Intelligent dispenser supporting dynamic liquid level tracking and closed-loop control method thereof

Abstract

The application discloses an intelligent liquid distributor supporting dynamic liquid level tracking and a closed-loop control method thereof, and the system comprises a multi-modal sensing module, a distributed pressure monitoring unit, an intelligent execution module, an adaptive control module and an interaction and verification module. By constructing the multi-modal sensing module and the distributed pressure monitoring unit, and fusing a dynamic threshold adaptive algorithm, the liquid level measurement can effectively penetrate steam, foam and suppress hanging material interference. By adopting an LSTM-RNN prediction model and a multi-factor dynamic compensation model built in an edge AI chip, and combining distributed pressure feedback, the system can predict the liquid level trend in advance and eliminate the influence of pipeline fluctuation in real time. Through the modular design of the intelligent execution module and the liquid surface dynamic following function, the equipment can adapt to different viscosity, corrosive liquid and containers. Through the two-dimensional verification and error tracing system of the interaction and verification module, the whole process data can be traced and has self-learning and iteration capability.

Description

Technical Field

[0001] This invention relates to the field of liquid handling equipment and control technology, and more specifically, to an intelligent liquid dispenser that supports dynamic liquid level tracking and its closed-loop control method. Background Technology

[0002] In the field of precision liquid handling, the dispenser is a core piece of equipment, and its performance directly determines the reliability of experimental data, the stability of production processes, and reagent utilization. Currently, dispensers on the market generally suffer from the following technical defects: 1. Low accuracy and weak anti-interference capability in liquid level measurement: Traditional liquid dispensers mostly use single sensing technology, such as single ultrasonic or infrared sensors. When faced with viscous liquids, container wall reflections, or steam foam interference, measurement errors are easily generated, and they cannot adapt to liquids with different transparency and dielectric constants. For example, when handling high-viscosity protein solutions, liquid easily adheres to the sensor surface, leading to misjudgment of the liquid level; for low dielectric constant media, the measurement resolution of a single radar sensor is insufficient to meet millimeter-level requirements.

[0003] 2. Lagging control and poor stability: Existing equipment mostly adopts single-loop open-loop or semi-closed-loop control, which relies solely on flow feedback adjustment. It cannot compensate for the interference caused by pipeline pressure fluctuations, temperature changes and container deformation in real time, resulting in lag in liquid level tracking during the liquid separation process. Especially in the micro-liquid separation μL level scenario, the volume error often exceeds ±3%, which is difficult to meet the stringent precision requirements of the biopharmaceutical industry.

[0004] 3. Insufficient adaptability and safety: Contact measurements are prone to sample contamination, especially unsuitable for biosensitive experiments; non-contact devices suffer from poor container compatibility and an inability to dynamically follow liquid level changes, leading to risks of air suction and spillage. Furthermore, most devices do not support modular replacement, making it difficult to handle common water-based solutions, highly corrosive organic solvents, and high-viscosity liquids.

[0005] 4. Lack of full-process verification and error correction mechanism: Traditional dispensers are only calibrated before dispensing and there is no real-time verification process. It is impossible to detect problems such as missing or over-dispensing in a timely manner. Manual verification is required afterward, which is inefficient and increases the risk of experimental errors, making it difficult to meet GMP compliance requirements.

[0006] In existing technologies, some high-end dispensers attempt to improve performance by optimizing the sealing structure and enhancing pump precision. For example, using precision ceramic piston technology can control the dispensing error to ±0.5%. However, the core challenges of dynamic level tracking, compensation for multiple interference factors, and closed-loop verification throughout the entire process remain unresolved. Therefore, there is an urgent need for a dispenser technology that combines high-precision dynamic tracking, strong anti-interference capabilities, a wide adaptability range, and intelligent closed-loop control to fill the performance gaps in existing equipment. Summary of the Invention

[0007] To address the aforementioned technical problems in related technologies, this invention proposes an intelligent liquid dispenser that supports dynamic liquid level tracking and its closed-loop control method, which solves the problems of inaccurate liquid level measurement, control lag, poor adaptability, and lack of real-time verification mechanism in existing liquid dispensers, and can overcome the above-mentioned shortcomings of the prior art.

[0008] To achieve the above-mentioned technical objectives, the technical solution of the present invention is implemented as follows: A smart dispenser supporting dynamic liquid level tracking includes: A multimodal sensing module, including a non-contact sensing array, is used to acquire initial liquid level data and disturbance characteristic data of the liquid in the container; The distributed pressure monitoring unit includes several pressure sensors distributed at the nodes of the liquid delivery pipeline to collect pressure data at each node of the pipeline in real time. The intelligent execution module includes a dispensing actuator for performing dispensing operations according to control commands; The adaptive control module is communicatively connected to the multimodal sensing module, the distributed pressure monitoring unit, and the intelligent execution module, respectively. It is used to receive the initial liquid level data, disturbance characteristic data, and pressure data, and generate and output control commands to the intelligent execution module based on the built-in prediction model and dynamic compensation model. The interaction and verification module is communicatively connected to the adaptive control module. It is used to receive the liquid separation task parameters input by the user and display the liquid separation process data. It also includes a verification unit, which is used to verify the liquid separation results after the liquid separation operation is completed and to feed back the verification results to the adaptive control module.

[0009] Furthermore, the multimodal sensing module includes a dual-frequency ultrasonic sensor and a laser sensor, which are coaxially mounted; the multimodal sensing module also includes a conical bionic probe with a hydrophobic coating on its surface; the adaptive control module has a built-in dynamic threshold adaptive algorithm that works in conjunction with the conical bionic probe to distinguish between the real liquid level signal and the material buildup interference signal.

[0010] Furthermore, the distributed pressure monitoring unit includes three pressure sensors, which are respectively located at the pump outlet, the valve front end, and the middle section of the pipeline, and are used to determine abnormal pipeline conditions through pressure gradient analysis.

