Multi-stage liquid separation treatment and self-adaptive liquid level detection system and control method

Abstract

The invention discloses a multi-stage liquid separation treatment and self-adaptive liquid level detection system and a control method. The system comprises a customized round bottle liquid separation funnel, a high-precision rotary positioning module, an AI visual liquid level detection module and a self-adaptive steering engine valve module. And the main control system is used for receiving real-time liquid level information of the AI visual liquid level detection module and controlling the steering engine to drive the valve element to adjust the opening degree through the ball screw transmission mechanism according to a preset liquid level-valve opening degree mapping relation. The design of the customized round bottle separating funnel with the gravity center coinciding with the rotating axis and the double discharging branches is adopted, so that the rotating operation of the equipment is more stable, and liquid residues are extremely few; by introducing an AI visual liquid level detection module based on an improved U-Net algorithm, non-contact and high-precision real-time identification of a layered interface is realized, and the defects that manual visual errors are large and contact detection is easy to pollute are overcome.

Description

Technical Field

[0001] This invention relates to the field of liquid separation technology, and more specifically, to a multi-stage liquid separation and adaptive liquid level detection system and control method. Background Technology

[0002] In fields such as chemical experiments, fine chemicals, and biomedicine, liquid-liquid separation is a fundamental and crucial operation. Its core requirement is to achieve precise separation of liquids with different densities, ensuring both separation efficiency and accuracy in volume. Current liquid-liquid separation technologies primarily rely on traditional separating funnels and are operated manually, which has several drawbacks: 1. Low degree of automation: The entire process from adding liquid, stirring, settling to separation requires manual intervention, such as manually controlling the start and stop of stirring, manually rotating the device to adjust the angle, and manually operating the valve to release liquid. This not only involves high labor intensity, but also makes it easy for human error to affect the separation effect.

[0003] 2. Disconnection between liquid level detection and valve control: In traditional liquid separation processes, liquid level observation relies on manual visual judgment, which cannot accurately identify the critical separation liquid level. The valve opening and closing angle is fixed and cannot be dynamically adjusted according to changes in liquid level. This results in either sacrificing separation accuracy for efficiency or reducing processing efficiency to ensure accuracy, making it difficult to achieve both.

[0004] 3. Poor equipment adaptability: Traditional separating funnels are mostly designed with a single discharge port, which cannot adapt to the needs of different separation volumes; and the center of gravity is prone to shift during rotation, resulting in unstable operation, large liquid residue, and affecting separation purity and material utilization.

[0005] 4. Insufficient system coordination: The existing separation equipment's stirring, rotation, settling, and separation modules are mostly independently controlled, lacking a unified core control system. Data interaction is lagging, and a closed-loop control system cannot be formed, resulting in poor connection between various links, which further reduces separation efficiency and accuracy.

[0006] Therefore, developing a liquid separation system that can achieve integrated control of stirring, rotation, settling, and liquid separation, has adaptive liquid level detection and dynamic valve adjustment functions, and is stable and highly adaptable has become a key requirement to solve the pain points of existing technologies. Summary of the Invention

[0007] To address the aforementioned technical problems in related technologies, this invention proposes a multi-stage liquid separation and adaptive liquid level detection system and control method. Through customized core component design, multi-module collaborative control, and adaptive adjustment strategies, it achieves full-process automation and high-precision control of liquid separation, balancing separation efficiency and separation volume accuracy, and improving the stability and adaptability of liquid separation processing. This overcomes the aforementioned shortcomings of existing technologies.

[0008] To achieve the above-mentioned technical objectives, the technical solution of the present invention is implemented as follows: A multi-stage liquid separation and adaptive liquid level detection system includes: A customized round bottle dispensing funnel serving as the core execution unit; A high-precision rotary positioning module for driving the separation funnel to rotate, the rotation axis of which coincides with the center of gravity of the separation funnel; An AI visual liquid level detection module for non-contact detection of liquid level and layering interface in the separating funnel; At least one adaptive servo valve module connected to the outlet of the separating funnel includes a servo motor, a ball screw transmission mechanism driven by the servo motor, a valve core, and a position sensor for detecting the valve opening degree. A main control system is also included, which is connected to the AI ​​visual liquid level detection module and the adaptive servo valve module respectively. The system receives real-time liquid level information from the AI ​​visual liquid level detection module and controls the servo motor to drive the valve core to adjust the opening according to the preset liquid level-valve opening mapping relationship.

[0009] Furthermore, the customized round bottle dispensing funnel includes a main bottle body, a dispensing section located at the lower part of the bottle body, and two symmetrically arranged discharge branches on the left and right sides of the upper part of the bottle body; each of the two discharge branches integrates an independent adaptive servo valve module, and the discharge interfaces of the two branches have different diameters.

[0010] Furthermore, the AI ​​visual liquid level detection module includes an industrial camera, a ring light source, and an AI image processing unit; the AI ​​image processing unit has a built-in deep learning model based on an improved U-Net architecture and an attention mechanism, which is used to process the images captured by the industrial camera to identify the liquid level layer interface and output liquid level height and fluctuation amplitude data.

[0011] Furthermore, the preset liquid level-valve opening mapping relationship of the main control system is a segmented mapping, which includes at least: when the real-time liquid level is higher than the first preset value, a larger opening is corresponding to the liquid level; when the real-time liquid level is between the first preset value and a lower second preset value, a smaller opening is corresponding to the liquid level; and when the real-time liquid level reaches the critical separation threshold, the valve is controlled to close.

[0012] Furthermore, the main control system is also configured to: when the liquid level fluctuation amplitude reported by the AI ​​visual liquid level detection module exceeds a preset fluctuation threshold, control the adaptive servo valve module to pause the opening adjustment.

[0013] Furthermore, it also includes a magnetically coupled stirring module with a stirring magnet placed inside the separating funnel; and an intelligent settling timer module; the main control system is configured to sequentially control the magnetically coupled stirring module, the high-precision rotary positioning module, the intelligent settling timer module, and the adaptive servo valve module to work together.

[0014] Furthermore, the system also includes a redundant power supply module, which includes a main power supply, a backup power supply, and a power monitoring unit, for automatically switching to the backup power supply when the main power supply fails.

