially 90°), and the transition arc radius at each interface.

[0083] Multiphysics Mesh Generation and Solver Setup: The geometric model was imported into the ANSYS Workbench platform. Computational Fluid Dynamics (CFD) and Finite Element Analysis (FEA) were used for co-simulation. The fluid domain (gas flow region) employed an unstructured mesh with boundary layer refinement; the thickness of the first mesh layer near the wall was verified using y+ to ensure accurate capture of convective heat transfer. The solid domain (pipe wall, wound heater) used a hexahedral mesh. Boundary conditions were set as follows: gas inlet at mass flow rate (based on process formulation), temperature at room temperature (25°C), and outlet at pressure; the heater outer wall was set to constant heat flux density, and natural convection heat dissipation was considered on the pipe wall outer surface.

[0084] Two-way fluid-structure interaction (FSI) solution: A two-way FSI thermal coupling process is built in ANSYS System Coupling. In each iteration step, Fluent (CFD solver) calculates the convective heat transfer coefficient and fluid temperature at the fluid-to-pipe wall interface and passes them to Mechanical (FEA solver). After Mechanical solves the solid temperature field, it returns the updated pipe wall temperature to Fluent as the new thermal boundary condition. This iteration continues until both the flow field and temperature field converge, resulting in a high-fidelity aerodynamic-thermal coupling model.

[0085] In some embodiments, the process of performing quantitative analysis using a coupled aerodynamic-thermal field model includes the following steps:

[0086] The steps for quantitative analysis of airflow uniformity are as follows: extract the velocity cloud map and streamline map of the pipeline section, identify and quantify the location, size and intensity of the vortex region; calculate the local airflow residence time caused by the structure, which is the difference between the average time of fluid particles passing through the target local area and the time of passage through the mainstream area;

[0087] Steps for quantitative analysis of thermal field uniformity: Draw temperature distribution cloud maps of the pipeline along the axial and radial directions, and calculate the axial temperature gradient and radial temperature non-uniformity of the target section from the gas heating start point to the inlet of the reaction chamber.

[0088] Thermal dynamic response characteristic analysis steps: In the simulation, set a step change in heating power or a step change in inlet flow rate to simulate transient conditions, calculate the time required for the system to reach the preset percentage of the new steady-state value from the start of the step change, and define it as the system thermal hysteresis time.

[0089] Parameter impact analysis steps: Continuously change the geometric parameters of the pipeline structure geometric model and re-analyze to establish a quantitative relationship graph between geometric parameters and airflow uniformity index, temperature gradient, and system thermal lag time, providing optimization objectives and design basis for the aerodynamic layout optimization design stage.

[0090] In this embodiment, the quantitative analysis process is as follows: Figure 3 As shown. Quantitative analysis and problem identification: Run the coupled model to perform quantitative analysis:

[0091] Airflow analysis: Velocity contour maps were extracted from key sections (downstream of each bend, before and after the splitter, and at the inlet of the reaction chamber). Analysis revealed significant flow separation and Dean's vortex behind the right-angle bend, and the Y-shaped splitter, due to its asymmetrical design, resulted in a 15% flow deviation between the two branches.

[0092] Thermal field analysis: The axial temperature distribution curve from the heating start point to the inlet of the reaction chamber was plotted, and the maximum axial temperature gradient was calculated to be 8℃ / m. The radial temperature contour plot of the inlet section of the reaction chamber was plotted, and the temperature non-uniformity (standard deviation / mean) was calculated to be 4.2%.

[0093] Dynamic response analysis: In the simulation, a step change in gas flow rate was simulated, and the system thermal hysteresis time (the time required for the temperature to reach 95% of the new steady-state value) was calculated to be approximately 9 seconds.

[0094] Based on the above, the pipeline analysis results indicate that right-angle elbows and asymmetrical distributors are the main sources of airflow disturbance and temperature unevenness.

[0095] In some embodiments, the solid domain includes a tube wall and a heating element; the thermal boundary conditions between the tube wall and the heater include a constant heat flux density and a convective heat transfer coefficient; and the physical properties of the process gas include density, viscosity, specific heat capacity, and thermal conductivity.

[0096] In some embodiments, the pipeline cross-section includes the cross-section downstream of each bend, before and after the distributor, and at the inlet of the reaction chamber.

[0097] In some embodiments, the aerodynamic layout optimization design phase further includes the following steps:

[0098] Elbow structure optimization steps: Replace right-angle elbows in the gas delivery pipeline with elbows with gradually decreasing curvature to eliminate flow separation and Dean vortices caused by sudden changes in airflow direction, and reduce local drag coefficient and energy loss.