[0011] Furthermore, the intelligent execution module includes a modular and detachable precision volumetric pump body, a corrosion-resistant electronically controlled valve, and a three-axis linkage dispensing needle mechanism; the pump body is provided with a micro pulsation suppression chamber for buffering pump body pulsation; the dispensing needle mechanism includes a liquid surface distance sensing submodule for real-time detection of the distance between the dispensing needle and the liquid surface to achieve dynamic following.

[0012] Furthermore, the adaptive control module includes a dual-core architecture of an industrial-grade PLC and an edge AI chip; the PLC is responsible for executing real-time control instructions; the edge AI chip has a built-in prediction model and a dynamic compensation model, the prediction model is an improved LSTM-RNN prediction model, used to predict the trend of liquid level change; the dynamic compensation model is a multi-factor dynamic compensation model, used to correct control parameters based on liquid characteristic data and pressure data.

[0013] Furthermore, the verification unit of the interaction and verification module includes an ultrasonic verification sensor and / or a high-precision weighing module, forming a two-dimensional verification system; the interaction and verification module also includes an error tracing system for recording and analyzing liquid separation errors.

[0014] A closed-loop control method for the intelligent liquid dispenser includes the following steps: S1 Task Initialization and Parameter Matching: Receives liquid separation task parameters, calls the corresponding container model and liquid characteristic data, performs pre-detection through the multimodal sensing module, and matches the initial control parameters by the adaptive control module; S2 Dynamic Liquid Level Acquisition and Interference Filtering: During the liquid distribution process, liquid level data is acquired in real time through a multimodal sensing module, and the actual liquid level value is obtained after processing by a fusion algorithm and a filtering algorithm. S3 Deviation Calculation and Multi-Objective Parameter Optimization: Calculates the deviation between the actual liquid level and the target liquid level, and optimizes the control parameters in real time by combining distributed pressure data and using the built-in prediction model and multi-objective optimization function; S4 Dynamic Execution and Time-Sharing Cooperative Control: Based on the optimized control parameters, a time-sharing cooperative control strategy is adopted to drive the actions of each actuator in the intelligent execution module, and real-time fine-tuning is performed based on the feedback from the distributed pressure monitoring unit; S5 Dual-Dimensional Verification and Deviation Correction: After the separation is completed, the separation result is verified through the verification unit of the interaction and verification module. If the error exceeds the preset threshold, the deviation correction process is automatically started. S6 Data Recording and Self-Learning Iteration: Records the data of the entire liquid separation process and stores it in the database to optimize the parameters of the prediction model and the dynamic compensation model.

[0015] Furthermore, the filtering algorithm in step S2 includes a liquid level mutation adaptive filtering algorithm, which is used to identify and filter false liquid level changes caused by bubble bursting or material detachment; the multi-objective optimization function in step S3 takes accuracy, efficiency and energy consumption as optimization objectives; the time-sharing collaborative control strategy in step S4 includes controlling the dispensing needle to maintain a preset distance from the liquid surface.

[0016] Furthermore, it also includes a fault self-diagnosis and graded recovery process: Real-time monitoring of the operating data of each module and comparison with preset normal threshold range; When the data exceeds the threshold, it is classified according to the degree of impact of the fault, and the fault point and fault type are located by combining multi-parameter correlation analysis algorithm. Based on the fault level, a corresponding graded early warning and handling mechanism is activated, which includes automatically starting a compensation program, attempting automatic recovery, or executing emergency shutdown protection. Record fault information to optimize algorithm parameters or generate equipment maintenance suggestions.

[0017] Furthermore, for multi-container dispensing scenarios, a dynamic sequential truncation algorithm is also included, which is used to dynamically adjust the control parameters of the container to be dispensed based on the verification results of the dispensed containers, so as to achieve synchronous and accurate dispensing of multiple containers.

[0018] The beneficial effects of this invention are as follows: By constructing a multimodal sensing module and a distributed pressure monitoring unit, and integrating a dynamic threshold adaptive algorithm, this invention enables liquid level measurement to effectively penetrate steam and foam while suppressing material buildup interference, improving dynamic measurement accuracy to ±0.01mm and solving the problems of low measurement accuracy and weak anti-interference in the prior art. By employing the LSTM-RNN prediction model and multi-factor dynamic compensation model built into the edge AI chip, combined with distributed pressure feedback, the system can predict liquid level trends in advance and eliminate the influence of pipeline fluctuations in real time, controlling the liquid volume error within ±1.5%, thus solving the problem of control lag. Through the modular design of the intelligent execution module and the dynamic liquid level following function, the equipment can adapt to liquids and containers of different viscosities and corrosiveness, solving the problems of insufficient adaptability and safety. Through the dual-dimensional verification and error traceability system of the interaction and verification module, the entire process data is traceable and has self-learning iteration capabilities, meeting GMP compliance requirements. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0020] A smart liquid dispenser supporting dynamic liquid level tracking according to an embodiment of the present invention includes: A multimodal sensing module, including a non-contact sensing array, is used to acquire initial liquid level data and disturbance characteristic data of the liquid in the container; The distributed pressure monitoring unit includes several pressure sensors distributed at the nodes of the liquid delivery pipeline to collect pressure data at each node of the pipeline in real time. The intelligent execution module includes a dispensing actuator for performing dispensing operations according to control commands; The adaptive control module is communicatively connected to the multimodal sensing module, the distributed pressure monitoring unit, and the intelligent execution module, respectively. It is used to receive the initial liquid level data, disturbance characteristic data, and pressure data, and generate and output control commands to the intelligent execution module based on the built-in prediction model and dynamic compensation model. The interaction and verification module is communicatively connected to the adaptive control module. It is used to receive the liquid separation task parameters input by the user and display the liquid separation process data. It also includes a verification unit, which is used to verify the liquid separation results after the liquid separation operation is completed and to feed back the verification results to the adaptive control module.

