BUB bubble conveying system for semiconductor production process and control method

Abstract

The invention discloses a BUB bubble conveying system for a semiconductor production process and a control method, and belongs to the technical field of bubble conveying control, and the method specifically comprises the steps: in a fluid conveying pipeline of the semiconductor production process, generating a bubble group through a bubble generator, and adjusting the bubble generation frequency and volume distribution based on a liquid impedance feedback signal, the flow velocity, pressure and bubble distribution parameters of process liquid are collected, a pre-trained flow field prediction model is used for analyzing the bubble gathering trend, when the transmission retardation risk is detected, the opening degree of a micro valve is automatically adjusted, a bubble path is shunted, and a micro-pulse air pressure signal is applied to the throttling section of the semiconductor production process fluid conveying pipeline. Dynamically matching the bubble rupture phase with the fluid viscosity, and switching to a reverse bubbling mode to perform self-cleaning on the pipeline when detecting that the bubble residue exceeds a threshold value; linkage regulation and control of the multi-node micro valve are achieved, and the uniformity, repeatability and safety of fluid conveying can be kept in long-term operation.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of bubble delivery control, and particularly relates to a BUB bubble delivery system and a control method for a semiconductor production process. BACKGROUND

[0002] In a semiconductor manufacturing process, chemical vapor deposition, wet etching, cleaning and CMP (chemical mechanical polishing) and other processes generally rely on the accurate delivery and mixing control of high-purity chemical liquids. Traditional liquid delivery systems mostly use constant-flow pumps or pressure-driven modes for fluid transmission, but in high-precision processes, micro-bubble retention, gas-liquid stratification and flow field instability and other phenomena easily occur at the fluid interface, resulting in fluctuations in the flow resistance of process liquids, thereby affecting the uniformity of the film layer and the particle control level. To alleviate this problem, some systems introduce a BUB bubble mechanism to maintain the dynamic stability of the liquid, but its control still stays at the fixed frequency and simple throttle adjustment stage, lacking real-time coordination of bubble characteristics and fluid state.

[0003] Existing BUB systems mostly generate bubbles and maintain flow field circulation through a single pressure source, and cannot achieve adaptive regulation of bubble generation frequency, volume distribution and rupture phase. When the production line is running under different temperature, pressure and fluid viscosity conditions, uneven bubble distribution or aggregation phenomenon easily causes transmission resistance and local pollution, especially in sub-micron process nodes, any small disturbance may cause yield fluctuations. In addition, the traditional control method cannot accurately model the complex coupled dynamics of the gas-liquid two-phase, and lacks intelligent prediction and multi-dimensional feedback regulation mechanism. SUMMARY

[0004] In view of the deficiencies of the prior art, the application provides a BUB bubble delivery system and control method for a semiconductor production process, which adaptively adjusts the bubble generation, splitting and rupture rhythm under dynamic process conditions, realizes fine control of the gas-liquid phase interface behavior, and provides a new fluid delivery control idea and implementation approach for semiconductor processes.

[0005] To achieve the above-mentioned purposes, the application provides the following technical solutions: A BUB bubble delivery control method for a semiconductor production process, comprising: In the semiconductor production process fluid delivery pipeline, a bubble group is generated by a bubble generator, and the bubble generation frequency and volume distribution are adjusted based on the liquid impedance feedback signal; The process liquid flow rate, pressure and bubble distribution parameters are collected, and a pre-trained flow field prediction model is used to analyze the bubble aggregation trend, and when the transmission resistance risk is detected, the micro-valve opening degree is automatically adjusted to split the bubble path; A micro-pulse gas pressure signal is applied to the throttling section of the fluid delivery pipeline in semiconductor manufacturing process to dynamically match the bubble burst phase with the fluid viscosity; When the detected residual bubbles exceed the threshold, switch to reverse bubbling mode to self-clean the pipeline.

[0006] Specifically, in the semiconductor manufacturing process fluid transport pipeline, a bubble swarm is generated by a bubble generator, and the bubble generation frequency and volume distribution are adjusted based on a liquid impedance feedback signal, including: Inject process fluid of predetermined purity into the semiconductor manufacturing process fluid delivery pipeline, and simultaneously establish an initial reference value for liquid resistance at the inlet of the bubble generator; The array of holes in the bubble generator is opened sequentially, and multi-point nucleation is triggered by the disturbance of the stacked gas-liquid interface, forming a bubble group on the same process fluid cross section. Impedance monitoring nodes are set up in the bubble nucleation region and the downstream process fluid section to collect complex impedance signals in the gas-liquid mixing state in real time, and extract their phase difference and amplitude characteristics as impedance feedback parameters. The frequency of bubble generation is automatically adjusted based on changes in impedance feedback parameters, and the bubble generation rhythm is adaptively controlled. During the bubble transport stage, bubbles of different sizes are grouped and identified, and the opening sequence of the array-type holes is adjusted based on the impedance feedback parameters of the previous cycle, and then updated in the next cycle.

[0007] Specifically, the automatic adjustment of the bubble generation frequency based on changes in impedance feedback parameters, and the adaptive control of the bubble generation rhythm, includes: During the bubble generation cycle, the impedance feedback parameters are divided into multiple stages according to the time series, and the signal data of each stage are collected in sequence to form a dynamic sequence reflecting the gas-liquid coupling state. The dynamic sequence is periodically compared, and the process fluid response type corresponding to the impedance change trend is determined by judging the direction and amplitude changes of signal differences between adjacent stages. Based on the process fluid response type, a matching adjustment path is selected within a preset frequency control domain, and a set of correction instructions containing frequency offset and phase synchronization parameters is generated. Based on the aforementioned modified instruction set, the bubble generation rhythm is adaptively controlled.

[0008] Specifically, the process involves collecting parameters such as liquid flow rate, pressure, and bubble distribution. A pre-trained flow field prediction model is used to analyze bubble aggregation trends. When a risk of transport obstruction is detected, the micro-valve opening is automatically adjusted to divert the bubble path, including: Monitoring nodes are set at different locations in the fluid delivery pipeline of semiconductor manufacturing process to collect flow rate, pressure and bubble distribution parameters, and form a time-synchronized multidimensional data stream. The multidimensional data stream is subjected to time-series arrangement and normalization processing to convert it into a continuous feature sequence representing the gas-liquid flow state, and the flow field disturbance section is marked. The continuous feature sequence is input into a pre-trained flow field prediction model. Based on the model output, the spatiotemporal positional relationship of bubble aggregation is identified, and a determination index for potential transport obstruction zones is generated. Based on the determination index and combined with the operating parameters of the current semiconductor manufacturing process fluid delivery pipeline, the corresponding flow path is selected in the virtual valve control matrix, and an adjustment command sequence for controlling the micro-valve opening is generated. The adjustment command sequence is sequentially transmitted to each microvalve control node, and the opening adjustment is performed according to the preset timing to divert the bubble path.

[0009] Specifically, the continuous feature sequence is input into a pre-trained flow field prediction model, and the spatiotemporal positional relationship of bubble aggregation is identified based on the model output results. A determination index for potential transport obstruction zones is generated, including: The continuous feature sequence is sequentially input into each level of the analytical layer of the pre-trained flow field prediction model. At each level, the coupling mode between the process fluid state and bubble aggregation is extracted, and an intermediate representation vector is generated. By performing correlation matching on the intermediate representation vector, the aggregation center of the bubble in the spatial distribution and its adjacent diffusion range are located, forming a spatiotemporal coordinate set representing the aggregation dynamics. Based on the continuous change relationship of the spatiotemporal coordinate group, the identification parameter corresponding to the bubble density threshold is calculated, and the identification parameter is used as a basis to establish a judgment index for potential transmission hindrance zones.

