Microfluidic droplet generation parameter adaptive closed-loop control method and system

Abstract

The application belongs to the technical field of droplet digital control, and discloses a microfluidic droplet generation parameter adaptive closed-loop control method and system. The system is internally provided with a step emulsion droplet size prediction model based on working condition partition. The step platform height, platform width and step drop difference and other chip geometric parameters are used as constraints, and the fluid viscosity, interfacial tension and other physical parameters are combined to establish a droplet balance volume correlation equation. The two-phase flow velocity reference value can be directly calculated according to the target droplet diameter, and the process debugging cycle is shortened. The system divides the fluid physical properties and environmental parameters into multiple groups of preset working condition intervals, each interval matches an independent model correction coefficient group, and when started, the same working condition historical optimal parameter in the iterative learning parameter library is called as the initial reference, and the continuous phase control flow rate is calibrated in combination with the flow field stability margin. After completing one production batch, the steady-state operation data is automatically extracted to update the parameter library, and the correction coefficient can also be online fine-tuned according to the real-time closed-loop feedback during operation.

Description

Technical Field

[0001] This invention belongs to the field of droplet digital control technology, specifically a microfluidic droplet generation parameter adaptive closed-loop control method and system. Background Technology

[0002] Droplet microfluidics has wide applications in drug delivery, single-cell analysis, digital PCR detection, material synthesis, and point-of-care diagnostics. The performance of its control system directly determines the quality, stability, and production efficiency of droplet generation. Current control technologies for microfluidic droplet generation systems mainly face the following technical challenges: Most existing droplet generation systems employ an open-loop programmable control mode, requiring operators to manually set parameters such as the injection pump flow rate and pressure based on experience, and then iteratively correct the settings by repeatedly sampling and detecting droplet size offline. In this mode, the system cannot automatically adjust parameters according to real-time operating conditions, resulting in long debugging cycles, and the droplet quality is significantly affected by the operator's skill level.

[0003] Traditional control systems pre-determine a fixed mapping relationship between flow velocity and droplet size. When fluid properties such as viscosity and interfacial tension change, or when environmental temperature fluctuations cause fluid characteristics to drift, the pre-determined control parameters become inapplicable, and the droplet size deviation continues to increase. Existing systems lack the ability to adaptively calculate based on fluid properties and target parameters, and therefore cannot correct the control model online.

[0004] Some systems with closed-loop functionality only employ conventional PID feedback control, directly correcting the driving flow velocity based on droplet size deviations, with control parameters remaining constant. When the system is in a large deviation startup phase or subjected to strong disturbances, fixed-parameter PID is prone to overshoot or response lag. Existing systems do not achieve multi-stage decoupled control of the droplet generation process; flow velocity fluctuations are directly transmitted to droplet size, resulting in a heavy burden on closed-loop regulation and slow convergence. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive closed-loop control method and system for microfluidic droplet generation parameters to solve one or more problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an adaptive closed-loop control system for microfluidic droplet generation parameters, comprising a task configuration module, a parameter calculation module, a drive module, a control module, an execution module, a feature detection module, and an adjustment module; Furthermore, the task configuration module receives user input via a touch screen, including the target droplet diameter, upper limit of the coefficient of variation of size, required throughput, as well as dispersed phase viscosity, continuous phase viscosity, interfacial tension coefficient between the two phases, and ambient temperature parameters. The task configuration module has built-in parameter validity verification logic. Combining chip geometric structure parameters and CNC program running boundaries, it determines whether the input parameters are within the allowable working range and outputs warning prompts for parameters that exceed the boundaries. The process numerical control program storage and recall unit stores process numerical control program packages. Each program package contains a set of control parameters for the corresponding fluid formula and droplet specifications, operating condition event response logic, and iterative learning history data. When the system starts, it receives the process scheme selected by the user, sends the configuration parameters and corresponding control program segments to the parameter calculation module, starts the CNC self-test process, and verifies the communication link, drive timing and running status of each functional module in sequence according to the preset program; During operation, the system displays the flow rate, droplet generation frequency, size statistics, and fault alarm information of each channel in real time, and automatically records the operating data according to the numerical control clock cycle.

[0007] Furthermore, the parameter calculation module incorporates a stepped emulsion droplet size prediction model based on working condition partitioning. Using the height, width, and drop of the stepped platform as geometric constraints, and combining fluid viscosity and interfacial tension parameters, it establishes a correlation equation between the droplet equilibrium volume and geometric and flow velocity parameters. The fluid properties and environmental parameters are divided into multiple preset working condition intervals, each of which corresponds to an independent set of model correction coefficients, which are automatically matched and called by the CNC program. Equation for predicting the equilibrium volume of stepped emulsion droplets: This represents the steady-state equilibrium volume of the droplet, which is the theoretical calculation output value of the stepped emulsion droplet size prediction model. It corresponds to the theoretical volume of a single droplet when it reaches the interfacial tension equilibrium state after the dispersed phase filaments complete regular fracture under the combined action of the stepped structure and interfacial tension under ideal steady-state conditions. It does not include deviations caused by actual operational disturbances.

[0008] It represents the equivalent spherical diameter of the droplet in steady-state equilibrium, a theoretical size characterization value obtained by converting the droplet's steady-state equilibrium volume; it only represents the ideal steady-state generation size predicted by the model, used in the parameter solution process to back-calculate the reference value of the two-phase flow velocity, and is the set reference for the open-loop prediction of the system.

[0009] This represents the geometric correction coefficient, one of the correction coefficients in the working condition partitioning model, used to linearly correct the influence of microchannel processing tolerances and stepped structure forming deviations on droplet volume.

[0010] This indicates the height of the stepped platform, specifically the vertical flow channel height of the micro-nozzle stepped platform section.

[0011] This indicates the width of the stepped platform, specifically the horizontal width of the flow channel in the stepped platform section of the micro-nozzle.

[0012] This indicates the height difference between the stepped platform section of the flow channel and the downstream extended flow channel.

[0013] This represents the geometric constraint function, which describes the fundamental constraint relationship between the geometric structural parameters of the stepped flow channel and the droplet forming volume.

[0014] This represents the viscosity correction factor, one of the correction factors in the operating condition partitioning model, used to linearly correct the influence of fluid property parameter fluctuations on droplet breakage volume.

[0015] This indicates the dynamic viscosity of the dispersed phase fluid.

[0016] This represents the dynamic viscosity of a continuous phase fluid.

[0017] It represents the interfacial tension coefficient of a two-phase fluid.

[0018] This represents the property correlation function, which describes the quantitative influence between fluid viscosity, interfacial tension, and droplet fracture volume.

[0019] This represents the flow rate correction coefficient, one of the correction coefficients in the operating condition zoning model, used to linearly correct the influence of flow rate control deviation on droplet volume.

[0020] This represents the reference velocity of the dispersed phase, which is the theoretical set velocity of the dispersed phase channel obtained from parameter calculation.

[0021] This represents the reference velocity of the continuous phase, which is the theoretical set velocity of the continuous phase channel obtained from parameter calculation.

[0022] This represents the velocity influence function, describing the relationship between the two-phase velocity ratio and the shear intensity on the droplet fracture volume.

[0023] After receiving the target droplet diameter, the historical best parameters under the same working conditions in the iterative learning parameter library are used as the initial benchmark. The theoretical equilibrium volume is calculated by combining the chip's fixed geometric parameters, and the benchmark flow rate of the dispersed phase is obtained by reverse calculation. The two-phase flow velocity reference value is calculated using a step-by-step iterative method. First, based on the target droplet diameter and the model correction coefficient corresponding to the current operating condition, the droplet equilibrium volume under the corresponding operating condition is calculated. Then, combined with the cross-sectional area of ​​the micro-nozzle channel and the target generation flux requirement, the preliminary calculation of the dispersed phase reference flow velocity is obtained.

[0024] After the initial calculation is completed, immediately check whether the dispersed phase reference flow rate is within the operating boundary range set by the CNC program. If it exceeds the upper and lower limits of the flow rate, adjust the target generation flux or droplet size parameters simultaneously and recalculate until the dispersed phase flow rate meets all operating constraints.

[0025] Once the dispersed phase velocity is determined, the critical velocity for droplet breakage under the corresponding operating condition is solved based on the balance between the flow field shear force and the interfacial tension between the two phases.