[0021] Preferably, the multimodal sensing module includes a dual-frequency ultrasonic sensor and a laser sensor, which are coaxially mounted; the multimodal sensing module also includes a conical bionic probe with a hydrophobic coating on its surface; the adaptive control module has a built-in dynamic threshold adaptive algorithm that works in conjunction with the conical bionic probe to distinguish between the real liquid level signal and the material buildup interference signal.

[0022] Preferably, the distributed pressure monitoring unit includes three pressure sensors, which are respectively installed at the pump outlet, the valve front end, and the middle section of the pipeline, and are used to determine abnormal pipeline conditions through pressure gradient analysis.

[0023] Preferably, the intelligent execution module includes a modular and detachable precision volumetric pump body, a corrosion-resistant electronically controlled valve, and a three-axis linkage dispensing needle mechanism; the pump body is provided with a micro pulsation suppression chamber for buffering pump body pulsation; the dispensing needle mechanism includes a liquid surface distance sensing submodule for real-time detection of the distance between the dispensing needle and the liquid surface to achieve dynamic following.

[0024] Preferably, the adaptive control module includes a dual-core architecture of an industrial-grade PLC and an edge AI chip; the PLC is responsible for executing real-time control instructions; the edge AI chip has a built-in prediction model and a dynamic compensation model, the prediction model is an improved LSTM-RNN prediction model, used to predict the trend of liquid level change; the dynamic compensation model is a multi-factor dynamic compensation model, used to correct control parameters based on liquid characteristic data and pressure data.

[0025] Preferably, the verification unit of the interaction and verification module includes an ultrasonic verification sensor and / or a high-precision weighing module, forming a two-dimensional verification system; the interaction and verification module also includes an error tracing system for recording and analyzing liquid separation errors.

[0026] A closed-loop control method for an intelligent dispenser supporting dynamic liquid level tracking includes the following steps: S1 Task Initialization and Parameter Matching: Receives liquid separation task parameters, calls the corresponding container model and liquid characteristic data, performs pre-detection through the multimodal sensing module, and matches the initial control parameters by the adaptive control module; S2 Dynamic Liquid Level Acquisition and Interference Filtering: During the liquid distribution process, liquid level data is acquired in real time through a multimodal sensing module, and the actual liquid level value is obtained after processing by a fusion algorithm and a filtering algorithm. S3 Deviation Calculation and Multi-Objective Parameter Optimization: Calculates the deviation between the actual liquid level and the target liquid level, and optimizes the control parameters in real time by combining distributed pressure data and using the built-in prediction model and multi-objective optimization function; S4 Dynamic Execution and Time-Sharing Cooperative Control: Based on the optimized control parameters, a time-sharing cooperative control strategy is adopted to drive the actions of each actuator in the intelligent execution module, and real-time fine-tuning is performed based on the feedback from the distributed pressure monitoring unit; S5 Dual-Dimensional Verification and Deviation Correction: After the separation is completed, the separation result is verified through the verification unit of the interaction and verification module. If the error exceeds the preset threshold, the deviation correction process is automatically started. S6 Data Recording and Self-Learning Iteration: Records the data of the entire liquid separation process and stores it in the database to optimize the parameters of the prediction model and the dynamic compensation model.

[0027] Preferably, the filtering algorithm in step S2 includes a liquid level change adaptive filtering algorithm, which is used to identify and filter false liquid level changes caused by bubble bursting or material detachment; the multi-objective optimization function in step S3 has accuracy, efficiency and energy consumption as optimization objectives; the time-sharing collaborative control strategy in step S4 includes controlling the dispensing needle to maintain a preset distance from the liquid surface.

[0028] Preferably, it also includes a fault self-diagnosis and graded recovery process: Real-time monitoring of the operating data of each module and comparison with preset normal threshold range; When the data exceeds the threshold, it is classified according to the degree of impact of the fault, and the fault point and fault type are located by combining multi-parameter correlation analysis algorithm. Based on the fault level, a corresponding graded early warning and handling mechanism is activated, which includes automatically starting a compensation program, attempting automatic recovery, or executing emergency shutdown protection. Record fault information to optimize algorithm parameters or generate equipment maintenance suggestions.

[0029] Preferably, for multi-container liquid dispensing scenarios, a dynamic sequential truncation algorithm is also included, which is used to dynamically adjust the control parameters of the container to be dispensed based on the verification results of the dispensed containers, so as to achieve synchronous and accurate liquid dispensing of multiple containers.

[0030] To facilitate understanding of the above technical solutions of the present invention, the following detailed description of the above technical solutions of the present invention will be provided through specific usage methods.

[0031] In practical application, the intelligent liquid dispenser supporting dynamic liquid level tracking according to the present invention adopts a full-link collaborative architecture of "sensing-control-execution-verification," including a multimodal sensing module, an adaptive control module, an intelligent execution module, an interaction and verification module, and a distributed pressure monitoring unit. Each module communicates via an EtherCAT high-speed industrial bus with a communication cycle ≤1ms and uses IEEE 754 floating-point format, achieving millisecond-level data interaction and high-precision control. The liquid dispenser adopts a full-link collaborative architecture of "sensing-control-execution-verification," including a multimodal sensing module, an adaptive control module, an intelligent execution module, an interaction and verification module, and a distributed pressure monitoring unit. Each module communicates via a high-speed bus, achieving millisecond-level data interaction and high-precision control, as detailed below: Multimodal sensing module: Employing a non-contact sensing array combining dual-frequency ultrasonic waves and laser coaxial calibration, this complementary detection system consists of an 80GHz high-frequency ultrasonic sensor that penetrates the surface of steam, foam, and viscous liquids, and a 1550nm near-infrared laser sensor that precisely locates the liquid level reference. Both sensors are coaxially mounted with a coaxiality error ≤0.02mm, eliminating measurement errors caused by installation deviations. The module innovatively uses a conical biomimetic probe structure with a cone angle of 30°~45°. The probe substrate is made of Hastelloy C-276 with a 50~80μm thick hydrophobic PTFE coating. Combined with a dynamic threshold adaptive algorithm, it distinguishes between real liquid level reflection signals and residual material signals by real-time identification of signal amplitude, frequency, and rise slope, reducing the false alarm rate to below 15%. Furthermore, the reference is automatically calibrated after material removal, requiring no manual intervention. Furthermore, an integrated dielectric constant auxiliary detection unit is used to acquire liquid dielectric constant data in real time, providing a basis for correcting the data fusion algorithm. It is compatible with various media with dielectric constants εr = 1.4~80 (covering water-based, organic solvents, high-viscosity colloids, etc.) and supports penetration detection through glass and ceramic container walls up to 10mm thick, with a detection response time ≤0.08s and a dynamic liquid level measurement resolution of 0.01mm. Data fusion employs a weighted adaptive fusion algorithm, dynamically adjusting the weights of each sensor based on the intensity of environmental interference (e.g., 70% weight for ultrasonic sensors in a steam environment and 80% weight for laser sensors in a strong light environment). Simultaneously, Kalman filtering is used to eliminate random noise, ensuring measurement stability under complex operating conditions.