[0010] Specifically, based on the determination index and the operating parameters of the current semiconductor manufacturing process fluid delivery pipeline, a corresponding flow path is selected in the virtual valve control matrix, and a sequence of adjustment commands for controlling the micro-valve opening is generated, including: The operating parameters of the fluid delivery pipeline in the current semiconductor manufacturing process are obtained and synchronously formed into an operating status benchmark. The operating parameters include flow load, pressure gradient and valve control response delay. The judgment index is correlated and compared with the running status benchmark. A virtual valve control matrix is ​​constructed based on the topological relationship of the fluid delivery pipeline in the semiconductor manufacturing process, and the set of bubble diversion paths is identified in the virtual valve control matrix. The set of diversion paths is optimized by calculation, a path priority sequence is generated according to the principle of fluid dynamic balance, and then converted into a corresponding valve control operation mapping table. Based on the valve control operation mapping table, a sequence of adjustment commands for controlling the opening degree of the micro-valve is generated and distributed to the corresponding valve control nodes through hierarchical transmission.

[0011] Specifically, the adjustment command sequence is sequentially transmitted to each microvalve control node, and the opening adjustment is executed according to a preset timing sequence to divert the bubble path, including: The adjustment command sequence is timestamped and numbered to establish an execution queue consistent with the reference clock, and the node target and opening target value range are marked for each adjustment command. The execution queue is divided into multiple segments according to a preset time slot window. An assignment list is generated based on the node mapping relationship of the virtual valve control matrix to determine the transmission order and acknowledgment path of each adjustment command on the micro-valve control node. Based on the transmission order and acknowledgment path, each microvalve control node adjusts its opening degree and records the feedback information, including the execution frame identifier and local sampling snapshot. The returned information is compared frame by frame with the execution queue. Based on the comparison results, the next round of dispatch list and time slot window are updated. Unexecuted or deviated instructions are resent and the queue positions are rearranged.

[0012] Specifically, the step of applying a micro-pulse gas pressure signal to the throttling section of the semiconductor manufacturing process fluid delivery pipeline to dynamically match the bubble burst phase with the fluid viscosity includes: After a stable fluid channel is formed in the throttling section of the semiconductor manufacturing process fluid delivery pipeline, process fluid data within the throttling section is collected, and an initial reference table for gas pressure pulse modulation is established based on the fluid data. The process fluid data includes fluid density, viscosity, and flow rate variation data. Under the constraints of the initial reference table, a micro-pulse pressure signal containing a main pulse and an auxiliary pulse is generated; Before the micro-pulse air pressure signal is applied, the fluid viscosity change range is continuously monitored, and the time phase of bubble rupture is mapped and compared with the viscosity change range to generate a synchronous trigger timetable; Micro-pulse air pressure signals are applied sequentially according to the synchronous triggering timetable, and fluid parameters are re-acquired and the initial reference table is updated after each cycle.

[0013] Specifically, before the micro-pulse gas pressure signal is applied, the fluid viscosity change range is continuously monitored, and the time phase of bubble rupture is mapped and compared with the viscosity change range to generate a synchronous triggering timetable, including: Set a sampling period, collect instantaneous change data of fluid viscosity in a time series, and organize the instantaneous change data into piecewise curves; During the monitoring period of bubble generation and transport, the time series of bubble rupture events are recorded, and each rupture moment is mapped as a phase node to form a phase trajectory table representing the bubble rupture rhythm. The phase trajectory table and the piecewise curve are compared in time, and the interrelated segments are calculated based on the overlapping intervals of the two, and the correspondence between the bubble bursting rhythm and viscosity change is determined. Based on the aforementioned correspondence, a synchronization triggering timetable is constructed.

[0014] A BUB bubble conveying system for semiconductor manufacturing processes, comprising a bubble conveying control method for semiconductor manufacturing processes, including: a bubble generation module, an adjustment module, a dynamic matching module, and a self-cleaning module; The bubble generation module is used to generate a bubble group in the fluid transport pipeline of semiconductor manufacturing process through a bubble generator, and to adjust the bubble generation frequency and volume distribution based on the liquid impedance feedback signal. The adjustment module is used to collect process liquid flow rate, pressure and bubble distribution parameters, and to analyze bubble aggregation trend using a pre-trained flow field prediction model. When a risk of transmission obstruction is detected, the micro-valve opening is automatically adjusted to divert the bubble path. The dynamic matching module is used to apply a micro-pulse air pressure signal to the throttling section of the fluid transport pipeline in semiconductor manufacturing process to dynamically match the bubble burst phase with the fluid viscosity. The self-cleaning module is used to switch to reverse bubbling mode to self-clean the pipeline when the detected bubble residue exceeds the threshold.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention proposes a BUB bubbling conveying system and control method for semiconductor manufacturing processes. By constructing an adaptive control system based on liquid impedance feedback, intelligent prediction, and multi-dimensional pulse modulation, it achieves dynamic and coordinated adjustment of bubble generation frequency, volume distribution, flow field distribution, and rupture phase, thereby maintaining a stable gas-liquid interface conveying state under complex process conditions. The method automatically adjusts the bubbling rhythm and flow path when fluid parameters change in real time, avoiding local stagnation or particle carrying problems caused by bubble aggregation. By introducing a virtual valve control matrix and synchronous triggering logic, it achieves linkage control of multi-node micro-valves, making the bubble conveying path programmable and learnable, while possessing self-updating and self-correcting capabilities. It can maintain the uniformity, repeatability, and safety of fluid conveying during long-term operation. Compared with traditional fixed-frequency bubbling systems, this invention has significant improvements in fluid uniformity, response accuracy, and reliability. Attached Figure Description

[0016] Figure 1A flowchart of a BUB bubble conveying control method for semiconductor manufacturing processes provided by the present invention; Figure 2 This is a schematic diagram of path splitting provided by the present invention; Figure 3 This invention provides an architecture diagram of a BUB bubbling conveyor system for semiconductor manufacturing processes. Detailed Implementation

[0017] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] It should be noted that, unless there is conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be performed in a different order than the module division in the device or the order in the flowchart. In addition, the " The terms "first," "second," and "third" do not limit the data or execution order; they are merely used to distinguish identical or similar items with essentially the same function and purpose.

[0020] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0021] Example 1 Please see Figures 1-2 The present invention provides an embodiment of a BUB (bubble conveying) control method for semiconductor manufacturing processes, comprising the following specific steps: Step S1: In the fluid delivery pipeline of semiconductor manufacturing process, a bubble group is generated by a bubble generator, and the bubble generation frequency and volume distribution are adjusted based on the liquid impedance feedback signal.

[0022] The specific steps of step S1 are as follows: Step S101: Inject process fluid of predetermined purity into the semiconductor manufacturing process fluid delivery pipeline, and simultaneously establish an initial reference value of liquid impedance at the inlet of the bubble generator.

[0023] In this embodiment, based on the control requirements of the target process section for metal ion, particle concentration, and organic residue, a process fluid with a predetermined purity range is selected to ensure that the fluid reaches stable and repeatable physical properties before entering the delivery pipeline. Subsequently, a liquid impedance monitoring node is arranged at the pipeline position near the bubble generator inlet to continuously sample the impedance of the process fluid in the non-bubbling state for a period of time. By statistically processing the impedance amplitude, phase, and fluctuation range in different time slices, abnormal data points introduced by instantaneous flow disturbances or small temperature fluctuations are screened out, so that the remaining data converges to a narrow range distribution in a statistical sense. On this basis, the impedance characteristic set corresponding to the narrow range distribution is defined as the initial reference value of the liquid impedance of this batch of process fluid, and this reference value is associated with the pipeline label and process formula number for storage. In order to judge and quantify the influence of fluid state and bubble introduction on the flow field by the deviation of real-time impedance measurement from the initial reference during the bubble generation and control process.

[0024] Step S102: Sequentially open the array of holes in the bubble generator, and use the disturbance of the stacked gas-liquid interface to trigger multi-point nucleation, forming a bubble group on the same process fluid cross section.