[0099] Optimization steps for split / merge structure: Optimize the asymmetric split or merge nodes in the gas delivery pipeline into a symmetrical structure with a streamlined transition; for splitters that supply gas to multiple process chambers or heating zones, adopt a Y-type or improved T-type symmetrical design, and set a DC stabilization section upstream of the splitter to ensure that the airflow is fully developed and uniform, and reduce the flow distribution deviation of each branch.

[0100] Pipeline inner surface and connection optimization steps: Electropolishing is applied to the inner wall of the pipeline, and the interface at the pipeline connection is set to a structure with a flush inner wall and no step abrupt change, in order to minimize flow friction resistance and prevent secondary flow and particle deposition caused by abrupt change in inner diameter.

[0101] In some embodiments, the aerodynamic layout optimization design stage further includes local flow guidance and rectification steps: for areas where there is still a risk of uneven airflow, fixed guide vanes or rectification grids are added inside the pipe. The installation angle and number of the guide vanes or rectification grids are determined based on the streamline analysis results of the aerodynamic-thermal field coupling model, which are used to further guide the airflow, break up large-scale vortices, and achieve uniformity of the velocity field.

[0102] In this embodiment, the aerodynamic layout optimization design process is as follows: Figure 4 As shown. Based on the quantitative analysis results, the physical structure of the pipeline is optimized to improve the flow field from the source.

[0103] Elbow structure optimization: All right-angle elbows were replaced with gradually curvature elbows, and the radius of curvature ratio R / D was optimized to 3.0. CFD verification shows that this design can effectively eliminate flow separation, reduce local pressure drop at elbows by 70%, and significantly weaken vortex intensity.

[0104] Streamline / Builtuplet Structure Optimization: The asymmetric Y-type splitter was redesigned into a symmetric streamlined structure, the branch angle θ was optimized to 60°, and an upstream DC stabilization section (length ≥ 10D) was added. After optimization, the flow distribution deviation between the two branches was reduced to within 3%.

[0105] Pipeline inner surface and connection optimization: The inner wall of the pipeline is electropolished to achieve a surface roughness Ra < 0.4 μm. All VCR joints are installed with the inner wall flush and without steps to minimize flow resistance and the risk of particle deposition.

[0106] Local guide vanes added: In the pipe section before the gas finally converges to the inlet of the reaction chamber, two fixed guide vanes are installed based on the streamline analysis results. These are used to further guide and mix the two gas streams, break up any large-scale vortices that may remain, and improve the uniformity of velocity distribution by 40%.

[0107] In some embodiments, the pneumatic control strategy is based on a fuzzy PID control algorithm; the thermal field control strategy is based on a model predictive control algorithm.

[0108] In some embodiments, the dual-domain collaborative dynamic control stage further includes the following steps:

[0109] Sensor network deployment steps: Deploy a multi-type sensor network at target locations in the optimized piping structure. The target locations in the optimized piping structure include at least the downstream of each gradually curvature bend, the outlets of each branch of the symmetrical splitter, and the inlet of the process reaction chamber. The multi-type sensor network includes at least a pressure sensor and a mass flow controller for monitoring airflow status, and a temperature sensor for monitoring thermal field temperature.

[0110] The pneumatic control strategy implementation steps are as follows: real-time pressure and flow data collected by pressure sensors and mass flow controllers are used as inputs. The opening degree of the upstream pneumatic proportional valve or mass flow controller is dynamically adjusted by a fuzzy PID control algorithm. The fuzzy PID controller takes the deviation between the pressure setpoint and the measured value and its rate of change as two-dimensional inputs. It infers online through a preset fuzzy rule base and outputs the correction amount of the proportional, integral and derivative parameters to suppress pipeline pressure fluctuations.

[0111] The thermal field control strategy implementation steps are as follows: using the stabilized flow and pressure data as feedforward input and temperature sensor data as feedback, a model predictive control algorithm is used for rolling optimization. The model predictive control algorithm is based on the simplified state-space equation of the aerodynamic-thermal field coupling model as the internal prediction model. In each control cycle, it predicts the temperature response of each zone of the wound heater in multiple future steps. By solving the constrained optimization problem, it calculates the optimal heating power sequence that makes the predicted temperature trajectory smoothest and closest to the target temperature. The control quantity is output to the actuator to compensate for the system thermal hysteresis, thereby shortening the overall thermal response time.

[0112] Dual-domain collaboration and signal interaction steps: The aerodynamic control strategy and the thermal field control strategy run in parallel on the same industrial control platform and interact with each other through shared memory or real-time communication bus; the model predictive control algorithm incorporates the stabilized flow and pressure data as feedforward disturbances into the internal prediction model, thereby adjusting the heating power in advance before the expected changes in airflow state, and realizing the collaborative control of aerodynamics and thermal field.