[0032] Adaptive control module: It adopts a dual-core architecture of "industrial-grade PLC + edge AI chip". The PLC is responsible for real-time control command execution, while the edge AI chip is responsible for data prediction and algorithm optimization. It has a built-in container 3D model library + liquid property database, supports the call of 100+ standard container models and the input of custom container parameters. It can quickly build non-standard container 3D models through laser scanning with a modeling error ≤0.1mm. The core innovative algorithms include: ① An improved LSTM-RNN prediction model, which introduces an attention mechanism to focus on capturing the characteristics of sudden changes in liquid level, such as liquid level fluctuations caused by bubble bursting and liquid splashing. The prediction accuracy is over 98.5%, and it can predict the trend of liquid level change 50ms in advance to achieve feedforward control and eliminate hysteresis error; ② A real-time container deformation recognition algorithm, which uses multiple sets of sensor data to infer the small deformation of the container caused by temperature and pressure. The deformation detection accuracy is ≤0.005mm, and the liquid level measurement benchmark is dynamically corrected; ③ A multi-factor dynamic compensation model, which establishes a nonlinear compensation equation based on liquid viscosity (1~1000mPa·s), temperature (-20℃~+80℃), and pipeline pressure (0~10bar) data to correct the liquid level measurement value and control parameters in real time and solve the problem of interference from environmental variables on accuracy.

[0033] Intelligent Execution Module: Connected to the adaptive control module via servo drive, this module features an innovative modular and detachable design, including a precision positive displacement pump body, corrosion-resistant electronically controlled valves, a three-axis linkage dispensing needle mechanism, and anti-bubble components. It supports quick plug-and-play replacement and is adaptable to different dispensing scenarios: ① The pump body is a ceramic piston positive displacement pump with an innovative addition of a micro-pulsation suppression chamber (inner diameter 8~12mm, volume 5~10mL, elastic diaphragm thickness 0.3~0.5mm, made of fluororubber). The elastic diaphragm buffers pump body pulsation, controlling the flow pulsation coefficient within 0.5%, with a flow rate adjustment range of 0.1μL / min~10mL / min, and also features self-priming and venting functions. ① The flow pulsation coefficient is controlled within 0.5%, the flow rate adjustment range is 0.1μL / min~10mL / min, and it has a self-priming and venting function; ② The valve adopts a PTFE valve core + fluororubber sealing structure, which is resistant to strong acid and alkali corrosion, has a response time ≤0.05s, and has a built-in flow monitoring subunit to provide real-time feedback on the actual valve opening and flow rate; ③ The dispensing needle mechanism is driven by a three-axis servo motor and innovatively integrates a liquid surface distance sensing submodule. It uses high-frequency ultrasonic waves to detect the distance between the dispensing needle and the liquid surface in real time (detection range 0~50mm, accuracy 0.01mm) to achieve dynamic tracking. At the same time, it adopts a beveled needle (bevel angle 30°~60°, needle inner diameter 0.1~2mm, material 316L stainless steel) + PTFE anti-fouling coating design to reduce viscous liquid residue. Combined with the reverse pumping function, it avoids drip contamination; ④ The anti-bubble component has a built-in negative pressure monitoring unit. When the negative pressure value in the pipeline exceeds the threshold, the venting valve is automatically opened to prevent bubbles from entering the dispensing channel.

[0034] Interaction and Verification Module: Equipped with a 7-inch high-definition touchscreen, supporting visualized setting of task parameters, real-time display of liquid level curves, and error data statistics. It also integrates a dual-dimensional verification system: ① Primary Verification: After separation, the liquid level is detected a second time using an ultrasonic sensor to calculate the actual separated volume, with an error accuracy of ±0.05mm; ② Secondary Verification: An optional high-precision weighing module is available to verify the weight of the separated sample, achieving dual verification of "liquid level-weight" to meet GMP compliance requirements. The module innovatively incorporates an "error traceability system," automatically recording error types (such as system errors and environmental interference errors) and corresponding parameters, providing data support for subsequent algorithm optimization. It also supports data export (PDF / Excel format) and cloud storage, enabling full-process data traceability.

[0035] Distributed pressure monitoring unit: Unlike traditional single-point pressure monitoring, it adopts a three-point distributed pressure sensor layout of "post-pump + pre-valve + mid-pipeline" to collect pressure data at different nodes in the pipeline in real time. Through pressure gradient analysis, it identifies abnormal operating conditions such as pipeline blockage and leakage, and issues an early warning 100ms in advance. At the same time, it links pressure data with flow data to establish a "pressure-flow" coupled model, dynamically correcting pump speed and valve opening, eliminating the interference of pipeline pressure fluctuations on liquid distribution accuracy, and providing accurate feedback for dual closed-loop control.