[0026] The critical velocity value for droplet breakage is solved by the flow field shear force calculation model. Based on the principle of reserving the flow field stability margin, the continuous phase control velocity is set to a value lower than the critical velocity. The flow field stability margin in the pre-stretching zone is calculated. When the fluctuation of physical property parameters leads to insufficient flow field stability margin, the continuous phase velocity is adjusted according to the preset adjustment step size of the program. After each production batch is completed, the system automatically extracts the parameter matching data from the steady-state operation phase within the batch, updates the model correction coefficients and iterative learning parameter library for the corresponding operating condition range, and fine-tunes the model correction coefficients under the current operating condition online based on the real-time closed-loop feedback statistical results.

[0027] Furthermore, the drive module adopts stepper motor microstepping drive technology, and all channel drive units are synchronously triggered by a unified numerical control clock, receiving flow rate setpoints and timing control commands, and smoothly driving the injection pump to run according to a preset program acceleration curve; During operation, the displacement of the injection pump push rod and the pressure signal of the pipeline are collected in real time according to a fixed sampling rhythm. The displacement feedback is used to compensate for the stepper motor's step loss error, and the pressure feedback is used to monitor pipeline blockage and abnormal fluctuations. When an abnormal increase in pressure is detected, the graded deceleration protection and fault alarm are executed according to the preset program, and the working condition event identifier is triggered and uploaded to the main control unit. The pressure anomaly protection is set with three response thresholds. The first threshold corresponds to slight overpressure, which is caused by a slight increase in pipeline resistance or a change in fluid viscosity. The system triggers an early warning and at the same time slightly increases the drive output to compensate for the change in resistance and maintain a stable flow rate in the channel.

[0028] The secondary threshold corresponds to moderate overpressure, caused by partial blockage of the pipeline or nozzle. The system performs graded deceleration, gradually reducing the flow velocity in the corresponding channel according to the preset slope. At the same time, the anomaly investigation logic is activated, and the location and degree of blockage are determined by combining the outlet droplet detection data.

[0029] Level 3 threshold corresponds to severe overpressure, caused by complete blockage of the pipeline or nozzle. The system immediately stops the drive output of the corresponding channel, closes the pipeline control valve, triggers a severe fault alarm, and simultaneously disconnects the faulty channel and activates the backup channel to maintain overall production operation. After each level of protection is triggered, the system automatically records the complete pressure change curve and corresponding operating condition data.

[0030] The drive module has a built-in programmed automatic fluid change control program. When a syringe is advanced to the end of its stroke, the pre-filling process of the backup syringe is started in advance according to a preset timing sequence. The flow rate is connected during the fluid change process through a flow rate smooth transition algorithm.

[0031] Furthermore, the control module, corresponding to the flow control focusing or co-current coaxial sheath flow configuration of the pre-stretching zone, receives the continuous phase flow velocity setpoint and numerical control adjustment command output by the parameter calculation module, and outputs control commands to the drive module to make the continuous phase injection pump run in a synchronous sequence. The stability of the flow field is assessed by detecting the diameter fluctuation of the fluid filament downstream of the pre-stretching zone. When changes in ambient temperature or fluid property drift cause changes in the stretching state, a small incremental adjustment of the continuous phase flow velocity is performed. When the diameter of the fluid filament is detected to be too large and irregular fluctuations are observed, the continuous phase flow velocity is increased by a step size set according to the program; when the diameter of the filament is detected to be close to the critical value for breakage, the continuous phase flow velocity is decreased by a step size set according to the program; the adjustment process adopts incremental numerical control. The flow rate deviation data of each branch channel are collected synchronously, and the driving parameters of the corresponding channel are finely adjusted according to the preset distribution coefficient of the program to adjust the uniformity of the feed flow distribution of the parallel nozzles.

[0032] Furthermore, the execution module includes a parallel integrated stepped micro-nozzle array and corresponding channel numerical control logic. Each nozzle is equipped with a dispersed phase inlet, a stepped platform, a continuous phase pre-filled area, and a stepped drop structure. The dispersed phase fluid enters the stepped platform channel of the nozzle after forming filaments in the pre-stretching area. The feeding sequence and pressure state of the nozzle are controlled by a unified numerical control cycle, so that the nozzle enters a steady-state generation state synchronously. The droplet cutting and peeling process is dominated by the stepped geometry and interfacial tension. During steady-state operation, sensor signals at the outlet end are collected at a fixed sampling rate. Abnormal conditions such as nozzle blockage and droplet coalescence are judged by a programmed feature recognition algorithm, and the number of working nozzles can be dynamically configured. The start-up and shutdown process of a single channel adopts a gradual flow rate adjustment, and the start-up and shutdown sequence is matched with the peak-shaving of adjacent channels.

[0033] The dynamic configuration of the number of working nozzles supports two triggering modes: manual setting and abnormal triggering. In manual mode, the user sets the target number of working nozzles according to the production throughput requirements. In abnormal triggering mode, when a single nozzle experiences an unrecoverable abnormality such as blockage, the system automatically disconnects the faulty channel and activates the corresponding backup channel. After the configuration command is issued, the system first recalculates the main pipe flow distribution coefficient based on the target number of working nozzles, and synchronously updates the flow velocity reference value and drive parameter threshold of each working channel.

[0034] New channels are started up one at a time following the staggered start-up and shutdown rule, while disconnected channels are gradually taken out of operation according to the gradual deceleration rule. During the adjustment process, small flow compensation is performed on the remaining steady-state working channels to maintain the stability of the total flow rate of the main pipe and the overall droplet generation flux.

[0035] After configuration, the system automatically verifies the flow rate deviation and droplet generation status of all working channels. Once it confirms that all channels have entered the steady-state generation range, the configuration process is completed, and the final configuration parameters are synchronously updated to the corresponding process CNC program package.

[0036] Furthermore, the feature detection module adopts a sensing scheme that combines capacitance detection and optical detection. The capacitance detection unit is integrated on both sides of the microchannel at the nozzle outlet and adopts a differential capacitance sensing structure. All detection channels are sampled synchronously by a unified numerical control clock. When a droplet passes through the detection area, the acquisition circuit captures the capacitive pulse signal, and calculates the droplet volume based on the pulse width and the real-time flow rate of the corresponding channel, thus completing the online counting and size measurement of each droplet in each channel. The optical detection unit adopts a combination of microscopic imaging and image processing. The optical acquisition is synchronously triggered by the CNC program according to preset triggering conditions, including the cumulative number of generated droplets, the ambient temperature fluctuation threshold, and the continuous running time. High-speed cameras are used to acquire droplet images of the exit area, and edge detection algorithms are used to extract the droplet contours and calculate the equivalent diameter and roundness parameters. During the calibration process, the CNC system synchronously latches the capacitance detection data at the corresponding moment and updates the capacitance and size conversion coefficients through a programmed fitting algorithm. Statistical calculations are performed on continuously collected droplet samples at a fixed statistical rhythm to output characteristic parameters such as average droplet diameter, coefficient of variation, and generation frequency. The data is transmitted to the adjustment module in real time. A programmed feature recognition algorithm is used to screen out droplets with out-of-tolerance size and irregular shape, and to count the abnormal proportion, triggering the corresponding operating condition event identifier.

[0037] Furthermore, the adjustment module adopts an event-triggered numerical control cascade adaptive control algorithm and uses an outer loop and inner loop cascade control architecture. The inner loop is a flow rate control loop, which realizes flow rate control based on dual feedback of injection pump pressure and displacement, and the response cycle is synchronized with the numerical control clock of the drive unit to suppress underlying disturbances. The outer loop is a droplet mass control loop, which corrects the flow rate setpoint of the inner loop based on droplet size detection feedback. The outer loop control adopts an event-driven variable parameter PID algorithm with a built-in preset operating condition event library. The CNC program identifies the current operating condition event in real time and automatically switches the matching PID parameter group and adjustment strategy. The variable parameter PID algorithm optimizes the PID parameter group of each operating condition online based on the historical adjustment effect and updates it to the iterative learning parameter library synchronously. When the adjustment module identifies changes in operating conditions, it adjusts the control parameters in advance based on the parameter calculation model and iterative learning data; when it detects that the droplet size is continuously out of tolerance and cannot be corrected by conventional adjustment, it triggers a system calibration condition event and starts the system calibration process according to a preset procedure.