[0113] In this embodiment, the dual-domain collaborative dynamic control process is as follows: Figure 5 As shown. Based on the optimized hardware, an intelligent control system is deployed to achieve rapid and coordinated adjustment of the aerodynamic and thermal fields.

[0114] Sensor network deployment: Deploy sensors at key points in the optimized pipeline: Install high-precision PT1000 platinum resistance temperature sensors (accuracy ±0.1℃) downstream of each gradually curvature bend, at the outlet of each branch of the symmetrical splitter, and at the inlet of the reaction chamber; install pressure sensors at the downstream end of the mass flow controller (MFC); the MFC itself provides high-precision flow feedback.

[0115] Pneumatic control strategy implementation: A fuzzy PID control algorithm is used to stabilize pipeline pressure. The deviation (e) and rate of change (ec) between the pressure setpoint (e.g., 2 bar) and the measured value are used as two-dimensional inputs. Through online inference using a preset fuzzy rule base (e.g., "if e is positive and ec is negative, then increase the proportional gain Kp"), the PID parameters are dynamically adjusted, and the output signal is sent to the upstream pneumatic proportional valve. Experiments show that this algorithm is more effective than traditional PID in suppressing pressure fluctuations caused by rapid switching of the ALD pulse valve and reducing the amplitude of pressure fluctuations.

[0116] Thermal field control strategy implementation: Precise temperature control is achieved using a Model Predictive Control (MPC) algorithm. First, the aerodynamic-thermal field coupling model is linearized and simplified to obtain the state-space equations for real-time control, which serve as the internal predictive model for MPC. Within each control cycle (e.g., 200ms), the MPC controller receives stable flow and pressure data as feedforward inputs and temperature sensor data as feedback. Based on the model, the controller predicts the heater temperature response over the next 10 steps and calculates the optimal heating power sequence that most smoothly and quickly tracks the target temperature of 150℃ by solving a constrained quadratic programming problem. This sequence is immediately output to the solid-state relays (SSRs) of each zone heater. This "feedforward + feedback" mode reduces the system thermal response time from 9 seconds to 1.5 seconds, perfectly matching the fast cyclic timing of ALD.

[0117] In some embodiments, the disturbance region includes at least the final confluence pipe section before multiple gases enter the process reaction chamber. When the gas flow disturbance causes the pipe wall temperature in the disturbance region to rise instantaneously, the heat is conducted to the phase change material through the pipe wall, causing it to undergo a solid-liquid phase change and absorb a large amount of latent heat, thereby suppressing the gas temperature spike. At the same time, the nanofluid flowing through the microchannel continuously carries away the heat stored in the phase change material through forced convection, allowing it to periodically solidify and regenerate, thereby providing sustainable buffering.

[0118] In some embodiments, the phase change material is a paraffin-based composite material, which is encapsulated in a microchannel heat exchanger in the form of microencapsulation or in combination with a high thermal conductivity metal foam; the nanofluid is an Al2O3-water-based fluid.

[0119] In this embodiment, to cope with extreme instantaneous disturbances (such as a sudden failure of a gas valve causing a drastic change in flow), a passive buffer unit is integrated on the outer wall of the final gas confluence pipe section where the airflow disturbance is most severe.

[0120] Unit Design and Integration: A microchannel heat exchange unit was designed, filled with paraffin-based composite phase change material (PCM). To address the poor thermal conductivity of pure PCM, PCM was combined with high-porosity copper metal foam. The shell contains a network of microchannels with triangular cross-sections, within which Al2O3 nanofluid (1.0% by volume) circulates as a forced convection working fluid.

[0121] Working principle: When an unexpected transient thermal shock occurs, causing the wall temperature of the manifold section to rise instantaneously above the PCM phase transition point, the PCM rapidly melts, absorbing a large amount of latent heat and preventing the formation of gas temperature spikes. Simultaneously, the circulating nanofluid efficiently carries away the heat stored in the PCM, allowing it to re-solidify during process breaks and restore its buffering capacity. This unit acts as a "safety net" for the control system, suppressing transient temperature spikes that cannot be completely suppressed by closed-loop control within ±0.5℃.