[0015] A multi-stage liquid separation control method using the system includes the following steps: S1 Stirring Step: Control the magnetically coupled stirring module to stir the material added to the separatory funnel; S2 Rotation and Settling Step: Control the high-precision rotary positioning module to drive the separating funnel to rotate to the specified angle and allow it to settle so that the material separates into layers; S3 Adaptive Liquid Separation Steps: The AI ​​visual liquid level detection module is activated to perform real-time liquid level detection. The main control system dynamically queries the preset liquid level-valve opening mapping table based on the detected real-time liquid level information, and controls the opening of the adaptive servo valve module accordingly to perform liquid separation until the liquid level reaches the critical separation threshold and the valve is closed.

[0016] Furthermore, in the adaptive liquid separation step, dynamically controlling the valve opening specifically includes: When the real-time liquid level is greater than the first threshold and the liquid level is stable, the control valve operates at the first maximum opening degree. When the real-time liquid level drops to between the first threshold and the second threshold, the valve opening is gradually reduced to the second smaller opening, and the opening adjustment step size is reduced. When the real-time liquid level reaches or approaches the critical separation threshold, the control valve is closed or kept at a small opening and low flow rate.

[0017] Furthermore, during the liquid separation process, if the liquid level fluctuation is detected to exceed the preset value, the valve opening adjustment is paused, and adjustment is resumed only after the liquid level returns to stability.

[0018] The beneficial effects of this invention are as follows: By employing a customized round bottle-shaped dispensing funnel with its center of gravity coinciding with the axis of rotation and a dual-discharge branch design, this invention achieves more stable equipment rotation and minimal liquid residue, while flexibly adapting to different dispensing volumes. This solves the problems of traditional equipment, such as center of gravity shift, excessive residue, and poor adaptability. Furthermore, by introducing an AI visual liquid level detection module based on an improved U-Net algorithm, non-contact, high-precision (≤0.5mm) real-time identification of layered interfaces is achieved, overcoming the shortcomings of large errors in manual visual inspection and easy contamination in contact detection. By constructing a closed-loop control system composed of a ball screw servo valve and a PLC, and dynamically adjusting according to a pre-calibrated liquid level-valve opening mapping relationship, the dispensing process can be automatically optimized between high efficiency and high precision, completely changing the rigid control mode of traditional fixed valve angles that cannot balance efficiency and precision. Finally, by integrating multiple modules such as stirring, rotation, settling, and dispensing under PLC control, fully automated closed-loop operation is achieved, significantly reducing labor intensity and human error, and improving the overall reliability and separation effect of the system. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the overall structure of the multi-stage liquid separation and adaptive liquid level detection system according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a customized round bottle separating funnel for the multi-stage liquid separation and adaptive liquid level detection system according to an embodiment of the present invention; Figure 3 This is a flowchart of the multi-stage liquid separation control process of the multi-stage liquid separation and adaptive liquid level detection system according to an embodiment of the present invention. Figure 4 This is an AI visual liquid level detection and valve adaptive adjustment logic diagram of the multi-stage liquid separation and adaptive liquid level detection system according to an embodiment of the present invention; In the diagram: 1. Main bottle body; 2. Dispensing section. Detailed Implementation

[0021] 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. 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.

[0022] like Figure 1-4 As shown, a multi-stage liquid separation and adaptive liquid level detection system according to an embodiment of the present invention includes: A customized round bottle dispensing funnel serving as the core execution unit; A high-precision rotary positioning module for driving the separation funnel to rotate, the rotation axis of which coincides with the center of gravity of the separation funnel; An AI visual liquid level detection module for non-contact detection of liquid level and layering interface in the separating funnel; At least one adaptive servo valve module connected to the outlet of the separating funnel includes a servo motor, a ball screw transmission mechanism driven by the servo motor, a valve core, and a position sensor for detecting the valve opening degree. A main control system is also included, which is connected to the AI ​​visual liquid level detection module and the adaptive servo valve module respectively. The system receives real-time liquid level information from the AI ​​visual liquid level detection module and controls the servo motor to drive the valve core to adjust the opening according to the preset liquid level-valve opening mapping relationship.

[0023] Preferably, the customized round bottle dispensing funnel includes a main bottle body, a dispensing section located at the lower part of the bottle body, and two symmetrically arranged discharge branches on the left and right sides of the upper part of the bottle body; each of the two discharge branches integrates an independent adaptive servo valve module, and the discharge interfaces of the two branches have different diameters.

[0024] Preferably, the AI ​​visual liquid level detection module includes an industrial camera, a ring light source, and an AI image processing unit; the AI ​​image processing unit has a built-in deep learning model based on an improved U-Net architecture and an attention mechanism, which is used to process the images acquired by the industrial camera to identify the liquid level layer interface and output liquid level height and fluctuation amplitude data.

[0025] Preferably, the liquid level-valve opening mapping relationship preset by the main control system is a segmented mapping, which includes at least: when the real-time liquid level is higher than a first preset value, a larger opening is corresponding to the liquid level; when the real-time liquid level is between the first preset value and a lower second preset value, a smaller opening is corresponding to the liquid level; and when the real-time liquid level reaches a critical separation threshold, the valve is controlled to close.

[0026] Preferably, the main control system is further configured to: when the liquid level fluctuation amplitude reported by the AI ​​visual liquid level detection module exceeds a preset fluctuation threshold, control the adaptive servo valve module to pause the opening adjustment.

[0027] Preferably, it also includes a magnetically coupled stirring module with a stirring magnet placed inside the separating funnel; and an intelligent settling timer module; the main control system is configured to sequentially control the magnetically coupled stirring module, the high-precision rotary positioning module, the intelligent settling timer module and the adaptive servo valve module to work together.

[0028] Preferably, the system further includes a redundant power supply module, which includes a main power supply, a backup power supply and a power monitoring unit, for automatically switching to the backup power supply when the main power supply fails.

[0029] A multi-stage liquid separation control method using the system includes the following steps: S1 Stirring Step: Control the magnetically coupled stirring module to stir the material added to the separatory funnel; S2 Rotation and Settling Step: Control the high-precision rotary positioning module to drive the separating funnel to rotate to the specified angle and allow it to settle so that the material separates into layers; S3 Adaptive Liquid Separation Steps: The AI ​​visual liquid level detection module is activated to perform real-time liquid level detection. The main control system dynamically queries the preset liquid level-valve opening mapping table based on the detected real-time liquid level information, and controls the opening of the adaptive servo valve module accordingly to perform liquid separation until the liquid level reaches the critical separation threshold and the valve is closed.