[0025] In this embodiment, the bubble generator is equipped with an array of micropores arranged along the main axis of the fluid. The opening size, spacing, and angle of each micropore are pre-calibrated experimentally to ensure a controllable local gas-liquid interface response under different flow rates and pressures. During this step, the array of micropores is opened one group at a time according to a timing signal, causing the fluid to form a periodic interface disturbance superposition zone at the micropore outlet. The superposition effect of this disturbance causes the microscale pressure waves inside the liquid phase to superimpose and generate a local supersaturation zone, thereby inducing the instantaneous precipitation of the gas phase at multiple interfaces and achieving multi-point nucleation. Since the sequential opening of the micropores introduces a spatial phase delay, the bubble nucleation process forms a nonlinearly distributed bubble group on the pipe cross-section, which avoids single-point aggregation and maintains the continuity of the overall gas-liquid interface. By fine-tuning the micropore opening sequence, a stable bubble density and uniform volume distribution are obtained in the early stage of nucleation.

[0026] Step S103: Set up impedance monitoring nodes in the bubble nucleation region and the downstream process fluid section to collect complex impedance signals in the gas-liquid mixing state in real time, and extract their phase difference and amplitude characteristics as impedance feedback parameters.

[0027] In this embodiment, several impedance monitoring nodes are deployed along the bubble nucleation region and its downstream fluid channel. Each node forms a local measurement loop with the fluid medium via electrode coupling to record the electrical response signal of the gas-liquid mixture in the instantaneous flow state. After bubble generation, the conductive path in the fluid medium exhibits periodic disturbances, and its equivalent impedance changes with the bubble ratio and spatial distribution, resulting in phase drift and amplitude fluctuations. By performing multi-band modulation sampling on the acquired complex impedance signal and performing discrete averaging processing with a sliding time window, the gradual variation components caused by liquid viscosity, temperature, and flow velocity are separated, retaining only the high-frequency characteristics reflecting the dynamic behavior of the bubbles. Subsequently, the phase difference of different nodes is compared to determine the delay characteristics of bubble propagation along the pipeline, and the relative change trend of bubble swarm density is estimated by the amplitude variation. Through this data processing, the transient impedance signal is converted into a set of feedback parameters characterizing the gas-liquid coupling state.

[0028] Step S104: Automatically adjust the frequency of bubble generation based on the change of impedance feedback parameters to adaptively control the bubble generation rhythm.

[0029] The specific steps of step S104 are as follows: Step S1041: During the bubble generation cycle, the impedance feedback parameters are divided into multiple stages according to the time series, and the signal data of each stage are collected in sequence to form a dynamic sequence reflecting the gas-liquid coupling state.

[0030] In this embodiment, within the complete cycle of bubble generation and transport, a time reference is first established based on the average flow velocity of the process fluid and the bubble generation frequency. The entire cycle is then divided into several continuous stages, each corresponding to different flow dynamics of bubbles from nucleation and expansion to initial rupture. Subsequently, impedance acquisition units are synchronously triggered to sample signals within each stage to ensure the temporal continuity of the data. To avoid signal jitter caused by random bubble distribution, the sampled signals are processed using time-weighted filtering and window integration to ensure that the impedance parameters of each stage statistically represent a stable gas-liquid coupling state. By combining these stage parameters in chronological order, a dynamic sequence with phase characteristics is formed. This sequence not only reflects the generation rhythm and spatial propagation process of bubbles in the fluid but also serves as the input basis for subsequent frequency control and model prediction. Finally, by analyzing the change patterns of the dynamic sequence, key temporal information reflecting the transient flow characteristics of the process fluid is extracted, achieving a time-domain structured expression of the impedance feedback signal.

[0031] Step S1042: Periodically compare the dynamic sequence, and determine the process fluid response type corresponding to the impedance change trend by judging the direction and amplitude changes of signal differences between adjacent stages.

[0032] In this embodiment, for the impedance dynamic sequence formed in step S1041, the control unit calls the comparison algorithm after each bubble generation cycle to perform differential analysis on the signal amplitude and phase data between adjacent stages. By performing time series expansion on the signal change direction, change rate, and amplitude gradient of each stage, typical characteristic segments such as the rising region, the stable region, and the falling region of the impedance curve are identified. Subsequently, using these characteristic segments as indexes, the impedance change trend is matched with a preset process fluid response pattern library. This pattern library is obtained through multiple experimental calibrations and covers typical response types of fluids under different flow rates, pressures, and viscosity conditions. After matching, the process fluid state corresponding to the current impedance change is determined, such as changes in bubble nucleation density, enhanced gas-liquid interface disturbance, or flow field stagnation tendency. By periodically executing this comparison process, the stability and repeatability of the impedance trend are captured in the time dimension, thereby establishing a mapping relationship between gas-liquid coupling characteristics and process fluid response types.

[0033] Step S1043: Based on the process fluid response type, select a matching adjustment path within the preset frequency control domain, and generate a correction instruction set containing frequency offset and phase synchronization parameters.

[0034] In this embodiment, after identifying the process fluid response type, a preset frequency control domain parameter table is invoked. This parameter table is divided into several adjustment paths according to the correspondence between fluid impedance characteristics and bubble dynamic response. Each path corresponds to a set of adjustable bubble generation frequency and phase synchronization intervals. First, the path node closest to the current response type is retrieved, and the impedance offset trend of the previous cycle is compared to determine the direction and magnitude of frequency correction. Based on this, the phase correction parameter for synchronizing the bubble generation rhythm is generated by calculating the correspondence between impedance phase drift and fluid cycle delay. Subsequently, the frequency offset and phase correction parameter are combined and encoded into a set of correction instructions, and the execution order is marked with a timestamp to ensure dynamic closed-loop adjustment is achieved in the next cycle. Through this logic chain, the frequency control is flexibly corrected within the predetermined adjustable domain, keeping the bubble generation rhythm coordinated with the transient state of the fluid, and providing basic control instructions for subsequent closed-loop iterative updates.

[0035] Step S1044: Based on the modified instruction set, adaptively adjust the bubble generation rhythm.

[0036] In this embodiment, after obtaining the correction instruction set containing frequency offset and phase synchronization parameters, the parameters are loaded into the bubble generation control unit according to the time identifier sequence, so that the trigger interval, phase offset, and duty cycle of its output pulse signal are dynamically corrected according to the real-time feedback of the process fluid. During this process, the difference between the impedance change rate and the frequency before correction is calculated in real time within each bubble generation cycle to determine the degree of lead or lag in the gas-liquid interface response, and the timing distribution of the trigger signal is finely adjusted on a microsecond time scale accordingly, so that the rhythm of adjacent bubble generation events gradually tends to a dynamic equilibrium state that matches the fluid viscosity fluctuation. At the same time, to prevent the cumulative error of frequency adjustment, the control logic periodically compares the execution result with the target frequency of the previous cycle. When the frequency drift is detected to exceed the set range, the reference parameters in the instruction set are immediately called back for correction. Through the above iterative correction mechanism, the bubble generation rhythm achieves adaptive balance in the time domain, thereby forming a continuous, stable, and predictable bubble generation sequence.

[0037] Step S105: During the bubble transport stage, bubbles of different sizes are grouped and identified, and the opening sequence of the array-type holes is adjusted based on the impedance feedback parameters of the previous cycle, and then updated in the next cycle.

[0038] In this embodiment, after the bubbles enter the transport stage, the particle size distribution of the passing bubble group is identified by optical scattering nodes or acoustic monitoring nodes set in the middle of the pipeline. During the identification process, the bubbles are divided into several particle size intervals according to the variation law of scattering signal intensity or acoustic echo time delay, and the statistical distribution characteristics of the bubble group are output in the form of data sequence. Subsequently, the distribution characteristics are compared with the impedance feedback parameters obtained in the previous cycle. By analyzing the correspondence between impedance phase drift and bubble volume distribution, it is determined whether there is a deviation in the timing control of the bubble generation stage. If a specific particle size group is detected to be over-concentrated or unevenly distributed, the opening order of the array micropores is rearranged according to the comparison results so that the triggering time interval of each micropore in the next cycle matches the local viscosity range of the fluid. This adjustment strategy realizes closed-loop iterative optimization between bubble grouping characteristics and nucleation timing through continuous multi-cycle data association, so that the bubble group generated in subsequent cycles is more evenly distributed in space, thereby maintaining a stable gas-liquid coupling state during fluid transport.