[0122] In some embodiments, the thermal field equalization performance is evaluated using a digital twin, and the real-time monitoring and self-optimization stage of the thermal field temperature further includes the following steps:

[0123] Multi-source data real-time acquisition steps: Real-time acquisition of multi-dimensional operational data including gas temperature, pressure, mass flow rate and heating element power of pipeline cross-section; at the same time, acquisition of the currently executed process recipe information, including target temperature, gas flow rate setpoint and process step sequence;

[0124] Digital twin construction steps: Based on the aerodynamic-thermal field coupling model, a digital twin of the gas delivery system is constructed; the digital twin is synchronized with the physical system in real time through a data interface, and receives real-time data collected in the multi-source data real-time acquisition step as boundary conditions and initial conditions to drive simulation calculations, and virtually reproduces the airflow state and three-dimensional temperature field distribution of the physical system in real time.

[0125] Performance evaluation and anomaly detection steps: Use a digital twin to calculate and output target performance indicators, including at least the temperature non-uniformity of the reaction chamber inlet section, the maximum axial temperature gradient of the pipeline, and the system thermal response time; set a safety threshold for the target performance indicators. When any performance indicator calculated in real time deviates from its threshold range for more than a preset time, or when abnormal sensor data is detected, it is determined that the system performance has deviated or the equipment status has changed, triggering the self-optimization process.

[0126] Self-optimization steps: Execute control parameter optimization procedures and / or model optimization procedures to ensure that the gas delivery system always maintains its current optimal operating state for the current equipment status and process requirements.

[0127] In this embodiment, the real-time monitoring and self-optimization process of the thermal field temperature is as follows: Figure 6 As shown. To achieve long-term performance stability, a digital twin system is built to endow the device with self-sensing and self-optimization capabilities.

[0128] The control parameter optimization process includes the following steps: when the performance deviation is mainly caused by the mismatch of adjustable control parameters (such as the control rules of fuzzy PID and the weight matrix of MPC), a parameter optimization algorithm based on Bayesian optimization is started; the parameter optimization algorithm uses a digital twin as a simulation evaluator, takes the comprehensive score of the target performance index as the objective function, and performs an iterative search within a preset parameter space to find a new set of control parameters that makes the performance index return to the optimal range, and updates it online.

[0129] The model optimization process includes the following steps: When performance deviations are caused by system characteristic drift (such as aging of heating elements or changes in heat transfer coefficient due to pipe fouling), resulting in inaccurate prediction models of the digital twin, the online model calibration process is initiated; using the error between actual operating data and digital twin prediction data within a historical operating time range (a recent period, which can be preset or manually set), the key physical parameters (such as convective heat transfer coefficient and material thermal conductivity) in the aerodynamic-thermal field coupling model are dynamically corrected through recursive least squares or neural network online learning algorithms, so that the prediction accuracy of the digital twin remains consistently high.

[0130] The control parameter optimization process and the model optimization process are not executed in isolation, but rather through a higher-level coordinator. This higher-level coordinator has a built-in rule engine that, based on the type and magnitude of performance deviation and historical optimization records, decides to initiate the control parameter optimization process, the model optimization process, or a combination of both. When a slight, slow performance drift is detected, the control parameter optimization process is initiated first. When a sudden performance change is detected or the parameter optimization effect does not meet expectations, the model optimization process is initiated. After optimization, the new parameters or model will be loaded before the start of the next process batch to ensure that the gas delivery system always maintains the optimal operating point for the current equipment status and process requirements.

[0131] The digital twin employs a reduced-order model technique to improve real-time simulation speed. Its core is a low-dimensional state-space model extracted from the aerodynamic-thermal field coupling model through intrinsic orthogonal decomposition or dynamic mode decomposition. This reduced-order model can complete a single thermal field prediction calculation within milliseconds, meeting the timing requirements of real-time monitoring. The synchronization period between the digital twin and the physical system is no greater than the control period of the model predictive control algorithm in the dual-domain collaborative dynamic control stage, ensuring the consistency between the virtual state and the physical state.

[0132] Digital Twin Construction and Synchronization: Based on the coupled model, the intrinsic orthogonal decomposition (POD) method is used to reduce the order, resulting in a low-dimensional state-space model that can run on an industrial PC in milliseconds, serving as the core of the high-fidelity digital twin. This twin synchronizes with the sensors and actuators of the physical system in real time via the OPC UA protocol, reproducing the three-dimensional temperature field of the entire gas delivery system in real time in virtual space.

[0133] Performance Evaluation and Anomaly Decision-Making: The digital twin calculates key performance indicators (KPIs) in real time, such as reaction chamber inlet temperature non-uniformity, maximum axial temperature gradient, and thermal response time. A safety threshold is set for each KPI (e.g., non-uniformity <1.5%). After one month of system operation, if the twin detects that the "thermal response time" indicator is slowly drifting from 1.5 seconds to 2.2 seconds, consistently exceeding the threshold, the upper-level coordinator initiates a control parameter optimization process based on a rule engine (e.g., "slow drift priority parameter tuning").