[0030] Preferably, in the adaptive liquid separation step, dynamically controlling the valve opening specifically includes: When the real-time liquid level is greater than the first threshold and the liquid level is stable, the control valve operates at the first maximum opening degree. When the real-time liquid level drops to between the first threshold and the second threshold, the valve opening is gradually reduced to the second smaller opening, and the opening adjustment step size is reduced. When the real-time liquid level reaches or approaches the critical separation threshold, the control valve is closed or kept at a small opening and low flow rate.

[0031] Preferably, during the liquid separation process, if the liquid level fluctuation is detected to exceed the preset value, the valve opening adjustment is paused, and adjustment is resumed after the liquid level returns to stability.

[0032] 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.

[0033] The core technical solution of this invention is to construct a multi-stage liquid separation system integrating a "customized dispensing flask + multi-functional execution module + AI vision detection + PLC closed-loop control". Through preset control logic, it achieves the orderly connection of stirring, rotation, settling, and liquid separation. Simultaneously, it utilizes AI vision liquid level detection results to dynamically adjust the valve opening angle, precisely controlling the liquid separation process. The specific technical solution is as follows: I. System Overall Architecture This system includes a customized round bottle dispensing funnel (core actuator), a magnetically coupled stirring module, a high-precision rotary positioning module, an intelligent static timing module, an adaptive servo valve module, an AI vision liquid level detection module, a PLC main control system, and a redundant power supply module. Each module interacts with the other via an industrial bus to form a closed-loop control system, working together to complete integrated dispensing and adaptive liquid level control.

[0034] II. Design of Core Components and Modules 1. Customized round bottle separatory funnel As the core execution unit of the entire system, it integrates functions of liquid addition, stirring, liquid separation, and rotation. The specific design is as follows: Main body: Designed according to standard glass single-necked round-bottom flask, with a standard 1000ml capacity and 24 / 29# connector; made of high borosilicate 3.3 glass, which has a low coefficient of linear expansion (3.3×10⁻⁻⁴). 6 It can withstand temperature changes from -30℃ to 200℃, has strong chemical stability, and is suitable for various corrosive separation systems such as acids and alkalis; the volume scale is etched on the outside of the bottle with an accuracy of ±2%, making it easy to observe the amount of liquid added.

[0035] Separation section: Designed with reference to the standard pear-shaped separation funnel, it has a built-in PTFE piston and a high-elasticity fluororubber sealing ring to achieve a seal without dead angles; the piston stroke is 15mm, the liquid flow cross section can be finely adjusted, the specification is 500ml, the interface type is 24#, and the corrosion resistance and sealing performance are superior to traditional rubber pistons.

[0036] Branch structure: A liquid inlet (32mm diameter) is located at the top, equipped with a ground glass stopper to prevent liquid splashing during stirring; two funnel branches (180° angle between branches) are symmetrically arranged perpendicular to the liquid inlet in the left and right directions. Each branch integrates a valve mechanism driven by a "servo motor + ball screw", which are defined as "left branch servo motor valve" and "right branch servo motor valve" respectively; the servo motor is an MG996R high-pressure servo motor with a torque ≥13kg·cm, which can realize stepless adjustment of the valve from 0° to 90° and a response time ≤0.2s; the two branches are adapted to different diameter discharge interfaces (16mm for the left branch and 20mm for the right branch), which can be switched according to the liquid distribution requirements to achieve bidirectional selective liquid distribution.

[0037] Key parameters: After rotating 90° around the central axis, the volume of the colored area is 315mL±5mL, which meets the specific liquid separation requirements; the rotation axis coincides with the center of gravity of the funnel, and the center of gravity offset during rotation is ≤2mm, ensuring rotational stability; the bottom of the funnel is equipped with an arc-shaped flow guide structure with a flow guide angle of 30°, which can reduce the amount of liquid residue (residual amount ≤0.5ml).

[0038] 2. Magnetic coupling stirring module It adopts a magnetic coupling drive method to avoid the leakage problem of traditional stirring shaft seals; the stirring magnetic particle is placed inside the customized round bottle separating funnel, and the stirring function is achieved by driving the magnetic particle to rotate through an external magnetic field; the stirring speed can be adjusted by the PLC control system according to the needs of the separation system (adjustment range 50-500r / min) to ensure uniform mixing of materials and improve the subsequent separation effect.

[0039] 3. High-precision rotary positioning module Composed of a rotary motor, reducer, angle encoder, and positioning bracket, it drives a customized round bottle dispensing funnel to rotate around its central axis. The rotary motor is a stepper motor, which, together with the reducer, achieves high-precision angle control with an angle positioning accuracy of ±0.1°. The angle encoder collects the rotation angle signal in real time and feeds it back to the PLC main control system, forming a closed-loop control of the rotation angle. It can achieve precise positioning of the funnel at three key positions: 0° (initial position), 90° to the left, and 90° to the right, meeting the posture requirements of different dispensing stages.

[0040] 4. Intelligent static timing module It integrates a timing chip and a signal triggering unit, and controls the PLC to preset the settling time (adjustable range 1-60min). After stirring stops and the funnel rotates to the correct position, the timing function is triggered. After the settling time ends, it automatically sends a signal to the PLC to start the subsequent liquid separation process. It also has a timing abnormality alarm function to improve system reliability.

[0041] 5. Adaptive servo valve module The adaptive servo valve module is the core actuator for precise liquid distribution control. It employs an integrated transmission design of "servo motor-ball screw-valve core," boasting advantages such as high response, high precision, and adaptive adjustment. It can dynamically match the valve opening and closing angle according to changes in liquid level, achieving synergistic optimization of efficiency and accuracy. Its specific structure and working principle are as follows: (1) Core Components: The module consists of an MG996R high-voltage servo motor, a precision ball screw transmission mechanism, a corrosion-resistant valve body, a high-precision position sensor, and a signal conditioning unit. The MG996R high-voltage servo motor serves as the power source, employing a metal gear structure with a torque ≥13 kg·cm. It supports a 180° angle adjustment range, optimized to an effective adjustment range of 0°-90° for this system, with a response time ≤0.2s and a repeatability accuracy of ±0.1°, enabling precise response to control commands. The precision ball screw transmission mechanism uses a high-precision ball screw with a lead of 2mm, converting the servo motor's rotational motion into the valve core's linear reciprocating motion. The transmission efficiency is ≥95%, with no backlash, ensuring the linearity and stability of the valve opening adjustment. The valve body adopts... The valve core is made of PTFE and matched with a high borosilicate glass valve seat, equipped with a high-elasticity fluororubber sealing ring, which is resistant to acid and alkali corrosion, and has a sealing pressure ≥0.3MPa, with no risk of liquid residue or leakage. The high-precision position sensor uses a miniature photoelectric encoder with a resolution of 1024 lines to collect valve core displacement data in real time, convert it into valve opening and closing angle (detection accuracy ±0.05°), and feed it back to the PLC main control system after filtering and amplification by the signal conditioning unit. The signal conditioning unit integrates an anti-interference module, which can effectively suppress electromagnetic interference in the industrial environment and ensure the stability of signal transmission.