[0039] Step S2: Collect process liquid flow rate, pressure and bubble distribution parameters, and use a pre-trained flow field prediction model to analyze the bubble aggregation trend. When a risk of transmission obstruction is detected, automatically adjust the micro-valve opening to divert the bubble path.

[0040] like Figure 2 As shown, the specific steps of step S2 are as follows: Step S201: Set up monitoring nodes at different locations in the semiconductor manufacturing process fluid delivery pipeline to collect flow rate, pressure and bubble distribution parameters, and form a time-synchronized multidimensional data stream.

[0041] In this embodiment, multiple types of monitoring nodes are deployed at key locations along the semiconductor process fluid transport pipeline. The layout of each node is calculated based on the fluid flow direction, cross-sectional changes, and throttling section distribution, ensuring that the acquired signals are spatially representative and temporally continuous. Each node is equipped with a flow velocity sensing unit, a pressure sampling module, and an optical or acoustic bubble distribution recognition module, and a unified time reference is achieved through a clock synchronization bus. During operation, each node acquires raw signal data with the same sampling period, and the sampling frequency is dynamically set according to the fluid Reynolds number range to ensure the capture of instantaneous disturbances at the gas-liquid interface. After preliminary filtering and timestamp calibration at local nodes, the acquired data is multidimensionally aligned according to the time series to form a multidimensional data stream containing the flow velocity change rate, pressure fluctuation trend, and bubble spatial distribution matrix.

[0042] Step S202: Perform time-series arrangement and normalization processing on the multidimensional data stream, convert it into a continuous feature sequence representing the gas-liquid flow state, and mark the flow field disturbance section.

[0043] In this embodiment, after obtaining multidimensional data streams from various monitoring nodes, the sampled data from different nodes are first arranged in time sequence according to a unified time reference. Interpolation and resampling are used to eliminate time misalignment caused by signal delay or sampling offset, ensuring that the three types of data—flow velocity, pressure, and bubble distribution—maintain a corresponding relationship on the same time axis. Subsequently, each type of data is normalized, converting signals with different dimensions into dimensionless feature values ​​so that subsequent model calculations can directly compare the dynamic change amplitudes between different physical quantities. An interval adaptive algorithm is used during the normalization process, allowing the feature value range to automatically expand or contract according to the current process state, thereby maintaining sensitivity to abnormal disturbances. The processed data is constructed into a continuous feature sequence according to the time series, with each time slice in the sequence containing composite indicators such as flow velocity gradient, pressure oscillation frequency, and bubble density ratio. To identify sections in the flow field where unstable behavior exists, the control logic calculates the volatility of the feature sequence within a sliding time window and marks time periods exceeding a threshold as flow field disturbance sections. In this way, a time-series feature sequence that reflects the evolution of gas-liquid flow states is formed.

[0044] Step S203: Input the continuous feature sequence into the pre-trained flow field prediction model, identify the spatiotemporal position relationship of bubble aggregation based on the model output results, and generate a determination index for potential transport obstruction zones.

[0045] The specific steps of step S203 are as follows: Step S2031: The continuous feature sequence is sequentially input into each level of the analytical layer of the pre-trained flow field prediction model. The coupling mode between the process fluid state and bubble aggregation is extracted in each layer, and an intermediate representation vector is generated.

[0046] In this embodiment, the continuous feature sequence after normalization and temporal processing is sequentially input into the pre-trained flow field prediction model in chronological order. The model adopts a hierarchical analytical structure, with each analytical layer corresponding to different scale response features in fluid dynamics. The first layer performs primary correlation analysis on the velocity gradient and pressure oscillation features to capture the inertial coupling behavior of the macroscopic flow field. The second layer extracts the spatial co-occurrence pattern of bubble density change and local viscosity fluctuation through convolutional temporal mapping to characterize the flow disturbance at the microscopic level. The third layer further utilizes a recursive memory mechanism to analyze the extension trend of bubble aggregation over time and extracts periodic and nonlinear correlations from adjacent segments of the feature sequence. Between each layer, the analytical unit adaptively adjusts the weight allocation according to the change amplitude of the features to strengthen important dynamic coupling patterns. The feature results output by each layer are nested and aggregated to generate an intermediate representation vector, which numerically reflects the multi-layer coupling relationship between the process fluid state, bubble swarm distribution, and fluid disturbance features.

[0047] Step S2032: By performing correlation matching on the intermediate representation vector, locate the aggregation center of the bubble in the spatial distribution and its adjacent diffusion range, forming a spatiotemporal coordinate set representing the aggregation dynamics.

[0048] In this embodiment, the intermediate representation vector generated in step S2031 is subjected to spatial correlation analysis. By calculating the similarity and time delay distribution between different feature vectors, a spatial mapping model of bubble aggregation features in the fluid channel is constructed. This model performs cluster analysis on the combination patterns of bubble density, flow velocity gradient, and pressure disturbance based on the correlation matrix of feature vectors, thereby identifying the region with the highest concentration of feature vectors as the bubble aggregation center. Subsequently, the feature distribution around the aggregation center is expanded and scanned, and the boundaries of adjacent diffusion ranges are determined based on the correlation decay threshold. To ensure positioning accuracy, the analysis process repeatedly calculates the spatial correlation matrix under multiple time windows, and eliminates the bias caused by single sampling noise through iterative weighting. Finally, the coordinates of the aggregation center and the boundary points of its diffusion range are arranged in chronological order into a spatiotemporal coordinate group, which dynamically reflects the aggregation and migration law of the bubble group inside the fluid.

[0049] Step S2033: Based on the continuous change relationship of the spatiotemporal coordinate group, calculate the identification parameter corresponding to the bubble density threshold, and establish a determination index for potential transmission hindrance zones based on the identification parameter.

[0050] In this embodiment, after obtaining the spatiotemporal coordinate set of the bubble aggregation center and diffusion range, the density of spatial coordinate points in each time period is first calculated to form a time-series density curve of bubble distribution. Subsequently, by integrating and smoothing the density change rate within a continuous time segment, the fluctuation period of bubble aggregation intensity is identified, and the local extreme value segment within this period is used as a candidate region for high-density aggregation. In the candidate region, the synergy index of bubble volume fraction and local velocity change is further calculated to comprehensively characterize the strength of the bubble's tendency to stagnate in space. By normalizing the synergy index under multiple time windows, its statistically significant critical turning point is extracted and defined as the bubble density threshold. Subsequently, an identification parameter is generated based on this threshold to mark the critical transition condition of bubble distribution from a flowable dynamic to a stagnant state. Finally, a transmission stagnation determination index is constructed with this identification parameter as the core. The index simultaneously records the correspondence between density threshold, time segment, and spatial coordinate, enabling subsequent control logic to quickly identify potential stagnation areas when the bubble group characteristics reach the determination threshold.

[0051] Step S204: Based on the determination index and combined with the operating parameters of the current semiconductor manufacturing process fluid delivery pipeline, select the corresponding flow path in the virtual valve control matrix and generate an adjustment command sequence for controlling the micro-valve opening.

[0052] The specific steps of step S204 are as follows: Step S2041: Obtain the operating parameters of the current semiconductor manufacturing process fluid delivery pipeline and synchronously form an operating status benchmark. The operating parameters include flow load, pressure gradient, and valve control response delay.

[0053] In this embodiment, before the gas-liquid transport control unit is started, the operating parameters under the current process state are first synchronously acquired from the pipeline monitoring nodes and the valve control drive end. The flow load is calculated by the output end of the main transport pump and the differential pressure sensor in the middle section of the pipeline to reflect the dynamic change of the liquid throughput. The pressure gradient is determined based on the linear fitting results of the pressure values ​​at multiple points along the pipeline to characterize the energy distribution of the fluid between the throttling section and the diffusion section. The valve control response delay is statistically estimated by the time difference between the previous microvalve trigger signal and the flow change response to characterize the inertial characteristics of the actuator. After the sampling time of each parameter is uniformly calibrated, they are merged according to priority so that the data have a relatively consistent response scale under the same time base. Subsequently, by constructing a parameter correlation matrix, the coupling trend between flow, pressure and valve control delay is calculated at once to extract the key feature set representing the current transport state.