[0134] Bayesian optimization self-optimization execution: The control parameter optimization process is initiated. Using a digital twin as the simulator, and the weighted sum of "thermal response time" and "temperature overshoot" as the optimization objective function, the rule table weights of the fuzzy PID controller and the prediction time domain of the MPC are used as optimization variables. The Bayesian optimization (BO) algorithm is employed to intelligently search within a pre-defined parameter space. After approximately 50 iterations (all completed in virtual space, without affecting actual production), BO finds a new set of optimal control parameters. Before the next process batch begins, the system automatically downloads the new parameters to the PLC controller. After optimization, the thermal response time recovers to 1.6 seconds, and the system returns to its optimal operating state.

[0135] Online model calibration: If the performance deviation is caused by changes in the heat transfer coefficient due to heater aging, the BO optimization effect will be limited, and the coordinator will decide to initiate the model optimization process. Utilizing the error between recent actual temperature data and twin prediction data, the key convective heat transfer coefficients in the twin's internal model are dynamically corrected online using the recursive least squares (RLS) method, ensuring that the prediction model maintains high accuracy and providing a reliable foundation for subsequent control optimization.

[0136] In some embodiments, the semiconductor thin film deposition process is an atomic layer deposition process.

[0137] In some embodiments, the semiconductor thin film deposition process is a chemical vapor deposition process or a metal-organic chemical vapor deposition process.

[0138] A second aspect of the present invention provides: a thermal field temperature equalization control system based on aerodynamic layout optimization, used to implement any of the above-mentioned thermal field temperature equalization control methods based on aerodynamic layout optimization, comprising:

[0139] The aerodynamic-thermal field coupled modeling and quantitative analysis module is used to construct an aerodynamic-thermal field coupled model based on the pipeline structure of the gas delivery system by using computational fluid dynamics and finite element analysis for joint simulation. The aerodynamic-thermal field coupled model is used to quantitatively analyze the influence of pipeline geometric features on airflow distribution and thermal field uniformity, identify airflow vortices, stagnation zones and temperature gradient exceeding the standard areas to obtain pipeline analysis results.

[0140] The aerodynamic layout optimization design module is used to physically optimize the pipeline structure based on pipeline analysis results, thereby reducing airflow disturbance and thermal unevenness.

[0141] The dual-domain collaborative dynamic control module is used to deploy a sensor network in the optimized pipeline structure and execute pneumatic control strategy and thermal field control strategy; the pneumatic control strategy is used to dynamically adjust pneumatic components to stabilize airflow; the thermal field control strategy is used to pre-adjust heating power in a feedforward manner to compensate for thermal hysteresis.

[0142] The local phase change energy storage buffer module is used to embed a microchannel heat exchanger filled with phase change material and built with nanofluid in the disturbance area that still has the risk of instantaneous thermal disturbance after dual-domain collaborative dynamic control. It uses the latent heat of phase change to absorb or release instantaneous thermal shock and suppress temperature peaks caused by airflow disturbance.

[0143] The real-time monitoring and self-optimization module for thermal field temperature is used to monitor the operating status of the gas transport system in real time, evaluate the thermal field equalization performance, and automatically trigger the control parameter optimization process or model optimization process when performance deviation or equipment status change is detected, so as to keep the gas transport system in the current optimal working state.

[0144] A third aspect of the present invention provides: a computer-readable storage medium storing computer-executable instructions, wherein when the computer-executable instructions are loaded and executed by a processor, the above-described thermal field temperature equalization control method based on aerodynamic layout optimization is implemented.

[0145] A fourth aspect of the present invention provides: a computer program product containing instructions, which, when run on a terminal, causes the terminal to execute any of the above-described thermal field temperature equalization control methods based on aerodynamic layout optimization.