[0042] (2) Adaptive adjustment principle: The module adopts a closed-loop control logic of "instruction-feedback-correction" and works in synergy with the AI ​​vision liquid level detection module and the PLC main control system. The specific process is as follows: The PLC receives the real-time liquid level height data transmitted by the AI ​​vision liquid level detection module, and combines it with the preset liquid level-valve angle mapping relationship (established through experimental calibration, with different liquid level ranges corresponding to the optimal valve opening), and sends an angle control pulse signal to the servo motor; the servo motor drives the ball screw to move the valve core and adjust the valve opening angle (0° is fully closed, 90° is fully open); the position sensor collects the actual valve opening data in real time and feeds it back to the PLC; the PLC compares the actual opening with the target opening, and if there is a deviation (the deviation threshold is set to ±0.5°), it outputs a correction command until the valve reaches the target opening. For example, when the liquid level is high (e.g., more than 120mm from the bottom of the funnel), the PLC controls the valve opening to be 60°-90° to increase the liquid flow cross-section and improve the separation efficiency; when the liquid level drops to near the critical separation level (e.g., 50-60mm from the bottom), the valve opening is automatically adjusted to 10°-30° to reduce the flow rate, avoid disturbance at the stratification interface, and ensure separation accuracy; if a liquid level fluctuation is detected, the module can complete the opening correction within 0.2s to achieve dynamic adaptive adjustment.

[0043] Experimental calibration method for the mapping relationship between liquid level and valve angle (specifically addressing the problems of "no basis for mapping relationship and no guarantee of adjustment accuracy"): This mapping relationship is established through "multi-condition comparative experiments + data fitting optimization". The core is to find the correspondence between different liquid level ranges and the optimal valve opening, ensuring accurate matching of "liquid level change - opening adjustment", balancing efficiency and accuracy. The specific steps are as follows: The core principle of calibration is that the valve opening determines the flow cross-sectional area, which in turn affects the flow rate; the liquid level determines the static pressure of the liquid (the higher the liquid level, the greater the static pressure, and the faster the flow rate at the same opening). The calibration goal is to determine the optimal opening for different liquid level ranges that "both ensures liquid separation efficiency and avoids disturbance at the layering interface," while adapting to different material properties (density, viscosity) and establishing a universal mapping library.

[0044] Preparation for calibration experiment: 1) Equipment and materials: The complete liquid separation system of this invention; three types of calibration materials (low viscosity oil-water mixture, medium viscosity ethanol-glycerol mixture, corrosive dilute hydrochloric acid-organic solvent mixture); electronic balance (0.1g accuracy), stopwatch (0.01s accuracy), constant temperature room (25±1℃, to avoid temperature affecting viscosity).

[0045] 2) Parameter presets: Liquid level division (effective range of 40-120mm, core range of 40-60mm with 1 node every 1mm, and other ranges with 1 node every 5mm, for a total of 230 nodes); Valve opening range (0°-90°, 0.1° adjustment accuracy); Evaluation indicators (dispensing efficiency: mL / s; accuracy: actual deviation from target ≤ ±1%; interface fluctuation ≤ 1mm).

[0046] 3) Step-by-step calibration process: Step 1: Baseline Calibration (Low-viscosity oil-water mixture): ① Follow the system procedure to complete stirring and settling to form stable stratification; ② The AI ​​vision module locates the first liquid level node (e.g., 120mm); ③ Test 10 opening gradients (60°-90°) at this node, and record the separation time, actual volume, and fluctuation amplitude; ④ Select the opening value with "fluctuation ≤ 1mm and highest efficiency" as the optimal value (e.g., 120mm corresponds to 80°); ⑤ Traverse 230 liquid level nodes to form the original dataset.

[0047] Step 2: Multi-material adaptation calibration: Repeat step 1 for medium viscosity and corrosive materials, adjust the opening gradient (for medium viscosity materials with poor flowability, increase the opening by 10°-20° at the same liquid level) and establish a dedicated dataset for different materials.

[0048] Step 3: Critical range fine calibration (40-60mm): At 1mm node intervals, test 15 low opening degrees (5°-35°) at each node, prioritizing accuracy ≤±0.5%, and determine the opening degree of 15° corresponding to the critical liquid level of 50mm (flow rate 0.3mL / s, fluctuation 0.3mm).

[0049] 4) Data processing and mapping establishment: ① Clean abnormal data (validity ≥ 99%); ② Segmented fitting (efficient interval linear, transition interval quadratic polynomial, accurate interval cubic polynomial, goodness of fit R² ≥ 0.98); ③ Convert into discrete lookup table (0.1mm resolution), store in PLC according to material category to ensure fast retrieval.

[0050] 5) Verification and optimization: ① Static verification (random 10 nodes, accuracy ≤ ±1%); ② Dynamic verification (liquid level continuously decreases from 120-40mm, total time ≤ 120s, deviation ≤ ±0.5%); ③ Online optimization (when the accuracy is ≥ ±1.5% for 3 consecutive times, retest and update the mapping table.

[0051] (3) Redundant protection design: The module has built-in overcurrent protection, overload protection and limit protection mechanisms. When the valve encounters abnormal conditions such as jamming and the load on the servo motor exceeds 15 kg·cm, the overload protection is triggered, the servo motor immediately stops running and sends an alarm signal to the PLC; the limit switch is set at the valve fully closed (0°) and fully open (90°) positions to prevent the valve core from moving excessively and causing damage to the mechanism; the overcurrent protection module can cut off the power supply when the power supply voltage is abnormal (exceeding 24V±1V) to protect the safety of the core components.