[0054] Step S2042: The judgment index is compared with the running status benchmark, a virtual valve control matrix is ​​constructed based on the topological relationship of the fluid delivery pipeline in the semiconductor manufacturing process, and the set of bubble diversion paths is identified in the virtual valve control matrix.

[0055] In this embodiment, after obtaining the determination index and operating status benchmark of potential transmission hindrance zones, the two sets of data are first compared and correlated in terms of time series and parameter dimensions. By identifying the correlation between the location of the hindrance zone and the current flow rate, pressure gradient, and valve control response delay, the degree of flow restriction of the process fluid in different branches is determined. Subsequently, based on the geometric topology of the semiconductor manufacturing process fluid delivery pipeline, the connection relationships of each node, throttling section, and branch valve are extracted, and a virtual valve control matrix with nodes as rows and valves as columns is constructed. This matrix is ​​used to map the adjustable channels and flow paths of the gas and liquid phases inside the pipeline. After the matrix is ​​established, the control logic performs a correlation search on the spatial location of each potential hindrance zone and the valve distribution, and filters out the set of bubble splitting paths that maintain gas-liquid separation and flow velocity balance by traversing the path units in the matrix that satisfy the continuity and pressure gradient constraints.

[0056] Step S2043: Perform optimization calculations on the set of diversion paths, generate a path priority sequence according to the principle of fluid dynamic balance, and convert it into a corresponding valve control operation mapping table.

[0057] In this embodiment, after obtaining the set of bubble diversion paths, the fluid dynamic characteristics of each path are quantitatively calculated based on parameters such as flow load, local pressure difference, bubble density distribution, and valve response time. First, a balance equation is established using the pressure gradient and flow rate change rate of nodes within the path, representing the fluid impedance and bubble passage capacity of each path at the current moment as comparable quantitative indicators. Subsequently, iterative evaluation is performed on each channel in the path set to calculate its flow balance deviation under steady-state conditions and its recovery rate under dynamic disturbances, and a comprehensive path score is constructed based on these indicators. To ensure the timeliness of the evaluation, the score calculation is updated using a sliding time window, allowing the path selection decision to reflect the transient changes in the process environment. After calculation, a path priority sequence is generated according to the score, with the highest-scoring path corresponding to the main channel for controllable bubble flow, and low-priority paths designated as backup diversion channels. Finally, the priority sequence is converted into a valve control operation mapping table, which records the target opening degree, execution sequence, and linkage relationship of each valve.

[0058] Step S2044: Based on the valve control operation mapping table, generate a sequence of adjustment instructions for controlling the opening of the micro valve, and distribute them to the corresponding valve control nodes through hierarchical transmission.

[0059] In this embodiment, after the valve control operation mapping table is generated, the control logic unit first parses the target opening degree, action sequence, and valve linkage relationship recorded in the table, and constructs a multi-level instruction cache structure based on the response delay of each microvalve and the bandwidth allocation results of the communication channel. Subsequently, according to the priority and timing relationship in the mapping table, the action parameters of each valve are encoded into independent adjustment instruction units, and assembled into a complete instruction sequence containing time identifiers, node addresses, and execution parameters in the upper layer. The instruction sequence is transmitted hierarchically, with the upper-level control node responsible for allocating instruction blocks and the lower-level nodes performing time corrections based on local clock offsets to ensure that each microvalve responds at a predetermined rhythm on the global timeline. During the instruction distribution process, the transmission path is also periodically redundantly checked to prevent execution offset caused by accumulated delays or data loss. Finally, each valve control node receives and parses the corresponding adjustment instruction unit, driving the microvalve opening adjustment action to keep the bubble diversion path synchronized with the real-time fluid state. Through this hierarchical instruction distribution mechanism, high-precision dynamic control of multiple valves in complex pipeline topologies is achieved.

[0060] Step S205: The adjustment command sequence is sequentially transmitted to each microvalve control node, and the opening adjustment is performed according to the preset timing to divert the bubble path.

[0061] The specific steps of step S205 are as follows: Step S2051: Timestamp and sequence the adjustment command sequence, establish an execution queue consistent with the reference clock, and mark the node target and opening target value range for each adjustment command.

[0062] In this embodiment, after obtaining the sequence of adjustment instructions for controlling the opening of the micro-valve, the global time base synchronization unit is first invoked to compare the generation time and transmission delay of each instruction, calculate its relative offset from the system reference clock, and assign a precise timestamp to each instruction accordingly. Subsequently, by sorting the timestamps and arranging the sequence numbers, an execution queue strictly consistent with the reference clock is established, so that the instructions maintain a non-overlapping and continuous execution order in the time domain. On this basis, according to the path index and target valve parameters recorded in the valve control operation mapping table, the execution target node of each instruction is bound to the opening target value range, forming an instruction identifier set with hierarchical attributes. To prevent time drift from causing queue misalignment, the control logic dynamically detects the matching relationship between the time interval of adjacent instructions and the minimum trigger cycle during the arranging process, and automatically inserts placeholder instructions when necessary to maintain the integrity of the rhythm. The calibrated and arranged execution queue has precise timing consistency and node orientation, and directly serves as the input carrier in the next level time slot allocation stage to realize the synchronized scheduling of multi-node valve control actions.

[0063] Step S2052: Divide the execution queue into multiple segments according to a preset time slot window, generate a dispatch list based on the node mapping relationship of the virtual valve control matrix, and determine the transmission order and acknowledgment path of each adjustment command on the micro-valve control node.

[0064] In this embodiment, after completing the timestamp calibration of the adjustment instructions and the construction of the execution queue, the time slot partitioning algorithm is first invoked. Based on the preset control cycle and microvalve response characteristics, the continuous execution queue is divided into multiple time slot windows. Each window represents a set of instructions that can be executed in parallel within the same time period. During the partitioning process, the time interval between adjacent instructions and the valve control delay coefficient are calculated to ensure that the execution density in each window can meet the fluid synchronization requirements without causing signal overlap. Subsequently, based on the node mapping relationship in the virtual valve control matrix, the instructions in each time slot window are mapped to the corresponding microvalve control nodes. The connection topology between nodes is reconstructed to form a preliminary instruction dispatch list. To prevent communication blockage caused by multiple path intersections, the execution order between nodes in the list is further sorted topologically, the shortest data return path is calculated, and the transmission order and acknowledgment link of each instruction are marked in the path table.

[0065] Step S2053: Based on the transmission order and acknowledgment path, each microvalve control node adjusts its opening degree and records the feedback information, including the execution frame identifier and local sampling snapshot.

[0066] In this embodiment, after the instruction is dispatched, each microvalve control node receives the adjustment instruction according to its corresponding transmission order and sets up a bidirectional channel for the communication link according to the acknowledgment path. After receiving the instruction belonging to its own time slot window, each node first reads the target opening range and execution time stamp contained in the instruction, and performs time correction in conjunction with the local clock to ensure that the valve action is consistent with the global scheduling rhythm. Then, the opening is adjusted by driving the valve plate to move through a micro-step signal, so that the valve port cross-sectional area transitions within the target range. During the execution of the action, the node simultaneously monitors the change in current load and the valve port flow feedback signal to verify the completion accuracy of the opening instruction, and generates a corresponding execution frame identifier at the end of the instruction to mark the timing position of the action in the global execution queue. To achieve state tracking, the node collects a snapshot of the local pressure and flow velocity of the fluid at the moment the valve action is completed, encapsulates it together with the execution frame identifier into a return data packet, and reports it to the upper-level coordination along the predetermined acknowledgment path. Through this mechanism, each microvalve node forms a quantifiable closed-loop information chain while independently executing control tasks.

[0067] Step S2054: Compare the returned information with the execution queue frame by frame, update the next round of dispatch list and time slot window based on the comparison results, and resend and rearrange the queue positions of unexecuted or deviated instructions.