[0038] This invention also provides an adaptive closed-loop control method for microfluidic droplet generation parameters, based on the above-described system, comprising the following specific steps: S1. Receives the target droplet process parameters and fluid properties input by the user, performs parameter validity verification in conjunction with the chip geometric boundaries, calls the process CNC program package and sends out configuration parameters, and completes full-link communication and driver timing self-check; displays the running status and droplet statistics in real time during operation, records running data according to the CNC clock cycle and supports exporting running logs; S2. Call the stepped emulsified droplet size prediction model of the working condition partition, establish the droplet equilibrium volume correlation equation with chip geometric parameters and fluid properties as constraints, and match the model correction coefficients for the corresponding working conditions; calculate the two-phase reference flow velocity according to the target droplet diameter, adjust the continuous phase flow velocity in combination with the flow field stability margin, and iteratively update the model correction coefficients based on the operating data. S3. The injection pumps in each channel are synchronously triggered by a unified numerical control clock to run smoothly. The drive error is compensated by displacement and pressure dual feedback, and pipeline abnormalities are monitored and protection is implemented. The flow field stability is evaluated by detecting the fluctuation of the fluid filament diameter in the pre-stretching zone. The continuous phase flow velocity is adjusted incrementally to balance the feed flow distribution of each branch channel. S4. Control each stepped micro-nozzle to feed synchronously at a uniform pace, and dominate the formation of droplet breakage through the stepped geometry and interfacial tension; collect outlet sensor signals during steady-state operation, identify abnormal working conditions, and support dynamic configuration of the number of nozzles and gradual staggered start and stop. S5. Differential capacitance sensing is used to realize the counting and size measurement of droplets in the whole sample. Microscopic optical detection is started according to the preset trigger conditions. The capacitance and size conversion coefficients are updated based on the optical results. The droplet characteristic parameters and abnormality ratio data are statistically output. S6. An outer loop and inner loop cascade control architecture is adopted. The inner loop realizes flow rate control based on pressure and displacement dual feedback. The outer loop corrects the flow rate setpoint using an event-driven variable parameter PID algorithm based on droplet detection data. Combined with working condition identification, feedforward compensation is performed. When continuous out-of-tolerance is detected, the system calibration process is triggered.

[0039] The beneficial effects of this invention are as follows: 1. The system of this invention incorporates a stepped emulsion droplet size prediction model based on working condition partitioning. Using chip geometric parameters such as stepped platform height, platform width, and step drop as constraints, and combining fluid viscosity, interfacial tension, and other physical properties, it establishes a droplet equilibrium volume correlation equation. This allows for direct calculation of the two-phase flow velocity baseline value based on the target droplet diameter, shortening the process debugging cycle and lowering the system's operational threshold. The system divides fluid properties and environmental parameters into multiple preset working condition intervals, each matched with an independent set of model correction coefficients. Upon startup, it calls the historically optimal parameters from the iterative learning parameter library under the same working condition as the initial baseline, while simultaneously calibrating the continuous phase control flow rate based on the flow field stability margin. After each production batch is completed, it automatically extracts steady-state operating data to update the parameter library. During operation, it can also fine-tune the correction coefficients online based on real-time closed-loop feedback, adaptively compensating for size deviations caused by fluid property drift and environmental temperature fluctuations.

[0040] 2. The inner loop of this invention is a flow rate control loop, which achieves precise flow rate control based on dual feedback of injection pump push rod displacement and pipeline pressure. The response cycle is synchronized with the CNC clock of the drive unit, which can quickly suppress underlying disturbances such as stepper motor step loss and pipeline pressure fluctuations, and prevent flow rate fluctuations from being directly transmitted to droplet size. The outer loop is a droplet quality control loop, which adopts an event-driven variable parameter PID algorithm and has a built-in preset operating condition event library. It can identify the current operating state in real time and automatically switch the matching control parameter group: a large proportional gain is used to accelerate the convergence speed during the large deviation start-up stage, the proportional gain is reduced and the integral action is enhanced in the steady-state operating range to ensure control accuracy, and the derivative action is enhanced to suppress fluctuations when oscillations occur. At the same time, combined with incremental flow rate adjustment based on fluid filament state detection in the pre-stretching zone, parallel channel flow equalization adjustment, and feedforward compensation mechanism when operating conditions change, it can effectively reduce the risk of control overshoot and response lag, and improve the droplet size control accuracy under complex disturbances.

[0041] 3. The execution module of this invention adopts a parallel integrated stepped micro-nozzle array structure. A single chip can integrate multiple independent nozzles, supporting dynamic configuration of the number of working nozzles. Combined with single-channel gradual flow rate adjustment and staggered start-stop control of adjacent channels, it can flexibly adapt to different production throughput requirements, and the start-stop process will not generate flow impact to disturb the steady flow field. The detection end adopts a sensing scheme combining differential capacitance and microscopic optics. The differential capacitance detection unit is integrated at the outlet of each nozzle, which can realize online droplet counting and size measurement of the entire sample. Microscopic optical detection is triggered by preset conditions such as cumulative droplet count and temperature fluctuation threshold. The conversion coefficient between capacitance and size is periodically calibrated through image recognition results, balancing detection efficiency and measurement accuracy. At the same time, the system has the ability to automatically identify abnormal operating conditions such as nozzle blockage and droplet aggregation. Combined with pipeline pressure abnormality grade protection and programmed automatic liquid replacement function, it can promptly handle operational abnormalities and achieve uninterrupted liquid replacement. Attached Figure Description

[0042] Figure 1 This is the main flowchart of the adaptive closed-loop control for microfluidic droplet generation in this invention; Figure 2 This is a flowchart of the cascade adaptive closed-loop regulator of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0044] like Figures 1 to 2 As shown, this embodiment of the invention provides an adaptive closed-loop control system for microfluidic droplet generation parameters, including a task configuration module, a parameter calculation module, a drive module, a control module, an execution module, a feature detection module, and an adjustment module; In this embodiment of the invention, the task configuration module receives user input via a touch screen, including the target droplet diameter, upper limit of the coefficient of variation of size, required throughput, and parameters such as dispersed phase viscosity, continuous phase viscosity, interfacial tension coefficient between the two phases, and ambient temperature. The task configuration module has built-in parameter validity verification logic. Combining chip geometric structure parameters and CNC program running boundaries, it determines whether the input parameters are within the allowable working range and outputs warning prompts for parameters that exceed the boundaries. The process numerical control program storage and recall unit stores process numerical control program packages. Each program package contains a set of control parameters for the corresponding fluid formula and droplet specifications, operating condition event response logic, and iterative learning history data, and supports one-click recall and program version management. When the system starts, it receives the process scheme selected by the user, sends the configuration parameters and corresponding control program segments to the parameter calculation module, starts the CNC self-test process, and verifies the communication link, drive timing and running status of each functional module in sequence according to the preset program; During operation, the system displays the flow rate, droplet generation frequency, size statistics, and fault alarm information of each channel in real time. It automatically records the operating data according to the CNC clock cycle and supports exporting the full cycle operation log.

[0045] In this embodiment of the invention, the parameter calculation module incorporates a stepped emulsion droplet size prediction model based on working condition partitioning. Using the height, width, and drop of the stepped platform as geometric constraints, and combining fluid viscosity and interfacial tension properties, it establishes a correlation equation between the droplet equilibrium volume and geometric and flow velocity parameters. The fluid properties and environmental parameters are divided into multiple preset working condition intervals, each corresponding to an independent set of model correction coefficients, which are automatically matched and called by the CNC program. The input parameters of the stepped emulsion droplet size prediction model include three types of chip geometric parameters: stepped platform height, platform width, and stepped drop; three types of fluid physical property parameters: dispersed phase viscosity, continuous phase viscosity, and interfacial tension between the two phases; and two types of control parameters: dispersed phase velocity and continuous phase velocity. The output parameters of the stepped emulsion droplet size prediction model are the droplet equilibrium volume and the corresponding equivalent droplet diameter.

[0046] The operating conditions are divided into three dimensions: dispersed phase viscosity, continuous phase viscosity, and ambient temperature. Each dimension is further divided into 3 to 5 segments, which are then combined to form multiple independent operating condition ranges. Each operating condition range corresponds to an independent set of model correction coefficients, which include three types: geometric correction coefficients, viscosity correction coefficients, and flow rate correction coefficients. These three types of coefficients linearly correct the calculation results for the geometric, physical property, and flow rate terms, respectively.