[0052] 6. AI Visual Liquid Level Detection Module The AI-powered visual liquid level detection module is the core innovation of this invention. Its core objective is to address the technical pain points of traditional liquid separation technology: large errors in manual visual liquid level detection, easy contamination of materials by contact detection, and difficulty in identifying layered interfaces under complex conditions (similar material colors / fluid surface changes). Through an innovative solution of "non-contact data acquisition + deep learning for accurate recognition," it provides reliable data support for adaptive valve regulation, achieving precise coordination between liquid level detection and liquid separation control. Its specific structure, targeted solutions, and working principle are as follows: (1) Core components and hardware selection: The module consists of an industrial camera, a customized ring light source, an image acquisition card, an AI image processing unit, and a data transmission module. The selection and design of each component are optimized around the core requirements of "high precision, anti-interference, and real-time performance". 1) Industrial Camera: Hikvision MV-CA013-21GM Gigabit Ethernet industrial camera is selected, with a resolution of 1.3 million pixels (1280×1024), a frame rate of ≥30fps, and support for global shutter, which can effectively avoid image blurring caused by liquid flow; the lens is an 8mm fixed-focus industrial lens with an adjustable focal length range of 5-10mm, and the field of view is adapted to the observation range of the customized round bottle separating funnel, ensuring that the entire liquid level in the funnel can be captured completely; the camera is installed at a 45° angle above the side of the funnel and fixed by an adjustable bracket, and the installation angle can be finely adjusted according to the specifications of the funnel to ensure image quality.

[0053] 2) Customized Ring Light Source: Employs a high-brightness LED ring light source with a wavelength of 500-600nm (visible light band) and an 80mm diameter emitting surface, compatible with industrial camera lens sizes. The light source features a diffuse reflection design and is equipped with a softbox, providing uniform and soft illumination to avoid interference from liquid surface reflection caused by direct light. It supports stepless brightness adjustment (adjustment range 10%-100%) and can automatically adjust the brightness via PLC according to the ambient light intensity and material color characteristics to ensure image clarity under different working conditions. The light source and camera are coaxially mounted to form the optimal imaging path of "light source-funnel-camera," highlighting the contrast of the liquid level stratification interface.

[0054] 3) Image acquisition card and data transmission module: A PCI-E interface image acquisition card is selected, with a transmission rate of ≥1GB / s. It can acquire image data output by the camera in real time and transmit it to the AI ​​image processing unit without delay. The data transmission adopts a dual-link design of industrial Ethernet + RS485. Ethernet is used to transmit high-definition image data (transmission delay ≤10ms), and RS485 is used to transmit detection results (liquid level height, critical liquid level status). Dual-link redundancy ensures the reliability of data transmission and avoids detection interruption caused by single link failure.

[0055] 4) AI Image Processing Unit: It adopts the NVIDIA Jetson Nano development board as the core computing unit, integrating a Quad-Core ARM Cortex-A57 CPU and a 128-core NVIDIA Maxwell GPU, which has powerful deep learning computing capabilities and can realize real-time image processing and algorithm inference; it has 16GB of built-in eMMC storage space for storing pre-trained models, image cache and detection logs; it supports edge computing, without relying on cloud servers, ensuring the real-time performance of detection and data security.

[0056] Core Working Principle (Algorithm Innovation - Pain Point Solution and Precise Matching): Through fully automated processing of "image acquisition - preprocessing - feature extraction - liquid level recognition - result output", the core innovation lies in addressing the pain point of low recognition accuracy of traditional edge detection algorithms (Canny, Sobel) in conditions of "similar material colors and fluctuating liquid levels". An improved U-Net deep learning algorithm is designed to achieve accurate recognition and dynamic tracking of layered interfaces. The specific process is as follows: 1) Image Acquisition and Preprocessing (specifically addressing the issues of "numerous image interference and distortion affecting accuracy"): Industrial cameras acquire images at a frame rate of 30fps. The preprocessing stage incorporates multiple optimization algorithms: ① Grayscale conversion reduces data volume; ② 5×5 Gaussian filtering removes noise such as ambient light fluctuations and glass reflections; ③ Histogram equalization enhances the contrast of layered interfaces and resolves interface blurring caused by color differences in different materials; ④ ROI cropping and perspective transformation correction remove irrelevant backgrounds such as supports and light sources, correct image distortion caused by installation angles, and ensure the accuracy of the detection area.

[0057] 2) Improved Feature Extraction and Recognition of the U-Net Algorithm (Specifically addressing the issues of inaccurate layered interface recognition and poor adaptability to multiple materials): An attention mechanism (SE module) is introduced to enhance the ability to extract features from layered interfaces, solving the problem of traditional algorithms being insensitive to weak interface signals; a transfer learning method is adopted, and the feature extraction network is initialized based on a pre-trained ResNet50 model, reducing the amount of training data and improving the model's generalization ability to different materials (oil, water, alcohols, corrosive liquids), addressing the pain point of "poor adaptability of a single algorithm". After training on 10,000+ multi-condition images, the model achieves a layered interface recognition accuracy of 99.2%, a liquid level height detection error of ≤0.5mm, and accurately outputs real-time liquid level height (0.1mm accuracy), layered interface coordinates, and liquid level stability judgment (whether the fluctuation amplitude exceeds 1mm). The specific implementation is as follows: Model Training: A dataset of over 10,000 liquid level images under different working conditions (different materials, different liquid levels, and different lighting conditions) was constructed. Each image was labeled, including the liquid level and the location of the layered interface. The dataset was divided into training and testing sets in an 8:2 ratio and input into the improved U-Net model for training. For model optimization, an attention mechanism (SE module) was introduced to enhance the ability to extract features from the layered interface. At the same time, a transfer learning method was adopted, and the feature extraction network was initialized based on the pre-trained ResNet50 model to reduce the amount of training data and improve the model's convergence speed and generalization ability. The trained model achieved a layered interface recognition accuracy of 99.2% and a liquid level detection error of ≤0.5mm.

[0058] Real-time recognition: The pre-processed image is input into the trained deep learning model. The model extracts image features (such as grayscale gradient, edge contour, and color difference) through the encoder, and restores the feature map through the decoder to output the liquid level recognition result, including real-time liquid level height (accurate to 0.1mm), layer interface coordinates, and liquid level stability judgment (whether the fluctuation amplitude exceeds 1mm). For slight fluctuations that may occur in the liquid surface during the liquid distribution process, a Kalman filter algorithm is introduced to smooth the recognition result, eliminate the impact of fluctuations on the detection accuracy, and ensure the stability of the output data.