[0068] In this embodiment, after each microvalve control node completes its action and uploads feedback information, the feedback data is compared frame by frame with the previously set execution queue according to the execution frame identifier order. During the comparison, the offset between the instruction timestamp and the receipt timestamp is checked to identify nodes with valve action delays, execution omissions, or feedback anomalies. Combined with the flow and pressure data from the local sampling snapshot, the nature and range of the action deviation are determined. Subsequently, the comparison results are mapped to the previous round of dispatch list. The time slot occupancy rate and signal conflict risk of nodes with deviations are recalculated, and a corrected time slot window structure is regenerated in the scheduling logic to ensure that subsequent instruction allocation avoids potential delay segments. For instructions that are not executed or have missing receipts, priority is rearranged according to path dependency, and a retransmission mark is inserted into the new round of dispatch list so that they are preferentially allocated to available channels for execution in the next cycle. At the same time, the control logic confirms the successfully executed instruction segments and removes them from the active queue to release bandwidth, thereby forming a dynamically self-correcting execution queue structure.

[0069] Figure 2 The diagram shows a bubble splitting structure in the BUB bubble conveying control method used in semiconductor manufacturing processes, illustrating how bubble splitting control based on flow field prediction and microvalve regulation can be achieved in fluid delivery pipelines.

[0070] Specifically, such as Figure 2 As shown, the process fluid transport pipeline is the main flow channel, equipped with multiple monitoring nodes to collect parameters such as flow velocity, pressure, and bubble distribution. These nodes synchronously transmit the collected signals to the central processing unit in a time-series fashion to form a multi-dimensional data stream reflecting the pipeline's operating status. The collected data is first input into a flow field prediction model, which calculates the velocity gradient and bubble density field based on a pre-trained algorithm to identify spatiotemporal regions where bubble aggregation may occur in the fluid. The prediction results are then fed into a decision index, which calculates potential transmission hindrance zones based on the model output and generates a decision index corresponding to the risk of bubble aggregation. Subsequently, the decision index is compared with... The current operating parameters of the pipeline are input into the virtual valve control matrix, which logically corresponds to the actual multi-microvalve control network. It is used to select a set of feasible diversion paths based on the flow field topology. After calculation, a set of adjustment command sequences for bubble diversion is generated and transmitted to each microvalve control node. At the microvalve control node, the valve opening is adjusted according to the received command, so that some fluid is discharged along the diversion path, forming a new bubble migration direction. After the bubble diversion, the local gas-liquid interface disturbance is weakened, the risk of flow stagnation in the main passage is eliminated, and some gas flows back to the main fluid channel, realizing the dynamic restoration of flow balance.

[0071] Step S3: Apply a micro-pulse gas pressure signal to the throttling section of the semiconductor manufacturing process fluid delivery pipeline to dynamically match the bubble burst phase with the fluid viscosity.

[0072] The specific steps of step S3 are as follows: Step S301: After a stable fluid channel is formed in the throttling section of the semiconductor manufacturing process fluid delivery pipeline, process fluid data within the throttling section is collected, and an initial reference table for gas pressure pulse modulation is established based on the fluid data. The process fluid data includes fluid density, viscosity, and flow rate change data.

[0073] In this embodiment, after a stable flow state is established in the throttling section of the semiconductor manufacturing process fluid delivery pipeline, the density, viscosity, and velocity of the fluid within the throttling area are sampled in real time. The sampling period is dynamically adjusted according to the fluid viscosity fluctuation rate to ensure that minute changes in flow characteristics are captured under steady-state conditions. Subsequently, the control logic combines the density data with the velocity gradient to calculate the fluid dynamic pressure distribution trend and uses the viscosity change sequence to identify the energy dissipation characteristics of the throttling area. The stable parameter range obtained through multiple cyclic sampling is defined as the reference fluid feature set of the throttling section after time-weighted averaging. Based on this feature set, the control algorithm constructs an initial reference table for gas pressure pulse modulation according to the empirical correspondence between pressure response sensitivity and bubble generation rate. This reference table uses the viscosity change range as the main index and corresponds to the minimum pulse amplitude, pulse duration, and frequency range that can achieve stable bubble rupture, providing a reference template for subsequent pulse loading control, so that the gas-liquid interaction process can be synchronously modulated within a predictable range of flow parameters.

[0074] Step S302: Under the constraints of the initial reference table, generate a micro-pulse pressure signal containing a main pulse and an auxiliary pulse.

[0075] In this embodiment, after obtaining the initial reference table of gas pressure pulse modulation for the throttling section, a multi-layer time-domain signal structure is constructed based on the target pulse amplitude, duration, and frequency range recorded in the table. Specifically, a main pulse sequence is first generated with the reference frequency as the core. This sequence is used to maintain the periodic disturbance of the gas-liquid interface in the throttling section. Then, an auxiliary pulse signal is superimposed within the main pulse period. The frequency and phase of the auxiliary pulse are dynamically adjusted according to the differential relationship between fluid viscosity and flow velocity changes, forming energy differential compensation under local resonance conditions. During the pulse synthesis process, a micro-phase delay is introduced on the time axis so that the overlap interval between the main and auxiliary pulses corresponds to the optimal propagation window of the fluid pressure wave, thereby achieving controllable excitation of the bubble rupture rhythm. After multiple iterative calculations, the combined signal of the main pulse and the auxiliary pulse forms a set of micro-pulse gas pressure signals that satisfy the fluid stability constraints in terms of amplitude, phase, and interval.

[0076] Step S303: Before the micro-pulse air pressure signal is applied, the fluid viscosity change range is continuously monitored, and the time phase of bubble rupture is mapped and compared with the viscosity change range to generate a synchronous trigger timetable.

[0077] The specific steps of step S303 are as follows: Step S3031: Set the sampling period, collect instantaneous change data of fluid viscosity according to the time series, and organize the instantaneous change data into piecewise curves.

[0078] In this embodiment, before the bubble transport stage begins, the viscosity sampling period is set based on the fluid stability parameters within the throttling section. The sampling period is automatically calculated by comparing the rate of change of flow velocity with the viscosity fluctuation amplitude of the previous period, ensuring that the sampling frequency is sufficient to capture the transient change characteristics of fluid viscosity. Subsequently, within the sampling period, viscosity sensing nodes are triggered sequentially to collect data, with each acquisition accompanied by a precise timestamp. To eliminate high-frequency noise interference, the collected data undergoes digital filtering and sliding weighting processing to extract the effective viscosity change trend. The control logic divides the continuous instantaneous viscosity data into several segments according to time slices, with each segment representing the internal friction change trajectory of the fluid within adjacent time windows. By fitting and interpolating the data of each segment, the discretely measured viscosity points are converted into continuous piecewise curves, and the inflection points, stable regions, and abrupt change regions of the curves are marked.

[0079] Step S3032: During the monitoring period of bubble generation and transport process, record the time series of bubble rupture events and map each rupture moment as a phase node to form a phase trajectory table representing the bubble rupture rhythm.

[0080] In this embodiment, during the monitoring period of bubble generation and transmission, bubble rupture events are captured in real time using optical scattering, acoustic echo, or resistance abrupt change signals, and the timestamp of each rupture moment is recorded with a fixed time resolution. To ensure the accuracy of the recording, the monitoring algorithm distinguishes real rupture events from background disturbance noise by identifying signal amplitude abrupt changes and phase reversal features, and introduces redundant nodes for cross-validation to reduce single-point detection errors. After all rupture events are arranged in chronological order, they are normalized to map each rupture moment to a periodic phase interval, forming a phase node set. This set is transformed into a phase trajectory table through time axis folding mapping to describe the periodic distribution and rhythmic characteristics of bubble rupture behavior in the time dimension. By comprehensively analyzing the node spacing, phase offset, and periodic stability in the phase trajectory table, the time series pattern of bubble rupture rhythm is obtained.

[0081] Step S3033: Compare the phase trajectory table with the piecewise curve in time sequence, calculate the interrelated segments based on the overlapping intervals of the two, and determine the correspondence between the bubble bursting rhythm and viscosity change.