[0047] During batch updates, the deviation between the measured average diameter of at least 1000 consecutive effective droplets in the steady-state phase and the predicted value from the step emulsion droplet size prediction model is extracted. The adjustment amount of the three types of coefficients is then fitted and corrected using the least squares method to complete the coefficient update. During online fine-tuning, the average size deviation of 100 consecutive effective droplets is taken, and the correction coefficient for the current operating condition is slightly corrected according to a preset ratio, with each correction not exceeding 5% of the original coefficient value.

[0048] After receiving the target droplet diameter, the historical best parameters under the same working conditions in the iterative learning parameter library are used as the initial benchmark. The theoretical equilibrium volume is calculated by combining the chip's fixed geometric parameters, and the benchmark flow rate of the dispersed phase is obtained by reverse calculation. Operating condition matching uses three parameters—dispersed phase viscosity, continuous phase viscosity, and ambient temperature—as matching dimensions. It calculates the deviation between the current input parameter and the center parameter value of each operating condition interval. When the deviations of the three parameters all fall within the corresponding interval range, it is determined to be a perfect match, and the historical optimal parameter of that operating condition interval is directly called.

[0049] When only some parameters fall within the range, the two adjacent operating condition ranges with the smallest deviation are selected, and parameter interpolation is performed according to the deviation ratio to obtain the initial benchmark value that adapts to the current parameters. The historical best parameters in the iterative learning parameter library are all selected and extracted from the operating batches that meet the steady-state running time requirements and whose droplet size variation coefficient is better than the process indicators. Multiple sets of historical parameters are retained for each operating condition range, and the set with the best effect is selected as the default call value.

[0050] When there are no matching historical parameters, the system uses a general basic correction coefficient to complete the initial calculation and stores the current running data as the first batch of historical data for this operating condition range into the parameter library.

[0051] The critical velocity value for droplet breakage is solved by the flow field shear force calculation model. Based on the principle of reserving the flow field stability margin, the continuous phase control velocity is set to a value lower than the critical velocity. The flow field stability margin in the pre-stretching zone is calculated. When the fluctuation of physical property parameters leads to insufficient flow field stability margin, the continuous phase velocity is adjusted according to the preset adjustment step size of the program. The critical flow velocity for droplet breakage corresponds to the flow velocity threshold at which the dispersed phase filaments undergo uncontrollable breakage under the shearing action of the continuous phase. Once this threshold is exceeded, the droplet breakage location will randomly shift forward, and the droplet size dispersion will increase significantly.

[0052] The flow field stability margin is defined as the ratio of the critical velocity for droplet breakage to the actual continuous phase control velocity. A larger margin value indicates a more sufficient steady-state margin in the flow field and a more stable droplet formation state. The system presets a reasonable operating range for the stability margin. During operation, it calculates the actual margin value in real time based on the current fluid properties and velocity parameters, and dynamically corrects the margin calculation results by combining the fluid filament diameter fluctuation data uploaded by the control module.

[0053] When the actual margin value is lower than the preset lower threshold, it is determined that the flow field stability margin is insufficient, posing a risk of premature droplet breakage and undersized droplets. When the actual margin value is higher than the preset upper threshold, it is determined that the flow field shearing effect is insufficient, posing a problem of oversized droplets and low generation frequency. The margin determination result directly serves as the triggering basis for continuous phase velocity adjustment. During the adjustment process, incremental adjustment is performed with the goal of returning the margin to the middle of the reasonable range.

[0054] Formula for calculating flow field stability margin: This represents the stability margin of the flow field, used to quantify the steady-state margin of the current flow field; the larger the value, the farther the flow field deviates from the critical state of fracture, and the higher the steady-state reliability of droplet generation.

[0055] This represents the critical flow velocity for droplet breakage, the continuous phase flow velocity threshold at which the dispersed phase filaments undergo uncontrollable random breakage under the shearing action of the continuous phase; after exceeding this threshold, the droplet breakage location shifts irregularly forward, and the size dispersion increases significantly.

[0056] This represents the actual continuous phase control velocity, which is the real-time operating velocity of the continuous phase channel currently output by the system, and is different from the theoretical reference velocity in the parameter calculation stage.

[0057] After each production batch is completed, the system automatically extracts the parameter matching data from the steady-state operation phase within the batch, updates the model correction coefficients and iterative learning parameter library for the corresponding operating condition range, and fine-tunes the model correction coefficients under the current operating condition online based on the real-time closed-loop feedback statistical results.

[0058] In this embodiment of the invention, the driving module adopts stepper motor microstepping driving technology, with the propulsion control accuracy set to 0.1 micrometers per step. All channel driving units are synchronously triggered by a unified numerical control clock, receive flow rate setpoints and timing control commands, and smoothly drive the injection pump to run according to a preset program acceleration curve. During operation, the displacement of the injection pump push rod and the pressure signal of the pipeline are collected in real time according to a fixed sampling rhythm. The displacement feedback is used to compensate for the stepper motor's step loss error, and the pressure feedback is used to monitor pipeline blockage and abnormal fluctuations. When an abnormal increase in pressure is detected, the graded deceleration protection and fault alarm are executed according to the preset program, and the working condition event identifier is triggered and uploaded to the main control unit. The displacement feedback uses a high-precision displacement sensor to collect the actual displacement of the injection pump push rod in real time. In each sampling period, the theoretical displacement value is compared with the actual displacement value, and the difference between the two is the cumulative step loss deviation of the stepper motor.

[0059] When the cumulative step loss deviation reaches the displacement corresponding to a single step pulse, the system inserts a compensation pulse into the subsequent drive pulse sequence. The compensation pulses are inserted in multiple drive cycles, and the number of compensation pulses inserted in a single cycle does not exceed the set upper limit to avoid instantaneous changes in flow velocity.

[0060] The system sets an upper limit for cumulative step loss compensation. When the step loss deviation continues to increase after continuous compensation, it is determined that there is a mechanical fault or pipeline blockage in the drive system, and the compensation action is stopped and the pressure abnormality check logic is triggered.

[0061] The drive module has a built-in programmed automatic fluid change control program. When a syringe is advanced to the end of its stroke, the backup syringe pre-filling process is started in advance according to a preset timing sequence. The flow rate transition algorithm is used to achieve the flow connection of the fluid change process. The timing of the fluid change process is controlled by a numerical control program.

[0062] The stroke switching threshold of the main syringe is set at a buffer stroke position reserved at the end of the stroke. The length of the buffer stroke is calculated based on the liquid change transition time and the set flow rate to ensure that the main syringe will not advance to the physical limit position before the liquid change process is completed.

[0063] During the pre-charging phase, the backup syringe is advanced at a low constant speed to completely expel the air at the front end of the tubing and fill it with the corresponding fluid. The pre-charging time is calculated and determined based on the tubing volume and the pre-charging flow rate. After the pre-charging is completed, the backup syringe remains in standby mode.

[0064] When entering the fluid change transition phase, the main syringe gradually reduces the propulsion flow rate according to the preset linear slope, while the backup syringe simultaneously increases the propulsion flow rate according to the matching linear slope. Throughout the transition process, the sum of the flow rates of the main and backup syringes remains at the set working flow rate to avoid step fluctuations in the total flow rate.

[0065] During the transition phase, the pipeline pressure changes are continuously monitored. If the pressure fluctuation exceeds the allowable range, the flow rate change slope is automatically adjusted to extend the transition time. When the flow rate of the main syringe drops to zero and the standby syringe reaches the set working flow rate, the fluid replacement process is completed and the system enters normal operation.

[0066] In this embodiment of the invention, the control module corresponds to the flow control focusing or co-flow coaxial sheath flow configuration of the pre-stretching zone, receives the continuous phase flow velocity setpoint and numerical control adjustment command output by the parameter calculation module, and outputs control command to the drive module to make the continuous phase injection pump run in a synchronous sequence. The stability of the flow field is assessed by detecting the diameter fluctuation of the fluid filament downstream of the pre-stretching zone. When changes in ambient temperature or fluid property drift cause changes in the stretching state, a small incremental adjustment of the continuous phase flow velocity is performed. The flow field stability assessment algorithm takes continuously sampled values ​​of the fluid filament diameter as input and evaluates two metrics: the coefficient of variation (COP) and the diameter fluctuation frequency. The COP is calculated for 50 consecutive sampling points. A steady-state flow is defined as a COP below a preset stability threshold, while an unstable flow is defined as a COP exceeding a preset fluctuation threshold and having a fluctuation frequency higher than a set value. The flow field stability assessment results are directly used as the trigger for fine-tuning the continuous phase velocity and are simultaneously output to the parameter calculation module for real-time calibration of the flow field stability margin.