[0146] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A thermal field temperature equalization control method based on aerodynamic layout optimization, characterized in that: A gas delivery system for semiconductor thin film deposition processes includes the following steps: Aerodynamic-thermal coupling modeling and quantitative analysis stage: Based on the pipeline structure of the gas delivery system, computational fluid dynamics and finite element analysis are used for joint simulation to construct an aerodynamic-thermal coupling model; the aerodynamic-thermal coupling model is used to quantitatively analyze the influence of pipeline geometric characteristics on airflow distribution and thermal field uniformity, and identify airflow vortices, stagnation zones and temperature gradient exceeding the standard areas to obtain pipeline analysis results; Aerodynamic layout optimization design stage: Based on the pipeline analysis results, the pipeline structure is physically optimized to reduce airflow disturbance and thermal unevenness; Dual-domain collaborative dynamic control stage: In the optimized pipeline structure, a sensor network is deployed, and a pneumatic control strategy and a thermal field control strategy are executed; the pneumatic control strategy is used to dynamically adjust pneumatic components to stabilize airflow; the thermal field control strategy is used to pre-adjust heating power in a feedforward manner to compensate for thermal hysteresis; the pneumatic control strategy is based on a fuzzy PID control algorithm; the thermal field control strategy is based on a model predictive control algorithm. The dual-domain collaborative dynamic control stage further includes the following steps: Sensor network deployment steps: Deploy a multi-type sensor network at target locations in the optimized piping structure. The target locations in the optimized piping structure include at least the downstream of each gradually curvature bend, the outlets of each branch of the symmetrical splitter, and the inlet of the process reaction chamber. The multi-type sensor network includes at least a pressure sensor and a mass flow controller for monitoring airflow status, and a temperature sensor for monitoring thermal field temperature. The pneumatic control strategy implementation steps are as follows: real-time pressure and flow data collected by pressure sensors and mass flow controllers are used as inputs. The opening degree of the upstream pneumatic proportional valve or mass flow controller is dynamically adjusted by a fuzzy PID control algorithm. The fuzzy PID controller takes the deviation between the pressure setpoint and the measured value and its rate of change as two-dimensional inputs. It infers online through a preset fuzzy rule base and outputs the correction amount of the proportional, integral and derivative parameters to suppress pipeline pressure fluctuations. The thermal field control strategy implementation steps are as follows: using the stabilized flow and pressure data as feedforward input and temperature sensor data as feedback, a model predictive control algorithm is used for rolling optimization. The model predictive control algorithm is based on the simplified state-space equation of the aerodynamic-thermal field coupling model as the internal prediction model. In each control cycle, it predicts the temperature response of each zone of the wound heater in multiple future steps. By solving the constrained optimization problem, it calculates the optimal heating power sequence that makes the predicted temperature trajectory smoothest and closest to the target temperature. The control quantity is output to the actuator to compensate for the system thermal hysteresis, thereby shortening the overall thermal response time. Dual-domain collaboration and signal interaction steps: The aerodynamic control strategy and the thermal field control strategy run in parallel on the same industrial control platform and interact with each other through shared memory or real-time communication bus; the model predictive control algorithm incorporates the stabilized flow and pressure data as feedforward disturbances into the internal prediction model, so as to adjust the heating power in advance before the expected change in airflow state, thereby realizing the collaborative control of aerodynamics and thermal field. Local phase change energy storage buffer stage: For the disturbance region that still has the risk of instantaneous thermal disturbance after dual-domain collaborative dynamic control, a microchannel heat exchanger filled with phase change material and built with nanofluid is embedded to absorb or release instantaneous thermal shock by utilizing the latent heat of phase change, thereby suppressing the temperature peak caused by airflow disturbance. Real-time monitoring and self-optimization stage of thermal field temperature: This stage involves real-time monitoring of the gas delivery system's operating status, evaluation of thermal field equalization performance, and automatic triggering of control parameter optimization or model optimization processes when performance deviations or equipment status changes are detected, ensuring the gas delivery system remains in its optimal operating state. The evaluation of thermal field equalization performance using a digital twin also includes the following steps: Multi-source data real-time acquisition steps: Real-time acquisition of multi-dimensional operational data including gas temperature, pressure, mass flow rate and heating element power of pipeline cross-section; at the same time, acquisition of the currently executed process recipe information, including target temperature, gas flow rate setpoint and process step sequence; Digital twin construction steps: Based on the aerodynamic-thermal field coupling model, a digital twin of the gas delivery system is constructed; the digital twin is synchronized with the physical system in real time through a data interface, and receives real-time data collected in the multi-source data real-time acquisition step as boundary conditions and initial conditions to drive simulation calculations, and virtually reproduces the airflow state and three-dimensional temperature field distribution of the physical system in real time. Performance evaluation and anomaly detection steps: Use a digital twin to calculate and output target performance indicators, including at least the temperature non-uniformity of the reaction chamber inlet section, the maximum axial temperature gradient of the pipeline, and the system thermal response time; set a safety threshold for the target performance indicators. When any performance indicator calculated in real time deviates from its threshold range for more than a preset time, or when abnormal sensor data is detected, it is determined that the system performance has deviated or the equipment status has changed, triggering the self-optimization process. Self-optimization steps: Execute control parameter optimization process and / or model optimization process to ensure that the gas delivery system always maintains the current optimal operating state for the current equipment state and process requirements; The control parameter optimization process includes the following steps: when the performance deviation is mainly caused by the mismatch of adjustable control parameters, a parameter optimization algorithm based on Bayesian optimization is started; the parameter optimization algorithm uses a digital twin as a simulation evaluator, takes the comprehensive score of the target performance index as the objective function, performs an iterative search in a preset parameter space, finds a new set of control parameters that makes the performance index return to the optimal range, and updates it online. The model optimization process includes the following steps: when the performance deviation is caused by the drift of system characteristics, resulting in the inaccuracy of the prediction model of the digital twin, the online calibration process of the model is initiated; using the error between the actual running data and the prediction data of the digital twin within the historical running time range, the physical parameters in the aerodynamic-thermal field coupling model are dynamically corrected through recursive least squares method or neural network online learning algorithm, so that the prediction accuracy of the digital twin is kept at a high level. The execution of the control parameter optimization process and the model optimization process is decided by the upper-level coordinator: the upper-level coordinator has a built-in rule engine, which decides to start the control parameter optimization process, the model optimization process, or a combination of both based on the type, magnitude, and historical optimization records of the performance deviation; when a slight, slow performance drift is detected, the control parameter optimization process is started first; when a sudden performance change is detected or the parameter optimization effect does not meet the expected effect, the model optimization process is started; after optimization is completed, the new parameters or model will be loaded before the start of the next process batch to ensure that the gas delivery system always maintains the optimal operating point for the current equipment status and process requirements.

2. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 1, characterized in that: The construction of the aerodynamic-thermal field coupling model includes the following steps: Parametric geometric modeling steps: Based on the pipeline structure of the gas delivery system, the geometric features that affect airflow and heat transfer are parametrically defined, including at least the radius of curvature of the elbow, the bending angle of the elbow, the branching angle of the splitter, and the interface transition radius of the splitter. Multiphysics mesh generation steps: Perform multi-domain mesh generation on the parameterized geometric model, where the fluid domain uses an unstructured tetrahedral mesh with boundary layer refinement to capture near-wall flow and heat transfer, and the solid domain uses a hexahedral mesh; Steps for setting boundary conditions and physical property parameters: Define simulation boundary conditions, including the mass of the gas inlet, the flow rate of the gas inlet, the temperature of the gas inlet, the pressure of the outlet, and the thermal boundary conditions between the pipe wall and the heater; and set the physical property parameters of the process gas as functions of temperature. Two-way fluid-structure interaction solution: The computational fluid dynamics solver and the finite element analysis solver are used for co-simulation. In each time step or iteration step, the computational fluid dynamics solver transmits the calculated convective heat transfer coefficient at the interface between the fluid and the pipe wall to the finite element analysis solver, and the finite element analysis solver transmits the calculated pipe wall temperature field back to the computational fluid dynamics solver as a new thermal boundary. This process is repeated until convergence to obtain the aerodynamic-thermal field coupled model.

3. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 2, characterized in that: The process of performing quantitative analysis using a coupled aerodynamic-thermal model includes the following steps: The steps for quantitative analysis of airflow uniformity are as follows: extract the velocity cloud map and streamline map of the pipeline section, identify and quantify the location, size and intensity of the vortex region; calculate the local airflow residence time caused by the structure, which is the difference between the average time of fluid particles passing through the target local area and the time of passage through the mainstream area; Steps for quantitative analysis of thermal field uniformity: Draw temperature distribution cloud maps of the pipeline along the axial and radial directions, and calculate the axial temperature gradient and radial temperature non-uniformity of the target section from the gas heating start point to the inlet of the reaction chamber. Thermal dynamic response characteristic analysis steps: In the simulation, set a step change in heating power or a step change in inlet flow rate to simulate transient conditions, calculate the time required for the system to reach the preset percentage of the new steady-state value from the start of the step change, and define it as the system thermal hysteresis time. Parameter impact analysis steps: Continuously change the geometric parameters of the pipeline structure geometric model and re-analyze to establish a quantitative relationship graph between geometric parameters and airflow uniformity index, temperature gradient, and system thermal lag time, providing optimization objectives and design basis for the aerodynamic layout optimization design stage.

4. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 2, characterized in that: The solid domain includes the tube wall and the heating element; the thermal boundary conditions between the tube wall and the heater include constant heat flux density and convective heat transfer coefficient; the physical properties of the process gas include density, viscosity, specific heat capacity and thermal conductivity.

5. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 3, characterized in that: The pipeline cross-section includes the cross-section downstream of each bend, before and after the distributor, and at the inlet of the reaction chamber.

6. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 1, characterized in that: The aerodynamic layout optimization design phase also includes the following steps: Elbow structure optimization steps: Replace right-angle elbows in the gas delivery pipeline with elbows with gradually decreasing curvature to eliminate flow separation and Dean vortices caused by sudden changes in airflow direction, and reduce local drag coefficient and energy loss. Optimization steps for split / merge structure: Optimize the asymmetric split or merge nodes in the gas delivery pipeline into a symmetrical structure with a streamlined transition; for splitters that supply gas to multiple process chambers or heating zones, adopt a Y-type or improved T-type symmetrical design, and set a DC stabilization section upstream of the splitter to ensure that the airflow is fully developed and uniform, and reduce the flow distribution deviation of each branch. Pipeline inner surface and connection optimization steps: Electropolishing is applied to the inner wall of the pipeline, and the interface at the pipeline connection is set to a structure with a flush inner wall and no step abrupt change, in order to minimize flow friction resistance and prevent secondary flow and particle deposition caused by abrupt change in inner diameter.

7. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 6, characterized in that: The aerodynamic layout optimization design stage also includes local flow guidance and rectification steps: for areas where there is still a risk of uneven airflow, fixed guide vanes or rectification grids are added inside the pipe. The installation angle and number of the guide vanes or rectification grids are determined based on the streamline analysis results of the aerodynamic-thermal field coupling model, which are used to further guide the airflow, break up large-scale vortices, and achieve uniformity of the velocity field.

8. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 1, characterized in that: The disturbance region includes at least the final confluence pipe section before multiple gases enter the process reaction chamber. When the airflow disturbance causes the pipe wall temperature in the disturbance region to rise instantaneously, the heat is conducted to the phase change material through the pipe wall, causing it to undergo a solid-liquid phase change and absorb a large amount of latent heat, thereby suppressing the gas temperature spike. At the same time, the nanofluid flowing through the microchannel continuously carries away the heat stored in the phase change material through forced convection, allowing it to periodically solidify and regenerate, thus providing sustainable buffering.

9. The thermal field temperature equalization control method based on aerodynamic layout optimization according to claim 8, characterized in that: The phase change material is a paraffin-based composite material, which is encapsulated in a microchannel heat exchanger in the form of microencapsulation or in combination with a high thermal conductivity metal foam; the nanofluid is an Al2O3-water-based fluid.

10. The thermal field temperature equalization control method based on aerodynamic layout optimization according to any one of claims 1-9, characterized in that: The semiconductor thin film deposition process is an atomic layer deposition process.

11. The thermal field temperature equalization control method based on aerodynamic layout optimization according to any one of claims 1-9, characterized in that: The semiconductor thin film deposition process is a chemical vapor deposition process or a metal-organic chemical vapor deposition process.

12. A thermal field temperature equalization control system based on aerodynamic layout optimization, characterized in that: The method for implementing thermal field temperature equalization control based on aerodynamic layout optimization as described in any one of claims 1-11 includes: The aerodynamic-thermal field coupled modeling and quantitative analysis module is used to construct an aerodynamic-thermal field coupled model based on the pipeline structure of the gas delivery system by using computational fluid dynamics and finite element analysis for joint simulation. The aerodynamic-thermal field coupled model is used to quantitatively analyze the influence of pipeline geometric features on airflow distribution and thermal field uniformity, identify airflow vortices, stagnation zones and temperature gradient exceeding the standard areas to obtain pipeline analysis results. The aerodynamic layout optimization design module is used to physically optimize the pipeline structure based on pipeline analysis results, thereby reducing airflow disturbance and thermal unevenness. The dual-domain collaborative dynamic control module is used to deploy a sensor network in the optimized pipeline structure and execute pneumatic control strategy and thermal field control strategy; the pneumatic control strategy is used to dynamically adjust pneumatic components to stabilize airflow; the thermal field control strategy is used to pre-adjust heating power in a feedforward manner to compensate for thermal hysteresis. The local phase change energy storage buffer module is used to embed a microchannel heat exchanger filled with phase change material and built with nanofluid in the disturbance area that still has the risk of instantaneous thermal disturbance after dual-domain collaborative dynamic control. It uses the latent heat of phase change to absorb or release instantaneous thermal shock and suppress temperature peaks caused by airflow disturbance. The real-time monitoring and self-optimization module for thermal field temperature is used to monitor the operating status of the gas transport system in real time, evaluate the thermal field equalization performance, and automatically trigger the control parameter optimization process or model optimization process when performance deviation or equipment status change is detected, so as to keep the gas transport system in the current optimal working state.

Images