[0082] In this embodiment, after obtaining the phase trajectory table of the bubble bursting rhythm and the piecewise curve of fluid viscosity, the two sets of data are first aligned on the time axis. The sampling period difference is eliminated by interpolation resampling and time normalization, making them comparable under the same time base. Subsequently, the time offset between adjacent bursting nodes and inflection points of viscosity curve changes in the phase trajectory table is calculated in a unified time domain. Based on this offset, the synchronous or lagging relationship between the bubble bursting event and the viscosity change is determined. By performing multiple rounds of comparison within a sliding time window, the overlapping intervals of the dense bursting node area and the abrupt viscosity curve change area are identified. The similarity operation is performed on the signal change direction within these intervals to screen out the associated segments with stable coupling characteristics. Finally, the statistical results of the associated segments are converted into a mapping table to clarify the bubble bursting frequency and phase delay parameters corresponding to different viscosity change intervals, thereby establishing a quantitative correspondence between the bubble bursting rhythm and fluid viscosity dynamics.

[0083] Step S3034: Construct a synchronization trigger timetable based on the aforementioned correspondence.

[0084] In this embodiment, after obtaining the correspondence between the bubble bursting rhythm and the fluid viscosity change, the time offset and phase difference parameters of each associated segment are first used as input variables to establish a synchronous triggering calculation model. This model uses the bubble bursting phase as the time axis and the peak and valley nodes of the viscosity change curve as control reference points. The optimal triggering window for each bubble event is calculated through interpolation. Subsequently, the phase delay distribution in different viscosity ranges is aggregated and analyzed to determine the set of triggering times corresponding to each bubble group. Based on the continuity of gas-liquid state changes, a periodic sequence is formed on the time axis. To ensure subsequent execution accuracy, the control logic constrains and corrects the minimum interval between adjacent times and the valve control response time during the generation process, so that the time distribution meets the system dynamic response boundary conditions. After multiple rounds of calculation, all triggering times are recorded in the form of timestamps to form a synchronous triggering time table. This time table uses viscosity segments as rows and phase nodes as columns, and marks the triggering priority and duration parameters accordingly. It is used to guide the loading rhythm of the micro-pulse gas pressure signal under different fluid states, so as to achieve precise synchronous control of the bubble bursting phase and the fluid viscosity change.

[0085] Step S304: Apply micro-pulse air pressure signals sequentially according to the synchronous triggering timetable, and re-acquire fluid parameters to update the initial reference table after each cycle.

[0086] In this embodiment, after the synchronous triggering timetable is established, the pressure modulation unit is triggered sequentially according to the time sequence recorded in the table, and a micro-pulse pressure signal composed of a main pulse and an auxiliary pulse is applied to the throttling section. The pulse loading process uses the phase node of the timetable as a reference, and the amplitude and duration of the pressure wave are precisely adjusted by the timing controller in each triggering cycle to keep the bubble bursting behavior and viscosity change node dynamically synchronized. During the loading process, fluid density, viscosity and local pressure changes are collected in real time to capture the feedback characteristics of the pulse response to the fluid state. After a complete pulse loading cycle is completed, the fluid parameters in this cycle are compared with the reference table of the previous cycle to calculate the deviation of viscosity fluctuation, flow velocity distribution and energy transfer. Through the normalization processing of the deviation, a corrected dataset reflecting the latest fluid state is regenerated, and the initial reference table is updated in the control logic so that it can participate in the pulse parameter calculation as a new reference baseline in the next cycle. Through continuous pulse application and reference update iteration, the gas-liquid interaction process forms a self-correcting dynamic control closed loop, realizing continuous matching and precise control between bubble rhythm and fluid characteristics in the fluid channel.

[0087] Step S4: When the detected bubble residue exceeds the threshold, switch to reverse bubbling mode to perform self-cleaning of the pipeline.

[0088] In this embodiment, during the process fluid transport, the residual bubble situation in the pipeline is continuously monitored. When the bubble volume fraction or local impedance fluctuation value exceeds a set threshold, the control logic triggers a mode switching command. Before the switching is executed, the downstream fluid passage is temporarily closed through the valve control matrix, and the gas inlet node near the upstream is opened in reverse, so that the airflow direction is opposite to the normal bubbling direction. Subsequently, the reverse airflow is injected into the bubble generating unit in the form of low frequency and high pressure, forming a reverse bubble chain, which triggers a local negative pressure pulse in the fluid channel, thereby disturbing the residual bubble group attached to the pipe wall and the throttling area. In order to avoid the impact causing fluid disturbance, the pulse interval of reverse bubbling is dynamically adjusted according to the real-time pressure feedback, so that the bubble desorption process is advanced segment by segment along the pipeline. After the cleaning cycle is completed, the monitoring node collects fluid impedance and bubble density data again and compares them with the previous cycle to confirm whether the residue has been removed. If the detection result shows that the pipeline impedance has recovered to the reference range, the reverse bubbling mode is exited and the normal transport state is restored. This self-cleaning process realizes the automatic recovery of the pipeline through fluid dynamic reversal and local disturbance control, providing long-term stable operating conditions for semiconductor process transport pipelines.

[0089] Example 2 Please see Figure 3 Another embodiment of the present invention provides a BUB bubble conveying system for semiconductor manufacturing processes, comprising: a bubble generation module, an adjustment module, a dynamic matching module, and a self-cleaning module; The bubble generation module is used to generate a bubble group in the fluid transport pipeline of semiconductor manufacturing process through a bubble generator, and to adjust the bubble generation frequency and volume distribution based on the liquid impedance feedback signal. The adjustment module is used to collect process liquid flow rate, pressure and bubble distribution parameters, and to analyze bubble aggregation trend using a pre-trained flow field prediction model. When a risk of transmission obstruction is detected, the micro-valve opening is automatically adjusted to divert the bubble path. The dynamic matching module is used to apply a micro-pulse air pressure signal to the throttling section of the fluid transport pipeline in semiconductor manufacturing process to dynamically match the bubble burst phase with the fluid viscosity. The self-cleaning module is used to switch to reverse bubbling mode to self-clean the pipeline when the detected bubble residue exceeds the threshold.

[0090] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0091] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A BUB bubbling delivery control method for a semiconductor production process, characterized by, The application relates to a semiconductor production process fluid conveying pipeline, and relates to a bubble generator and a bubble generation frequency and volume distribution adjusting method based on liquid impedance feedback signals. Process liquid flow rate, pressure and bubble distribution parameters are collected, and a pre-trained flow field prediction model is used to analyze the bubble aggregation trend; when the transmission blockage risk is detected, the micro valve opening degree is automatically adjusted to shunt the bubble path. A micro pulse air pressure signal is applied to the throttling section of the semiconductor production process fluid conveying pipeline to dynamically match the bubble breakup phase and fluid viscosity. When it is detected that the bubble residue exceeds a threshold value, the reverse bubbling mode is switched to for self-cleaning of the pipeline. The application relates to a semiconductor production process fluid conveying pipeline, and relates to a bubble generator and a bubble generation frequency and volume distribution adjusting method based on liquid impedance feedback signals.

2. The BUB sparging delivery control method for a semiconductor production process according to Claim 1, wherein, A predetermined purity process fluid is injected into the semiconductor production process fluid conveying pipeline, and an initial liquid impedance reference value is established at the inlet of the bubble generator; An array of holes in the bubble generator is sequentially opened, multi-point nucleation is triggered by using a laminated gas-liquid interface disturbance, and a bubble group is formed on the same process fluid section; Impedance monitoring nodes are arranged at the bubble nucleation region and the downstream process fluid section, real-time collection of complex impedance signals in the gas-liquid mixed state is carried out, and phase difference and amplitude characteristics are extracted as impedance feedback parameters; The frequency of bubble generation is automatically adjusted based on the change of the impedance feedback parameters, and the bubble generation rhythm is adaptively adjusted. Different particle size bubbles are identified in the bubble transmission stage, and the opening sequence of the array of holes is adjusted based on the impedance feedback parameters of the previous period, and is updated in the next cycle. The application relates to a semiconductor production process fluid conveying pipeline, and relates to a bubble generator and a bubble generation frequency and volume distribution adjusting method based on liquid impedance feedback signals.