[0067] The diameter of the fluid filament is obtained using a microscopic optical edge detection method. A microscopic imaging acquisition unit with a fixed field of view is set up in the transparent flow channel section downstream of the pre-stretching zone. Transmission images of the fluid filament in the flow channel are continuously acquired at a fixed sampling frequency. The fluid boundary contours on both sides of the filament are identified by a sub-pixel edge extraction algorithm, and the real-time diameter value of the filament perpendicular to the flow direction is calculated.

[0068] The raw diameter data is first processed by moving average filtering to remove image noise and abnormal jump values ​​caused by instantaneous flow field disturbances, and then output to the flow field stability evaluation logic. The sampling frequency of diameter detection is much higher than the adjustment frequency of the outer ring quality control loop to ensure that high-frequency small fluctuations of the filament can be captured.

[0069] When the diameter of the fluid filament is detected to be too large and irregular fluctuations are observed, the continuous phase flow rate is increased by a step size set according to the program; when the diameter of the filament is detected to be close to the critical value for breakage, the continuous phase flow rate is decreased by a step size set according to the program; the adjustment process adopts incremental numerical control, and the single adjustment amount does not exceed 2% of the set value. The adjustment actions of all branch channels are executed synchronously by a unified clock. The flow rate deviation data of each branch channel are collected synchronously, and the driving parameters of the corresponding channel are finely adjusted according to the preset distribution coefficient of the program to adjust the uniformity of the feed flow distribution of the parallel nozzles.

[0070] The flow balancing adjustment uses the average flow rate of all working channels within a fixed statistical period as the benchmark value. The flow rate data of each channel is the average value of multiple consecutive sampling periods, and the flow rate jump value caused by instantaneous disturbances is eliminated. The relative deviation between the actual average flow rate of a single channel and the benchmark flow rate is calculated. When the relative deviation exceeds the preset allowable range, the flow rate fine-tuning action of the corresponding channel is triggered.

[0071] The adjustment step size is adaptively adjusted according to the deviation. When the deviation is large, a larger single-step adjustment amount is used to accelerate the convergence speed, and when the deviation is small, a smaller single-step adjustment amount is used to avoid overshoot. The maximum single-step adjustment amount does not exceed 1% of the current drive value. The adjustment actions of all channels are synchronously triggered by a unified CNC clock to avoid repeated transmission of main pipe pressure fluctuations caused by individual channel adjustments.

[0072] When the relative deviation of the flow rate of all channels is within the allowable range for multiple consecutive statistical periods, the flow balance adjustment is determined to be converged, and the system maintains the current driving parameters.

[0073] In this embodiment of the invention, the execution module includes a parallel integrated stepped micro-nozzle array and corresponding channel numerical control logic. Eight stepped micro-nozzles are integrated in parallel on a single chip. Each nozzle is provided with a dispersed phase inlet, a stepped platform, a continuous phase pre-filled area and a stepped drop structure. The dispersed phase fluid enters the stepped platform channel of the nozzle after forming filaments in the pre-stretching area. The feeding sequence and pressure state of the nozzle are controlled by a unified numerical control cycle, so that the nozzle enters a steady-state generation state synchronously. The droplet cutting and peeling process is dominated by the stepped geometry and interfacial tension. After being treated in the pre-stretching zone, the dispersed phase filaments enter the stepped platform channel of the nozzle. The cross-section of the flow channel in the stepped platform section remains constant, and the continuous phase fluid uniformly coats the dispersed phase filaments along the channel wall. During this stage, the shear force distribution of the flow field is stable, which can maintain the morphological stability of the dispersed phase filaments and prevent the filaments from oscillating irregularly or breaking prematurely.

[0074] When the dispersed phase filaments flow with the fluid to the position of the stepped drop, the flow channel cross section suddenly expands outward, and the flow direction and velocity distribution of the continuous phase fluid change abruptly. Local eddies and shear force steps are formed at the step corners, generating additional shear force on the neck of the dispersed phase filaments. At the same time, the interfacial tension between the two phases continues to act on the free surface of the filaments, causing the neck of the filaments to continue to contract.

[0075] The shear force abrupt change brought about by the stepped structure, combined with the contraction effect of interfacial tension, accelerates the fracture process at the neck of the filament, causing the dispersed phase filament to break into uniform independent droplets at a fixed position, resulting in a higher consistency between the droplet formation location and the fracture timing.

[0076] During steady-state operation, sensor signals at the outlet end are collected at a fixed sampling rate. Abnormal conditions such as nozzle blockage and droplet coalescence are judged by a programmed feature recognition algorithm, and the number of working nozzles can be dynamically configured. The start-up and shutdown process of a single channel adopts a gradual flow rate adjustment, and the start-up and shutdown sequence is matched with the peak-shaving of adjacent channels.

[0077] The abnormal operating condition feature recognition algorithm takes into account three types of feature data from the outlet capacitive sensing: pulse amplitude, pulse interval, and pulse width. The nozzle blockage recognition logic is as follows: when there are no valid capacitive pulses in a single channel for three consecutive detection cycles, and the corresponding channel pipeline pressure is higher than the normal pressure threshold, it is determined to be a nozzle blockage condition.

[0078] The droplet coalescence identification logic is as follows: when a single pulse width is detected to exceed 1.5 times the normal droplet pulse width, and the pulse amplitude does not decrease significantly, it is determined to be a droplet coalescence condition. The abnormal condition feature identification algorithm outputs the abnormal condition identifier and the time of occurrence data of the corresponding channel, and synchronously triggers the corresponding level of alarm and handling logic.

[0079] When starting a single channel, the flow rate gradually increases from zero to the set working flow rate at a preset linear slope, and gradually decreases to zero at the same slope when stopping. The slope of the flow rate change is set according to the pipeline pressure fluctuation tolerance threshold to avoid flow shock caused by flow rate step.

[0080] The start-up and shutdown sequence of a single channel is staggered from that of the adjacent channels by a fixed time interval. The staggered duration is longer than the total duration of the start-up and shutdown process of a single channel and is also longer than the natural decay period of pipeline pressure fluctuations. This ensures that the pressure fluctuations caused by the start-up and shutdown of a single channel are completely decayed before the start-up and shutdown of the adjacent channels are executed.

[0081] During the start-up and shutdown process, the other channels that are in steady-state operation will simultaneously perform minor flow compensation adjustments to offset the flow rate fluctuations caused by changes in the main pipe pressure. When it is necessary to adjust the working status of multiple channels in batches, the system will perform start-up and shutdown operations in staggered order according to the channel arrangement, and only a single channel is allowed to be in the start-up and shutdown transition state at the same time.

[0082] In this embodiment of the invention, the feature detection module adopts a sensing scheme that combines capacitance detection and optical detection. The capacitance detection unit is integrated on both sides of the microchannel at the nozzle outlet and adopts a differential capacitance sensing structure. All detection channels are sampled synchronously by a unified numerical control clock. The multi-channel capacitance detection adopts a synchronous trigger time-division reading working mode. The sampling trigger signals of all channels are issued synchronously by a unified numerical control clock to ensure that the detection time base of each channel is completely consistent and the timing correspondence between droplet counting and size calculation is accurate.

[0083] Each channel's detection electrode is surrounded by an independent shielding ring structure. The shielding ring and the corresponding detection electrode are at the same potential, suppressing electric field crosstalk between adjacent channels and environmental common-mode interference. A multi-stage filtering network is set at the front end of the acquisition circuit to filter out high-frequency noise introduced by the stepper motor drive and power supply, improving the signal-to-noise ratio of the capacitor pulse signal. The detection zero point of each channel is automatically calibrated during the system self-test phase, and baseline dynamic tracking is performed periodically during operation to compensate for detection baseline drift caused by changes in ambient temperature and humidity.