[0059] 3) Critical Liquid Level Judgment and Signal Feedback (Specifically addressing the issues of "lagging critical liquid level response and low separation accuracy"): The PLC presets critical separation liquid level thresholds for different materials, and the AI ​​unit compares the measured liquid level with the threshold in real time: When the liquid level drops to 5mm above the critical threshold (early warning threshold), a "approaching critical liquid level" signal is sent, triggering the valve to gradually reduce its opening; when the critical threshold is reached, a "reached critical liquid level" signal is immediately sent, triggering the valve to close; if an interface fluctuation is detected to exceed 2mm, a "liquid level unstable" signal is sent, controlling the valve to pause adjustment and restarting after stabilization, avoiding stratification disorder caused by fluctuations and ensuring separation accuracy.

[0060] 7. PLC main control system An industrial-grade PLC controller (Siemens S7-1200 series) is selected as the core control unit of the system. It has a built-in preset control logic program to achieve coordinated control of various modules. Specific functions include: receiving signals from each module (stirring speed, rotation angle, settling time, liquid level, valve angle, etc.); outputting control commands according to preset procedures (stirring start / stop and speed adjustment, rotary motor start / stop and angle control, settling timer start, valve opening / closing angle adjustment, alarm signal output, etc.); and interacting with each module via an industrial bus to ensure the real-time performance and accuracy of control commands.

[0061] 8. Redundant power supply module It adopts a dual power supply design (main power supply + backup power supply). The main power supply is an AC220V to DC24V switching power supply, and the backup power supply is a lithium battery pack. It is equipped with a power monitoring unit to monitor the voltage status of the main power supply in real time. When the main power supply fails, it automatically switches to the backup power supply with a switching time of ≤10ms, ensuring continuous and stable operation of the system and avoiding interruption of the liquid separation process due to power failure.

[0062] III. Control Methods (Multi-stage Separation Process) This invention achieves multi-stage automated control of "liquid addition → stirring → rotation → settling → separation → reset" through preset control logic in the PLC main control system. The specific process is as follows: 1. Liquid addition stage: Open the ground glass stopper at the top of the customized round bottle separatory funnel, add the material to be separated into the funnel through the liquid addition bottle mouth, control the liquid addition amount according to the scale line on the bottle body, and close the ground glass stopper after the liquid addition is completed.

[0063] 2. Stirring stage: The PLC main control system sends a command to start the magnetic coupling stirring module, adjusts the stirring speed according to the material characteristics, drives the stirring magnet to rotate to achieve material mixing, and the stirring time is preset to 5-30 minutes (can be adjusted according to needs).

[0064] 3. Rotation-Station Stage: After stirring stops, the PLC sends a command to start the high-precision rotation positioning module, which controls the rotation motor of the separation device (customized round bottle separating funnel) to rotate 90° to the left and accurately position it; then the upward servo valve (corresponding branch valve) is opened and closed after 3 seconds, waiting for the intelligent standing timer module to complete the preset standing time (to achieve material stratification).

[0065] 4. Liquid Distribution Stage: The core of this stage is based on AI-based visual liquid level detection feedback. A PLC is used to achieve dynamic closed-loop control of "liquid level judgment - opening adjustment," clearly defining the judgment conditions and corresponding actions to ensure a balance between efficiency and accuracy. The specific process is as follows: Startup detection: After settling, the PLC sends a command to start the AI ​​vision liquid level detection module (camera + ring light source). The industrial camera collects liquid level images in the funnel in real time at a frame rate of 30fps. After being identified by the AI ​​processing unit, the three core parameters of "real-time liquid level height, fluctuation amplitude of the layer interface, and whether it is close to the critical liquid level" are fed back to the PLC (transmission delay ≤10ms).

[0066] Valve start-up logic: The PLC first controls the "upward steering valve" to open (to balance the air pressure in the funnel and avoid negative pressure causing the liquid separation to be interrupted), and then selects the corresponding discharge branch according to the preset liquid separation volume requirement (select the 20mm right branch for large liquid separation volume and the 16mm left branch for small liquid separation volume), and controls the "downward steering valve" of that branch to start the liquid separation.

[0067] Dynamic adjustment rules (judgment conditions → specific actions): ① High-efficiency liquid separation range judgment: real-time liquid level > 60mm (far from the critical separation interface), and fluctuation range ≤ 1mm (liquid level stable); corresponding action: PLC calls the liquid level-valve mapping table to control the valve opening to 60°-90° (the higher the liquid level, the larger the opening, such as 80° opening when 120mm, and 60° opening when 80mm), to maximize liquid separation efficiency; ② Judgment of transition adjustment range: The real-time liquid level is between 50-60mm (close to the critical stratification interface), or the fluctuation range is 1-2mm (slight fluctuation of liquid level); corresponding action: PLC controls the valve opening to gradually decrease to 30°-60° (e.g., 50° opening when 60mm, 30° opening when 55mm), while reducing the adjustment step size (from 5° / time to 1° / time) to reduce the disturbance of flow rate to the stratification interface; ③ Precise volume control range judgment: Real-time liquid level ≤ 50mm (near the critical stratification interface), or receive the AI ​​module's "approaching critical liquid level" warning signal (liquid level = critical threshold + 5mm); Corresponding action: Adjust the valve opening to 10°-30° (e.g., 15° opening when 50mm), and reduce the flow rate to 0.3-0.5mL / s to ensure the stability of the stratification interface; ④ Emergency Pause Judgment: Fluctuation amplitude > 2mm (violent liquid level fluctuation); Corresponding Action: The PLC immediately sends a "valve pause adjustment" command, maintains the current opening, and resumes adjustment after the AI ​​module reports "fluctuation amplitude ≤ 1mm"; ⑤ Liquid separation stop judgment: Real-time liquid level = critical liquid level threshold (e.g., 50mm for oil-water system), or receive the "reached critical liquid level" signal from the AI ​​module; Corresponding action: PLC immediately sends a command to close the "downward servo valve" to stop liquid separation, and at the same time closes the "upward servo valve".

[0068] 5. Reset Phase: Close the upward servo valve, and the PLC controls the rotary motor to return to the zero position (initial position); then control the rotary motor to rotate 90° to the right, opening the upward servo valve and the corresponding downward servo valve to discharge the residual liquid (preset discharge time); after the discharge is completed, close all servo valves, and the rotary motor returns to the zero position, completing one complete liquid separation process.

[0069] The core highlights are: the AI ​​vision liquid level detection module collects liquid level data in real time, and the PLC adaptively adjusts the opening and closing angle of the servo valve according to the liquid level height, so as to achieve both "high-efficiency liquid distribution" and "high-precision volume control"; at the same time, by switching between two branches with different diameters on the left and right, it can adapt to different liquid distribution volume requirements.