[0036] According to the closed-loop control method of the intelligent liquid dispenser supporting dynamic liquid level tracking of the present invention, predictive control and multi-objective optimization logic are innovatively introduced to solve the problems of lag and weak anti-interference in traditional control. The specific steps are as follows: 1. Task Initialization and Parameter Matching (Pre-detection Phase): The user inputs the liquid dispensing task parameters (target liquid level, dispensing volume, container model, liquid name) through the interaction and verification module. The system automatically calls up the corresponding container 3D model and liquid characteristic data. At the same time, the multimodal sensing module starts pre-detection, quickly collecting liquid dielectric constant, surface tension and initial liquid level data. Based on the pre-detection results, the adaptive control module matches the optimal sensing weights, control algorithm and execution parameters. For example, for high-viscosity liquids, the dispensing needle following speed is reduced and the pump pressure is increased, without the need for manual adjustment.

[0037] 2. Dynamic Liquid Level Acquisition and Interference Filtering (Filtering Algorithm): The multi-modal sensor array synchronously acquires liquid level data, which is processed by a weighted adaptive fusion algorithm to obtain the initial liquid level value. At the same time, the "liquid level mutation adaptive filtering algorithm" is activated. By identifying the rate of liquid level change (normal drop rate 0.1~1mm / s, mutation rate >5mm / s), it distinguishes between false liquid level changes caused by bubble bursting and material detachment and the real liquid level drop, filters false signals and marks abnormal conditions to ensure the accuracy of liquid level data.

[0038] 3. Deviation Calculation and Multi-Objective Parameter Optimization: The adaptive control module calculates the deviation (Δh) between the actual liquid level and the target liquid level. Combining distributed pressure data, liquid temperature, and viscosity changes, it adjusts the control parameters through a "multi-objective optimization function" (with accuracy, efficiency, and energy consumption as optimization objectives): When Δh > 0.1 mm, accuracy is prioritized, and the LSTM-RNN prediction model is activated to optimize the pump flow rate and the dispensing needle following speed; when Δh ≤ 0.1 mm, accuracy and efficiency are balanced, and the dispensing speed is appropriately increased. Simultaneously, the parameter self-tuning PID algorithm adjusts the proportional coefficient (P), integral coefficient (I), and derivative coefficient (D) in real time to adapt to the dynamic characteristics of different liquids without manual intervention.

[0039] 4. Dynamic Execution and Time-Sharing Coordinated Control (Timing Logic): The execution module's actions are driven by a "time-sharing coordinated control strategy": ① The pump starts at the optimized flow rate, and the valve slowly opens to the preset opening degree; ② The dispensing needle adjusts its height in advance according to the predicted liquid level trajectory, maintaining a stable distance of 2~5mm from the liquid surface, and can adaptively adjust according to the liquid surface tension; ③ The distributed pressure monitoring unit provides real-time feedback of pressure data, and fine-tunes the valve opening through the secondary loop control to eliminate pressure fluctuation interference. Throughout the entire execution process, the timing error of each component's actions is ≤0.1s, avoiding accuracy degradation caused by coupling interference.

[0040] 5. Dual-Dimensional Verification and Deviation Correction (Verification Mechanism): After liquid separation is completed, the interaction and verification module initiates dual-dimensional verification: ① The ultrasonic sensor detects the liquid level and calculates the actual separated volume; ② The weighing module (optional) verifies the sample weight and cross-verifies the accuracy. If the error exceeds the preset threshold (±1.5%), the system automatically initiates the deviation correction process: First, the error tracing system determines the cause of the error, and then the control parameters are adjusted accordingly. For example, if the error is caused by pressure fluctuations, the pressure compensation coefficient is increased; if the error is caused by material buildup, the probe self-cleaning program is initiated and the reference is recalibrated. Liquid separation is supplemented or subsequent liquid separation parameters are adjusted until the accuracy requirements are met.

[0041] 6. Data Recording and Self-Learning Iteration (Iterative Mechanism): After the liquid separation task is completed, the system automatically records all process data, including liquid level tracking curves, control parameters, pressure changes, error data, and liquid characteristic information, and stores them in the cloud-edge collaborative database. For similar liquid tasks, the system calls the historical optimal parameters as initial values, combines them with the data from this task to optimize the model parameters, and achieves self-learning iteration. The debugging time for subsequent similar tasks can be shortened by more than 60%.

[0042] For multi-container dispensing scenarios, this method additionally integrates a "dynamic sequential truncation algorithm," which combines weighing feedback and liquid level data to achieve synchronous and accurate dispensing of multiple containers: the system sorts containers by position, controls the movement of dispensing needles in a time-sharing manner, monitors the error data of already dispensed containers, and dynamically adjusts the dispensing parameters of subsequent containers. It supports the dispensing needs of unequal spacing orifices (such as 96-well plates and 384-well plates) and irregular container arrays. The overall plate dispensing efficiency is 40% higher than that of traditional equipment, and the error of a single container is ≤±1.5%.

[0043] The specific implementation is as follows: 1. Equipment selection and assembly The selection and assembly requirements for each component of the intelligent liquid dispenser in this embodiment are as follows: Multimodal sensing module: It adopts a combination of laser sensor (resolution 0.01mm, response time ≤0.05s) and 80GHz high-frequency ultrasonic sensor (measurement range 0~500mm, accuracy ±0.01mm). The cone-shaped bionic probe has a cone angle of 35°, the substrate is Hastelloy C-276, and the surface is coated with a 60μm polytetrafluoroethylene hydrophobic coating. The data fusion algorithm adopts DSP digital signal processing technology, which can simultaneously track 8 echo signals and filter false reflections.