[0084] When a droplet passes through the detection area, the acquisition circuit captures the capacitive pulse signal, and calculates the droplet volume based on the pulse width and the real-time flow rate of the corresponding channel, thus completing the online counting and size measurement of each droplet in each channel. The differential capacitance detection unit identifies droplets based on the difference in dielectric constant between the two phase fluids. There is a fixed difference in dielectric constant between the continuous phase fluid and the dispersed phase droplets. When the droplets flow through the detection electrode area, the capacitance value of the differential capacitor will generate a pulse change of corresponding amplitude.

[0085] The detection circuit determines the effective droplet pulse by setting a preset amplitude threshold, and records the start time, end time and peak amplitude of the pulse. The time it takes for a single droplet to pass through the detection area is calculated by the difference between the end time and the start time.

[0086] By combining the real-time flow rate of the corresponding channel with the fixed cross-sectional dimensions of the microchannel, the axial length of the droplet in the channel can be calculated. This length is then converted into the actual volume of the droplet by combining the cross-sectional constraints of the microchannel. Finally, the volume parameter is converted into the equivalent spherical diameter parameter.

[0087] Droplet counting is based on the number of valid pulses. Each pulse corresponds to the counting of only one droplet. Continuous abnormal pulses with an interval less than a set threshold will be marked as samples to be reviewed and will not be included in regular counting and size statistics.

[0088] Formula for converting droplet size in capacitance detection: It represents the measured volume of a single droplet, which is the actual volume of the generated droplet calculated based on the sensor signal; it includes the comprehensive influence of actual operating factors such as flow channel processing errors, driving flow velocity fluctuations, and environmental condition drift.

[0089] It represents the measured equivalent spherical diameter of the droplet, which is the actual size characterization value obtained by converting the measured volume of a single droplet; it represents the true size of the droplets generated in actual operation, and is the core feedback input of the outer loop quality control loop, used to calculate the size deviation and drive the closed loop adjustment.

[0090] This represents the capacitance size conversion factor, a proportional coefficient that is calibrated and updated based on optical detection results. It is used to compensate for system measurement errors caused by differences in the dielectric constants of the two phases and electrode processing deviations.

[0091] This indicates the real-time flow velocity of the detection channel. The actual operating flow velocity of the fluid in the channel at the detection moment is provided by the displacement feedback data of the drive module.

[0092] This represents the capacitance pulse width, the time it takes for a single droplet to completely pass through the capacitance detection area, which is measured by the capacitance acquisition circuit.

[0093] This represents the cross-sectional area of ​​the detection segment microchannel, which is a fixed geometric parameter of the chip.

[0094] The optical detection unit adopts a combination of microscopic imaging and image processing. The optical acquisition is synchronously triggered by the CNC program according to preset triggering conditions, including the cumulative number of generated droplets, the ambient temperature fluctuation threshold, and the continuous running time. High-speed cameras are used to acquire droplet images of the exit area, and edge detection algorithms are used to extract the droplet contours and calculate the equivalent diameter and roundness parameters. During the calibration process, the CNC system synchronously latches the capacitance detection data at the corresponding moment and updates the capacitance and size conversion coefficients through a programmed fitting algorithm. Before starting the calibration process, the system's operating status must be determined. Calibration is only allowed when the droplet generation frequency and size fluctuations are both within the steady-state range and the continuous stable running time reaches the preset value.

[0095] During calibration, at least 200 steadily generated droplets are selected as the calibration sample set. Using the timestamp of a unified numerical control clock as the reference, the equivalent diameter data of each droplet obtained by optical detection is matched one by one with the original data of capacitance detection at the same time, and abnormal samples with mismatched timing or out-of-roundness are eliminated.

[0096] Using the equivalent diameter obtained from optical detection as the true value of the size reference, the calculated value of the capacitance detection is linearly fitted to obtain the updated size conversion coefficient and zero offset. After the fitting is completed, the calibration effect is verified by residual verification. When the average size deviation of the calibrated sample is greater than the preset allowable value, the calibration sample size is automatically increased and the calibration process is repeated until the deviation meets the process requirements.

[0097] Statistical calculations are performed on continuously collected droplet samples at a fixed statistical rhythm to output characteristic parameters such as average droplet diameter, coefficient of variation, and generation frequency. The data is transmitted to the adjustment module in real time. A programmed feature recognition algorithm is used to screen out droplets with out-of-tolerance size and irregular shape, and to count the abnormal proportion, triggering the corresponding operating condition event identifier.

[0098] The determination of out-of-tolerance droplets is based on the target droplet diameter. When the relative deviation between the equivalent diameter of a single droplet and the target value exceeds the preset allowable range, it is marked as an out-of-tolerance sample. The determination of irregularly shaped droplets combines two characteristics: the droplet roundness is lower than the preset threshold under optical detection, or the symmetry of the rising and falling edges of the pulse waveform exceeds the set range under capacitive detection. If either condition is met, it is marked as an irregularly shaped droplet.

[0099] The anomaly percentage statistics employ a sliding sample window mechanism, using a fixed number of consecutive droplet samples as the statistical window to calculate the proportion of abnormal droplets within the window relative to the total number of samples. The system sets two levels of anomaly trigger thresholds: a low threshold corresponds to minor anomalies, triggering only the operating condition record and fine-tuning strategy; a high threshold corresponds to severe anomalies, directly triggering the operating condition event switch of the regulation module and initiating targeted balancing strategies. Different types of anomalies correspond to different operating condition event identifiers, each matched with independent adjustment parameter groups and handling logic.

[0100] In this embodiment of the invention, the adjustment module adopts an event-triggered numerical control cascade adaptive control algorithm and uses an outer loop and inner loop cascade control architecture. The inner loop is a flow rate control loop, which realizes flow rate control based on dual feedback of injection pump pressure and displacement, and the response cycle is synchronized with the numerical control clock of the drive unit to suppress underlying disturbances such as pressure fluctuations and step loss. The outer loop is a droplet mass control loop, which corrects the flow rate setpoint of the inner loop based on droplet size detection feedback. The flow rate control loop uses a sampling cycle that is synchronized with the drive clock. Each drive pulse cycle completes one sampling of displacement and pressure data and deviation calculation. The deviation directly affects the drive pulse output of the next cycle.

[0101] The droplet quality control loop uses a sampling and statistical period much shorter than that of the inner loop. It uses the statistical values ​​of a fixed number of continuous droplet samples as the input for one adjustment. The data interaction between the inner and outer loops adopts a register caching mechanism. The inner loop reads the flow rate setpoint register updated by the outer loop in real time, while the outer loop only updates the register value after completing one statistical calculation. An upper limit constraint is set on the adjustment amount of a single update to prevent large deviation adjustments in the outer loop from causing drastic fluctuations in the flow rate of the inner loop.

[0102] The outer loop control employs an event-driven variable-parameter PID algorithm with a built-in preset operating condition event library. The CNC program identifies current operating condition events in real time and automatically switches between matching PID parameter sets and adjustment strategies. A large proportional gain is used under large deviation conditions; the proportional gain is reduced and the integral action is enhanced within the steady-state range; and the derivative action is enhanced when oscillations occur. The variable-parameter PID algorithm optimizes the PID parameter sets for each operating condition online based on historical adjustment effects and updates them synchronously to the iterative learning parameter library. The operating condition event library includes four basic operating condition events: large deviation at startup, steady-state operation, oscillation trend, and property drift. The input parameters for the operating condition identification logic are four types of real-time data: absolute value of droplet size deviation, rate of change of size deviation, number of consecutive out-of-tolerance droplets, and change in ambient temperature.

[0103] When the absolute value of the droplet size deviation exceeds a preset threshold and the number of consecutive droplets is less than a set value, a large deviation event is detected, and the PID parameter group with high proportional gain and weak integral action is switched to the set value. When the absolute value of the droplet size deviation is continuously within the allowable range and the duration reaches a set value, a steady-state operation event is detected, and the PID parameter group with low proportional gain and strong integral action is switched to the set value. When the rate of change of dimensional deviation alternates between positive and negative values ​​and the amplitude exceeds the set threshold, it is determined to be an oscillation trend event, and the PID parameter group with enhanced derivative action is switched. When the change in ambient temperature exceeds the fluctuation threshold, it is determined to be a property drift event, and the initial value of the PID parameters is adjusted synchronously in conjunction with the feedforward compensation. The parameter group switching adopts a gradual transition mechanism to avoid abrupt changes in parameters that could lead to sudden changes in control output, thereby causing system flow velocity oscillations and droplet size fluctuations. After the operating condition event is identified and it is confirmed that the parameter group needs to be switched, the system will synchronously and gradually adjust the three PID parameters of the current operation (proportional, integral, and derivative) to the corresponding values ​​of the target parameter group according to the linear slope, according to the preset transition period. The transition period is adaptively adjusted according to the steady state of the current operating condition. The higher the steady state, the shorter the transition period. When the deviation is large, the transition period is appropriately extended.