[0070] The specific implementation is as follows: (I) Equipment selection and assembly 1. Customized round bottle separatory funnel: Customized according to design parameters, the main body is a 1000ml high borosilicate 3.3 glass standard mouth round bottom flask (24 / 29# interface), the separatory part is a 500ml pear-shaped structure with a built-in PTFE piston; the left and right branches are adapted to 16mm and 20mm discharge interfaces, each equipped with an MG996R high-pressure servo motor, which, together with the ball screw, forms a valve drive mechanism; the bottom is equipped with an arc-shaped flow guide structure (30° flow guide angle).

[0071] 2. Magnetic coupling stirring module: An adjustable speed magnetic stirrer (model: IKA RCT basic) is selected. The stirring magnet is made of polytetrafluoroethylene (20mm in diameter) and is installed in the center of the funnel. The external stirrer is coaxially aligned with the funnel.

[0072] 3. High-precision rotary positioning module: 28 stepper motors (torque 2.5 N·m) are selected, along with a 50:1 reducer. The angle encoder is an incremental encoder (resolution 1024 lines). The positioning bracket is made of aluminum alloy to ensure that the rotation axis coincides with the center of gravity of the funnel.

[0073] 4. AI Visual Liquid Level Detection Module: The industrial camera used is Hikvision MV-CA013-21UM (1.3 million pixels), the ring light source is an LED ring light source (wavelength 500-600nm), and the AI ​​image processing unit uses an NVIDIA Jetson Nano development board with a pre-trained liquid level recognition model.

[0074] 5. PLC main control system: Siemens S7-1214C PLC is selected, equipped with Ethernet module and industrial bus interface to realize data interaction with each module; the redundant power supply module is Mean Well AC220V to DC24V switching power supply (output power 100W), and the backup lithium battery pack is 12V / 10Ah.

[0075] 6. System Assembly: Fix each module in its designed position, and connect the PLC to the magnetically coupled stirring module, rotary positioning module, servo valve module, AI vision inspection module, and redundant power supply module via industrial bus; debug the communication links of each module to ensure normal data transmission.

[0076] (II) Parameter Preset and Adjustment Preset control parameters using PLC programming software: stirring speed 300 r / min, stirring time 10 min; left rotation angle 90°, settling time 15 min; right rotation angle 90°, residual liquid discharge time 2 min; critical liquid level height threshold (set according to the stratification characteristics of the separated materials, such as 50 mm from the bottom of the funnel).

[0077] Debug the AI ​​visual liquid level detection module: Start the industrial camera and ring light source to take standard images of different liquid level heights, calibrate the liquid level recognition model to ensure that the liquid level detection accuracy is ≤0.5mm; debug the critical liquid level recognition algorithm to ensure accurate identification of material layer interfaces.

[0078] Debug the servo valve module: Send control commands at different angles via PLC to test the servo response speed (ensure ≤0.2s) and angle accuracy (ensure ±0.5°); test the sealing performance of the valve opening and closing process to ensure no leakage.

[0079] Debug the rotary positioning module: control the rotary motor to rotate 90° to the left and right, calibrate the rotation angle through the feedback signal of the angle encoder, and ensure the positioning accuracy is ±0.1°; test the stability of the rotation process to ensure that the center of gravity offset is ≤2mm.

[0080] (III) Actual Operation Process Adding liquid: Open the ground glass stopper and add 800ml of the material to be separated (such as an oil-water mixture) into the customized round bottle separatory funnel, then close the glass stopper.

[0081] Stirring: Start the system, and the PLC sends a command to start the magnetically coupled stirring module, which stirs for 10 minutes at a preset speed of 300 r / min to achieve thorough mixing of oil and water.

[0082] Rotation-Station: After stirring stops, the PLC controls the rotary motor to rotate 90° to the left and position it; the upward servo valve opens and closes after 3 seconds; the intelligent stationary timing module starts, and the system is stationary for 15 minutes to achieve oil-water separation.

[0083] Liquid separation: After settling, the PLC activates the AI ​​vision detection module, turns on the ring light source, and the industrial camera captures real-time liquid level images. The AI ​​image processing unit identifies the liquid level height (the initial liquid level is relatively high, such as 120mm from the bottom) and feeds it back to the PLC. The PLC controls the opening of the upward servo valve and the downward right branch servo valve (20mm interface, suitable for larger liquid separation volumes), and controls the valve opening angle to 80° to improve liquid separation efficiency. As the liquid level drops, the AI ​​vision module provides real-time feedback on the liquid level data, and the PLC gradually reduces the valve opening angle (e.g., when the liquid level drops to 80mm, the angle is adjusted to 50°; when the liquid level drops to 60mm, the angle is adjusted to 30°). When the liquid level is detected to reach the critical level (50mm, oil-water separation interface), the PLC immediately sends a command to close the downward servo valve and stop liquid separation. At this time, the liquid separation volume is 310ml (meeting the design requirement of 315mL±5mL).

[0084] Reset: Close the upward servo valve, and the PLC controls the rotary motor to return to the zero position; then control the rotary motor to rotate 90° to the right, open the upward servo valve and the downward left branch servo valve (16mm interface), and discharge the residual liquid for 2 minutes; after the discharge is completed, close all valves, the rotary motor returns to the zero position, and one liquid separation process ends.

[0085] (iv) Verification of operational effectiveness This operation achieved precise separation of the oil-water mixture, with a separation volume of 310ml, meeting the design requirements; the residual liquid volume was 0.3ml (≤0.5ml); the purity of the oil phase after separation was ≥99.5%, and the purity of the water phase was ≥99.8%; the entire process was automated, requiring no manual intervention; the system operated stably, with no leaks or jamming; during the power outage test, the redundant power supply module automatically switched, and the system operated without interruption, verifying the reliability and practicality of the system.