[0044] Adaptive control module: It adopts Siemens S7-1500 industrial-grade PLC controller, paired with NVIDIA Jetson Nano edge AI chip, with built-in RNN prediction model and self-tuning PID algorithm, stores 100+ common container 3D models, and supports custom container parameter input; temperature compensation range -20℃~+80℃, can automatically correct thermal expansion error, and control cycle ≤1ms.

[0045] Intelligent execution module: Utilizes a precision ceramic piston pump with a flow rate adjustment range of 0.1μL / min to 10mL / min and an accuracy of ±0.5%; features a micro-pulse suppression chamber with an inner diameter of 10mm and a volume of 8mL, and a 0.4mm thick fluororubber elastic diaphragm; PTFE-coated corrosion-resistant valves with a valve core stroke of 0~5mm and a response time ≤0.05s; and 316L stainless steel dispensing needles with a 45° bevel angle and an inner diameter of 0.2~1.5mm, coated with a PTFE anti-fouling coating, supporting manual quick disassembly and replacement with a replacement time ≤2min.

[0046] Distributed pressure monitoring unit: Selects Hydac high-precision pressure sensors with a measurement range of 0~10 bar and an error of ≤±0.01 bar. They are installed at three nodes: the pump outlet, the valve front end, and the middle section of the pipeline, to provide real-time feedback on pipeline pressure fluctuations. The data update frequency is 100Hz.

[0047] Interaction and verification module: Equipped with a 7-inch high-definition capacitive touch screen, supporting data export and printing, and ultrasonic verification sensor accuracy ±0.05mm; optional Mettler Toledo high-precision weighing module with accuracy 0.1mg and weighing range 0~500g, realizing dual verification of liquid level and weight.

[0048] Communication and power supply: Each module is connected via EtherCAT high-speed industrial bus with a communication cycle of ≤1ms and data format of IEEE 754 floating point. The equipment is powered by dual 220V AC / 24V DC power supply and has a built-in UPS backup power supply. After a power outage, it can maintain the operation of critical modules for ≥5 minutes.

[0049] Fault self-diagnosis and recovery process The device of this invention integrates a full-link fault self-diagnosis and hierarchical recovery mechanism. Through the linkage analysis of sensor data from various modules, it achieves real-time fault detection, precise location, hierarchical early warning, and automatic / manual recovery. The specific process is as follows: (1) Real-time fault monitoring: During the entire liquid separation process, the system continuously collects multimodal sensor data, distributed pressure data, execution module status data (pump speed, valve opening, motor position), and temperature / humidity environmental data, sets normal threshold ranges for each parameter, and triggers the fault monitoring mechanism if the threshold is exceeded.

[0050] (2) Fault Classification and Location: Based on the degree of fault impact, three levels are defined, and the fault point and fault type are accurately located using the "multi-parameter correlation analysis algorithm": Level 1 fault (minor interference, not affecting liquid separation): such as slight material buildup on the probe, minor fluctuations in pipeline pressure (±0.02~±0.05 bar), and slight changes in ambient temperature (±2℃); the fault is located to the specific sensor / monitoring point, such as "material buildup on the #1 ultrasonic probe" or "pressure fluctuation in the middle section of the pipeline".

[0051] Level 2 faults (moderate interference, affecting dispensing accuracy): such as sensor signal attenuation, excessive pump flow pulsation (coefficient > 0.5%), excessive dispensing needle following error (> 0.1 mm), and slight pipeline leakage (pressure gradient change 0.01~0.03 bar / m); the fault is located to a specific module, such as "signal attenuation of multimodal sensing module" or "pump abnormality of intelligent execution module".

[0052] Level 3 fault (serious fault, unable to dispense liquid normally): such as no signal from the sensor, pump stuck, valve unable to open or close, severe blockage / leakage in the pipeline (pressure gradient change > 0.05 bar / m), motor overload; the fault is located to a specific component, such as "laser sensor failure" or "corrosion-resistant electronic valve stuck".

[0053] (3) Tiered early warning and handling: The system activates corresponding early warning and handling mechanisms according to the fault level. The early warning methods include touch screen pop-ups and audible and visual alarms (for level three faults). The handling mechanisms are divided into automatic recovery and manual prompts. Level 1 fault: The compensation / correction program is automatically started without manual intervention; if the probe is clogged with material, the "dynamic threshold adaptive algorithm" is activated to filter false signals; if the pressure fluctuates, the "pressure-flow" coupling model is activated to correct the pump speed; if the temperature changes, the multi-factor dynamic compensation model is activated to correct the parameters; after the processing is completed, the system will automatically report "fault cleared" and continue the liquid separation process.

[0054] Level 2 Fault: First, the system will automatically initiate a recovery attempt (lasting 5 seconds). If the sensor signal attenuates, the system will initiate signal gain adjustment. If the pump flow pulsation exceeds the standard, the system will adjust the pressure of the micro pulsation suppression chamber. If the dispensing needle following error exceeds the standard, the system will recalibrate the servo motor. If the automatic recovery is successful, the system will record the fault information and continue dispensing, while appropriately reducing the dispensing speed to ensure accuracy. If the automatic recovery fails, a pop-up warning will be triggered on the touch screen, indicating the fault type and location. Manual remote debugging (such as adjusting control parameters) is supported. After debugging, dispensing will continue.

[0055] Level 3 fault: Immediately triggers emergency shutdown protection, stopping all actions of the pump, valves, and dispensing needle, and closing the pipeline inlet and outlet to prevent sample leakage, cavitation, or equipment damage; simultaneously activates audible and visual alarms and touch screen pop-up warnings, clearly displaying the fault type, location, and handling suggestions, such as "Laser sensor fault, it is recommended to replace the sensor module" or "The pipeline is severely blocked, it is recommended to clean the pipeline"; after the fault is resolved, the system must complete module calibration, parameter reset, and no-stroke test, and the dispensing process can only be restarted after verification that it is normal.