[0104] During the transition, the system continuously monitors droplet size deviation and pipeline flow velocity fluctuations. If the deviation continues to increase or shows an oscillating trend during the transition, the parameter switching is immediately paused, and the transition action is resumed only after the deviation returns to a reasonable range. After the parameter switching is completed, the system automatically records the transition duration, steady-state deviation, and fluctuation amplitude data of this switch, and incorporates them into the iterative learning parameter library to optimize the transition period and slope parameters for subsequent similar operating conditions.

[0105] When the event-driven variable parameter PID algorithm iteratively learns and updates, it records the adjustment convergence time and steady-state deviation value under each type of working condition event. When the adjustment effect is better than the historical best value for three consecutive times under the same working condition, the corresponding PID parameters are overwritten and updated to the iterative learning parameter library.

[0106] When the adjustment module recognizes changes in operating conditions such as fluid replacement and ambient temperature changes, it adjusts the control parameters in advance based on the parameter calculation model and iterative learning data; when it detects that the droplet size is continuously out of tolerance and cannot be corrected by conventional adjustment, it triggers a system calibration condition event and starts the system calibration process according to the preset procedure.

[0107] Feedforward compensation performs pre-adjustments for predictable changes in operating conditions. The compensation inputs include three types of predictable changes: changes in ambient temperature, changes in fluid property parameters, and adjustments to the number of nozzles in operation. Feedforward compensation triggered by changes in ambient temperature calculates the pre-adjustment amount of the two-phase flow velocity in advance based on the magnitude of the temperature change and the droplet size temperature drift coefficient under the corresponding operating condition, thus offsetting the fluid viscosity and interfacial tension drift caused by temperature changes.

[0108] Feedforward compensation triggered by fluid property changes directly calls the stepped emulsion droplet size prediction model to recalculate the flow velocity baseline value under the corresponding properties, completing the pre-adjustment of the flow velocity setpoint before the fluid switching is completed. Feedforward compensation triggered by nozzle number adjustment adjusts the flow velocity baseline of each channel in advance according to the main pipe flow distribution relationship, avoiding flow deviation caused by main pipe pressure fluctuations.

[0109] All feedforward adjustment values ​​are subject to upper and lower limits to prevent excessive compensation from causing system oscillations. The feedforward adjustment value and the outer loop feedback correction value are superimposed and work together on the inner loop flow velocity setpoint. After each feedforward compensation is completed, the system records the compensation effect and updates the corresponding compensation coefficients in the iterative learning parameter library.

[0110] After the system calibration process is initiated, the adjustment of the outer ring quality control loop is first paused to maintain the steady-state operation of the inner ring flow rate control loop, thus avoiding repeated fluctuations in flow rate during the calibration process. Subsequently, the optical detection unit is triggered to perform continuous acquisition, acquiring no less than 500 continuously generated droplet images. The equivalent diameter obtained from image processing is used as the true value benchmark for size. Simultaneously, the raw capacitance detection data for the corresponding time period is acquired, and the capacitance size conversion coefficient and zero-point offset parameters are re-fitted and calculated.

[0111] After the coefficients are updated, the stepped emulsion droplet size prediction model is invoked, and the two-phase flow velocity baseline value is recalculated in conjunction with the current fluid properties and environmental parameters. The model correction coefficients for the corresponding operating range are updated synchronously. After the parameters are updated, the outer loop adjustment function is gradually restored to approach the target size in small steps to verify the calibration effect. If the size deviation still fails to meet the process requirements after verification, the calibration process is repeated once. If the second calibration is ineffective, a manual intervention alarm is triggered.

[0112] This invention also provides an adaptive closed-loop control method for microfluidic droplet generation parameters, based on the above-described system, including the following specific steps: S1. Receives the target droplet process parameters and fluid properties input by the user, performs parameter validity verification in conjunction with the chip geometric boundaries, calls the process CNC program package and sends out configuration parameters, and completes full-link communication and driver timing self-check; displays the running status and droplet statistics in real time during operation, records running data according to the CNC clock cycle and supports exporting running logs; S2. Call the stepped emulsified droplet size prediction model of the working condition partition, establish the droplet equilibrium volume correlation equation with chip geometric parameters and fluid properties as constraints, and match the model correction coefficients for the corresponding working conditions; calculate the two-phase reference flow velocity according to the target droplet diameter, adjust the continuous phase flow velocity in combination with the flow field stability margin, and iteratively update the model correction coefficients based on the operating data. S3. The injection pumps in each channel are synchronously triggered by a unified numerical control clock to run smoothly. The drive error is compensated by displacement and pressure dual feedback, and pipeline abnormalities are monitored and protection is implemented. The flow field stability is evaluated by detecting the fluctuation of the fluid filament diameter in the pre-stretching zone. The continuous phase flow velocity is adjusted incrementally to balance the feed flow distribution of each branch channel. S4. Control each stepped micro-nozzle to feed synchronously at a uniform pace, and dominate the formation of droplet breakage through the stepped geometry and interfacial tension; collect outlet sensor signals during steady-state operation, identify abnormal working conditions, and support dynamic configuration of the number of nozzles and gradual staggered start and stop. S5. Differential capacitance sensing is used to realize the counting and size measurement of droplets in the whole sample. Microscopic optical detection is started according to the preset trigger conditions. The capacitance and size conversion coefficients are updated based on the optical results. The droplet characteristic parameters and abnormality ratio data are statistically output. S6. An outer loop and inner loop cascade control architecture is adopted. The inner loop realizes flow rate control based on pressure and displacement dual feedback. The outer loop corrects the flow rate setpoint using an event-driven variable parameter PID algorithm based on droplet detection data. Combined with working condition identification, feedforward compensation is performed. When the droplet size is detected to be continuously out of tolerance, the system calibration process is triggered.

[0113] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0114] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A microfluidic droplet generation parameter adaptive closed-loop control system, characterized in that, Includes the following modules: The task configuration module receives process indicators and fluid property parameters, performs parameter validity verification and process numerical control program management, issues configuration parameters, and performs system self-checks, status displays, and operation log management. The parameter calculation module has a built-in droplet size prediction model. It calculates the two-phase flow velocity reference value based on the target droplet parameters, adjusts the control parameters in combination with the flow field stability, and iteratively updates the model correction coefficient based on the operation feedback. The drive module receives the flow rate setpoint and timing control command, drives the syringe pump to run, compensates for drive errors, and performs pipeline abnormality protection and automatic fluid replacement control. The control module adapts to the flow channel configuration of the pre-stretched zone, outputs control commands to drive the corresponding pump to operate synchronously, evaluates the stability of the flow field and fine-tunes the continuous phase flow velocity, and balances the flow distribution in the channel. The execution module includes a stepped micro-nozzle array and channel CNC logic, which controls the nozzle feeding sequence and pressure state, dominates the droplet breakage and generation process, and collects outlet sensor signals to identify abnormal working conditions. The feature detection module combines capacitance detection and optical detection sensing schemes to realize droplet counting and size measurement, calibrate the size conversion coefficient, and output droplet feature statistics and anomaly identification results; The adjustment module adopts an outer loop and inner loop cascade control architecture. The inner loop realizes flow rate control, and the outer loop adjusts the flow rate setpoint based on droplet feature detection data. Combined with operating condition identification, it performs feedforward compensation. When the droplet size is detected to be continuously out of tolerance, the system calibration process is triggered.