3. The BUB sparging delivery control method for a semiconductor production process of claim 2, wherein, A predetermined purity process fluid is injected into the semiconductor production process fluid conveying pipeline, and an initial liquid impedance reference value is established at the inlet of the bubble generator; An array of holes in the bubble generator is sequentially opened, multi-point nucleation is triggered by using a laminated gas-liquid interface disturbance, and a bubble group is formed on the same process fluid section; Impedance monitoring nodes are arranged at the bubble nucleation region and the downstream process fluid section, real-time collection of complex impedance signals in the gas-liquid mixed state is carried out, and phase difference and amplitude characteristics are extracted as impedance feedback parameters; The frequency of bubble generation is automatically adjusted based on the change of the impedance feedback parameters, and the bubble generation rhythm is adaptively adjusted. The application relates to a semiconductor production process fluid conveying pipeline, and relates to a bubble generator and a bubble generation frequency and volume distribution adjusting method based on liquid impedance feedback signals.

4. The BUB sparging delivery control method for a semiconductor production process as claimed in claim 3, wherein, Process liquid flow rate, pressure and bubble distribution parameters are collected, and a pre-trained flow field prediction model is used to analyze the bubble aggregation trend; when the transmission blockage risk is detected, the micro valve opening degree is automatically adjusted to shunt the bubble path. Monitoring nodes are arranged at different positions of the semiconductor production process fluid conveying pipeline, flow rate, pressure and bubble distribution parameters are collected, and time-synchronized multi-dimensional data streams are formed; The multi-dimensional data streams are time-sequentially arranged and normalized, are converted into continuous feature sequences representing the gas-liquid flow state, and flow field disturbance sections are marked. The continuous feature sequence is input into a pre-trained flow field prediction model. Based on the model output, the spatiotemporal positional relationship of bubble aggregation is identified, and a determination index for potential transport obstruction zones is generated. Based on the determination index and combined with the operating parameters of the current semiconductor manufacturing process fluid delivery pipeline, the corresponding flow path is selected in the virtual valve control matrix, and an adjustment command sequence for controlling the micro-valve opening is generated. The adjustment command sequence is sequentially transmitted to each microvalve control node, and the opening adjustment is performed according to the preset timing to divert the bubble path.

5. The BUB sparging delivery control method for a semiconductor production process as claimed in claim 4, wherein, The continuous feature sequence is input into a pre-trained flow field prediction model. Based on the model output, the spatiotemporal location relationship of bubble aggregation is identified, and a determination index for potential transport obstruction zones is generated, including: The continuous feature sequence is sequentially input into each level of the analytical layer of the pre-trained flow field prediction model. At each level, the coupling mode between the process fluid state and bubble aggregation is extracted, and an intermediate representation vector is generated. By performing correlation matching on the intermediate representation vector, the aggregation center of the bubble in the spatial distribution and its adjacent diffusion range are located, forming a spatiotemporal coordinate set representing the aggregation dynamics. Based on the continuous change relationship of the spatiotemporal coordinate group, the identification parameter corresponding to the bubble density threshold is calculated, and the identification parameter is used as a basis to establish a judgment index for potential transmission hindrance zones.

6. The BUB sparging delivery control method for a semiconductor production process as claimed in claim 5, wherein, Based on the determination index and the operating parameters of the current semiconductor manufacturing process fluid delivery pipeline, the corresponding flow path is selected in the virtual valve control matrix, and a sequence of adjustment commands for controlling the micro-valve opening is generated, including: The operating parameters of the fluid delivery pipeline in the current semiconductor manufacturing process are obtained and synchronously formed into an operating status benchmark. The operating parameters include flow load, pressure gradient and valve control response delay. The judgment index is correlated and compared with the running status benchmark. A virtual valve control matrix is ​​constructed based on the topological relationship of the fluid delivery pipeline in the semiconductor manufacturing process, and the set of bubble diversion paths is identified in the virtual valve control matrix. The set of diversion paths is optimized by calculation, a path priority sequence is generated according to the principle of fluid dynamic balance, and then converted into a corresponding valve control operation mapping table. Based on the valve control operation mapping table, a sequence of adjustment commands for controlling the opening degree of the micro-valve is generated and distributed to the corresponding valve control nodes through hierarchical transmission.

7. The BUB sparging delivery control method for a semiconductor production process as claimed in claim 6, wherein, The adjustment command sequence is sequentially transmitted to each microvalve control node, and the opening adjustment is executed according to a preset timing sequence to divert the bubble path, including: The adjustment command sequence is timestamped and numbered to establish an execution queue consistent with the reference clock, and the node target and opening target value range are marked for each adjustment command. The execution queue is divided into multiple segments according to a preset time slot window. An assignment list is generated based on the node mapping relationship of the virtual valve control matrix to determine the transmission order and acknowledgment path of each adjustment command on the micro-valve control node. Based on the transmission order and acknowledgment path, each microvalve control node adjusts its opening degree and records the feedback information, including the execution frame identifier and local sampling snapshot. The backhaul information is compared with the execution queue frame by frame, the next round of dispatch list and time slot window are updated based on the comparison result, and the instructions that are not executed or deviated are retransmitted and the queue position is rearranged.

8. The BUB sparging delivery control method for a semiconductor production process according to Claim 7, wherein, The micro-pulse air pressure signal is applied to the throttling section of the semiconductor production process fluid conveying pipeline, and the bubble breaking phase and fluid viscosity are dynamically matched, including: After the stable fluid channel state is formed in the throttling section of the semiconductor production process fluid conveying pipeline, process fluid data in the throttling section is collected, and an initial reference table of air pressure pulse modulation is established based on the fluid data, the process fluid data including fluid density, viscosity and flow rate change data; Under the constraint of the initial reference table, a micro-pulse air pressure signal containing a main pulse and an auxiliary pulse is generated; Before the micro-pulse air pressure signal is loaded, the fluid viscosity change interval is continuously monitored, and the time phase of bubble breaking is mapped and compared with the viscosity change section to generate a synchronous trigger time table; According to the synchronous trigger time table, the micro-pulse air pressure signal is applied in sequence, and the initial reference table is updated after each cycle is completed.

9. The BUB sparging delivery control method for a semiconductor production process of claim 8, wherein, Before the micro-pulse air pressure signal is loaded, the fluid viscosity change interval is continuously monitored, and the time phase of bubble breaking is mapped and compared with the viscosity change section to generate a synchronous trigger time table, including: A sampling period is set, and the instantaneous change data of fluid viscosity is collected in time sequence, and the instantaneous change data is arranged into a segmented curve; During the monitoring period of the bubble generation and transmission process, the time sequence of the bubble breaking event is recorded, and each breaking time is mapped as a phase node to form a phase trajectory table representing the bubble breaking rhythm; The phase trajectory table and the segmented curve are time series compared, the correlation section is calculated according to the overlapping interval of the two, and the corresponding relationship between the bubble breaking rhythm and the viscosity change is determined; According to the corresponding relationship, a synchronous trigger time table is constructed.

10. A BUB bubbling transport system for a semiconductor production process for implementing the BUB bubbling transport control method for a semiconductor production process according to any one of claims 1 to 9, characterized by It includes: The bubble generation module, the adjustment module, the dynamic matching module and the self-cleaning module; The bubble generation module is used to generate bubble groups in the semiconductor production process fluid conveying pipeline through a bubble generator, and adjust the bubble generation frequency and volume distribution based on liquid impedance feedback signals; The adjustment module is used to collect process liquid flow rate, pressure and bubble distribution parameters, and analyze bubble aggregation trend using a pre-trained flow field prediction model, and automatically adjust the micro-valve opening degree when detecting transmission blockage risk to split the bubble path; The dynamic matching module is used to apply a micro-pulse air pressure signal to the throttling section of the semiconductor production process fluid conveying pipeline to dynamically match the bubble breaking phase and fluid viscosity; The self-cleaning module is used to switch to a reverse bubbling mode for self-cleaning of the pipeline when detecting that the bubble residue exceeds a threshold.

Images