[0056] (4) Fault recording and iterative optimization: All fault information (level, type, location, handling method, recovery result) is automatically recorded to the error tracing system and stored in conjunction with the liquid separation data; for recurring faults, such as frequent material accumulation on the same probe or multiple blockages in a certain section of pipeline, the system automatically analyzes the cause of the fault, optimizes the corresponding algorithm parameters (such as adjusting the probe cleaning cycle, increasing the pipeline pressure compensation coefficient) or provides equipment maintenance suggestions, realizes fault self-learning iteration, and reduces the subsequent fault occurrence rate.

[0057] 3. Separation Test and Results The above-mentioned equipment was used to perform three-component liquid analysis, processing water-based buffer, 1 mg / ml IgG protein solution (surfactant-free), and 50% ethanol solution respectively. Each test group had 50 replicates, with a target liquid volume of 100 μL. Different interference conditions were simulated during the test (such as slight steam, probe residue buildup, and minor temperature fluctuations). The test results are as follows: (1) Water-based buffer solution: liquid volume error ±1.2%, repeatability CV value <2%, liquid level tracking response time <0.1s, the system automatically filters false signals when the simulated probe is attached to the material, there is no misjudgment of material attachment, and the first-level fault automatic recovery success rate is 100%.

[0058] (2) IgG protein solution: the volume error of the liquid separation is ±1.4%, the repeatability CV value is <2.5%, the liquid separation can be stabilized without the addition of surfactant, there is no sample contamination, and the sterility requirements of biological experiments are met; when the simulated temperature fluctuation is ±2℃, the multi-factor compensation model automatically corrects the parameters, and the accuracy does not decrease significantly.

[0059] (3) 50% ethanol solution: the volume error of liquid separation is ±1.3%, there is no leakage of corrosive media, the pressure fluctuation compensation effect is significant, and the pipeline pressure deviation is ≤0.02 bar; when the simulated pipeline has slight pressure fluctuations, the distributed pressure monitoring unit provides real-time feedback, and the secondary loop control quickly corrects the valve opening without any loss of accuracy.

[0060] Test results show that the equipment and method of the present invention have high precision and high stability in the processing of liquids with different characteristics. The fault self-diagnosis and recovery mechanism can effectively handle various interference conditions and ensure the smooth progress of the liquid separation process without manual intervention.

[0061] 4. Multi-container collaborative liquid dispensing application For 96-well plate dispensing scenarios, a dynamic sequential truncation algorithm combined with weighing feedback is adopted to achieve synchronous and accurate dispensing, with a target volume of 50μL per well. During the dispensing process, the dispensing error of the first 10 wells is monitored in real time, and the control parameters of the subsequent 86 wells are dynamically adjusted. After dispensing, the error of a single well is ≤±1.5%, and the dispensing time of the entire plate is ≤3min, which is 40% faster than traditional equipment. It supports unequal well spacing adjustment (well spacing 5~20mm) and is compatible with different specifications of multi-well plates such as 96-well and 384-well plates as well as irregular container arrays, meeting the needs of high-throughput experiments.

[0062] 5. Fault simulation test To verify the effectiveness of the fault self-diagnosis and recovery mechanism, three typical faults were simulated for testing, and the results are as follows: (1) Level 1 fault (material adhering to ultrasonic probe): During the test, a small amount of high viscosity protein solution is attached to the probe by human. The system identifies the fault within 50ms, starts the dynamic threshold adaptive algorithm to filter false signals, the liquid level measurement is error-free, the liquid separation accuracy is maintained at ±1.3%, the liquid separation is completed without any intervention, and the fault automatic recovery success rate is 100%.

[0063] (2) Level II fault (excessive pump flow pulsation): When the simulated pump flow pulsation coefficient rises to 0.8%, the system immediately identifies the fault, automatically adjusts the pressure of the elastic diaphragm in the micro pulsation suppression chamber, and reduces the pulsation coefficient to below 0.4% within 3 seconds. The liquid separation accuracy is restored from ±1.8% to ±1.4%, without the need for manual intervention.

[0064] (3) Level 3 fault (severe pipeline blockage): simulates blockage in the middle section of the pipeline. The distributed pressure monitoring unit detects a sudden change in pressure gradient > 0.05 bar / m. The system starts emergency shutdown protection within 100ms, closes all valves, triggers audible and visual alarms and pop-up prompts, accurately locates "blockage in the middle section of the pipeline", and provides cleaning suggestions. After the fault is cleared, the system completes calibration and testing. After restarting, the liquid separation accuracy returns to normal.

[0065] Fault simulation tests show that the fault self-diagnosis and recovery mechanism of the present invention can realize real-time fault detection, accurate location and graded processing, effectively improve the reliability and fault tolerance of equipment and reduce manual maintenance costs.

[0066] In summary, by utilizing the technical solutions described above in this invention, and through the construction of a multimodal sensing module and a distributed pressure monitoring unit, combined with a dynamic threshold adaptive algorithm, the liquid level measurement can effectively penetrate steam and foam while suppressing material buildup interference, improving the dynamic measurement accuracy to ±0.01mm. This solves the problems of low measurement accuracy and weak anti-interference in the prior art. By employing the LSTM-RNN prediction model and multi-factor dynamic compensation model built into the edge AI chip, combined with distributed pressure feedback, the system can predict liquid level trends in advance and eliminate the impact of pipeline fluctuations in real time, controlling the liquid volume error within ±1.5%, thus solving the problem of control lag. Through the modular design of the intelligent execution module and the dynamic liquid level following function, the equipment can adapt to liquids and containers of different viscosities and corrosiveness, solving the problems of insufficient adaptability and safety. Through the dual-dimensional verification and error traceability system of the interaction and verification module, the entire process data is traceable and has self-learning iteration capabilities, meeting GMP compliance requirements.