2. The microfluidic droplet generation parameter adaptive closed-loop control system according to claim 1, characterized in that, The task configuration module receives the target droplet diameter, upper limit of the coefficient of variation of size, generation flux requirement, as well as dispersed phase viscosity, continuous phase viscosity, interfacial tension coefficient between the two phases, and working environment temperature parameters through a touch screen. Built-in parameter validity verification logic combines chip geometric structure parameters and CNC program running boundaries to determine the validity of input parameter values ​​and outputs warnings for out-of-bounds parameters; The process CNC program storage unit stores process CNC program packages. Each program package contains the corresponding fluid formula, droplet specification control parameter set, operating condition event response logic, and iterative learning history data. When the system starts, it calls and sends the configuration parameters and control program segment corresponding to the selected process scheme, starts the CNC self-test process, and verifies the communication link, drive timing and running status of each module. During operation, the flow rate, generation frequency, size statistics, and alarm information are displayed in real time, and the operation data is recorded according to the CNC clock cycle.

3. The microfluidic droplet generation parameter adaptive closed-loop control system according to claim 2, characterized in that, The parameter calculation module has a built-in step emulsion droplet size prediction model based on working condition partitioning. It uses the height, width and drop of the step platform as geometric constraints, and combines physical properties such as fluid viscosity and interfacial tension to establish the correlation equation between droplet equilibrium volume and geometric and flow velocity parameters. The fluid properties and environmental parameters are divided into multiple preset working condition intervals. Each working condition interval corresponds to an independent set of model correction coefficients, which are automatically matched and called by the CNC program. After receiving the target droplet diameter, the historical best control parameters under the same operating conditions are used as the initial reference. The theoretical equilibrium volume is calculated by combining the chip geometric parameters, and the reference flow rate of the dispersed phase is deduced. Solve for the critical flow velocity value of droplet breakage under the corresponding working condition, set the continuous phase control flow velocity according to the flow field stability margin, and adjust the continuous phase flow velocity according to the preset step size when the stability margin is insufficient. After each production batch is completed, the parameter matching data of the steady-state operation phase is extracted, and the model correction coefficients of the corresponding operating condition range are updated; based on the real-time closed-loop feedback results, the model correction coefficients of the current operating condition are fine-tuned online.

4. The microfluidic droplet generation parameter adaptive closed-loop control system according to claim 3, characterized in that, The drive module adopts stepper motor microstepping drive technology. Each channel drive unit is synchronously triggered by a unified numerical control clock. After receiving the flow rate set value and timing control command, it smoothly drives the injection pump to run according to the preset curve. During operation, the displacement of the injection pump push rod and the pressure signal of the pipeline are collected at a fixed sampling rhythm. The displacement feedback is used to compensate for the step loss error, and the pressure feedback is used to monitor the pipeline abnormality. When the pressure is abnormal, the system will implement graded deceleration protection and alarm, and simultaneously upload the operating condition event identifier; The built-in automatic fluid change control program starts the backup pump in advance when the syringe reaches the end, and achieves flow connection through smooth flow rate transition.

5. The microfluidic droplet generation parameter adaptive closed-loop control system according to claim 4, characterized in that, The control module corresponds to the flow control focusing or co-flow coaxial sheath flow configuration of the pre-stretching zone. After receiving the flow rate set value and adjustment command, it drives the continuous phase injection pump to run synchronously. The stability of the flow field is assessed by detecting fluctuations in the diameter of the fluid filaments, and the continuous phase velocity is slightly adjusted when the stability of the flow field changes. When the filament diameter is too large and irregular fluctuations occur, the flow rate is increased; when it approaches the fracture critical value, the flow rate is decreased, and incremental numerical control adjustment is adopted. The flow rate deviation of each branch channel is collected, and the driving parameters are fine-tuned to balance the flow rate distribution of the parallel nozzles.

6. The microfluidic droplet generation parameter adaptive closed-loop control system according to claim 5, characterized in that, The execution module includes a parallel stepped micro-nozzle array and channel numerical control logic. Each nozzle is equipped with a dispersed phase inlet, a stepped platform, a continuous phase pre-filling area and a stepped drop structure. The dispersed phase fluid enters the stepped platform channel after being pre-stretched to form filaments. The nozzle feeding sequence and pressure are controlled by a unified numerical control cycle, so that the nozzle enters a steady state synchronously, and the droplet formation is dominated by the stepped structure and interfacial tension. In steady state, it collects outlet sensor signals to identify abnormal operating conditions, including blockage and coalescence, and supports dynamic configuration of the number of working nozzles. The single channel uses gradual flow rate adjustment to start and stop, and the start and stop timing is matched with the adjacent channels to avoid peak periods.

7. The microfluidic droplet generation parameter adaptive closed-loop control system according to claim 6, characterized in that, The capacitance detection unit of the feature detection module is integrated on both sides of the nozzle outlet microchannel, and adopts a differential capacitance structure, with each channel sampling synchronously. Capacitive pulses are captured as droplets pass by, and the droplet volume is calculated by combining the flow rate, enabling online counting and size measurement; Optical detection employs microscopic imaging and image processing, triggering acquisition based on preset conditions, including the cumulative number of generated droplets, ambient temperature fluctuation threshold, and continuous running time. Acquire droplet images and extract contours to calculate equivalent diameter and roundness; during calibration, synchronously latch capacitance detection data at corresponding moments and update size conversion coefficients; Statistically analyze the average droplet diameter, coefficient of variation, and generation frequency based on the beat rate, screen out out-of-tolerance and irregularly shaped droplets, and count the abnormal proportion to trigger operating condition event indicators.

8. The microfluidic droplet generation parameter adaptive closed-loop control system according to claim 7, characterized in that, The adjustment module adopts an event-triggered cascade adaptive control algorithm. The inner loop is a flow rate control loop, which controls the flow rate based on dual feedback of pressure and displacement and is synchronized with the drive clock to suppress underlying disturbances. The outer loop is a mass control loop, which corrects the flow rate setpoint based on size detection feedback. The outer loop adopts an event-driven variable parameter PID algorithm with a built-in operating condition event library. It identifies operating condition events in real time and switches the corresponding control parameter group and adjustment strategy. It optimizes parameters online based on historical results and updates the parameter library iteratively. When operating conditions change, control parameters are adjusted in advance to achieve feedforward compensation; when dimensions continue to exceed tolerances and conventional adjustments are ineffective, the system calibration process is triggered.

9. A microfluidic droplet generation parameter adaptive closed-loop control method, based on the system described in any one of claims 1-8, characterized in that, The specific steps include the following: S1: Receive process parameters and fluid properties, verify parameter validity by combining chip geometric boundaries, call the process numerical control program package and distribute configuration, complete end-to-end communication and driver timing self-test; display status and statistical data in real time during operation, and record logs according to the cycle time. S2. Call the step emulsion droplet size prediction model of the working condition zone, establish the droplet equilibrium volume correlation equation with geometric parameters and fluid properties as constraints, and match the corresponding correction coefficients; calculate the two-phase reference flow velocity according to the target diameter, adjust the continuous phase flow velocity in combination with the flow field stability margin, and iteratively update the model correction coefficients. S3. Synchronously drive each channel injection pump to run smoothly, and compensate for drive error and monitor pipeline abnormalities through displacement and pressure dual feedback; Detect fluctuations in the diameter of the fluid filament to assess flow field stability, fine-tune the continuous phase velocity, and balance the flow distribution in the branch channels; S4. Control the synchronous feeding of stepped micro-nozzles, with droplet generation dominated by the stepped structure and interfacial tension; in steady state, collect outlet sensor signals to identify abnormal working conditions, and support dynamic configuration of the number of nozzles and staggered start and stop. S5. Combine capacitance and optical sensing to realize droplet counting and size measurement, start optical detection according to preset conditions and update the size conversion coefficient, and output characteristic parameters and abnormality ratio. S6. It adopts a cascade control architecture. The inner loop controls the flow rate based on dual feedback of displacement and pressure, and the outer loop corrects the flow rate setpoint based on detection data. Combined with working condition identification, it performs feedforward compensation and triggers the system calibration process when the droplet size is continuously out of tolerance.

10. The adaptive closed-loop control method for microfluidic droplet generation parameters according to claim 9, characterized in that, In step S6, the outer loop adopts an event-driven variable parameter PID algorithm with a built-in operating condition event library. It identifies operating condition events in real time and switches the corresponding control parameter group and adjustment strategy. It optimizes parameters based on historical adjustment effects and updates the parameter library iteratively. When operating conditions change, control parameters are adjusted in advance to achieve feedforward compensation.

Images