[0086] Precise separation of oil-water mixtures: To verify the system's versatility and stability, experiments were conducted on three typical material scenarios: low-viscosity non-corrosive, medium-viscosity non-corrosive, and corrosive. All experiments were run with the same preset parameters (stirring 300 rpm for 10 min, settling for 15 min, critical liquid level 50 mm). The verification results are shown below, fully demonstrating the system's adaptability: Experimental scenario (material type) Separation volume (target / actual) Liquid residue Separation purity (light phase / heavy phase) Low viscosity, non-corrosive (oil-water mixture: oil phase density 0.85 g / cm³, water phase density 1.0 g / cm³) 315mL / 310mL (deviation -1.6%) 0.3mL (≤0.5mL) ≥99.5% / ≥99.8% Medium viscosity, non-corrosive (ethanol-glycerol mixture: viscosity 35 mPa·s) 315mL / 313mL (deviation -0.6%) 0.4mL (≤0.5mL) ≥99.3% / ≥99.6% Corrosive system (5% dilute hydrochloric acid-organic solvent mixture) 315mL / 312mL (deviation -1.0%) 0.35mL (≤0.5mL) ≥99.4% / ≥99.7% Common validation conclusions: The entire process is fully automated, requiring no manual intervention; it meets the design requirements of "residual amount ≤0.5mL, separation purity ≥99.3%" under different material scenarios; in the redundant power supply test, the backup power supply switched within 10ms when the main power supply failed, and the system did not interrupt operation. The validation results show that the system is adaptable to the separation needs of materials with different viscosities and corrosiveness, demonstrating strong versatility and high reliability.

[0087] In summary, by employing the customized round bottle dispensing funnel with its center of gravity coinciding with the axis of rotation and a dual-discharge branch design, the equipment achieves more stable rotation and operation with minimal liquid residue. It can also flexibly adapt to different dispensing volumes, solving the problems of center of gravity shift, excessive residue, and poor adaptability inherent in traditional equipment. Furthermore, by introducing an AI visual liquid level detection module based on an improved U-Net algorithm, non-contact, high-precision (≤0.5mm) real-time identification of layered interfaces is achieved, overcoming the drawbacks of large errors in manual visual inspection and easy contamination in contact detection. Finally, by constructing a closed-loop control system composed of a ball screw servo valve and a PLC, and dynamically adjusting according to a pre-calibrated liquid level-valve opening mapping relationship, the dispensing process can be automatically optimized between high efficiency and high precision, completely changing the rigid control mode of traditional fixed valve angles that cannot balance efficiency and precision. Ultimately, by integrating multiple modules such as stirring, rotation, settling, and separation under PLC control, the entire process is automated and closed-loop, significantly reducing labor intensity and human error, and improving the overall reliability and separation effect of the system.

[0088] 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 multi-stage liquid separation and adaptive liquid level detection system, characterized in that, include: A customized round bottle dispensing funnel serving as the core execution unit; A high-precision rotary positioning module for driving the separation funnel to rotate, the rotation axis of which coincides with the center of gravity of the separation funnel; An AI visual liquid level detection module for non-contact detection of liquid level and layering interface in the separating funnel; At least one adaptive servo valve module connected to the outlet of the separating funnel includes a servo motor, a ball screw transmission mechanism driven by the servo motor, a valve core, and a position sensor for detecting the valve opening degree. A main control system is also included, which is connected to the AI ​​visual liquid level detection module and the adaptive servo valve module respectively. The system receives real-time liquid level information from the AI ​​visual liquid level detection module and controls the servo motor to drive the valve core to adjust the opening according to the preset liquid level-valve opening mapping relationship.

2. The multi-stage liquid separation and adaptive liquid level detection system according to claim 1, characterized in that, The customized round bottle dispensing funnel includes a main bottle body, a dispensing section located at the lower part of the bottle body, and two symmetrically arranged discharge branches on the left and right sides of the upper part of the bottle body; each of the two discharge branches integrates an independent adaptive servo valve module, and the discharge interfaces of the two branches have different diameters.

3. The multi-stage liquid separation and adaptive liquid level detection system according to claim 1, characterized in that, The AI ​​visual liquid level detection module includes an industrial camera, a ring light source, and an AI image processing unit. The AI ​​image processing unit has a built-in deep learning model based on an improved U-Net architecture and an attention mechanism, which is used to process the images captured by the industrial camera to identify the liquid level layer interface and output liquid level height and fluctuation amplitude data.

4. The multi-stage liquid separation and adaptive liquid level detection system according to claim 1, characterized in that, The main control system presets a segmented mapping relationship between liquid level and valve opening, which includes at least the following: when the real-time liquid level is higher than a first preset value, a larger opening is corresponding to the liquid level; when the real-time liquid level is between the first preset value and a lower second preset value, a smaller opening is corresponding to the liquid level; and when the real-time liquid level reaches a critical separation threshold, the valve is controlled to close.

5. The multi-stage liquid separation and adaptive liquid level detection system according to claim 1, characterized in that, The main control system is also configured to: when the liquid level fluctuation amplitude reported by the AI ​​visual liquid level detection module exceeds a preset fluctuation threshold, control the adaptive servo valve module to pause the opening adjustment.

6. The multi-stage liquid separation and adaptive liquid level detection system according to claim 1, characterized in that, It also includes a magnetically coupled stirring module with a stirring magnet placed inside the separating funnel; and an intelligent settling timer module; the main control system is configured to sequentially control the magnetically coupled stirring module, the high-precision rotary positioning module, the intelligent settling timer module and the adaptive servo valve module to work together.

7. The multi-stage liquid separation and adaptive liquid level detection system according to claim 1, characterized in that, The system also includes a redundant power supply module, which includes a main power supply, a backup power supply and a power monitoring unit, for automatically switching to the backup power supply when the main power supply fails.

8. A multi-stage liquid separation control method using the system described in any one of claims 1-7, characterized in that, Includes the following steps: S1 Stirring Step: Control the magnetically coupled stirring module to stir the material added to the separatory funnel; S2 Rotation and Settling Step: Control the high-precision rotary positioning module to drive the separating funnel to rotate to the specified angle and allow it to settle so that the material separates into layers; S3 Adaptive Liquid Separation Steps: The AI ​​visual liquid level detection module is activated to perform real-time liquid level detection. The main control system dynamically queries the preset liquid level-valve opening mapping table based on the detected real-time liquid level information, and controls the opening of the adaptive servo valve module accordingly to perform liquid separation until the liquid level reaches the critical separation threshold and the valve is closed.

9. The control method according to claim 8, characterized in that, In the adaptive liquid separation step, dynamically controlling the valve opening specifically includes: When the real-time liquid level is greater than the first threshold and the liquid level is stable, the control valve operates at the first maximum opening degree. When the real-time liquid level drops to between the first threshold and the second threshold, the valve opening is gradually reduced to the second smaller opening, and the opening adjustment step size is reduced. When the real-time liquid level reaches or approaches the critical separation threshold, the control valve is closed or kept at a small opening and low flow rate.

10. The control method according to claim 9, characterized in that, If the liquid level fluctuation exceeds the preset value during the liquid separation process, the valve opening adjustment will be paused and will continue to be adjusted after the liquid level returns to stability.

Images