[0067] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart liquid dispenser supporting dynamic liquid level tracking, characterized in that, include: A multimodal sensing module, including a non-contact sensing array, is used to acquire initial liquid level data and disturbance characteristic data of the liquid in the container; The distributed pressure monitoring unit includes several pressure sensors distributed at the nodes of the liquid delivery pipeline to collect pressure data at each node of the pipeline in real time. The intelligent execution module includes a dispensing actuator for performing dispensing operations according to control commands; The adaptive control module is communicatively connected to the multimodal sensing module, the distributed pressure monitoring unit, and the intelligent execution module, respectively. It is used to receive the initial liquid level data, disturbance characteristic data, and pressure data, and generate and output control commands to the intelligent execution module based on the built-in prediction model and dynamic compensation model. The interaction and verification module is communicatively connected to the adaptive control module. It is used to receive the liquid separation task parameters input by the user and display the liquid separation process data. It also includes a verification unit, which is used to verify the liquid separation results after the liquid separation operation is completed and to feed back the verification results to the adaptive control module.

2. The intelligent liquid dispenser supporting dynamic liquid level tracking according to claim 1, characterized in that, The multimodal sensing module includes a dual-frequency ultrasonic sensor and a laser sensor, which are coaxially mounted. The multimodal sensing module also includes a conical bionic probe with a hydrophobic coating on its surface. The adaptive control module has a built-in dynamic threshold adaptive algorithm that works in conjunction with the conical bionic probe to distinguish between the real liquid level signal and the material buildup interference signal.

3. The intelligent liquid dispenser supporting dynamic liquid level tracking according to claim 1, characterized in that, The distributed pressure monitoring unit includes three pressure sensors, which are respectively located at the pump outlet, the valve front end, and the middle section of the pipeline, and are used to determine abnormal pipeline conditions through pressure gradient analysis.

4. The intelligent liquid dispenser supporting dynamic liquid level tracking according to claim 1, characterized in that, The intelligent execution module includes a modular and detachable precision volumetric pump body, a corrosion-resistant electronically controlled valve, and a three-axis linkage dispensing needle mechanism; the pump body is equipped with a micro pulsation suppression chamber to buffer pump body pulsation; the dispensing needle mechanism includes a liquid surface distance sensing submodule to detect the distance between the dispensing needle and the liquid surface in real time to achieve dynamic following.

5. The intelligent liquid dispenser supporting dynamic liquid level tracking according to claim 1, characterized in that, The adaptive control module includes a dual-core architecture of an industrial-grade PLC and an edge AI chip; the PLC is responsible for executing real-time control commands; the edge AI chip has a built-in prediction model and a dynamic compensation model. The prediction model is an improved LSTM-RNN prediction model, used to predict the trend of liquid level changes; the dynamic compensation model is a multi-factor dynamic compensation model, used to correct control parameters based on liquid characteristic data and pressure data.

6. The intelligent liquid dispenser supporting dynamic liquid level tracking according to claim 1, characterized in that, The verification unit of the interaction and verification module includes an ultrasonic verification sensor and / or a high-precision weighing module, forming a two-dimensional verification system; the interaction and verification module also includes an error tracing system for recording and analyzing liquid separation errors.

7. A closed-loop control method for an intelligent liquid dispenser according to any one of claims 1-6, characterized in that, Includes the following steps: S1 Task Initialization and Parameter Matching: Receives liquid separation task parameters, calls the corresponding container model and liquid characteristic data, performs pre-detection through the multimodal sensing module, and matches the initial control parameters by the adaptive control module; S2 Dynamic Liquid Level Acquisition and Interference Filtering: During the liquid distribution process, liquid level data is acquired in real time through a multimodal sensing module, and the actual liquid level value is obtained after processing by a fusion algorithm and a filtering algorithm. S3 Deviation Calculation and Multi-Objective Parameter Optimization: Calculates the deviation between the actual liquid level and the target liquid level, and optimizes the control parameters in real time by combining distributed pressure data and using the built-in prediction model and multi-objective optimization function; S4 Dynamic Execution and Time-Sharing Cooperative Control: Based on the optimized control parameters, a time-sharing cooperative control strategy is adopted to drive the actions of each actuator in the intelligent execution module, and real-time fine-tuning is performed based on the feedback from the distributed pressure monitoring unit; S5 Dual-Dimensional Verification and Deviation Correction: After the separation is completed, the separation result is verified through the verification unit of the interaction and verification module. If the error exceeds the preset threshold, the deviation correction process is automatically started. S6 Data Recording and Self-Learning Iteration: Records the data of the entire liquid separation process and stores it in the database to optimize the parameters of the prediction model and the dynamic compensation model.

8. The closed-loop control method for the intelligent liquid dispenser according to claim 7, characterized in that, The filtering algorithm in step S2 includes a liquid level mutation adaptive filtering algorithm, which is used to identify and filter false liquid level changes caused by bubble bursting or material detachment; the multi-objective optimization function in step S3 takes accuracy, efficiency and energy consumption as optimization objectives; the time-sharing collaborative control strategy in step S4 includes controlling the dispensing needle to maintain a preset distance from the liquid surface.

9. The closed-loop control method for the intelligent liquid dispenser according to claim 7, characterized in that, It also includes fault self-diagnosis and graded recovery processes: Real-time monitoring of the operating data of each module and comparison with preset normal threshold range; When the data exceeds the threshold, it is classified according to the degree of impact of the fault, and the fault point and fault type are located by combining multi-parameter correlation analysis algorithm. Based on the fault level, a corresponding graded early warning and handling mechanism is activated, which includes automatically starting a compensation program, attempting automatic recovery, or executing emergency shutdown protection. Record fault information to optimize algorithm parameters or generate equipment maintenance suggestions.

10. The closed-loop control method for the intelligent liquid dispenser according to claim 7, characterized in that, For multi-container dispensing scenarios, a dynamic sequential truncation algorithm is also included, which dynamically adjusts the control parameters of the container to be dispensed based on the verification results of the already dispensed containers, so as to achieve synchronous and accurate dispensing of multiple containers.